1 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3 * gimplify.c (nonlocal_vlas): Delete.
4 (nonlocal_vla_vars): Likewise.
5 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
7 (gimplify_body): Do not create and destroy nonlocal_vlas.
8 * tree-nested.c: Include diagnostic.h.
9 (use_pointer_in_frame): Tweak.
10 (lookup_field_for_decl): Add assertion and declare the transformation.
11 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
12 internal error when the reference is in a wrong context. Do not
13 create a debug decl by default.
14 (note_nonlocal_block_vlas): Delete.
15 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
16 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
17 create a debug decl by default.
18 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
19 call is in a wrong context.
20 (fixup_vla_decls): New function.
21 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
22 debug variables were created.
23 * tree.c (decl_value_expr_lookup): Add checking assertion.
24 (decl_value_expr_insert): Likewise.
26 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
29 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
30 * tree-scalar-evolution.c (interpret_expr): Likewise.
31 (expression_expensive_p): Likewise.
32 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
33 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
34 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
35 (ssa_defined_by_minus_one_stmt_p): New.
37 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
40 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
41 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
42 * dojump.c (do_jump): Likewise.
43 * expr.c (expand_expr_real_2): Check operand type's sign.
44 * fold-const.c (const_unop): Handle ABSU_EXPR.
45 (fold_abs_const): Likewise.
46 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
47 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
48 (strip_sign_op_1): Likesise.
49 * match.pd: Add new pattern to generate ABSU_EXPR.
50 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
51 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
52 * tree-eh.c (operation_could_trap_helper_p): Likewise.
53 * tree-inline.c (estimate_operator_cost): Likewise.
54 * tree-pretty-print.c (dump_generic_node): Likewise.
55 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
56 * tree.def (ABSU_EXPR): New.
58 2018-06-16 Jakub Jelinek <jakub@redhat.com>
61 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
62 documented as preserved for backward compatibility only.
63 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
65 PR rtl-optimization/86108
66 * bb-reorder.c (create_forwarder_block): Renamed to ...
67 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
68 jump from new landing pad to the second part.
69 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
72 2018-06-15 Jakub Jelinek <jakub@redhat.com>
75 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
76 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
77 Only call store_expr for halves if the mode is the same.
80 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
81 Fix up comment formatting.
83 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
85 * typed-splay-tree.h (typed_splay_tree::remove): New function.
86 (typed_splay_tree::closure,
87 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
88 (typed_splay_tree::typed_splay_tree,
89 typed_splay_tree::operator =): Declared private.
90 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
91 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
92 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
93 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
94 typed_splay_tree::splay_tree_splay,
95 typed_splay_tree::splay_tree_foreach_helper,
96 typed_splay_tree::splay_tree_insert,
97 typed_splay_tree::splay_tree_remove,
98 typed_splay_tree::splay_tree_lookup,
99 typed_splay_tree::splay_tree_predecessor,
100 typed_splay_tree::splay_tree_successor,
101 typed_splay_tree::splay_tree_min,
102 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
103 (typed_splay_tree::root, typed_splay_tree::comp,
104 typed_splay_tree::delete_key,
105 typed_splay_tree::delete_value): New data members.
106 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
107 typed_splay_tree::remove.
109 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
111 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
112 -mginv and -mno-ginv to the assembler.
113 * config/mips/mips.opt (-mcrc): New option.
115 * doc/invoke.text (-mcrc): Document.
118 2018-06-15 Nick Clifton <nickc@redhat.com>
121 * tree.c (escaped_string): New class. Converts an unescaped
122 string into its escaped equivalent.
123 (warn_deprecated_use): Use the new class to convert the
124 deprecation message, if present.
125 (test_escaped_strings): New self test.
126 (test_c_tests): Add test_escaped_strings.
127 * doc/extend.texi (deprecated): Add a note that the
128 deprecation message is affected by the -fmessage-length
129 option, and that control characters will be escaped.
130 (#pragma GCC error): Document this pragma.
131 (#pragma GCC warning): Likewise.
132 * doc/invoke.texi (-fmessage-length): Document this option's
133 effect on the #warning and #error preprocessor directives and
134 the deprecated attribute.
136 2018-06-15 Richard Biener <rguenther@suse.de>
138 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
139 here, also noting vector size used.
140 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
141 size used in MSG_OPTIMIZED_LOCATIONS dump.
142 (pass_slp_vectorize::execute): Adjust.
144 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
147 * config/arc/arc.c (arc_return_address_register): Fix
150 2018-06-15 Richard Biener <rguenther@suse.de>
153 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
154 leave useless conversion stripping to force_gimple_operand_gsi.
155 (gimplify_build2): Likewise.
156 (gimplify_build1): Likewise.
158 2018-06-15 Richard Biener <rguenther@suse.de>
161 * tree-cfg.c (move_stmt_op): unshare invariant addresses
162 before adjusting their block.
164 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
166 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
167 multilibs for *-*-rtems*.
168 * config/riscv/t-rtems: New file.
170 2018-06-14 Jakub Jelinek <jakub@redhat.com>
173 * match.pd ((A +- CST1) +- CST2): Punt if last resort
174 unsigned_type_for returns NULL.
177 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
178 subregs of multi-word pseudos unless the float mode has word size.
180 2018-06-14 Richard Biener <rguenther@suse.de>
183 * tree-vect-generic.c (build_word_mode_vector_type): Remove
184 duplicate and harmful type_hash_canon.
185 * tree.c (type_hash_canon): Assert we didn't find ourselves.
187 2018-06-14 Richard Biener <rguenther@suse.de>
190 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
193 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
195 * config/rtems.h (STDINT_LONG32): Define.
197 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
198 Prachi Godbole <prachi.godbole@imgtec.com>
200 * config/mips/mips-cpus.def: Define P6600.
201 * config/mips/mips-tables.opt: Regenerate.
202 * config/mips/mips.c (mips_ucbranch_type): New enum.
203 (mips_rtx_cost_data): Add support for P6600.
204 (mips_issue_rate): Likewise.
205 (mips_multipass_dfa_lookahead): Likewise.
206 (mips_avoid_hazard): Likewise.
207 (mips_reorg_process_insns): Likewise.
208 (mips_classify_branch_p6600): New function.
209 * config/mips/mips.h (TUNE_P6600): New define.
210 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
211 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
212 * config/mips/mips.md: Include p6600.md.
213 (processor): Add p6600.
214 * config/mips/p6600.md: New file.
215 * doc/invoke.texi: Add p6600 to supported architectures.
217 2018-06-13 Martin Sebor <msebor@redhat.com>
219 PR tree-optimization/86114
220 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
222 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
224 2018-06-13 Richard Biener <rguenther@suse.de>
226 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
227 Properly set vector type of the intermediate stmt.
228 * tree-vect-stmts.c (vectorizable_operation): The destination
229 var always has vectype_out type.
231 2018-06-13 Jeff Law <law@redhat.com>
233 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
234 integer 0 for argument to print_rtl_with_bb.
235 (rl78_reorg): Likewise.
237 2018-06-13 David Malcolm <dmalcolm@redhat.com>
239 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
240 from rtx to rtx_insn *.
241 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
243 (add_sched_insns_for_speculation): Likewise for local "target",
244 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
245 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
246 from rtx_insn ** to rtx_code_label **.
247 (reorg_emit_nops): Likewise.
248 (c6x_reorg): Likewise for local "call_labels".
249 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
251 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
252 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
253 the loops over LABEL_REFS.
254 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
256 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
257 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
258 (split_branches): Strengthen local "olabel" from rtx to
259 rtx_insn *, adding a safe_as_a cast.
260 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
262 (add_insn_after): Likewise for first two params.
263 (add_insn_before): Likewise.
264 (remove_insn): Likewise for param.
265 (emit_pattern_before_noloc): Likewise for second and third params.
266 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
267 (emit_call_insn_before_noloc): Likewise.
268 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
270 (emit_barrier_before): Likewise.
271 (emit_label_before): Strengthen "label" param from "rtx" to
272 "rtx_code_label *". Strengthen "before" param from "rtx" to
274 (emit_insn_after_1): Strengthen "after" param from "rtx" to
276 (emit_pattern_after_noloc): Likewise.
277 (emit_insn_after_noloc): Likewise.
278 (emit_jump_insn_after_noloc): Likewise.
279 (emit_call_insn_after_noloc): Likewise.
280 (emit_debug_insn_after_noloc): Likewise.
281 (emit_barrier_after): Likewise.
282 (emit_label_after): Likewise for both params.
283 (emit_pattern_after_setloc): Likewise for "after" param. Convert
284 "loc" param from "int" to "location_t".
285 (emit_insn_after_setloc): Likewise.
286 (emit_jump_insn_after_setloc): Likewise.
287 (emit_call_insn_after_setloc): Likewise.
288 (emit_debug_insn_after_setloc): Likewise.
289 (emit_pattern_before_setloc): Likewise for "before" param. Convert
290 "loc" param from "int" to "location_t".
291 (emit_pattern_before): Convert NULL_RTX to NULL.
292 (emit_insn_before_setloc): Convert "loc" param from "int" to
294 (emit_jump_insn_before_setloc): Likewise.
295 (emit_call_insn_before_setloc): Likewise.
296 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
297 rtx_insn *. Convert "loc" param from "int" to "location_t".
298 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
299 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
300 Convert 3rd param from "int" to "location_t".
301 (emit_barrier_before, emit_barrier_after, next_real_insn):
302 Strengthen param from rtx to rtx_insn *.
303 (emit_label_before): Strengthen 1st param from "rtx" to
304 "rtx_code_label *". Strengthen 2nd param from "rtx" to
306 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
307 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
308 Strengthen 2nd param from "rtx" to "rtx_insn *".
309 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
310 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
311 Likewise. Convert 3rd param from "int" to "location_t".
312 (emit_label_after): Strengthen 1st param from "rtx" to
314 (next_real_insn, remove_insn): Strengthen param from "rtx" to
316 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
317 from "rtx" to "rtx_insn *".
319 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
321 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
322 bodies streamed in with -Q.
323 * dumpfile.c (dump_files): Add lto-stream-out dump file.
324 * dumpfile.h (tree_dump_index): Add lto_stream_out.
325 * gimple-streamer-out.c: Include gimple-pretty-print.h
326 (output_bb): Dump stmts streamed.
327 * lto-section-out.c: Include print-tree.h
328 (lto_begin_section): Dump sections created.
329 (lto_output_decl_index): Dump decl encoded.
330 * lto-streamer-out.c: Include print-tree.h
331 (create_output_block): Dump output block created.
332 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
333 (output_function): Dump function output.
334 (output_constructor): Dump constructor streamed.
335 (write_global_stream): Output indexes encoded.
336 (produce_asm_for_decls): Dump streams encoded.
337 * lto-streamer.c (streamer_dump_file): New global var.
338 * lto-streamer.h (streamer_dump_file): Declare.
339 * passes.c (ipa_write_summaries): Initialize streamer dump.
340 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
343 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
346 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
347 offsets for register save directives. Emit a second batch of save
348 directives, if need be, when the function accesses prior frames.
350 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
352 * config/arc/fpu.md (fmasf4): Force operand to register.
355 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
357 * config/arc/arc-protos.h (arc_pad_return): Remove.
358 * config/arc/arc.c (machine_function): Remove force_short_suffix
360 (arc_print_operand): Adjust printing of '&'.
361 (arc_verify_short): Remove conditional printing of short suffix.
362 (arc_final_prescan_insn): Remove reference to size_reason.
363 (pad_return): New function.
364 (arc_reorg): Call pad_return.
365 (arc_pad_return): Remove.
366 (arc_init_machine_status): Remove reference to force_short_suffix.
367 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
368 (attr length): When attribute iscompact is true force to 2
369 regardless; in the case of maybe check if we want to force the
370 instruction to have 4 bytes length.
371 (nopv): Change it to generate 4 byte long nop as well.
372 (blockage): New pattern.
373 (simple_return): Remove call to arc_pad_return.
374 (p_return_i): Likewise.
376 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
378 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
380 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
382 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
385 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
387 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
388 for ARC700 and ARCv2.
390 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
393 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
394 operands[2] instead of operands[1].
397 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
399 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
400 case, check whether the outer register overlaps an unallocatable
401 register, not just whether it fits the required class.
403 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
405 * poly-int.h (can_div_trunc_p): Add new overload in which all values
407 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
408 (memrefs_conflict_p): Likewise.
409 (init_alias_analysis): Likewise.
410 * cfgexpand.c (expand_debug_expr): Likewise.
411 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
412 * cse.c (fold_rtx): Likewise.
413 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
414 * expr.c (emit_block_move_hints): Likewise.
415 (clear_storage_hints, push_block, emit_push_insn): Likewise.
416 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
417 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
418 (emit_group_store): Likewise.
419 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
420 to read the PRE/POST_MODIFY increment.
421 * calls.c (store_one_arg): Use strip_offset.
422 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
424 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
426 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
427 (simplify_binary_operation_1): Extend CONST_INT handling to
429 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
430 than a HOST_WIDE_INT.
431 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
433 (adjust_mems, add_stores): Update accodingly.
434 (vt_canonicalize_addr): Track polynomial offsets.
435 (emit_note_insn_var_location): Likewise.
436 (vt_add_function_parameter): Likewise.
437 (vt_initialize): Likewise.
439 2018-06-12 Jeff Law <law@redhat.com>
441 * config.gcc (alpha*-*-freebsd*): Remove.
442 * config/alpha/freebsd.h: Remove.
444 2018-06-12 David Malcolm <dmalcolm@redhat.com>
447 * spellcheck-tree.c (levenshtein_distance): Rename to...
448 (get_edit_distance): ...this, and update for underlying renaming.
449 * spellcheck-tree.h (levenshtein_distance): Rename to...
450 (get_edit_distance): ...this.
451 * spellcheck.c (levenshtein_distance): Rename to...
452 (get_edit_distance): ...this. Convert from Levenshtein distance
453 to Damerau-Levenshtein distance by supporting transpositions of
454 adjacent characters. Rename "v1" to "v_next" and "v0" to
456 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
457 (selftest::test_edit_distance_unit_test_oneway): ...this, and
458 update for underlying renaming.
459 (selftest::levenshtein_distance_unit_test): Rename to...
460 (selftest::test_get_edit_distance_unit): ...this, and update for
462 (selftest::test_find_closest_string): Add example from PR 69968
463 where transposition helps
464 (selftest::test_metric_conditions): Update for renaming.
465 (selftest::test_metric_conditions): Likewise.
466 (selftest::spellcheck_c_tests): Likewise.
467 * spellcheck.h (levenshtein_distance): Rename both overloads to...
468 (get_edit_distance): ...this.
469 (best_match::consider): Update for renaming.
471 2018-06-12 Martin Sebor <msebor@redhat.com>
473 PR tree-optimization/85259
474 * builtins.c (compute_objsize): Handle constant offsets.
475 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
476 true iff a warning has been issued.
477 * gimple.h (gimple_nonartificial_location): New function.
478 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
479 gimple_nonartificial_location and handle -Wno-system-headers.
480 (handle_builtin_stxncpy): Same.
482 2018-06-12 Martin Sebor <msebor@redhat.com>
485 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
487 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
489 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
490 BUILTIN_VEC_XST entries for pointer to double and long long.
492 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
495 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
497 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
500 2018-06-12 Martin Liska <mliska@suse.cz>
502 * doc/options.texi: Document IntegerRange.
504 2018-06-12 Martin Liska <mliska@suse.cz>
506 * config/i386/i386.opt: Make MPX-related options as Deprecated.
507 * opt-functions.awk: Handle Deprecated flag.
508 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
510 (read_cmdline_option): Report warning for a deprecated option.
511 * opts.h (struct cl_option): Add new field cl_deprecated.
512 (CL_ERR_DEPRECATED): New.
514 2018-06-12 Martin Liska <mliska@suse.cz>
516 * doc/options.texi: Document Deprecated option flag.
518 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
520 * config/arc/arc-arch.h (arc_extras): New enum.
521 (arc_cpu_t):Add field extra.
522 (arc_cpu_types): Consider the extras.
523 * config/arc/arc-cpus.def: Add extras info.
524 * config/arc/arc-opts.h (processor_type): Consider extra field.
525 * config/arc/arc.c (arc_override_options): Handle extra field.
527 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
529 * config/arc/arc-arch.h: Update ARC_OPTX macro.
530 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
532 * config/arc/arc.c (arc_init): Update pic warning.
533 (irq_range): Update irq range parsing warnings.
534 (arc_override_options): Update various warning messages.
535 (arc_handle_aux_attribute): Likewise.
537 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
539 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
541 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
543 * doc/sourcebuild.texi: Document usage of line number 0 in verify
544 compiler messages directives.
546 2018-06-12 Matthew Fortune <mfortune@gmail.com>
548 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
549 * config/mips/mips-tables.opt: Regenerate.
550 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
552 * doc/invoke.texi: Document -march=i6500.
554 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
556 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
557 (i6400_gpmul): Add cpu_unit.
558 (i6400_gpdiv): Likewise.
559 (i6400_msa_add_d): Update reservations.
560 (i6400_msa_int_add) Likewise.
561 (i6400_msa_short_logic3) Likewise.
562 (i6400_msa_short_logic2) Likewise.
563 (i6400_msa_short_logic) Likewise.
564 (i6400_msa_move) Likewise.
565 (i6400_msa_cmp) Likewise.
566 (i6400_msa_short_float2) Likewise.
567 (i6400_msa_div_d) Likewise.
568 (i6400_msa_long_logic1) Likewise.
569 (i6400_msa_long_logic2) Likewise.
570 (i6400_msa_mult) Likewise.
571 (i6400_msa_long_float2) Likewise.
572 (i6400_msa_long_float4) Likewise.
573 (i6400_msa_long_float5) Likewise.
574 (i6400_msa_long_float8) Likewise.
575 (i6400_fpu_fadd): Include frint type.
576 (i6400_fpu_store): New define_insn_reservation.
577 (i6400_fpu_load): Likewise.
578 (i6400_fpu_move): Likewise.
579 (i6400_fpu_fcmp): Likewise.
580 (i6400_fpu_fmadd): Likewise.
581 (i6400_int_mult): Include imul3nc type and update reservation.
582 (i6400_int_div): Include idiv3 type and update reservation.
583 (i6400_int_load): Update to check type not move_type.
584 (i6400_int_store): Likewise.
585 (i6400_int_prefetch): Set zero latency.
587 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
589 * gcc.c: Document new %@{...} sequence.
590 (LINK_COMMAND_SPEC): Use it for the -L switches.
591 (cpp_unique_options): Use it for the -I switches.
592 (at_file_argbuf): New global variable.
593 (in_at_file): Likewise.
594 (alloc_args): Create at_file_argbuf.
595 (clear_args): Truncate at_file_argbuf.
596 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
597 (open_at_file): New function.
598 (close_at_file): Likewise.
599 (create_at_file): Delete.
600 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
603 (validate_switches_from_spec): Deal with %@{...} sequence.
604 (validate_switches): Likewise.
605 (driver::finalize): Call clear_args.
607 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
609 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
611 2018-06-11 Martin Sebor <msebor@redhat.com>
613 * doc/invoke.texi (-Wall): List -Wc++17-compat.
614 (Wno-class-memaccess): Add @opindex.
615 (Wno-templates, Wno-multiple-inheritance): Same.
616 (Wno-virtual-inheritance, Wno-namespaces): Same.
617 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
618 (Wno-format-overflow, Wno-format-truncation): Same.
619 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
620 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
621 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
622 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
623 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
624 (Wno-misspelled-isr): Same.
626 2018-06-11 Martin Sebor <msebor@redhat.com>
628 * PR tree-optimization/86083
629 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
631 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
633 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
635 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
638 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
639 on the correct operand.
640 (*movdi_internal64): Ditto.
642 2018-06-11 Martin Liska <mliska@suse.cz>
644 PR tree-optimization/86089
645 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
647 2018-06-11 Julia Koval <julia.koval@intel.com>
649 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
650 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
651 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
653 2018-06-11 Olivier Hainque <hainque@adacore.com>
655 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
656 for Ada with strict dwarf2.
658 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
661 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
664 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
666 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
668 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
670 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
673 2018-06-08 Martin Liska <mliska@suse.cz>
675 * tree-cfg.h (debug_function): Fix argument type to match
678 2018-06-08 Martin Liska <mliska@suse.cz>
680 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
681 Remove usage of MPX-related (and removed) fields.
682 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
684 2018-06-08 David Malcolm <dmalcolm@redhat.com>
686 * cfg.c (debug): Use TDF_NONE rather than 0.
687 * cfghooks.c (debug): Likewise.
688 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
689 (struct dump_option_value_info): Convert to...
690 (struct kv_pair): ...this template type.
691 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
693 (optinfo_verbosity_options): Likewise.
694 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
696 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
697 than int for "optgroup_flags" param.
698 (dump_generic_expr_loc): Use dump_flags_t rather than int for
700 (dump_dec): Likewise.
701 (dump_finish): Use TDF_NONE rather than 0.
702 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
703 rather than int for "optgroup_flags" param. Use TDF_NONE rather
704 than 0. Update for change to option_ptr.
705 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
706 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
707 0. Update for changes to optinfo_verbosity_options and
709 (opt_info_switch_p): Convert optgroup_flags from int to
711 (dump_basic_block): Use dump_flags_t rather than int
712 for "dump_kind" param.
713 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
714 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
715 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
716 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
717 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
718 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
719 TDF_NONE): Convert from macros to...
720 (enum dump_flag): ...this new enum.
721 (dump_flags_t): Update to use enum.
722 (operator|, operator&, operator~, operator|=, operator&=):
723 Implement for dump_flags_t.
724 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
725 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
726 Convert from macros to...
727 (enum optgroup_flag): ...this new enum.
728 (optgroup_flags_t): New typedef.
729 (operator|, operator|=): Implement for optgroup_flags_t.
730 (struct dump_file_info): Convert field "alt_flags" to
731 dump_flags_t. Convert field "optgroup_flags" to
733 (dump_basic_block): Use dump_flags_t rather than int for param.
734 (dump_generic_expr_loc): Likewise.
735 (dump_dec): Likewise.
736 (dump_register): Convert param "optgroup_flags" to
738 (opt_info_enable_passes): Likewise.
739 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
741 * gimple-pretty-print.c (debug): Likewise.
742 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
743 (merged_store_group::apply_stores): Likewise.
744 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
745 * gimple.c (verify_gimple_pp): Likewise.
746 * graphite-poly.c (print_pbb_body): Likewise.
747 * passes.c (pass_manager::register_one_dump_file): Convert
748 local "optgroup_flags" to optgroup_flags_t.
749 * print-tree.c (print_node): Use TDF_NONE rather than 0.
751 (debug_body): Likewise.
752 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
754 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
756 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
757 (convert_mult_to_fma): Likewise.
758 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
759 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
760 * tree-vect-data-refs.c (dump_lower_bound): Convert param
761 "dump_kind" to dump_flags_t.
763 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
765 * config/rs6000/rs6000.c (min, max): Delete.
767 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
769 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
770 -mabi=spe and -mabi=no-spe.
772 2018-06-08 Martin Liska <mliska@suse.cz>
774 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
775 where we expect an existing summary.
777 2018-06-08 Martin Liska <mliska@suse.cz>
779 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
780 * ipa-inline.h (estimate_edge_growth): Likewise.
782 2018-06-08 Martin Liska <mliska@suse.cz>
784 * cgraph.c (function_version_hasher::hash): Use
785 cgraph_node::get_uid ().
786 (function_version_hasher::equal):
787 * cgraph.h (cgraph_node::get_uid): New method.
788 * ipa-inline.c (update_caller_keys): Use
789 cgraph_node::get_uid ().
790 (update_callee_keys): Likewise.
791 * ipa-utils.c (searchc): Likewise.
792 (ipa_reduced_postorder): Likewise.
793 * lto-cgraph.c (input_node): Likewise.
794 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
795 * symbol-summary.h (symtab_insertion): Likewise.
796 (symtab_removal): Likewise.
797 (symtab_duplication): Likewise.
798 * tree-pretty-print.c (dump_function_header): Likewise.
799 * tree-sra.c (convert_callers_for_node): Likewise.
801 2018-06-08 Martin Liska <mliska@suse.cz>
803 * cgraph.c (symbol_table::create_edge): Always assign a new
805 (symbol_table::free_edge): Do not recycle numbers.
806 * cgraph.h (cgraph_edge::get): New method.
807 * symbol-summary.h (symtab_removal): Use it.
808 (symtab_duplication): Likewise.
809 (call_summary::hashable_uid): Remove.
811 2018-06-08 Martin Liska <mliska@suse.cz>
813 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
814 (initialize_growth_caches): Remove.
815 (free_growth_caches): Likewise.
816 (do_estimate_edge_time): Use edge_growth_cache.
817 (do_estimate_edge_size): Likewise.
818 (do_estimate_edge_hints): Likewise.
819 * ipa-inline.c (reset_edge_caches): Likewise.
820 (recursive_inlining): Likewise.
821 (inline_small_functions): Likewise.
822 * ipa-inline.h (initialize_growth_caches): Remove.
823 (estimate_edge_size): Likewise.
824 (estimate_edge_time): Likewise.
825 (estimate_edge_hints): Likewise.
826 (reset_edge_growth_cache): Likewise.
827 * symbol-summary.h (call_summary::remove): New method.
829 2018-06-08 Martin Liska <mliska@suse.cz>
831 * ipa-cp.c (class edge_clone_summary): New summary.
832 (grow_edge_clone_vectors): Remove.
833 (ipcp_edge_duplication_hook): Remove.
834 (class edge_clone_summary_t): New call_summary class.
835 (ipcp_edge_removal_hook): Remove.
836 (edge_clone_summary_t::duplicate): New function.
837 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
838 (create_specialized_node): Likewise.
839 (ipcp_driver): Initialize edge_clone_summaries and do not
842 2018-06-08 Martin Liska <mliska@suse.cz>
844 * symbol-summary.h (get): New function.
845 (call_summary::m_initialize_when_cloning): New class member.
847 2018-06-08 Martin Liska <mliska@suse.cz>
849 * cgraph.c (cgraph_node::remove): Do not recycle uid.
850 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
851 (symbol_table::allocate_cgraph_symbol): Do not set uid.
852 * passes.c (uid_hash_t): Record removed_nodes by their uids.
853 (remove_cgraph_node_from_order): Use the removed_nodes set.
854 (do_per_function_toporder): Likwise.
855 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
856 instead of summary_uid.
857 (symtab_removal): Likewise.
858 (symtab_duplication): Likewise.
860 2018-06-08 Martin Liska <mliska@suse.cz>
862 * ipa-cp.c (ipcp_store_bits_results): Use
863 ipcp_transformation_sum.
864 (ipcp_store_vr_results): Likewise.
865 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
867 (ipcp_transformation_initialize): ... this.
868 (ipa_set_node_agg_value_chain):
869 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
870 (write_ipcp_transformation_info): Likewise.
871 (read_ipcp_transformation_info): Likewise.
872 (ipcp_update_bits): Likewise.
873 (ipcp_update_vr): Likewise.
874 (ipcp_transform_function): Likewise.
875 * ipa-prop.h: Rename ipcp_transformation_summary to
877 (class ipcp_transformation_t): New function summary.
878 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
879 (ipa_get_agg_replacements_for_node): Likewise.
881 2018-06-08 Martin Liska <mliska@suse.cz>
883 * ipa-pure-const.c (struct funct_state_d): Do it class instead
885 (class funct_state_summary_t): New function_summary class.
886 (has_function_state): Remove.
887 (get_function_state): Likewise.
888 (set_function_state): Likewise.
889 (add_new_function): Likewise.
890 (funct_state_summary_t::insert): New function.
891 (duplicate_node_data): Remove.
892 (remove_node_data): Remove.
893 (funct_state_summary_t::duplicate): New function.
894 (register_hooks): Create new funct_state_summaries.
895 (pure_const_generate_summary): Use it.
896 (pure_const_write_summary): Likewise.
897 (pure_const_read_summary): Likewise.
898 (propagate_pure_const): Likewise.
899 (propagate_nothrow): Likewise.
900 (dump_malloc_lattice): Likewise.
901 (propagate_malloc): Likewise.
902 (execute): Do not register hooks, just remove summary
904 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
907 2018-06-08 Martin Liska <mliska@suse.cz>
909 * ipa-reference.c (remove_node_data): Remove.
910 (duplicate_node_data): Likewise.
911 (class ipa_ref_var_info_summary_t): New class.
912 (class ipa_ref_opt_summary_t): Likewise.
913 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
914 (get_reference_optimization_summary): Use
915 ipa_ref_opt_sum_summaries.
916 (set_reference_vars_info): Remove.
917 (set_reference_optimization_summary): Likewise.
918 (ipa_init): Create summaries.
919 (init_function_info): Use function summary.
920 (ipa_ref_opt_summary_t::duplicate): New function.
921 (ipa_ref_opt_summary_t::remove): New function.
922 (get_read_write_all_from_node): Fix GNU coding style.
923 (propagate): Use function summary.
924 (write_node_summary_p): Fix GNU coding style.
925 (stream_out_bitmap): Likewise.
926 (ipa_reference_read_optimization_summary): Use function summary.
927 (ipa_reference_c_finalize): Do not release hooks.
929 2018-06-08 Martin Liska <mliska@suse.cz>
931 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
932 (analyze_function_body): Extract multiple calls of get_create.
933 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
934 * ipa-inline.c (recursive_inlining): Use ::get method.
935 * ipa-inline.h (estimate_edge_growth): Likewise.
937 2018-06-08 Martin Liska <mliska@suse.cz>
939 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
941 (hsa_function_summary::hsa_function_summary): Use the new enum
943 (hsa_gpu_implementation_p): Use hsa_summaries::get.
944 * hsa-gen.c (hsa_get_host_function): Likewise.
945 (get_brig_function_name): Likewise.
946 * ipa-hsa.c (process_hsa_functions): Likewise.
947 (ipa_hsa_write_summary): Likewise.
948 * symbol-summary.h (symtab_duplication): Use ::get function/
951 2018-06-08 Martin Liska <mliska@suse.cz>
953 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
955 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
956 (hsa_register_kernel): Likewise.
957 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
958 * hsa-gen.c (hsa_get_host_function): Likewise.
959 (get_brig_function_name): Likewise.
960 (generate_hsa): Likewise.
961 (pass_gen_hsail::execute): Likewise.
962 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
963 (devirtualization_time_bonus): Likewise.
964 (ipcp_propagate_stage): Likewise.
965 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
966 (edge_set_predicate): Likewise.
967 (evaluate_conditions_for_known_args): Likewise.
968 (evaluate_properties_for_edge): Likewise.
969 (ipa_fn_summary::reset): Likewise.
970 (ipa_fn_summary_t::duplicate): Likewise.
971 (dump_ipa_call_summary): Likewise.
972 (ipa_dump_fn_summary): Likewise.
973 (analyze_function_body): Likewise.
974 (compute_fn_summary): Likewise.
975 (estimate_edge_devirt_benefit): Likewise.
976 (estimate_edge_size_and_time): Likewise.
977 (estimate_calls_size_and_time): Likewise.
978 (estimate_node_size_and_time): Likewise.
979 (inline_update_callee_summaries): Likewise.
980 (remap_edge_change_prob): Likewise.
981 (remap_edge_summaries): Likewise.
982 (ipa_merge_fn_summary_after_inlining): Likewise.
983 (ipa_update_overall_fn_summary): Likewise.
984 (read_ipa_call_summary): Likewise.
985 (inline_read_section): Likewise.
986 (write_ipa_call_summary): Likewise.
987 (ipa_fn_summary_write): Likewise.
988 (ipa_free_fn_summary): Likewise.
989 * ipa-hsa.c (process_hsa_functions): Likewise.
990 (ipa_hsa_write_summary): Likewise.
991 (ipa_hsa_read_section): Likewise.
992 * ipa-icf.c (sem_function::merge): Likewise.
993 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
994 (do_estimate_edge_time): Likewise.
995 (estimate_size_after_inlining): Likewise.
996 (estimate_growth): Likewise.
997 (growth_likely_positive): Likewise.
998 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
999 (inline_call): Likewise.
1000 * ipa-inline.c (caller_growth_limits): Likewise.
1001 (can_inline_edge_p): Likewise.
1002 (can_inline_edge_by_limits_p): Likewise.
1003 (compute_uninlined_call_time): Likewise.
1004 (compute_inlined_call_time): Likewise.
1005 (want_inline_small_function_p): Likewise.
1006 (edge_badness): Likewise.
1007 (update_caller_keys): Likewise.
1008 (update_callee_keys): Likewise.
1009 (recursive_inlining): Likewise.
1010 (inline_small_functions): Likewise.
1011 (inline_to_all_callers_1): Likewise.
1012 (dump_overall_stats): Likewise.
1013 (early_inline_small_functions): Likewise.
1014 (early_inliner): Likewise.
1015 * ipa-inline.h (estimate_edge_growth): Likewise.
1016 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1017 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1018 * ipa-prop.h (IPA_NODE_REF): Likewise.
1019 (IPA_EDGE_REF): Likewise.
1020 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1021 (propagate_malloc): Likewise.
1022 * ipa-split.c (execute_split_functions): Likewise.
1023 * symbol-summary.h: Rename get to get_create.
1025 (get_create): Likewise.
1026 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1028 2018-06-08 Martin Liska <mliska@suse.cz>
1030 * symbol-summary.h (release): Move definition out of class
1032 (symtab_removal): Likewise.
1033 (symtab_duplication): Likewise.
1035 2018-06-08 Martin Liska <mliska@suse.cz>
1037 * symbol-summary.h (function_summary): Move constructor
1038 implementation out of class declaration.
1039 (release): Likewise.
1040 (symtab_insertion): Likewise.
1041 (symtab_removal): Likewise.
1042 (symtab_duplication): Likewise.
1045 2018-06-08 Martin Liska <mliska@suse.cz>
1047 * Makefile.in: Remove support for MPX (macros, related functions,
1048 fields in cgraph_node, ...).
1049 * builtin-types.def (BT_BND): Likewise.
1050 (BT_FN_BND_CONST_PTR): Likewise.
1051 (BT_FN_CONST_PTR_BND): Likewise.
1052 (BT_FN_VOID_PTR_BND): Likewise.
1053 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1054 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1055 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1056 (expand_builtin_mempcpy_with_bounds): Likewise.
1057 (expand_builtin_memset_with_bounds): Likewise.
1058 (expand_builtin_memset_args): Likewise.
1059 (std_expand_builtin_va_start): Likewise.
1060 (expand_builtin): Likewise.
1061 (expand_builtin_with_bounds): Likewise.
1062 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1063 (DEF_LIB_BUILTIN_CHKP): Likewise.
1064 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1065 (DEF_CHKP_BUILTIN): Likewise.
1066 (BUILT_IN_MEMCPY): Likewise.
1067 (BUILT_IN_MEMMOVE): Likewise.
1068 (BUILT_IN_MEMPCPY): Likewise.
1069 (BUILT_IN_MEMSET): Likewise.
1070 (BUILT_IN_STPCPY): Likewise.
1071 (BUILT_IN_STRCAT): Likewise.
1072 (BUILT_IN_STRCHR): Likewise.
1073 (BUILT_IN_STRCPY): Likewise.
1074 (BUILT_IN_STRLEN): Likewise.
1075 (BUILT_IN_MEMCPY_CHK): Likewise.
1076 (BUILT_IN_MEMMOVE_CHK): Likewise.
1077 (BUILT_IN_MEMPCPY_CHK): Likewise.
1078 (BUILT_IN_MEMSET_CHK): Likewise.
1079 (BUILT_IN_STPCPY_CHK): Likewise.
1080 (BUILT_IN_STRCAT_CHK): Likewise.
1081 (BUILT_IN_STRCPY_CHK): Likewise.
1082 * calls.c (store_bounds): Likewise.
1083 (emit_call_1): Likewise.
1084 (special_function_p): Likewise.
1085 (maybe_warn_nonstring_arg): Likewise.
1086 (initialize_argument_information): Likewise.
1087 (finalize_must_preallocate): Likewise.
1088 (compute_argument_addresses): Likewise.
1089 (expand_call): Likewise.
1090 * cfgexpand.c (expand_call_stmt): Likewise.
1091 (expand_return): Likewise.
1092 (expand_gimple_stmt_1): Likewise.
1093 (pass_expand::execute): Likewise.
1094 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1095 (cgraph_node::remove): Likewise.
1096 (cgraph_node::dump): Likewise.
1097 (cgraph_node::verify_node): Likewise.
1098 * cgraph.h (chkp_function_instrumented_p): Likewise.
1099 (symtab_node::get_alias_target): Likewise.
1100 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1101 (cgraph_local_p): Likewise.
1102 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1103 (cgraph_edge::rebuild_references): Likewise.
1104 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1105 (walk_polymorphic_call_targets): Likewise.
1106 (cgraph_node::expand_thunk): Likewise.
1107 (symbol_table::output_weakrefs): Likewise.
1108 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1109 (ix86_handle_option): Likewise.
1110 * config/i386/constraints.md: Likewise.
1111 * config/i386/i386-builtin-types.def (BND): Likewise.
1115 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1116 (BDESC_FIRST): Likewise.
1118 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1119 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1120 * config/i386/i386.c (enum reg_class): Likewise.
1121 (ix86_target_string): Likewise.
1122 (ix86_option_override_internal): Likewise.
1123 (ix86_conditional_register_usage): Likewise.
1124 (ix86_valid_target_attribute_inner_p): Likewise.
1125 (ix86_set_indirect_branch_type): Likewise.
1126 (ix86_set_current_function): Likewise.
1127 (ix86_function_arg_regno_p): Likewise.
1128 (init_cumulative_args): Likewise.
1129 (ix86_function_arg_advance): Likewise.
1130 (ix86_function_arg): Likewise.
1131 (ix86_pass_by_reference): Likewise.
1132 (ix86_function_value_regno_p): Likewise.
1133 (ix86_function_value_1): Likewise.
1134 (ix86_function_value_bounds): Likewise.
1135 (ix86_return_in_memory): Likewise.
1136 (ix86_setup_incoming_vararg_bounds): Likewise.
1137 (ix86_va_start): Likewise.
1138 (indirect_thunk_need_prefix): Likewise.
1139 (print_reg): Likewise.
1140 (ix86_print_operand): Likewise.
1141 (ix86_expand_call): Likewise.
1142 (ix86_output_function_return): Likewise.
1143 (reg_encoded_number): Likewise.
1144 (BDESC_VERIFYS): Likewise.
1145 (ix86_init_mpx_builtins): Likewise.
1146 (ix86_init_builtins): Likewise.
1147 (ix86_emit_cmove): Likewise.
1148 (ix86_emit_move_max): Likewise.
1149 (ix86_expand_builtin): Likewise.
1150 (ix86_builtin_mpx_function): Likewise.
1151 (ix86_get_arg_address_for_bt): Likewise.
1152 (ix86_load_bounds): Likewise.
1153 (ix86_store_bounds): Likewise.
1154 (ix86_load_returned_bounds): Likewise.
1155 (ix86_store_returned_bounds): Likewise.
1156 (ix86_class_likely_spilled_p): Likewise.
1157 (ix86_hard_regno_mode_ok): Likewise.
1158 (x86_order_regs_for_local_alloc): Likewise.
1159 (ix86_mitigate_rop): Likewise.
1160 (ix86_bnd_prefixed_insn_p): Likewise.
1161 (ix86_mpx_bound_mode): Likewise.
1162 (ix86_make_bounds_constant): Likewise.
1163 (ix86_initialize_bounds): Likewise.
1164 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1165 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1166 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1167 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1168 (TARGET_CHKP_BOUND_MODE): Likewise.
1169 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1170 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1171 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1172 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1173 * config/i386/i386.h (TARGET_MPX): Likewise.
1174 (TARGET_MPX_P): Likewise.
1175 (VALID_BND_REG_MODE): Likewise.
1176 (FIRST_BND_REG): Likewise.
1177 (LAST_BND_REG): Likewise.
1178 (enum reg_class): Likewise.
1179 (BND_REG_P): Likewise.
1180 (BND_REGNO_P): Likewise.
1181 (BNDmode): Likewise.
1182 (ADJUST_INSN_LENGTH): Likewise.
1183 * config/i386/i386.md: Likewise.
1184 * config/i386/i386.opt: Likewise.
1185 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1186 (defined): Likewise.
1187 (LINK_MPX): Likewise.
1188 (MPX_SPEC): Likewise.
1189 (LIBMPX_SPEC): Likewise.
1190 (LIBMPXWRAPPERS_SPEC): Likewise.
1191 (CHKP_SPEC): Likewise.
1192 * config/i386/predicates.md: Likewise.
1193 * dbxout.c (dbxout_type): Likewise.
1194 * doc/extend.texi: Likewise.
1195 * doc/invoke.texi: Likewise.
1196 * doc/md.texi: Likewise.
1197 * doc/tm.texi: Likewise.
1198 * doc/tm.texi.in: Likewise.
1199 * dwarf2out.c (is_base_type): Likewise.
1200 (gen_formal_types_die): Likewise.
1201 (gen_subprogram_die): Likewise.
1202 (gen_type_die_with_usage): Likewise.
1203 (gen_decl_die): Likewise.
1204 (dwarf2out_late_global_decl): Likewise.
1205 * expr.c (expand_assignment): Likewise.
1206 (emit_storent_insn): Likewise.
1207 (store_expr_with_bounds): Likewise.
1208 (store_expr): Likewise.
1209 (expand_expr_real_1): Likewise.
1210 * expr.h (store_expr_with_bounds): Likewise.
1211 * function.c (use_register_for_decl): Likewise.
1212 (struct bounds_parm_data): Likewise.
1213 (assign_parms_augmented_arg_list): Likewise.
1214 (assign_parm_find_entry_rtl): Likewise.
1215 (assign_parm_is_stack_parm): Likewise.
1216 (assign_parm_load_bounds): Likewise.
1217 (assign_bounds): Likewise.
1218 (assign_parms): Likewise.
1219 (expand_function_start): Likewise.
1220 * gcc.c (CHKP_SPEC): Likewise.
1221 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1222 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1223 (wrestrict_dom_walker::check_call): Likewise.
1224 * gimple.c (gimple_build_call_from_tree): Likewise.
1225 * gimple.h (enum gf_mask): Likewise.
1226 (gimple_call_with_bounds_p): Likewise.
1227 (gimple_call_set_with_bounds): Likewise.
1228 * gimplify.c (gimplify_init_constructor): Likewise.
1229 * ipa-cp.c (initialize_node_lattices): Likewise.
1230 (propagate_constants_across_call): Likewise.
1231 (find_more_scalar_values_for_callers_subset): Likewise.
1232 * ipa-hsa.c (process_hsa_functions): Likewise.
1233 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1234 * ipa-icf.c (sem_function::merge): Likewise.
1235 * ipa-inline.c (early_inliner): Likewise.
1236 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1237 (warn_function_cold): Likewise.
1238 (propagate_pure_const): Likewise.
1239 * ipa-ref.h (enum GTY): Likewise.
1240 * ipa-split.c (find_retbnd): Likewise.
1241 (consider_split): Likewise.
1242 (split_function): Likewise.
1243 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1244 * ipa.c (walk_polymorphic_call_targets): Likewise.
1245 (symbol_table::remove_unreachable_nodes): Likewise.
1246 (process_references): Likewise.
1247 (cgraph_build_static_cdtor_1): Likewise.
1248 * lto-cgraph.c (lto_output_node): Likewise.
1249 (output_refs): Likewise.
1250 (compute_ltrans_boundary): Likewise.
1251 (input_overwrite_node): Likewise.
1252 (input_node): Likewise.
1253 (input_cgraph_1): Likewise.
1254 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1255 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1256 (class pass_chkp_instrumentation_passes): Likewise.
1257 (make_pass_chkp_instrumentation_passes): Likewise.
1258 * passes.def: Likewise.
1259 * rtl.h (struct GTY): Likewise.
1260 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1261 * stor-layout.c (layout_type): Likewise.
1262 * symtab.c: Likewise.
1263 * target.def: Likewise.
1264 * targhooks.c (default_chkp_bound_type): Likewise.
1265 (default_chkp_bound_mode): Likewise.
1266 (default_builtin_chkp_function): Likewise.
1267 (default_chkp_function_value_bounds): Likewise.
1268 (default_chkp_make_bounds_constant): Likewise.
1269 (default_chkp_initialize_bounds): Likewise.
1270 * targhooks.h (default_chkp_bound_type): Likewise.
1271 (default_chkp_bound_mode): Likewise.
1272 (default_builtin_chkp_function): Likewise.
1273 (default_chkp_function_value_bounds): Likewise.
1274 (default_chkp_make_bounds_constant): Likewise.
1275 (default_chkp_initialize_bounds): Likewise.
1276 * toplev.c (compile_file): Likewise.
1277 (process_options): Likewise.
1278 * tree-core.h (DEF_BUILTIN): Likewise.
1279 (DEF_BUILTIN_CHKP): Likewise.
1280 * tree-inline.c (declare_return_variable): Likewise.
1281 (remap_gimple_stmt): Likewise.
1282 (copy_bb): Likewise.
1283 (initialize_inlined_parameters): Likewise.
1284 (expand_call_inline): Likewise.
1285 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1286 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1287 (make_pass_ipa_chkp_produce_thunks): Likewise.
1288 (make_pass_chkp): Likewise.
1289 (make_pass_chkp_opt): Likewise.
1290 (make_pass_chkp_instrumentation_passes): Likewise.
1291 * tree-pretty-print.c (dump_generic_node): Likewise.
1292 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1293 * tree-ssa-dce.c (propagate_necessity): Likewise.
1294 (eliminate_unnecessary_stmts): Likewise.
1295 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1296 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1297 * tree-ssa-sccvn.h: Likewise.
1298 * tree-ssa-strlen.c (get_string_length): Likewise.
1299 (valid_builtin_call): Likewise.
1300 (adjust_last_stmt): Likewise.
1301 (handle_builtin_strchr): Likewise.
1302 (handle_builtin_strcpy): Likewise.
1303 (handle_builtin_stxncpy): Likewise.
1304 (handle_builtin_memcpy): Likewise.
1305 (handle_builtin_strcat): Likewise.
1306 (strlen_check_and_optimize_stmt): Likewise.
1307 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1308 * tree-streamer-in.c: Likewise.
1309 * tree-streamer.c (record_common_node): Likewise.
1310 * tree.c (tree_code_size): Likewise.
1311 (wide_int_to_tree_1): Likewise.
1312 (type_contains_placeholder_1): Likewise.
1313 (build_common_tree_nodes): Likewise.
1314 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1315 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1316 (POINTER_BOUNDS_P): Likewise.
1317 (BOUNDED_TYPE_P): Likewise.
1318 (BOUNDED_P): Likewise.
1319 (CALL_WITH_BOUNDS_P): Likewise.
1320 (pointer_bounds_type_node): Likewise.
1321 * value-prof.c (gimple_ic): Likewise.
1322 * var-tracking.c (vt_add_function_parameters): Likewise.
1323 * varasm.c (make_decl_rtl): Likewise.
1324 (assemble_start_function): Likewise.
1325 (output_constant): Likewise.
1326 (maybe_assemble_visibility): Likewise.
1327 * varpool.c (ctor_for_folding): Likewise.
1328 * chkp-builtins.def: Remove.
1329 * ipa-chkp.c: Remove.
1330 * ipa-chkp.h: Remove.
1331 * rtl-chkp.c: Remove.
1332 * rtl-chkp.h: Remove.
1333 * tree-chkp-opt.c: Remove.
1334 * tree-chkp.c: Remove.
1335 * tree-chkp.h: Remove.
1337 2018-06-07 Carl Love <cel@us.ibm.com>
1339 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1340 vextract_fp_from_shortl): Add BE support.
1342 2018-06-07 Paul Koning <ni1d@arrl.net>
1344 * compare-elim.c (try_merge_compare): Don't merge compare if
1345 address contains a side effect.
1346 (try_eliminate_compare): Likewise.
1348 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1350 * config.gcc: Support "tremont".
1351 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1352 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1354 * config/i386/i386.c (m_TREMONT): Define.
1355 (processor_target_table): Add "tremont".
1356 (PTA_TREMONT): Define.
1357 (ix86_lea_outperforms): Add TARGET_TREMONT.
1358 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1359 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1360 and M_INTEL_GOLDMONT_PLUS.
1361 (fold_builtin_cpu): Add "tremont".
1362 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1363 (ix86_option_override_internal): Add "tremont".
1364 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1365 (processor_type): Add PROCESSOR_TREMONT.
1366 * config/i386/x86-tune.def: Add m_TREMONT.
1367 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1369 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1371 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1372 symbol defined for msp430i* devices to be lower case.
1374 gcc/testsuite/gcc.target/msp430/
1375 * msp430i-device-symbol.c: New test.
1376 * msp430f-device-symbol.c: New test.
1377 * msp430.h: New test header file.
1379 2018-06-07 Richard Biener <rguenther@suse.de>
1381 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1382 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1383 Properly wrap signed arithmetic if overflow wraps.
1385 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1387 PR tree-optimization/69615
1388 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1389 of a cast from a same precision integral SSA_NAME in a bb dominated
1390 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1391 cast to utype if rhs2 has already a compatible type.
1393 2018-06-07 Richard Biener <rguenther@suse.de>
1395 PR tree-optimization/85935
1396 * graphite-scop-detection.c (find_params_in_bb): Analyze
1397 condition operands with respect to the correct loop. Assert
1398 the analysis doesn't fail.
1400 2018-06-04 Carl Love <cel@us.ibm.com>
1402 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1403 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1404 as it is slightly cheaper.
1405 (first_match_or_eos_index_<mode>):
1406 Calculate index using natural element order.
1407 (first_match_index_<mode>):
1408 Calculate index using natural element order.
1409 (first_match_or_eos_index_<mode>):
1410 Calculate index using natural order.
1411 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1412 for BE and LE modes.
1413 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1414 P9V_BUILTIN_VCLZLSBB_V16QI.
1415 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1418 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1420 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1421 indentation and line wrap for many prototypes. Add missing
1422 @smallexample directives around block of prototypes for vec_xl and
1425 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1427 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1428 track if we pass or return IEEE 128-bit floating point.
1429 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1430 C++ mangling that is compatible with GCC 8.1.
1431 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1432 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1433 floating point types.
1434 (rs6000_function_arg_advance_1): Likewise.
1435 (rs6000_mangle_type): Optionally generate mangled names that match
1436 what GCC 8.1 generated for IEEE 128-bit floating point types.
1437 (rs6000_globalize_decl_name): If we have an external function that
1438 passes or returns IEEE 128-bit types, generate a weak reference
1439 from the mangled name used in GCC 8.1 to the current mangled
1441 (rs6000_init_builtins): Make __ibm128 use the long double type if
1442 long double is IBM extended double. Make __float128 use the long
1443 double type if long double is IEEE 128-bit.
1446 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1447 macro for __ibm128 built-in functions.
1448 (PACK_IF): Add __ibm128 pack/unpack functions.
1449 (UNPACK_IF): Likewise.
1450 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1451 enable long double built-in functions if long double is IEEE
1452 128-bit floating point.
1453 (rs6000_invalid_builtin): Update long double built-in function
1455 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1456 functions, adjust the built-in function to use the long double
1457 built-in function if __ibm128 and long double are the same type.
1458 * doc/extend.texi (PowerPC builtins): Update documention for
1459 __builtin_{,un}pack_longdouble. Add documentation for
1460 __builtin_{,un}pack_ibm128.
1462 2018-06-06 Jim Wilson <jimw@sifive.com>
1464 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1465 (struct machine_function): New field interrupt_mode.
1466 (riscv_handle_type_attribute): New function. Add forward declaration.
1467 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1468 (riscv_expand_epilogue): Check interrupt_mode field.
1469 (riscv_set_current_function): Check interrupt attribute args and
1470 set interrupt_mode field.
1471 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1472 (riscv_sret, riscv_uret): New.
1473 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1474 new arguments to interrupt attribute.
1476 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1479 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1480 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1482 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1484 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1485 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1486 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1487 several redundant entries.
1489 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1491 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1492 type from "rtx" to "rtx_insn *".
1493 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1494 for local "call_insn", removing cast.
1495 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1497 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1499 PR tree-optimization/86066
1500 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1501 for BIT_INSERT_EXPR stores.
1503 2018-06-06 Richard Biener <rguenther@suse.de>
1505 PR tree-optimization/86062
1506 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1507 component refs ontop
1508 of to be offsetted base.
1510 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1512 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1513 to be static and remove check on interrupt attribute name.
1515 gcc/testsuite/gcc.target/msp430/
1516 * function-attributes-4.c: New test.
1517 * static-interrupts.c: New test.
1519 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1521 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1522 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1524 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1527 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1529 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1530 Remove second argument, change how error is called.
1531 (aarch64_layout_arg): Remove second argument from
1532 aarch64_err_no_fpadvsimd call.
1533 (aarch64_init_cumulative_args): Ditto.
1534 (aarch64_gimplify_va_arg_expr): Ditto.
1535 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1537 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1539 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1540 (*simple_return_indirect_internal<mode>): Rename from
1541 simple_return_indirect_internal. Use W mode iterator.
1542 (rstorssp): New expander.
1543 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1544 (clrssbsy): New expander.
1545 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1547 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1549 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1551 (cmse_check_pointed_object): Likewise.
1553 2018-06-05 Martin Liska <mliska@suse.cz>
1555 PR gcov-profile/47618
1556 * doc/invoke.texi: Document how -fprofile-dir format
1559 2018-06-05 Richard Biener <rguenther@suse.de>
1561 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1562 removal pretend DOM info isn't available so we do not update
1563 it and only remove edges, not dominated blocks. Actually free
1564 DOM info in case we removed something. Remove unreachable blocks.
1565 (mfb_keep_latches): Work with either DOM info or marked backedges.
1566 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1567 first. Mark backedges if DOM info isn't available.
1568 (Re-)compute DOM info after cleanup_control_flow_pre.
1570 2018-06-05 Richard Biener <rguenther@suse.de>
1572 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1573 (locus_discrim_hasher::hash): Adjust.
1574 (locus_discrim_hasher::equal): Likewise.
1575 (next_discriminator_for_locus): Work on line directly.
1576 (same_line_p): Pass in expanded locus1 as well.
1577 (assign_discriminators): Avoid redundant location expansions.
1579 2018-06-05 Richard Biener <rguenther@suse.de>
1581 PR tree-optimization/86046
1582 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1583 if required after clearing TREE_ADDRESSABLE.
1585 2018-06-05 Richard Biener <rguenther@suse.de>
1587 PR tree-optimization/86047
1588 * tree-ssa-loop.c (for_each_index): Glob handling of all
1589 decls and constants and really handle all of them.
1591 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1594 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1595 qualifier_void_pointer and qualifier_const_void_pointer.
1596 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1597 (arm_init_builtins): Handle the above.
1598 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1599 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1600 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1603 2018-06-05 Martin Liska <mliska@suse.cz>
1605 * auto-profile.c (read_autofdo_file): Do not use
1606 gcov_ctr_summary struct.
1607 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1608 * coverage.c (struct counts_entry): Likewise.
1609 (read_counts_file): Read just single summary entry.
1610 (get_coverage_counts): Use gcov_summary struct.
1611 * coverage.h (get_coverage_counts): Likewise.
1612 * gcov-dump.c (dump_working_sets): Likewise.
1613 (tag_summary): Dump just single summary.
1614 * gcov-io.c (gcov_write_summary): Write just histogram
1616 (gcov_read_summary): Read just single summary.
1617 (compute_working_sets): Use gcov_summary struct.
1618 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1619 of GCOV_COUNTERS_SUMMABLE.
1620 (GCOV_COUNTERS_SUMMABLE): Remove.
1621 (GCOV_FIRST_VALUE_COUNTER): Replace with
1622 GCOV_COUNTER_V_INTERVAL.
1623 (struct gcov_ctr_summary): Remove.
1624 (struct gcov_summary): Directly use fields of former
1626 (compute_working_sets): Use gcov_summary struct.
1627 * gcov.c (read_count_file): Do not use ctrs fields.
1628 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1630 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1632 * profile.c: Likewise.
1633 * profile.h: Likewise.
1635 2018-06-05 Martin Liska <mliska@suse.cz>
1637 PR gcov-profile/84846
1638 * gcov.c (output_lines): Print working directory only
1639 in intermediate format.
1641 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1643 * config/s390/s390-builtin-types.def: Add void function type.
1644 * config/s390/s390-builtins.def: Use the function type for the
1647 2018-06-04 Jim Wilson <jimw@sifive.com>
1649 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1651 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1652 and maybe_eh_return. Change regno to unsigned int. Use new args to
1653 handle EH_RETURN_DATA_REGNO registers properly.
1654 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1655 (riscv_expand_epilogue): Update comment. Change argument name and
1656 type. Update code to use new name and type. Pass new args to
1657 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1659 * config/riscv/riscv.md (NORMAL_RETURN): New.
1660 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1661 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1662 (eh_return): Call gen_eh_return_internal and emit barrier.
1663 (eh_return_internal): Call riscv_expand_epilogue.
1665 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1667 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1668 bit_insertion field and declare can_be_merged_into method.
1669 (merged_store_group::can_be_merged_into): New method.
1670 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1671 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1672 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1674 2018-06-04 Richard Biener <rguenther@suse.de>
1676 PR tree-optimization/85955
1677 * builtins.c (fold_builtin_sincos): Convert pointers to
1678 destination to appropriate type before dereferencing.
1680 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1682 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1684 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1686 * expr.c (expand_expr_real_1): Force the operand into memory if
1687 its TYPE_MODE is BLKmode and if there is no integer mode for
1688 the number of bits being extracted.
1690 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1694 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1695 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1697 2018-06-04 Richard Biener <rguenther@suse.de>
1699 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1700 (cleanup_tree_cfg_noloop): ... single caller. Do
1701 start_recording_case_labels later.
1703 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1705 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1706 to _IMMINTRIN_H_INCLUDED.
1707 * config/i386/pconfigintrin.h: Ditto.
1708 * config/i386/waitpkgintrin.h: Ditto.
1709 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1710 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1711 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1712 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1713 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1714 waitpkgintrin.h and cldemoteintrin.h.
1716 2018-06-04 Richard Biener <rguenther@suse.de>
1718 PR tree-optimization/86038
1719 * tracer.c (find_best_successor): Check probability for
1720 being initialized, bail out if not.
1722 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1725 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1726 of bits to ignore when comparing architectures.
1728 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1730 PR tree-optimization/69615
1731 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1732 maximum or minimum of the type, try to merge it also as if
1733 range1 is + [-, x - 1] or + [x + 1, -].
1736 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1738 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1740 PR tree-optimization/86034
1741 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1742 the unsigned bitfield type in a bit insertion sequence if it does not
1743 have a larger precision than the bitfield size.
1744 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1746 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1748 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1750 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1752 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1753 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1754 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1755 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1757 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1759 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1760 Disable -fdelete-null-pointer-checks for ELF toolchain.
1762 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1763 Kito Cheng <kito.cheng@gmail.com>
1765 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1766 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1767 (nds32*-*-*): Add glibc and uclibc conditions.
1768 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1769 (TARGET_EXCEPT_UNWIND_INFO): Define.
1770 * config/nds32/elf.h: New file.
1771 * config/nds32/linux.h: New file.
1772 * config/nds32/nds32-elf.opt: New file.
1773 * config/nds32/nds32-linux.opt: New file.
1774 * config/nds32/nds32-fp-as-gp.c
1775 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1776 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1778 (nds32_asm_file_end): Ditto.
1779 (nds32_print_operand): Ditto.
1780 (nds32_insert_attributes): Ditto.
1781 (nds32_init_libfuncs): New function.
1782 (TARGET_HAVE_TLS): Define.
1783 (TARGET_INIT_LIBFUNCS): Define.
1784 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1786 (TARGET_ELF): Apply different mcmodel setting.
1787 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1788 been migrated into elf.h and linux.h files.
1789 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1790 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1791 (mcmodel): The content has been migrated into nds32-elf.opt and
1792 nds32-linux.opt files.
1793 * config/nds32/t-elf: New file.
1794 * config/nds32/t-linux: New file.
1796 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1797 Shiva Chen <shiva0217@gmail.com>
1799 * config/nds32/constants.md (unspec_volatile_element): Add
1800 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1801 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1803 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1804 (make_pass_nds32_fp_as_gp): Declare.
1805 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1807 (nds32_asm_function_end_prologue): Remove unused asm output.
1808 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1809 (nds32_asm_file_start): Output necessary fp_as_gp information.
1810 (nds32_option_override): Adjust register usage.
1811 (nds32_expand_prologue): Consider fp_as_gp situation.
1812 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1813 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1816 (simple_return): Ditto.
1817 (omit_fp_begin): Output special directive for fp_as_gp.
1818 (omit_fp_end): Output special directive for fp_as_gp.
1819 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1820 mforbid-fp-as-gp): New options.
1822 2018-06-01 Mark Wielaard <mark@klomp.org>
1824 * dwarf2out.c (dwarf2out_finish): Remove generation of
1825 DW_AT_loclists_base.
1827 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1829 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1830 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1831 (struct merged_store_group): Add bit_insertion field.
1832 (dump_char_array): Use standard hexadecimal format.
1833 (merged_store_group::merged_store_group): Set bit_insertion to false.
1834 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1835 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1836 also print the mask in the dump file.
1837 (pass_store_merging::gate): Minor tweak.
1838 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1839 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1840 stores with INTEGER_CST stores.
1841 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1842 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1843 and use it throughout. Generate bit insertion sequences if need be.
1844 (pass_store_merging::process_store): Remove redundant condition.
1845 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1847 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1849 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1850 the 128-bit floating point types. Fix function comment.
1852 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1854 * config/aarch64/aarch64-simd.md
1855 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1857 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1860 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1862 PR tree-optimization/85989
1863 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1865 (backprop::intersect_uses): Check it when deciding whether this
1866 is a backedge reference.
1867 (backprop::process_block): Add each phi to m_visited_phis
1868 after visiting it, then clear it at the end.
1870 2018-06-01 Richard Biener <rguenther@suse.de>
1872 * tree-vectorizer.h (vect_dr_stmt): New function.
1873 (vect_get_load_cost): Adjust.
1874 (vect_get_store_cost): Likewise.
1875 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1876 Use vect_dr_stmt instead of DR_SMTT.
1877 (vect_record_base_alignments): Likewise.
1878 (vect_calculate_target_alignment): Likewise.
1879 (vect_compute_data_ref_alignment): Likewise and make static.
1880 (vect_update_misalignment_for_peel): Likewise.
1881 (vect_verify_datarefs_alignment): Likewise.
1882 (vector_alignment_reachable_p): Likewise.
1883 (vect_get_data_access_cost): Likewise. Pass down
1884 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1885 (vect_get_peeling_costs_all_drs): Likewise.
1886 (vect_peeling_hash_get_lowest_cost): Likewise.
1887 (vect_enhance_data_refs_alignment): Likewise.
1888 (vect_find_same_alignment_drs): Likewise.
1889 (vect_analyze_data_refs_alignment): Likewise.
1890 (vect_analyze_group_access_1): Likewise.
1891 (vect_analyze_group_access): Likewise.
1892 (vect_analyze_data_ref_access): Likewise.
1893 (vect_analyze_data_ref_accesses): Likewise.
1894 (vect_vfa_segment_size): Likewise.
1895 (vect_small_gap_p): Likewise.
1896 (vectorizable_with_step_bound_p): Likewise.
1897 (vect_prune_runtime_alias_test_list): Likewise.
1898 (vect_analyze_data_refs): Likewise.
1899 (vect_supportable_dr_alignment): Likewise.
1900 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1901 (vect_gen_prolog_loop_niters): Likewise.
1902 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1903 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1905 (vect_recog_mask_conversion_pattern): Likewise.
1906 (vect_try_gather_scatter_pattern): Likewise.
1907 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1908 to vect_get_store_cost.
1909 (vect_get_store_cost): Get stmt_info instead of DR.
1910 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1911 (vect_get_load_cost): Get stmt_info instead of DR.
1913 2018-06-01 Richard Biener <rguenther@suse.de>
1916 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1917 vars wrapped in MEM_REFs.
1919 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1921 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1922 Fix subreg tests so that we only return a choice between
1923 GENERAL_REGS and FP_REGS if the original classes included both.
1925 2018-06-01 Richard Biener <rguenther@suse.de>
1928 * tree-ssa-structalias.c (get_function_part_constraint):
1929 Handle NULL fi->decl.
1930 (find_func_aliases_for_call): Properly handle indirect
1931 fi from direct call.
1932 (find_func_clobbers): Likewise.
1933 (ipa_pta_execute): Likewise.
1934 (create_variable_info_for): For functions that are ifunc_resolver
1935 resolve to a varinfo that contains the result of the resolver call.
1936 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1939 2018-05-31 Michael Collison <michael.collison@arm.com>
1941 * config/aarch64/aarch64.md:
1942 (*fix_to_zero_extenddfdi2): New pattern.
1943 * gcc.target/aarch64/fix_extend1.c: New testcase.
1945 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1949 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1950 and BUILT_IN_STRNCMP_EQ.
1951 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1952 BUILT_IN_STRNCMP_EQ.
1953 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1954 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1955 (gimple_fold_builtin): Likewise.
1956 * tree-ssa-strlen.c (compute_string_length): New function.
1957 (determine_min_obsize): New function.
1958 (handle_builtin_string_cmp): New function to handle calls to
1959 string compare functions.
1960 (strlen_optimize_stmt): Add handling to builtin string compare
1962 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1963 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1964 * tree.c (build_common_builtin_nodes): Add new defines of
1965 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1967 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1970 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1971 functions with naked attribute.
1973 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1975 * config/i386/sse.md (avx_vec_concat<mode>):
1976 Substitute concat_tg_mode mode attribute with xtg_mode.
1977 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1978 (concat_tg_mode): Remove mode attribute.
1980 2018-05-31 Martin Sebor <msebor@redhat.com>
1983 * calls.c (alloc_max_size): Correct a logic error/typo.
1984 Treat excessive arguments as infinite. Warn for invalid arguments.
1985 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1987 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1990 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1991 and movx for Haswell.
1993 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1994 Cesar Philippidis <cesar@codesourcery.com>
1997 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1998 when emitting error on private/firstprivate reductions.
1999 * omp-low.c (lower_omp_target): Avoid reference-type processing
2000 on pointers for firstprivate clause.
2002 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2004 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2007 * config/aarch64/aarch64-simd.md
2008 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2009 (aarch64_ld1_x3_<mode>): Likewise
2010 (aarch64_st1x2<VALLDIF:mode>): Likewise
2011 (aarch64_st1_x2_<mode>): Likewise
2012 (aarch64_st1x3<VALLDIF:mode>): Likewise
2013 (aarch64_st1_x3_<mode>): Likewise
2014 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2015 (vld1_s8_x3): Likewise.
2016 (vld1_u16_x3): Likewise.
2017 (vld1_s16_x3): Likewise.
2018 (vld1_u32_x3): Likewise.
2019 (vld1_s32_x3): Likewise.
2020 (vld1_u64_x3): Likewise.
2021 (vld1_s64_x3): Likewise.
2022 (vld1_f16_x3): Likewise.
2023 (vld1_f32_x3): Likewise.
2024 (vld1_f64_x3): Likewise.
2025 (vld1_p8_x3): Likewise.
2026 (vld1_p16_x3): Likewise.
2027 (vld1_p64_x3): Likewise.
2028 (vld1q_u8_x3): Likewise.
2029 (vld1q_s8_x3): Likewise.
2030 (vld1q_u16_x3): Likewise.
2031 (vld1q_s16_x3): Likewise.
2032 (vld1q_u32_x3): Likewise.
2033 (vld1q_s32_x3): Likewise.
2034 (vld1q_u64_x3): Likewise.
2035 (vld1q_s64_x3): Likewise.
2036 (vld1q_f16_x3): Likewise.
2037 (vld1q_f32_x3): Likewise.
2038 (vld1q_f64_x3): Likewise.
2039 (vld1q_p8_x3): Likewise.
2040 (vld1q_p16_x3): Likewise.
2041 (vld1q_p64_x3): Likewise.
2042 (vst1_s64_x2): Likewise.
2043 (vst1_u64_x2): Likewise.
2044 (vst1_f64_x2): Likewise.
2045 (vst1_s8_x2): Likewise.
2046 (vst1_p8_x2): Likewise.
2047 (vst1_s16_x2): Likewise.
2048 (vst1_p16_x2): Likewise.
2049 (vst1_s32_x2): Likewise.
2050 (vst1_u8_x2): Likewise.
2051 (vst1_u16_x2): Likewise.
2052 (vst1_u32_x2): Likewise.
2053 (vst1_f16_x2): Likewise.
2054 (vst1_f32_x2): Likewise.
2055 (vst1_p64_x2): Likewise.
2056 (vst1q_s8_x2): Likewise.
2057 (vst1q_p8_x2): Likewise.
2058 (vst1q_s16_x2): Likewise.
2059 (vst1q_p16_x2): Likewise.
2060 (vst1q_s32_x2): Likewise.
2061 (vst1q_s64_x2): Likewise.
2062 (vst1q_u8_x2): Likewise.
2063 (vst1q_u16_x2): Likewise.
2064 (vst1q_u32_x2): Likewise.
2065 (vst1q_u64_x2): Likewise.
2066 (vst1q_f16_x2): Likewise.
2067 (vst1q_f32_x2): Likewise.
2068 (vst1q_f64_x2): Likewise.
2069 (vst1q_p64_x2): Likewise.
2070 (vst1_s64_x3): Likewise.
2071 (vst1_u64_x3): Likewise.
2072 (vst1_f64_x3): Likewise.
2073 (vst1_s8_x3): Likewise.
2074 (vst1_p8_x3): Likewise.
2075 (vst1_s16_x3): Likewise.
2076 (vst1_p16_x3): Likewise.
2077 (vst1_s32_x3): Likewise.
2078 (vst1_u8_x3): Likewise.
2079 (vst1_u16_x3): Likewise.
2080 (vst1_u32_x3): Likewise.
2081 (vst1_f16_x3): Likewise.
2082 (vst1_f32_x3): Likewise.
2083 (vst1_p64_x3): Likewise.
2084 (vst1q_s8_x3): Likewise.
2085 (vst1q_p8_x3): Likewise.
2086 (vst1q_s16_x3): Likewise.
2087 (vst1q_p16_x3): Likewise.
2088 (vst1q_s32_x3): Likewise.
2089 (vst1q_s64_x3): Likewise.
2090 (vst1q_u8_x3): Likewise.
2091 (vst1q_u16_x3): Likewise.
2092 (vst1q_u32_x3): Likewise.
2093 (vst1q_u64_x3): Likewise.
2094 (vst1q_f16_x3): Likewise.
2095 (vst1q_f32_x3): Likewise.
2096 (vst1q_f64_x3): Likewise.
2097 (vst1q_p64_x3): Likewise.
2099 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2101 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2102 user_label_prefix to name.
2104 * tree-core.h: Update comment about the format of NAME string
2105 passed to handler in attribute_spec.
2107 * config/msp430/msp430.md: Remove erroneous subreg expression from
2108 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2109 zero_extend{q,h}isi2.
2111 2018-05-30 Borislav Petkov <bp@suse.de>
2113 * doc/extend.texi: Document some architecture specific
2114 constraints and sort entries.
2116 2018-05-30 Martin Sebor <msebor@redhat.com>
2119 * builtins.c (expand_builtin_stpcpy_1): New function.
2120 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2121 only if the former succeeds.
2123 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2125 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2128 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2130 * doc/invoke.texi (-flinker-output): Document
2132 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2134 * passes.c (ipa_write_summaries): Only modify statements if body
2136 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2137 incrementally linking.
2138 (ipa_passes): Likewise.
2139 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2140 pass down resolution info.
2141 * common.opt (flag_incremental_link): Update info.
2142 * gcc.c (plugin specs): Turn flinker-output=* to
2143 -plugin-opt=-linker-output-known
2144 * toplev.c (compile_file): Also cut compilation when doing incremental
2146 * flag-types. (enum lto_partition_model): Add
2147 LTO_LINKER_OUTPUT_NOLTOREL.
2148 (invoke.texi): Add -flinker-output docs.
2149 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2150 link same way as WPA; do not stream in dead initializers.
2152 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2153 darf2out_register_external_decl): Support incremental link.
2155 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2157 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2159 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2161 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2162 it down to simple_object_copy_lto_debug_sections.
2163 (run_gcc): Determine incremental LTO link time and configure
2164 lto1 into non-wpa mode, disable renaming of debug sections.
2166 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2168 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2169 descriptions of various incorrectly documented functions.
2171 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2174 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2175 address check not strict.
2177 2018-05-30 Richard Biener <rguenther@suse.de>
2179 PR tree-optimization/85964
2180 * tracer.c (better_p): Drop initialized count check, we only
2181 call the function with initialized counts now.
2182 (find_best_successor): Do find a best edge if one
2183 has uninitialized count.
2184 (find_best_predecessor): Likewise. Do BB frequency check only
2185 if count is initialized.
2187 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2189 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2190 (aarch64_ldrstr_offset_compare): New.
2191 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2192 load/store orderings.
2193 (aarch64_gen_adjusted_ldpstp): Likewise.
2195 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2197 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2198 Check for subset of GENERAL_REGS and FP_REGS.
2199 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2202 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2204 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2205 and wi::to_poly_offset. Add the current offset and then check
2206 whether the sum fits, rather than using an unchecked addition of
2207 a checked term. Check for a shwi rather than a uhwi.
2208 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2209 (store_constructor): Use poly_int_tree_p.
2210 (expand_expr_real_1): Likewise.
2211 * function.c (assign_temp): Likewise.
2212 * fold-const.c (const_binop): Use poly_int_tree_p and
2214 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2216 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2217 to_poly_offset for MEM offsets.
2218 * ipa-icf.c (sem_variable::equals): Likewise.
2219 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2220 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2221 wi::to_poly_offset for BIT_FIELD_REF offsets.
2222 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2224 * var-tracking.c (emit_note_insn_var_location): Use
2225 tree_to_poly_uint64.
2227 2018-05-29 Jim Wilson <jimw@sifive.com>
2229 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2231 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2234 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2235 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2237 (sse4_1_round<mode>2): Use nonimmediate_operand
2238 for operand 1 predicate.
2240 2018-05-29 Martin Sebor <msebor@redhat.com>
2241 Richard Biener <rguenther@suse.de>
2244 * calls.c (get_size_range): Call determine_value_range instead
2245 of get_value_range..
2246 * tree-vrp.h (determine_value_range): Declared new function.
2247 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2249 2018-05-29 Richard Biener <rguenther@suse.de>
2251 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2252 sure to use non-pattern stmts for get_earlier_stmt arguments.
2253 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2254 called on pattern stmts.
2255 (get_later_stmt): Likewise.
2257 2018-05-29 Martin Liska <mliska@suse.cz>
2259 PR gcov-profile/85759
2260 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2263 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2265 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2267 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2271 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2272 VEC_PACK_FLOAT_EXPR): New tree codes.
2273 * tree-pretty-print.c (op_code_prio): Handle
2274 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2275 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2276 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2277 * tree-inline.c (estimate_operator_cost): Likewise.
2278 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2279 * fold-const.c (const_binop): Likewise.
2280 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2281 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2282 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2283 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2284 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2285 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2286 * expr.c (expand_expr_real_2): Likewise.
2287 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2288 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2289 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2291 * optabs.c (expand_widen_pattern_expr): For
2292 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2293 sign from result type rather than operand's type.
2294 (expand_binop_directly): For vec_packu_float_optab and
2295 vec_packs_float_optab allow result type to be different from operand's
2297 * optabs-tree.c (optab_for_tree_code): Handle
2298 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2299 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2300 * tree-vect-generic.c (expand_vector_operations_1): Handle
2301 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2302 VEC_PACK_FLOAT_EXPR.
2303 * tree-vect-stmts.c (supportable_widening_operation): Handle
2305 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2306 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2307 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2308 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2309 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2311 (vec_pack<floatprefix>_float_<mode>): New expander.
2312 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2314 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2315 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2316 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2317 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2318 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2320 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2321 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2322 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2323 VEC_PACK_FLOAT_EXPR): Document.
2325 2018-05-29 Richard Biener <rguenther@suse.de>
2327 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2329 (stmt_vec_info_vec): Make pointer.
2330 (init_stmt_vec_info_vec): Remove.
2331 (free_stmt_vec_info_vec): Likewise.
2332 (set_stmt_vec_info_vec): New function.
2333 (free_stmt_vec_infos): Likewise.
2334 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2335 (set_vinfo_for_stmt): Likewise.
2336 (get_earlier_stmt): Likewise.
2337 (get_later_stmt): Likewise.
2338 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2339 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2340 (vec_info::~vec_info): Free stmt_vec_infos.
2341 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2342 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2343 (pass_slp_vectorize::execute): Likewise.
2344 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2345 (free_stmt_vec_info_vec): Likewise.
2346 (set_stmt_vec_info_vec): New function.
2347 (free_stmt_vec_infos): Likewise.
2348 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2349 the global stmt_vec_info_vec.
2350 * tree-parloops.c (gather_scalar_reductions): Use
2351 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2354 2018-05-29 Richard Biener <rguenther@suse.de>
2356 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2358 2018-05-29 Martin Liska <mliska@suse.cz>
2359 David Malcolm <dmalcolm@redhat.com>
2361 * vec.c (test_reverse): New.
2362 (vec_c_tests): Add new test.
2363 * vec.h (vl_ptr>::reverse): New function.
2365 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2367 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2369 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2372 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2374 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2376 2018-05-28 Richard Biener <rguenther@suse.de>
2378 PR tree-optimization/85933
2379 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2380 look at stmts marked as vectorizable.
2382 2018-05-28 Richard Biener <rguenther@suse.de>
2384 PR tree-optimization/85934
2385 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2386 vector boolean check before scalar optimization.
2388 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2390 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2393 2018-05-28 Mark Wielaard <mark@klomp.org>
2395 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2396 if it is an expression containing a minus sign.
2398 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2400 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2402 2018-05-27 Paul Koning <ni1d@arrl.net>
2404 * config/pdp11/pdp11.md (truncsihi2): Remove.
2406 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2407 Chung-Ju Wu <jasonwucj@gmail.com>
2409 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2411 (unaligned_store_dw): Ditto.
2412 * config/nds32/nds32-memory-manipulation.c
2413 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2414 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2415 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2416 (emit_setmem_word_loop): Rename to ...
2417 (emit_setmem_doubleword_loop): ... this.
2418 (nds32_gen_dup_4_byte_to_word_value): New function.
2419 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2420 (nds32_expand_setmem_loop): Refine implementation.
2421 (nds32_expand_setmem_loop_v3m): Ditto.
2422 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2425 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2427 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2429 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2431 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2432 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2433 machine->attr_no_prologue_p.
2434 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2435 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2436 (nds32_expand_epilogue): Consider attr_naked_p.
2437 (nds32_expand_epilogue_v3pop): Likewise.
2438 (nds32_can_use_return_insn): Likewise.
2439 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2440 attr_no_prologue_p fields.
2441 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2443 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2446 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2448 * config/i386/sse.md
2449 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2451 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2453 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2455 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2457 (*<floatsuffix>floatv2div2sf2): Rename to ...
2458 (*float<floatunssuffix>v2div2sf2): ... this.
2459 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2460 (float<floatunssuffix>v2div2sf2_mask): ... this.
2461 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2462 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2463 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2465 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2467 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2469 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2471 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2473 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2475 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2476 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2477 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2478 gen_ufix_truncv8dfv8si2.
2479 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2480 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2481 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2482 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2483 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2484 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2485 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2486 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2488 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2492 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2494 2018-05-25 Jim Wilson <jimw@sifive.com>
2496 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2497 * config/riscv/riscv.c (struct machine_function): Add
2498 interrupt_handler_p and attribute_checked_p fields.
2499 (riscv_attribute_table): Add interrupt.
2500 (riscv_interrupt_type_p): New.
2501 (riscv_save_reg_p): Save extra regs for interrupt handler.
2502 (riscv_use_save_libcall): Return false for interrupt handler.
2503 (riscv_first_stack_step): Add forward declaration.
2504 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2505 for interrupt handler with large frame. Use it for saved reg list.
2506 (riscv_expand_prologue): Move flag_stack_usage_info support to
2507 eliminate duplication.
2508 (riscv_expand_epilogue): Generate mret for interrupt handler.
2509 (riscv_epilogue_uses): New.
2510 (riscv_can_use_return_insn): Return false for interrupt handler.
2511 (riscv_function_ok_for_sibcall): Likewise.
2512 (riscv_set_current_function): Add interrupt handler support.
2513 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2514 * config/riscv/riscv.md (UNSPECV_MRET): New.
2516 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2518 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2520 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2522 PR tree-optimization/85712
2523 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2524 this candidate has already been replaced in-situ by a copy.
2526 2018-05-25 Jason Merrill <jason@redhat.com>
2528 PR c++/80485 - inline function non-zero address.
2529 * symtab.c (nonzero_address): Check DECL_COMDAT.
2531 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2534 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2535 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2536 extension of SImode operation. Use const123_operand predicate.
2537 (*saddsi_1): Remove.
2538 (*saddl_se_1): Ditto.
2540 (*ssubl_se_1): Ditto.
2541 * config/alpha/predicates.md (const123_operand): New predicate.
2542 * config/alpha/constraints.md (P): Use IN_RANGE.
2544 2018-05-25 Richard Biener <rguenther@suse.de>
2546 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2548 (ref_maybe_used_by_stmt_p): Likewise.
2549 (stmt_may_clobber_ref_p): Likewise.
2550 (stmt_may_clobber_ref_p_1): Likewise.
2551 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2553 (ref_maybe_used_by_stmt_p): Likewise.
2554 (stmt_may_clobber_ref_p): Likewise.
2555 (stmt_may_clobber_ref_p_1): Likewise.
2556 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2557 the alias oracle to disambiguate DRs with stmts DR analysis
2559 (vect_analyze_data_refs): Do not give up on not analyzable
2560 DRs for BB vectorization. Remove code truncating the dataref
2563 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2566 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2567 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2568 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2570 2018-05-25 Richard Biener <rguenther@suse.de>
2572 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2573 function, combining stmt data ref gathering and fatal analysis
2575 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2576 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2577 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2578 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2579 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2581 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2583 PR tree-optimization/85720
2584 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2585 SCC if all partitions are builtins.
2586 (version_loop_by_alias_check): New parameter. Generate cancelable
2587 runtime alias check if all partitions are builtins.
2588 (distribute_loop): Update call to above function.
2590 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2592 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2593 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2594 (parm_default_def_partition_arg): Ditto.
2595 (set_parm_default_def_partition): Ditto.
2596 (get_parm_default_def_partitions): Ditto and make it static.
2597 (get_undefined_value_partitions): Ditto and make it static.
2598 (remove_ssa_form): Refactor call to init_var_map here.
2599 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2600 computation for loop region.
2601 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2602 (register_default_def): Delete.
2603 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2604 (parm_default_def_partition_arg): Ditto.
2605 (set_parm_default_def_partition): Ditto.
2606 (get_parm_default_def_partitions): Ditto and make it static.
2607 (get_undefined_value_partitions): Ditto and make it static.
2608 (coalesce_with_default, coalesce_with_default): Update comment.
2609 (create_coalesce_list_for_region): New func factored out from
2610 create_outofssa_var_map.
2611 (populate_coalesce_list_for_outofssa): New func factored out from
2612 create_outofssa_var_map and coalesce_ssa_name.
2613 (create_outofssa_var_map): Delete.
2614 (coalesce_ssa_name): Refactor to support live range computation.
2615 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2616 (get_parm_default_def_partitions): Delete.
2617 (get_undefined_value_partitions): Ditto.
2618 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2619 computation for loop region.
2620 (new_tree_live_info, loe_visit_block): Ditto.
2621 (live_worklist, set_var_live_on_entry): Ditto.
2622 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2623 * tree-ssa-live.h (struct _var_map): New fields.
2624 (init_var_map): Change decl.
2625 (region_contains_p): New.
2627 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2629 * tree-ssa-live.h (live_merge_and_clear): Delete.
2631 2018-05-25 Richard Biener <rguenther@suse.de>
2634 * tree-dump.c (dequeue_and_dump): Remove access to removed
2635 operand 2 of a SWITCH_EXPR.
2637 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2639 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2640 multiplication and division.
2641 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2642 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2643 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2644 (cond_udiv_optab, cond_umod_optab): New optabs.
2645 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2646 (IFN_COND_RDIV): New internal functions.
2647 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2648 TRUNC_MOD_EXPR and RDIV_EXPR.
2649 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2650 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2652 (SVE_INT_BINARY): Include mult.
2653 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2654 (optab, sve_int_op): Handle mult.
2655 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2657 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2658 for SVE_INT_BINARY_SD.
2660 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2662 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2663 (optab, sve_int_op): Handle div and udiv.
2664 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2665 for SVE_INT_BINARY_SD.
2666 (*<optab><mode>3): New insn for the same.
2668 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2670 * tree-vect-patterns.c: Include predict.h.
2671 (vect_recog_divmod_pattern): Restrict check for division support
2672 to when optimizing for size.
2674 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2676 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2677 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2678 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2679 (gimple_match_op::set_op): Likewise.
2680 (gimple_resimplify4): Declare.
2681 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2682 (expr::gen_transform): Likewise.
2683 (decision_tree::gen): Generate a simplification routine for 4 operands.
2684 * gimple-match-head.c (gimple_simplify): Add an overload for
2685 4 operands. In the top-level function, handle up to 4 call
2686 arguments and call gimple_resimplify4.
2687 (gimple_resimplify4): New function.
2688 (build_call_internal): Pass a fourth operand.
2689 (maybe_push_to_seq): Likewise.
2690 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2691 Fold VEC_COND_EXPRs of an operation and a default value into
2692 an IFN_COND_* function if possible.
2693 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2695 (SVE_COND_FP_BINARY): Include them.
2696 (optab, sve_fp_op): Handle them.
2697 (SVE_INT_BINARY_REV): New code iterator.
2698 (SVE_COND_FP_BINARY_REV): New int iterator.
2699 (commutative): New int attribute.
2700 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2702 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2704 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2705 (*cond_<optab><mode>): New patterns for reversed operands.
2707 2018-05-25 Richard Biener <rguenther@suse.de>
2709 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2710 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2711 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2712 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2713 * tree-vect-data-refs.c (everywhere): Adjust users.
2714 * tree-vect-loop.c (everywhere): Likewise.
2715 * tree-vect-slp.c (everywhere): Likewise.
2716 * tree-vect-stmts.c (everywhere): Likewise.
2717 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2719 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2721 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2723 (gcc_cv_as_section_exclude): ... this.
2724 Try Solaris as #exclude syntax.
2725 * configure: Regenerate.
2726 * config.in: Regenerate.
2727 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2729 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2730 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2732 * varasm.c (default_elf_asm_named_section): Don't check if
2733 HAVE_GAS_SECTION_EXCLUDE is defined.
2735 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2737 * doc/md.texi: Update the documentation of the cond_* optabs
2738 to mention the new final operand. Fix GET_MODE_NUNITS call.
2739 Describe the scalar case too.
2740 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2741 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2743 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2744 (get_conditional_internal_fn): Update comment.
2745 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2746 accumulator value as a final argument to conditional functions.
2747 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2748 a define_expand and add an "else" operand. Assert for now that
2749 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2750 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2751 (*cond_<optab><mode>): New patterns.
2752 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2753 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2754 (UNSPEC_COND_EOR): Delete.
2755 (optab): Remove associated mappings.
2756 (SVE_INT_BINARY): New code iterator.
2757 (sve_int_op): Remove int attribute and add "minus" to the code
2759 (SVE_COND_INT_OP): Delete.
2760 (SVE_COND_FP_OP): Rename to...
2761 (SVE_COND_FP_BINARY): ...this.
2763 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2765 * optabs.c (can_reuse_operands_p): New function.
2766 (maybe_legitimize_operands): Try to reuse the results for
2769 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2771 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2772 Add {q} suffix to insn mnemonic.
2774 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2776 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2777 (msp430_warn_func_return): New.
2779 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2781 * fold-const.c (tree_nonzero_bits): New function.
2782 * fold-const.h (tree_nonzero_bits): Likewise.
2783 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2784 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2786 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2790 * varasm.c (assemble_alias): Check ifunc_resolver only on
2793 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2796 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2797 when memory input operand is handled.
2799 2018-05-24 Luis Machado <luis.machado@linaro.org>
2801 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2803 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2805 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2807 * match.pd: Delay FMA folds until after vectorization.
2809 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2812 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2813 address check not strict.
2815 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2817 * gimple-match.h (gimple_match_op): New class.
2818 (mprts_hook): Replace parameters with a gimple_match_op *.
2819 (maybe_build_generic_op): Likewise.
2820 (gimple_simplified_result_is_gimple_val): Replace parameters with
2821 a const gimple_match_op *.
2822 (gimple_simplify): Replace code_helper * and tree * parameters with
2823 a gimple_match_op * parameter.
2824 (gimple_resimplify1): Replace code_helper *, tree and tree *
2825 parameters with a gimple_match_op * parameter.
2826 (gimple_resimplify2): Likewise.
2827 (gimple_resimplify3): Likewise.
2828 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2829 parameters with a gimple_match_op * parameter.
2830 * gimple-match-head.c (gimple_simplify): Change prototypes of
2831 auto-generated functions to take a gimple_match_op * instead of
2832 separate code_helper * and tree * parameters. Make the same
2833 change in the top-level overload and update calls to the
2834 gimple_resimplify routines. Update calls to the auto-generated
2835 functions and to maybe_push_res_to_seq in the publicly-facing
2836 operation-specific gimple_simplify overloads.
2837 (gimple_match_op::MAX_NUM_OPS): Define.
2838 (gimple_resimplify1): Replace rcode and ops with a single res_op
2839 parameter. Update call to gimple_simplify.
2840 (gimple_resimplify2): Likewise.
2841 (gimple_resimplify3): Likewise.
2842 (mprts_hook): Replace parameters with a gimple_match_op *.
2843 (maybe_build_generic_op): Likewise.
2844 (build_call_internal): Replace type, nargs and ops with
2845 a gimple_match_op *.
2846 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2847 with a single gimple_match_op *. Update calls to mprts_hook,
2848 build_call_internal and gimple_simplified_result_is_gimple_val.
2849 Factor out code that is common to the tree_code and combined_fn cases.
2850 * genmatch.c (expr::gen_transform): Replace tem_code and
2851 tem_ops with a gimple_match_op called tem_op. Update calls
2852 to the gimple_resimplify functions and maybe_push_res_to_seq.
2853 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2854 res_ops. Update call to the gimple_resimplify functions.
2855 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2856 (decision_tree::gen): Make the functions take a gimple_match_op *
2857 called res_op instead of separate res_code and res_ops parameters.
2858 Update call accordingly.
2859 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2860 and ops with a single res_op parameter. Update calls to
2861 maybe_build_generic_op and maybe_push_res_to_seq.
2862 (fold_stmt_1): Update calls to gimple_simplify and
2863 replace_stmt_with_simplification.
2864 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2865 and gimple_simplified_result_is_gimple_val.
2866 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2868 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2869 with a gimple_match_op *.
2870 (vn_nary_build_or_lookup): Likewise. Update call to
2871 vn_nary_build_or_lookup_1.
2872 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2873 gimple_match_op *. Update calls to the gimple_resimplify routines
2874 and to gimple_simplified_result_is_gimple_val.
2875 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2876 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2877 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2878 (visit_nary_op): Likewise.
2879 (visit_reference_op_load): Likewise.
2881 2018-05-23 Luis Machado <luis.machado@linaro.org>
2883 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2884 modifier for printing the step amount.
2886 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2889 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2892 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2894 * doc/sourcebuild.texi (Endianness): New subsubsection.
2896 2018-05-23 Luis Machado <luis.machado@linaro.org>
2898 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2899 <prefetch_dynamic_strides>: New const bool field.
2900 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2901 prefetch_dynamic_strides.
2902 (exynosm1_prefetch_tune): Likewise.
2903 (thunderxt88_prefetch_tune): Likewise.
2904 (thunderx_prefetch_tune): Likewise.
2905 (thunderx2t99_prefetch_tune): Likewise.
2906 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2908 (aarch64_override_options_internal): Update to set
2909 PARAM_PREFETCH_DYNAMIC_STRIDES.
2910 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2911 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2912 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2913 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2914 prefetch-dynamic-strides setting.
2916 2018-05-23 Luis Machado <luis.machado@linaro.org>
2918 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2919 <minimum_stride>: New const int field.
2920 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2921 minimum_stride field defaulting to -1.
2922 (exynosm1_prefetch_tune): Likewise.
2923 (thunderxt88_prefetch_tune): Likewise.
2924 (thunderx_prefetch_tune): Likewise.
2925 (thunderx2t99_prefetch_tune): Likewise.
2926 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2927 <default_opt_level>: Set to 3.
2928 (aarch64_override_options_internal): Update to set
2929 PARAM_PREFETCH_MINIMUM_STRIDE.
2930 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2931 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2932 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2933 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2934 stride is constant and is below the minimum stride threshold.
2936 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2938 * config/arm/arm-cpus.in (mode26): Delete.
2939 (armv4): Delete mode26 reference.
2940 * config/arm/arm.c (arm_configure_build_target): Delete use of
2943 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2945 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2947 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2948 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2950 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2951 (floatunsdidf2): Ditto.
2953 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2955 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2956 (fixuns_trunc<mode>si2_avx512f): Ditto.
2957 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2958 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2959 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2961 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2963 PR rtl-optimization/79985
2964 * df-scan.c (df_insn_refs_collect): Remove special case for
2965 global registers and asm statements.
2967 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2969 * extend.texi (Global Register Variables): Rewrite the bullet list.
2970 Note that the register is available for allocation. Note that access
2971 via inline asm must use constraints. Add note about async-signal
2972 handlers. Remove paragraph about automagic register selection.
2974 2018-05-23 Richard Biener <rguenther@suse.de>
2976 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2977 of fixed offset from memset VN.
2979 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2981 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2983 (alloc_cand_and_find_basis): Initialize first_interp field.
2984 (slsr_process_mul): Modify first_interp field.
2985 (slsr_process_add): Likewise.
2986 (slsr_process_cast): Modify first_interp field for each new
2988 (slsr_process_copy): Likewise.
2989 (dump_candidate): Dump first_interp field.
2990 (replace_mult_candidate): Process all interpretations, not just
2992 (replace_rhs_if_not_dup): Likewise.
2993 (replace_one_candidate): Likewise.
2995 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2997 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2999 (aarch64_needs_frame_chain): New function.
3000 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3002 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3005 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3006 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3007 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3008 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3009 as true for strict-align.
3010 (aarch64_can_inline_p): Perform checks even when callee has no
3011 attributes to check for strict alignment.
3012 * doc/extend.texi (AArch64 Function Attributes): Document
3014 * doc/invoke.texi: (AArch64 Options): Likewise.
3016 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3018 PR tree-optimization/85853
3019 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3020 the handling of the root of the node to...
3021 (vect_slp_analyze_node_operations_1): ...this new function,
3022 and run the whole thing with the child nodes' def types
3023 set according to their SLP node's def type.
3025 2018-05-23 Richard Biener <rguenther@suse.de>
3028 * tree-data-ref.c (create_runtime_alias_checks): Defer
3029 and ignore overflow warnings.
3031 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3033 PR tree-optimization/85822
3034 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3037 2018-05-23 Richard Biener <rguenther@suse.de>
3039 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3040 memset constants via native_interpret_expr.
3042 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3045 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3047 (cgraph_node::create_alias): Likewise.
3048 (cgraph_node::get_availability): Check ifunc_resolver instead
3049 of looking up ifunc attribute.
3050 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3051 * varasm.c (do_assemble_alias): Likewise.
3052 (assemble_alias): Likewise.
3053 (default_binds_local_p_3): Likewise.
3054 * cgraph.h (cgraph_node): Add ifunc_resolver.
3055 (cgraph_node::only_called_directly_or_aliased_p): Return false
3057 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3059 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3060 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3061 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3062 instead of looking up ifunc attribute.
3064 2018-05-22 Luis Machado <luis.machado@linaro.org>
3066 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3068 2018-05-22 Martin Sebor <msebor@redhat.com>
3071 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3072 only when expasion succeeds.
3073 (expand_builtin_strcmp): Same.
3074 (expand_builtin_strncmp): Same.
3076 2018-05-22 Martin Sebor <msebor@redhat.com>
3078 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3080 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3081 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3083 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3084 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3086 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3087 Add check for legitimate_address.
3088 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3089 (aarch64_swap_ldrstr_operands): New.
3090 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3093 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3094 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3096 * config/aarch64/aarch64.md: New patterns to generate stp
3098 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3099 single words and double words.
3100 (load_pair_sw, load_pair_dw): Likewise.
3101 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3103 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3105 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3106 for different mode ldpstp and add peephole for merged zero stores.
3108 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3110 (aarch64_gen_store_pair): Rename calls to match new patterns.
3111 (aarch64_gen_load_pair): Rename calls to match new patterns.
3112 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3113 (load_pair<DREG:mode><DREG2:mode>): ... This.
3114 (store_pair<mode>): Rename to...
3115 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3116 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3118 (V_INT_EQUIV): Handle SImode.
3119 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3122 2018-05-22 Martin Sebor <msebor@redhat.com>
3125 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3126 or ajust the presumed bound on an operation to avoid unnecessary
3129 2018-05-22 Martin Sebor <msebor@redhat.com>
3131 PR tree-optimization/85826
3132 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3133 assuming that a DECL necesarily has a constant size.
3135 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3138 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3140 2018-05-22 Richard Biener <rguenther@suse.de>
3142 PR tree-optimization/85834
3143 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3144 non-constant and non-zero memset arguments.
3146 2018-05-22 Martin Liska <mliska@suse.cz>
3149 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3151 2018-05-22 Richard Biener <rguenther@suse.de>
3153 PR tree-optimization/85863
3154 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3155 comparisons when vectype is specified.
3156 (vectorizable_condition): Do not specify vectype for
3157 vect_is_simple_cond when SLP vectorizing.
3159 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3162 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3163 define __ibm128 as long double.
3164 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3165 as a distinct type when IEEE 128-bit support is enabled.
3166 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3167 128-bit types to use the correct functions.
3168 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3169 convert between 128-bit floating point types that have different
3170 modes but the same representation, instead of using gen_lowpart to
3172 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3174 (IFKF_reg): New attributes to give the register constraints for
3176 (extend<mode>tf2_internal): New insns to mark an explicit
3177 conversion between 128-bit floating point types that have a
3178 different mode but share the same representation.
3180 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3182 PR tree-optimization/85814
3183 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3184 a null return from get_strinfo when unsharing the next
3185 strinfo in the chain.
3187 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3190 * varasm.c (weak_finish): Clean up weak_decls.
3192 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3194 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3195 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3196 UNSPEC_UADALP values.
3197 * config/aarch64/iterators.md (ABAL): New int iterator.
3200 (sur): Add mappings for the above.
3201 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3203 (aarch64_<sur>abal<mode>_4): Likewise.
3204 (aarch64_<sur>adalp<mode>_3): Likewise.
3205 (<sur>sadv16qi): New define_expand.
3207 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3209 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3210 (*movdf_internal): Ditto.
3211 (*rcpsf2_sse): Ditto.
3212 (*rsqrtsf2_sse): Ditto.
3213 (*sqrt<mode>2_sse): Ditto.
3215 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3217 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3219 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3220 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3221 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3222 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3224 * config/aarch64/arm_neon.h: Likewise.
3225 * config/aarch64/iterators.md (VQ_I): New.
3227 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3229 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3231 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3234 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3236 * config/nds32/constraints.md (S): New constraint.
3237 * config/nds32/nds32.md (call_internal): Use constraint S.
3238 (call_value_internal): Likewise.
3239 (sibcall_internal): Likewise.
3240 (sibcall_value_internal): Likewise.
3242 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3243 Chung-Ju Wu <jasonwucj@gmail.com>
3245 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3248 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3249 Chung-Ju Wu <jasonwucj@gmail.com>
3251 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3252 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3253 (nds32_rtx_costs_impl): Simplify.
3254 (nds32_address_cost_impl): Simplify.
3255 (nds32_init_rtx_costs): New function.
3256 (nds32_rtx_costs_speed_prefer): Likewise.
3257 (nds32_rtx_costs_size_prefer): Likewise.
3258 (nds32_address_cost_speed_prefer): Likewise.
3259 (nds32_address_cost_speed_fwprop): Likewise.
3260 (nds32_address_cost_size_prefer): Likewise.
3261 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3262 * config/nds32/nds32.c (nds32_option_override): Use
3263 nds32_init_rtx_costs function.
3265 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3267 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3268 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3269 (TARGET_PIPELINE_N8): Likewise.
3270 (TARGET_PIPELINE_N10): Likewise.
3271 (TARGET_PIPELINE_N13): Likewise.
3272 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3274 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3276 * config/nds32/nds32-fpu.md: Update copyright year.
3278 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3280 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3282 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3284 * config/nds32/nds32.c
3285 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3286 * config/nds32/nds32.opt (minline-asm-r15): New option.
3288 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3290 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3292 * config/nds32/nds32.md (abssi2): New pattern.
3294 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3296 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3297 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3298 Merge insn pattern from sse_cvtsi2ss<round_name> and
3299 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3300 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3301 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3302 using SWI48 mode iterator.
3303 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3304 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3305 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3306 pattern from sse_cvttss2si<round_saeonly_name>
3307 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3308 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3309 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3310 using SWI48 mode iterator.
3311 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3312 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3313 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3314 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3315 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3316 using SWI48 mode iterator.
3317 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3318 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3319 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3320 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3321 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3322 SWI48 mode iterator.
3323 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3324 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3325 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3326 pattern from sse_cvttsd2si<round_saeonly_name>
3327 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3329 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3331 * config/nds32/nds32-md-auxiliary.c
3332 (nds32_valid_smw_lwm_base_p): Refine.
3333 (nds32_output_smw_single_word): Refine.
3334 (nds32_output_smw_double_word): New.
3335 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3337 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3339 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3340 (nds32_output_stack_pop): Refine.
3341 (nds32_expand_unaligned_load): Refine.
3342 (nds32_expand_unaligned_store): Refine.
3344 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3345 Chung-Ju Wu <jasonwucj@gmail.com>
3347 * config/nds32/constants.md: Add TP_REGNUM constant.
3348 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3349 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3351 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3352 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3353 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3354 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3355 and PIC code generation.
3356 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3358 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3360 * config/nds32/nds32.md: Support TLS and PIC.
3361 * config/nds32/nds32.c: Support TLS and PIC.
3362 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3363 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3366 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3368 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3369 mode with E_ prefix.
3371 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3372 Chung-Ju Wu <jasonwucj@gmail.com>
3374 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3375 * config/nds32/nds32-md-auxiliary.c
3376 (symbolic_reference_mentioned_p): New.
3377 (nds32_legitimize_ict_address): New.
3378 (nds32_expand_ict_move): New.
3379 (nds32_indirect_call_referenced_p): New.
3380 (nds32_symbol_binds_local_p): Delete.
3381 (nds32_long_call_p): Modify.
3382 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3383 * config/nds32/nds32-protos.h
3384 (symbolic_reference_mentioned_p): Declare.
3385 (nds32_legitimize_ict_address): Declare.
3386 (nds32_expand_ict_move): Declare.
3387 (nds32_indirect_call_referenced_p): Declare.
3388 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3389 (nds32_relax_group): Use nds32_ict_const_p as condition.
3390 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3391 (nds32_asm_file_start): Output ict_model directive in asm code.
3392 (nds32_legitimate_address_p): Consider indirect call.
3393 (nds32_print_operand): Consider indirect call.
3394 (nds32_print_operand_address): Consider indirect call.
3395 (nds32_insert_attributes): Handle "indirect_call" attribute.
3396 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3397 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3398 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3399 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3400 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3401 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3402 (TARGET_ICT_MODEL_SMALL): Define.
3403 (TARGET_ICT_MODEL_LARGE): Define.
3404 * config/nds32/nds32.md (movsi): Consider ict model.
3405 (call, call_value): Consider ict model.
3406 (sibcall, sibcall_value): Consider ict model.
3407 * config/nds32/nds32.opt (mict-model): New option.
3408 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3411 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3412 Monk Chiang <sh.chiang04@gmail.com>
3413 Jim Wilson <jimw@sifive.com>
3415 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3416 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3417 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3418 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3419 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3420 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3421 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3422 compute save_libcall_adjustment properly.
3423 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3424 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3425 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3426 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3427 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3428 (ABI_SPEC): Handle mabi=ilp32e.
3429 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3430 (RVE): Add RVE mask.
3431 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3432 <-march>: Add rv32e as an example.
3434 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3437 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3438 (intra_create_variable_infos): Handle C++ constructors.
3440 2018-05-18 Martin Liska <mliska@suse.cz>
3442 * passes.def: Remove a redundant pass.
3444 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3447 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3449 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3451 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3453 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3455 (armv2, armv2a, armv3, armv3m): Delete architectures.
3456 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3457 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3458 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3460 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3461 (*mulsidi3adddi): Likewise.
3462 (mulsidi3): Likewise.
3463 (*mulsidi3_nov6): Likewise.
3464 (umulsidi3): Likewise.
3465 (umulsidi3_nov6): Likewise.
3466 (umaddsidi4): Likewise.
3467 (*umulsidi3adddi): Likewise.
3468 (smulsi3_highpart): Likewise.
3469 (*smulsi3_highpart_nov6): Likewise.
3470 (umulsi3_highpart): Likewise.
3471 (*umulsi3_highpart_nov6): Likewise.
3472 * config/arm/arm.h (arm_arch3m): Delete.
3473 * config/arm/arm.c (arm_arch3m): Delete.
3474 (arm_option_override_internal): Update armv3-related comment.
3475 (arm_configure_build_target): Delete use of isa_bit_mode32.
3476 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3477 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3478 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3480 (mulusa3): Likewise.
3481 * config/arm/arm-protos.h (arm_arch3m): Delete.
3482 * config/arm/arm-tables.opt: Regenerate.
3483 * config/arm/arm-tune.md: Likewise.
3484 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3485 deleted architectures.
3487 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3489 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3490 (armv5t, armv5te): New features.
3491 (ARMv5, ARMv5e): Delete fgroups.
3492 (ARMv5t, ARMv5te): Adjust for above changes.
3494 (armv5, armv5e): Delete arches.
3495 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3497 (*call_reg_arm): Likewise.
3498 (*call_value_reg_armv5): Likewise.
3499 (*call_value_reg_arm): Likewise.
3500 (*call_symbol): Likewise.
3501 (*call_value_symbol): Likewise.
3502 (*sibcall_insn): Likewise.
3503 (*sibcall_value_insn): Likewise.
3505 (prefetch): Likewise.
3506 (define_split and define_peephole2 dependent on arm_arch5):
3508 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3510 (TARGET_ARM_QBIT): Likewise.
3511 (TARGET_DSP_MULTIPLY): Likewise.
3512 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3513 (arm_arch5, arm_arch5e): Delete.
3514 (arm_arch5t, arm_arch5te): Declare.
3515 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3516 (arm_arch5t): Declare.
3517 (arm_option_reconfigure_globals): Update for the above.
3518 (arm_options_perform_arch_sanity_checks): Update comment, replace
3519 use of arm_arch5 with arm_arch5t.
3520 (use_return_insn): Likewise.
3521 (arm_emit_call_insn): Likewise.
3522 (output_return_instruction): Likewise.
3523 (arm_final_prescan_insn): Likewise.
3524 (arm_coproc_builtin_available): Likewise.
3525 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3526 arm_arch5e with arm_arch5t and arm_arch5te.
3527 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3528 (arm_arch5t, arm_arch5te): Declare.
3529 * config/arm/arm-tables.opt: Regenerate.
3530 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3531 * config/arm/t-multilib: Likewise.
3532 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3533 instead of arm_arch5.
3534 (*call_reg_thumb1): Likewise.
3535 (*call_value_reg_thumb1_v5): Likewise.
3536 (*call_value_reg_thumb1): Likewise.
3537 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3539 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3541 2018-05-18 Martin Liska <mliska@suse.cz>
3543 PR gcov-profile/84846
3544 * doc/gcov.texi: Document -t option of gcov tool.
3546 2018-05-18 Martin Liska <mliska@suse.cz>
3548 PR gcov-profile/84846
3549 * gcov.c (print_usage): Add new -t option.
3550 (process_args): Handle the option.
3551 (generate_results): Use stdout as output when requested by
3554 2018-05-18 Martin Liska <mliska@suse.cz>
3556 PR gcov-profile/84846
3557 * coverage.c (coverage_init): Write PWD to .gcno file.
3558 * doc/gcov.texi: Document how working directory is printed.
3559 * gcov-dump.c (dump_gcov_file): Print PWD.
3560 * gcov.c (output_intermediate_file): Likewise.
3561 (read_graph_file): Read PWD string.
3562 (output_lines): Print PWD.
3564 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3567 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3568 for retval and return false if all args to phi are zero.
3570 2018-05-18 Richard Biener <rguenther@suse.de>
3572 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3574 (evrp_dom_walker::before_dom_children): Call it.
3576 2018-05-18 Richard Biener <rguenther@suse.de>
3578 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3579 results when processing array refs with variable index.
3581 2018-05-18 Toon Moene <toon@moene.org>
3583 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3584 directly after that of -floop-interchange. Indicate that both
3585 options are enabled by default when specifying -O3.
3587 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3589 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3590 iterator. Delete separate integer-mode vec_set<mode> expander.
3591 (aarch64_simd_vec_setv2di): Delete.
3592 (vec_setv2di): Delete.
3593 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3594 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3595 the "w, r" alternative.
3597 2018-05-18 Martin Liska <mliska@suse.cz>
3599 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3600 * tree-pass.h (make_pass_lower_switch_O0): New function.
3601 * tree-switch-conversion.c (node_has_low_bound): Remove.
3602 (node_has_high_bound): Likewise.
3603 (node_is_bounded): Likewise.
3604 (class pass_lower_switch): Make it a template type and create
3606 (pass_lower_switch::execute): Add template argument.
3607 (make_pass_lower_switch): New function.
3608 (make_pass_lower_switch_O0): New function.
3609 (do_jump_if_equal): Remove.
3610 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3611 all the hard work to tree optimization passes.
3613 2018-05-18 Martin Liska <mliska@suse.cz>
3615 * dbgcnt.c (limit_low): Renamed from limit.
3616 (limit_high): New variable.
3617 (dbg_cnt_is_enabled): Check for upper limit.
3618 (dbg_cnt): Adjust dumping.
3619 (dbg_cnt_set_limit_by_index): Add new argument for high
3621 (dbg_cnt_set_limit_by_name): Likewise.
3622 (dbg_cnt_process_single_pair): Parse new format.
3623 (dbg_cnt_process_opt): Use strtok.
3624 (dbg_cnt_list_all_counters): Remove 'value' and add
3626 * doc/invoke.texi: Document changes.
3628 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3630 * doc/sourcebuild.texi (scalar_all_fma): Document.
3631 * tree.def (FMA_EXPR): Delete.
3632 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3633 * internal-fn.c (ternary_direct): New macro.
3634 (expand_ternary_optab_fn): Likewise.
3635 (direct_ternary_optab_supported_p): Likewise.
3636 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3637 * builtins.c (fold_builtin_fma): Delete.
3638 (fold_builtin_3): Don't call it.
3639 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3640 * expr.c (expand_expr_real_2): Likewise.
3641 * fold-const.c (operand_equal_p): Likewise.
3642 (fold_ternary_loc): Likewise.
3643 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3644 * gimple.c (DEFTREECODE): Likewise.
3645 * gimplify.c (gimplify_expr): Likewise.
3646 * optabs-tree.c (optab_for_tree_code): Likewise.
3647 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3648 * tree-eh.c (operation_could_trap_p): Likewise.
3649 (stmt_could_throw_1_p): Likewise.
3650 * tree-inline.c (estimate_operator_cost): Likewise.
3651 * tree-pretty-print.c (dump_generic_node): Likewise.
3652 (op_code_prio): Likewise.
3653 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3654 * tree-ssa-operands.c (get_expr_operands): Likewise.
3655 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3656 * fold-const-call.h (fold_fma): Delete.
3657 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3658 CFN_FNMA and CFN_FNMS.
3660 * genmatch.c (combined_fn): New enum.
3661 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3662 (commutative_op): New function.
3663 (commutate): Use it. Handle more than 2 operands.
3664 (dt_operand::gen_gimple_expr): Use commutative_op.
3665 (parser::parse_expr): Allow :c to be used with non-binary
3666 operators if the commutative operand is known.
3667 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3668 CFN_FMS, CFN_FNMA and CFN_FNMS.
3669 (backprop::process_assign_use): Remove FMA_EXPR handling.
3670 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3671 (gen_hsa_fma): New function.
3672 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3673 IFN_FNMA and IFN_FNMS.
3674 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3675 * gimple-fold.h (follow_all_ssa_edges): Declare.
3676 * gimple-fold.c (follow_all_ssa_edges): New function.
3677 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3678 gimple_build interface and use follow_all_ssa_edges to fold the result.
3679 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3680 instead of checking for optabs directly.
3681 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3682 rather than FMA_EXPRs.
3683 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3684 call to IFN_FMA instead of an FMA_EXPR.
3686 2018-05-17 Jim Wilson <jimw@sifive.com>
3688 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3689 sign extended promoted subreg, and the range does not have the sign bit
3690 set, then do a sign extend.
3692 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3693 test, check for sign extended subreg and/or constant operands, and
3694 do a sign extend in that case.
3696 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3698 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3699 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3701 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3702 Change logics_shift_reg to logics_shift_imm.
3703 (thunderx2t99_fp_loadpair_basic): Delete.
3704 (thunderx2t99_fp_storepair_basic): Delete.
3705 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3706 (thunderx2t99_asimd_polynomial): Delete.
3707 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3708 and neon_fp_mul_d_scalar_q.
3709 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3710 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3711 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3712 (thunderx2t99_asimd_lut): Add missing tbl types.
3713 (thunderx2t99_asimd_ext): Delete.
3714 (thunderx2t99_asimd_load1_1_mult): Delete.
3715 (thunderx2t99_asimd_load1_2_mult): Delete.
3716 (thunderx2t99_asimd_load1_ldp): New.
3717 (thunderx2t99_asimd_load1): New.
3718 (thunderx2t99_asimd_load2): Add missing *load2* types.
3719 (thunderx2t99_asimd_load3): New.
3720 (thunderx2t99_asimd_load4): New.
3721 (thunderx2t99_asimd_store1_1_mult): Delete.
3722 (thunderx2t99_asimd_store1_2_mult): Delete.
3723 (thunderx2t99_asimd_store2_mult): Delete.
3724 (thunderx2t99_asimd_store2_onelane): Delete.
3725 (thunderx2t99_asimd_store_stp): New.
3726 (thunderx2t99_asimd_store1): New.
3727 (thunderx2t99_asimd_store2): New.
3728 (thunderx2t99_asimd_store3): New.
3729 (thunderx2t99_asimd_store4): New.
3731 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3733 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3734 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3736 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3737 Segher Boessenkool <segher@kernel.crashing.org>
3740 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3743 2018-05-17 Richard Biener <rguenther@suse.de>
3745 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3746 for pruning loop and prune defs feeding only already visited PHIs.
3748 2018-05-17 Richard Biener <rguenther@suse.de>
3750 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3752 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3753 Richard Biener <rguenther@suse.de>
3755 PR tree-optimization/85793
3756 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3757 for VMAT_ELEMENTWISE.
3759 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3761 * internal-fn.h (lookup_internal_fn): Declare
3762 * internal-fn.c (lookup_internal_fn): New function.
3763 * gimple.c (gimple_build_call_from_tree): Handle calls to
3765 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3766 internal function names.
3767 * tree-pretty-print.c (dump_generic_node): Likewise.
3768 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3770 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3772 * gimple-fold.h (gimple_build): Make the function forms take
3773 combined_fn rather than built_in_function.
3774 (gimple_simplify): Likewise.
3775 * gimple-match-head.c (gimple_simplify): Likewise.
3776 * gimple-fold.c (gimple_build): Likewise.
3777 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3778 rather than gimple_build_call_internal.
3779 (get_initial_defs_for_reduction): Likewise.
3780 (vect_create_epilog_for_reduction): Likewise.
3781 (vectorizable_live_operation): Likewise.
3783 2018-05-17 Martin Liska <mliska@suse.cz>
3785 * gimple-ssa-sprintf.c (format_directive): Do not use
3786 space in between 'G_' and '('.
3788 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3791 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3792 even if the mask is not all ones.
3795 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3797 (ix86_gimple_fold_builtin): Likewise.
3800 * config/i386/i386.c: Include tree-vector-builder.h.
3801 (ix86_vector_shift_count): New function.
3802 (ix86_fold_builtin): Fold shift builtins by scalar count.
3803 (ix86_gimple_fold_builtin): Likewise.
3805 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3806 _mm512_setzero): New intrinsics.
3808 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3809 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3811 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3812 code generation for cases where splatting a value is not useful.
3813 * simplify-rtx.c (simplify_ternary_operation): Simplify
3814 vec_merge across a vec_duplicate and a paradoxical subreg forming
3815 a vector mode to a vec_concat.
3817 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3819 * config.gcc: Support "goldmont-plus".
3820 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3822 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3823 PROCESSOR_GOLDMONT_PLUS.
3824 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3825 (processor_target_table): Add "goldmont-plus".
3826 (PTA_GOLDMONT_PLUS): Define.
3827 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3828 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3829 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3830 (fold_builtin_cpu): Add "goldmont-plus".
3831 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3832 (ix86_option_override_internal): Add "goldmont-plus".
3833 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3834 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3835 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3836 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3838 2018-05-17 Richard Biener <rguenther@suse.de>
3840 PR tree-optimization/85757
3841 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3842 remove defs that only feed that PHI from further processing.
3844 2018-05-16 Jim Wilson <jimw@sifive.com>
3846 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3848 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3850 2018-05-16 Mark Wielaard <mark@klomp.org>
3852 * dwarf2out.c (count_index_strings): New function.
3853 (output_indirect_strings): Call count_index_strings and generate
3854 header for dwarf_version >= 5.
3856 2018-05-16 Mark Wielaard <mark@klomp.org>
3858 * dwarf2out.c (dwarf_FORM): New function.
3859 (set_indirect_string): Use dwarf_FORM.
3860 (reset_indirect_string): Likewise.
3861 (size_of_die): Likewise.
3862 (value_format): Likewise.
3863 (output_die): Likewise.
3864 (add_skeleton_AT_string): Likewise.
3865 (output_macinfo_op): Likewise.
3866 (index_string): Likewise.
3867 (output_index_string_offset): Likewise.
3868 (output_index_string): Likewise.
3869 (count_index_strings): Likewise.
3871 2018-05-16 Carl Love <cel@us.ibm.com>
3873 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3874 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3876 2018-05-16 Martin Jambor <mjambor@suse.cz>
3878 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3879 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3881 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3883 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3884 (fnma<mode>4): Likewise.
3885 (fms<mode>4): Likewise.
3886 (fnms<mode>4): Likewise.
3887 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3888 (aarch64_fnma<mode>4): Likewise.
3889 (aarch64_fms<mode>4): Likewise.
3890 (aarch64_fnms<mode>4): Likewise.
3891 (aarch64_fnmadd<mode>4): Likewise.
3893 2018-05-16 Jason Merrill <jason@redhat.com>
3895 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3897 2018-05-16 Richard Biener <rguenther@suse.de>
3899 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3900 (dump_stmt_cost): Declare.
3901 (add_stmt_cost): Dump cost we add.
3902 (add_stmt_costs): New function.
3903 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3905 (vect_analyze_stmt): Adjust prototype.
3906 (vectorizable_condition): Likewise.
3907 (vectorizable_live_operation): Likewise.
3908 (vectorizable_reduction): Likewise.
3909 (vectorizable_induction): Likewise.
3910 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3911 cost vector to pass to vectorizable_ and record afterwards.
3912 (vect_model_reduction_cost): Take cost vector argument and adjust.
3913 (vect_model_induction_cost): Likewise.
3914 (vectorizable_reduction): Likewise.
3915 (vectorizable_induction): Likewise.
3916 (vectorizable_live_operation): Likewise.
3917 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3918 SLP_TREE_NUMBER_OF_VEC_STMTS.
3919 (vect_analyze_slp_cost_1): Remove.
3920 (vect_analyze_slp_cost): Likewise.
3921 (vect_slp_analyze_node_operations): Take visited args and
3922 a target cost vector. Avoid processing already visited stmt sets.
3923 (vect_slp_analyze_operations): Use a local cost vector to gather
3924 costs and register those of non-discarded instances.
3925 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3926 (vect_schedule_slp_instance): Remove copying of
3927 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3929 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3930 adding cost. Record cost entry location.
3931 (vect_prologue_cost_for_slp_op): Function to compute cost of
3932 a constant or invariant generated for SLP vect in the prologue,
3933 split out from vect_analyze_slp_cost_1.
3934 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3935 (vect_model_promotion_demotion_cost): Likewise.
3936 (vect_model_store_cost): Likewise, make static.
3937 (vect_model_load_cost): Likewise.
3938 (vectorizable_bswap): Add cost vector arg and adjust.
3939 (vectorizable_call): Likewise.
3940 (vectorizable_simd_clone_call): Likewise.
3941 (vectorizable_conversion): Likewise.
3942 (vectorizable_assignment): Likewise.
3943 (vectorizable_shift): Likewise.
3944 (vectorizable_operation): Likewise.
3945 (vectorizable_store): Likewise.
3946 (vectorizable_load): Likewise.
3947 (vectorizable_condition): Likewise.
3948 (vectorizable_comparison): Likewise.
3949 (can_vectorize_live_stmts): Likewise.
3950 (vect_analyze_stmt): Likewise.
3951 (vect_transform_stmt): Adjust calls to vectorizable_*.
3952 * tree-vectorizer.c: Include gimple-pretty-print.h.
3953 (dump_stmt_cost): New function.
3955 2018-05-16 Richard Biener <rguenther@suse.de>
3957 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3958 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3959 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3960 (check_name): New callback.
3961 (dse_classify_store): Track cycles via a visited bitmap of PHI
3962 defs and simplify handling of in-loop and across loop dead stores
3963 and properly fail for loop-variant refs. Handle byte-tracking with
3964 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3967 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3969 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3970 (vect_get_mask_type_for_stmt): Likewise.
3971 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3973 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3974 to determine the statement's vector type and the vector type that
3975 should be used for calculating nunits. Deal with cases in which
3976 the type has to be deferred.
3977 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3978 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3979 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3980 (vect_determine_vf_for_stmt): New functions, split out from...
3981 (vect_determine_vectorization_factor): ...here.
3982 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3983 (vect_get_mask_type_for_stmt): New functions, split out from
3984 vect_determine_vectorization_factor.
3986 2018-05-16 Richard Biener <rguenther@suse.de>
3988 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3989 verify the [VEC_]COND_EXPR embedded comparison.
3991 2018-05-15 Martin Sebor <msebor@redhat.com>
3993 PR tree-optimization/85753
3994 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3995 RECORD_TYPE in addition to ARRAY_TYPE.
3997 2018-05-15 Martin Sebor <msebor@redhat.com>
4000 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4002 2018-05-15 Richard Biener <rguenther@suse.de>
4004 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4005 add by_clobber_p one. Change algorithm to collect all defs
4006 representing uses we need to walk and try reducing them to
4007 a single one before failing.
4008 (dse_dom_walker::dse_optimize_stmt): Adjust.
4010 2018-05-13 Mark Wielaard <mark@klomp.org>
4012 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4013 (size_of_loc_descr): Likewise.
4014 (output_loc_operands): Likewise.
4015 (output_loc_operands_raw): Likewise.
4016 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4017 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4018 (hash_loc_operands): Likewise.
4019 (compare_loc_operands): Likewise.
4021 2018-05-14 Mark Wielaard <mark@klomp.org>
4023 * dwarf2out.c (count_index_addrs): New function.
4024 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4026 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4028 PR tree-optimization/83648
4029 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4030 return value as malloc candidate.
4032 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4035 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4036 param as true in call to suggest_attribute.
4038 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4040 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4041 -mreadonly-in-sdata.
4043 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4045 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4047 (aarch64_crypto_aesd_fused): Likewise.
4049 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4051 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4052 (movsi_aarch64): Likewise.
4053 (load_pairsi): Likewise.
4054 (load_pairdi): Likewise.
4055 (store_pairsi): Likewise.
4056 (store_pairdi): Likewise.
4057 (load_pairsf): Likewise.
4058 (load_pairdf): Likewise.
4059 (store_pairsf): Likewise.
4060 (store_pairdf): Likewise.
4061 (zero_extend): Likewise.
4062 (trunc): Swap alternatives.
4063 (fcvt_target): Add '?' to prefer w over r.
4065 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4068 * config/i386/i386.md: Disallow non-commutative arithmetics in
4069 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4070 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4071 in the peephole2 before it.
4073 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4075 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4076 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4077 (ix86_handle_option): Handle -mcldemote.
4078 * config.gcc: New header.
4079 * config/i386/cldemoteintrin.h: New file.
4080 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4081 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4083 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4084 OPTION_MASK_ISA_CLDEMOTE.
4085 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4086 (ix86_valid_target_attribute_inner_p): Ditto.
4087 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4088 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4089 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4090 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4091 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4093 * config/i386/i386.opt: Add -mcldemote.
4094 * config/i386/x86intrin.h: New header.
4095 * doc/invoke.texi: Add -mcldemote.
4097 2018-05-14 Richard Biener <rguenther@suse.de>
4099 * doc/match-and-simplify.texi: Adjust :s documentation.
4101 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4103 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4104 intended memcpy size.
4105 (REORDER_45): Likewise.
4107 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4109 * sort.cc: New file.
4110 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4111 * vec.c (qsort_chk): Use gcc_qsort.
4112 * Makefile.in (OBJS-libcommon): Add sort.o.
4113 (build/sort.o): New target. Use it...
4114 (BUILD_RTL): ... here, and...
4115 (build/gencfn-macros): ... here, and...
4116 (build/genmatch): ... here.
4118 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4119 Chung-Ju Wu <jasonwucj@gmail.com>
4121 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4122 * config/nds32/nds32-graywolf.md: New file.
4123 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4124 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4126 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4127 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4128 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4129 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4130 * config/nds32/pipelines.md: Include n15 settings.
4132 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4133 Chung-Ju Wu <jasonwucj@gmail.com>
4135 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4136 * config/nds32/nds32-n13.md: New file.
4137 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4138 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4140 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4141 * config/nds32/nds32.md (pipeline_model): Add n13.
4142 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4143 * config/nds32/pipelines.md: Include n13 settings.
4145 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4146 Chung-Ju Wu <jasonwucj@gmail.com>
4148 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4149 * config/nds32/nds32-n10.md: New file.
4150 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4151 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4153 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4154 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4155 * config/nds32/nds32.md (pipeline_model): Add n10.
4156 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4157 * config/nds32/pipelines.md: Include n10 settings.
4159 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4160 Kito Cheng <kito.cheng@gmail.com>
4161 Chung-Ju Wu <jasonwucj@gmail.com>
4163 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4164 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4165 Add enum values for DSP extension instructions.
4166 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4168 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4169 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4171 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4172 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4173 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4174 * config/nds32/nds32-intrinsic.md: Likewise.
4175 * config/nds32/nds32_intrinsic.h: Likewise.
4176 * config/nds32/nds32-md-auxiliary.c: Likewise.
4177 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4178 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4179 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4180 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4181 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4182 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4183 TYPE_DMAC in switch statement.
4184 * config/nds32/nds32.c: New checking and implementation for DSP
4185 extension instructions.
4186 * config/nds32/nds32.h: Likewise.
4187 * config/nds32/nds32.md: Likewise.
4188 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4189 * config/nds32/predicates.md: Implement new predicates for DSP
4192 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4194 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4195 Reformat alternatives and attributes so it is easier to identify
4196 which constraints/attributes go with which instruction.
4197 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4198 (mov<mode>_softfloat32, FMOVE64): Likewise.
4199 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4200 (mov<mode>_softfloat64, FMOVE64): Likewise.
4202 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4204 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4206 (Basic PowerPC Built-in Functions): The new name of the
4207 subsection previously known as "PowerPC Built-in Functions".
4208 (Basic PowerPC Built-in Functions Available on all Configurations):
4210 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4211 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4212 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4213 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4215 2018-05-11 Martin Jambor <mjambor@suse.cz>
4218 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4221 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4224 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4226 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4228 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4229 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4230 (ix86_handle_option): Handle -mwaitpkg.
4231 * config.gcc: New header.
4232 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4233 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4234 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4236 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4237 OPTION_MASK_ISA_WAITPKG.
4238 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4239 (ix86_option_override_internal): Add PTA_WAITPKG.
4240 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4241 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4242 IX86_BUILTIN_TPAUSE.
4243 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4244 __builtin_ia32_umwait and __builtin_ia32_tpause.
4245 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4246 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4247 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4248 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4249 UNSPECV_TPAUSE): New.
4250 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4251 * config/i386/i386.opt: Add -mwaitpkg.
4252 * config/i386/waitpkgintrin.h: New file.
4253 * config/i386/x86intrin.h: New header.
4254 * doc/invoke.texi: Add -mwaitpkg.
4256 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4259 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4261 (cortex-m0): Use armv6s-m isa.
4262 (cortex-m0plus): Likewise.
4263 (cortex-m1): Likewise.
4264 (cortex-m0.small-multiply): Likewise.
4265 (cortex-m0plus.small-multiply): Likewise.
4266 (cortex-m1.small-multiply): Likewise.
4268 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4269 Jakub Jelinek <jakub@redhat.com>
4271 PR tree-optimization/85692
4272 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4273 source permute as well.
4275 2018-05-11 Martin Liska <mliska@suse.cz>
4278 * doc/extend.texi: Document LLVM style format for no_sanitize
4281 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4283 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4284 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4285 (mode_supports_vsx_dform_quad): Likewise.
4286 (mode_supports_vmx_dform): Move these functions to be next to the
4287 other mode_supports functions.
4288 (mode_supports_dq_form): Likewise.
4289 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4290 mode_supports_dq_form.
4291 (reg_offset_addressing_ok_p): Likewise.
4292 (offsettable_ok_by_alignment): Likewise.
4293 (rs6000_legitimate_offset_address_p): Likewise.
4294 (legitimate_lo_sum_address_p): Likewise.
4295 (rs6000_legitimize_address): Likewise.
4296 (rs6000_legitimize_reload_address): Likewise.
4297 (rs6000_secondary_reload_inner): Likewise.
4298 (rs6000_preferred_reload_class): Likewise.
4299 (rs6000_output_move_128bit): Likewise.
4301 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4303 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4304 Generate SImode target register for null target.
4305 <case IX86_BUILTIN_XGETBV>: Ditto.
4306 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4307 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4309 2018-05-10 Carl Love <cel@us.ibm.com>
4311 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4312 dcbtt and dcbtstt if operands[2] is 0.
4314 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4317 * config/i386/sse.md (usadv64qi): New expander.
4319 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4321 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4322 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4323 -maltivec=be support.
4324 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4325 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4326 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4327 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4328 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4329 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4330 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4331 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4332 altivec_vsumsws): Adjust.
4333 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4334 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4335 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4337 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4338 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4339 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4340 (altivec_lve<VI_char>x): Delete expand.
4341 (*altivec_lve<VI_char>x_internal): Rename to...
4342 (altivec_lve<VI_char>x): ... this.
4343 (altivec_lvxl_<mode>): Delete expand.
4344 (*altivec_lvxl_<mode>_internal): Rename to ...
4345 (altivec_lvxl_<mode>): ... this.
4346 (altivec_stvxl_<mode>): Delete expand.
4347 (*altivec_stvxl_<mode>_internal): Rename to ...
4348 (altivec_stvxl_<mode>): ... this.
4349 (altivec_stve<VI_char>x): Delete expand.
4350 (*altivec_stve<VI_char>x_internal): Rename to ...
4351 (altivec_stve<VI_char>x): ... this.
4352 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4353 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4354 reduc_plus_scal_<mode>): Adjust.
4355 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4357 (rs6000_cpu_cpp_builtins): Adjust.
4358 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4359 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4360 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4361 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4362 -maltivec=be support.
4363 (rs6000_split_vec_extract_var): Adjust.
4364 (rs6000_split_v4si_init): Adjust.
4365 (swap_selector_for_mode): Delete.
4366 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4367 altivec_expand_stvex_be): Delete.
4368 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4369 -maltivec=be support.
4370 (rs6000_gimple_fold_builtin): Ditto.
4371 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4373 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4374 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4375 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4376 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4377 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4378 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4379 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4380 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4381 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4382 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4383 anonymous split): Adjust.
4384 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4385 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4387 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4389 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4390 when --with-gxx-include-dir is also specified.
4391 * configure: Regenerate.
4393 2018-05-09 Jim Wilson <jimw@sifive.com>
4396 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4397 * config/riscv/t-withmultilib: New.
4398 * config/riscv/withmultilib.h: New.
4399 * doc/install.texi: Document RISC-V --with-multilib-list support.
4401 2018-05-09 Richard Biener <rguenther@suse.de>
4403 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4405 (vect_bb_vectorization_profitable_p): Adjust. Compute
4406 actual scalar cost using the cost vector and the add_stmt_cost
4409 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4411 PR rtl-optimization/85645
4412 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4413 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4415 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4417 PR rtl-optimization/85645
4418 * shrink-wrap.c (spread_components): Return a boolean saying if
4419 anything was changed.
4420 (try_shrink_wrapping_separate): Iterate spread_components until
4421 nothing changes anymore.
4423 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4425 PR rtl-optimization/85645
4426 * regrename.c (build_def_use): Also kill the chains that include the
4427 destination of a REG_CFA_REGISTER note.
4429 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4431 PR rtl-optimization/85645
4432 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4433 insn that has a REG_CFA_REGISTER note.
4435 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4437 * cfgexpand.c (expand_clobber): New function.
4438 (expand_gimple_stmt_1): Use it.
4439 * tree-vect-stmts.c (vect_clobber_variable): New function,
4441 (vectorizable_simd_clone_call): ...here.
4442 (vectorizable_store): Emit a clobber either side of an
4443 IFN_STORE_LANES sequence.
4444 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4446 2018-05-09 Tom de Vries <tom@codesourcery.com>
4449 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4450 (define_insn "trap_if_false"): Add exit after trap.
4452 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4454 PR rtl-optimization/85638
4455 * bb-reorder.c: Include common/common-target.h.
4456 (create_forwarder_block): New function extracted from...
4457 (fix_up_crossing_landing_pad): ...here. Rename into...
4458 (dw2_fix_up_crossing_landing_pad): ...this.
4459 (sjlj_fix_up_crossing_landing_pad): New function.
4460 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4461 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4462 from both partitions and exit the loop after one iteration.
4464 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4467 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4469 (Basic PowerPC Built-in Functions): The new name of the
4470 subsection previously known as "PowerPC Built-in Functions".
4471 (Basic PowerPC Built-in Functions Available on all Configurations):
4473 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4475 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4477 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4479 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4482 2018-05-08 Jim Wilson <jimw@sifive.com>
4484 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4485 (LD_EMUL_SUFFIX): New.
4486 (LINK_SPEC): Use it.
4488 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4490 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4492 (Basic PowerPC Built-in Functions): The new name of the
4493 subsection previously known as "PowerPC Built-in Functions".
4494 (Basic PowerPC Built-in Functions Available on all Configurations):
4496 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4498 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4500 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4502 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4505 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4508 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4509 after cmpelim optimization.
4511 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4513 * config.gcc: Support "goldmont".
4514 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4515 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4517 * config/i386/i386.c (m_GOLDMONT): Define.
4518 (processor_target_table): Add "goldmont".
4519 (PTA_GOLDMONT): Define.
4520 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4521 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4522 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4523 (fold_builtin_cpu): Add "goldmont".
4524 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4525 (ix86_option_override_internal): Add "goldmont".
4526 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4527 (processor_type): Add PROCESSOR_GOLDMONT.
4528 * config/i386/i386.md: Add CPU "glm".
4529 * config/i386/glm.md: New file.
4530 * config/i386/x86-tune.def: Add m_GOLDMONT.
4531 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4533 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4536 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4538 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4539 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4540 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4543 * config/i386/i386.c (ix86_fold_builtin): Handle
4544 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4547 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4548 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4550 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4553 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4554 (check_arch): Likewise.
4555 (check_fpu): Return the result rather than printing it.
4556 (end arch): Fix operator precedence.
4557 (end cpu): Likewise.
4558 (END): Print the result from check_fpu.
4560 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4561 Alan Hayward <alan.hayward@arm.com>
4562 David Sherwood <david.sherwood@arm.com>
4564 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4565 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4566 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4567 (*fcmuo<mode>_and): New patterns.
4569 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4571 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4572 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4573 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4574 (cmp_op, sve_imm_con): New code attributes.
4575 (SVE_COND_INT_CMP, imm_con): Delete.
4576 (cmp_op): Remove above unspecs from int attribute.
4577 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4579 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4580 comparison-specific unspecs.
4581 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4582 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4583 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4584 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4585 (*vec_fcm<cmp_op><mode>): Rename to...
4586 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4587 (*vec_fcmuo<mode>): Rename to...
4588 (*fcmuo<mode>): ...this and adjust likewise.
4589 (*pred_fcm<cmp_op><mode>): New pattern.
4590 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4591 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4593 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4595 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4596 (aarch64_emit_sve_predicated_cond): New function.
4597 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4598 (aarch64_emit_unspec_cond_or): Replace with...
4599 (aarch64_emit_sve_or_conds): ...this new function. Use
4600 aarch64_emit_sve_ptrue_op for the individual comparisons and
4601 aarch64_emit_binop to OR them together.
4602 (aarch64_emit_inverted_unspec_cond): Replace with...
4603 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4604 aarch64_emit_sve_ptrue_op for the comparison and
4605 aarch64_emit_unop to invert the result.
4606 (aarch64_expand_sve_vec_cmp_float): Update after the above
4607 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4609 2018-05-07 Nathan Sidwell <nathan@acm.org>
4611 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4612 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4613 (Backwards Compatibility): Likewise.
4615 2018-05-07 Luis Machado <luis.machado@linaro.org>
4619 2018-05-07 Luis Machado <luis.machado@linaro.org>
4621 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4622 <prefetch_dynamic_strides>: New const bool field.
4623 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4624 prefetch_dynamic_strides.
4625 (exynosm1_prefetch_tune): Likewise.
4626 (thunderxt88_prefetch_tune): Likewise.
4627 (thunderx_prefetch_tune): Likewise.
4628 (thunderx2t99_prefetch_tune): Likewise.
4629 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4631 (aarch64_override_options_internal): Update to set
4632 PARAM_PREFETCH_DYNAMIC_STRIDES.
4633 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4634 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4635 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4636 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4637 prefetch-dynamic-strides setting.
4639 2018-05-07 Luis Machado <luis.machado@linaro.org>
4641 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4642 <minimum_stride>: New const int field.
4643 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4644 minimum_stride field.
4645 (exynosm1_prefetch_tune): Likewise.
4646 (thunderxt88_prefetch_tune): Likewise.
4647 (thunderx_prefetch_tune): Likewise.
4648 (thunderx2t99_prefetch_tune): Likewise.
4649 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4650 (aarch64_override_options_internal): Update to set
4651 PARAM_PREFETCH_MINIMUM_STRIDE.
4652 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4653 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4654 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4655 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4656 stride is constant and is below the minimum stride threshold.
4658 2018-05-07 Luis Machado <luis.machado@linaro.org>
4660 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4663 2018-05-07 Luis Machado <luis.machado@linaro.org>
4665 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4666 <prefetch_dynamic_strides>: New const bool field.
4667 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4668 prefetch_dynamic_strides.
4669 (exynosm1_prefetch_tune): Likewise.
4670 (thunderxt88_prefetch_tune): Likewise.
4671 (thunderx_prefetch_tune): Likewise.
4672 (thunderx2t99_prefetch_tune): Likewise.
4673 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4675 (aarch64_override_options_internal): Update to set
4676 PARAM_PREFETCH_DYNAMIC_STRIDES.
4677 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4678 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4679 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4680 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4681 prefetch-dynamic-strides setting.
4683 2018-05-07 Luis Machado <luis.machado@linaro.org>
4685 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4686 <minimum_stride>: New const int field.
4687 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4688 minimum_stride field.
4689 (exynosm1_prefetch_tune): Likewise.
4690 (thunderxt88_prefetch_tune): Likewise.
4691 (thunderx_prefetch_tune): Likewise.
4692 (thunderx2t99_prefetch_tune): Likewise.
4693 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4694 (aarch64_override_options_internal): Update to set
4695 PARAM_PREFETCH_MINIMUM_STRIDE.
4696 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4697 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4698 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4699 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4700 stride is constant and is below the minimum stride threshold.
4702 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4705 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4706 the type is addressable. Don't force op into register if it has
4709 2018-05-05 Roland McGrath <mcgrathr@google.com>
4712 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4713 any section for which we don't know a specific type it should have,
4714 regardless of name. Previously this was done only for the exact
4715 names ".init_array", ".fini_array", and ".preinit_array".
4716 (default_elf_asm_named_section): Add comment about
4717 relationship with default_section_type_flags and SECTION_NOTYPE.
4718 (get_section): Don't consider it a type conflict if one side has
4719 SECTION_NOTYPE and the other doesn't, as long as neither has the
4720 SECTION_BSS et al used in the default_section_type_flags logic.
4722 2018-05-05 Tom de Vries <tom@codesourcery.com>
4725 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4726 (workaround_barsyncs): New function.
4727 (nvptx_reorg): Use workaround_barsyncs.
4728 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4729 (define_expand "nvptx_membar_cta"): New define_expand.
4730 (define_insn "*nvptx_membar_cta"): New insn.
4732 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4734 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4735 To improve optimization opportunities.
4736 * builtin-types.def: The new needed builtin types for the above.
4738 2018-05-04 Richard Biener <rguenther@suse.de>
4740 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4741 * gimple-ssa-store-merging.c
4742 (imm_store_chain_info::output_merged_store): Remove redundant create,
4743 release split_store vector contents on failure.
4744 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4745 scalar stmt vector on cache hit.
4747 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4749 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4751 * config.gcc (powerpc-xilinx-eabi*): Remove.
4752 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4754 (fusion_addis_mem_combo_load): Ditto.
4755 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4757 (rs6000_cpu_cpp_builtins): Ditto.
4758 * config/rs6000/rs6000-linux.c
4759 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4760 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4761 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4763 (rs6000_setup_reg_addr_masks): Ditto.
4764 (rs6000_init_hard_regno_mode_ok): Ditto.
4765 (rs6000_option_override_internal): Ditto.
4766 (legitimate_lo_sum_address_p): Ditto.
4767 (rs6000_legitimize_address): Ditto.
4768 (rs6000_legitimize_reload_address): Ditto.
4769 (rs6000_legitimate_address_p): Ditto.
4770 (abi_v4_pass_in_fpr): Ditto.
4771 (setup_incoming_varargs): Ditto.
4772 (rs6000_gimplify_va_arg): Ditto.
4773 (rs6000_split_multireg_move): Ditto.
4774 (rs6000_savres_strategy): Ditto.
4775 (rs6000_emit_prologue_components): Ditto.
4776 (rs6000_emit_epilogue_components): Ditto.
4777 (rs6000_emit_prologue): Ditto.
4778 (rs6000_emit_epilogue): Ditto.
4779 (rs6000_elf_file_end): Ditto.
4780 (rs6000_function_value): Ditto.
4781 (rs6000_libcall_value): Ditto.
4782 * config/rs6000/rs6000.h: Ditto.
4783 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4784 (TARGET_MINMAX): ... this. New.
4785 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4786 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4787 (*movsi_internal1_single): Delete.
4788 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4789 mfpu=, mxilinx-fpu): Delete.
4790 * config/rs6000/singlefp.h: Delete.
4791 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4792 * config/rs6000/t-rs6000: Ditto.
4793 * config/rs6000/t-xilinx: Delete.
4794 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4795 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4796 (VStype_simple): Delete.
4797 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4798 * config/rs6000/xfpu.h: Delete.
4799 * config/rs6000/xfpu.md: Delete.
4800 * config/rs6000/xilinx.h: Delete.
4801 * config/rs6000/xilinx.opt: Delete.
4802 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4803 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4805 2018-05-04 Tom de Vries <tom@codesourcery.com>
4808 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4811 2018-05-04 Richard Biener <rguenther@suse.de>
4814 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4815 (expand_complex_div_wide): Likewise.
4816 (expand_complex_operations_1): Likewise.
4817 (expand_complex_libcall): Preserve EH info of the original stmt.
4818 (tree_lower_complex): Handle removed blocks.
4819 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4820 on complex multiplication and division libcall builtins.
4822 2018-05-04 Richard Biener <rguenther@suse.de>
4825 * fold-const.c (negate_expr_p): Restrict negation of operand
4826 zero of a division to when we know that can happen without
4828 (fold_negate_expr_1): Likewise.
4830 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4833 * real.h (real_nextafter): Declare.
4834 * real.c (real_nextafter): New function.
4835 * fold-const-call.c (fold_const_nextafter): New function.
4836 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4837 CASE_CFN_NEXTTOWARD.
4838 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4839 even when arg1_mode is different from arg0_mode.
4841 2018-05-03 Nathan Sidwell <nathan@acm.org>
4843 * doc/extend.texi (Deprecated Features): Remove
4845 (Backwards Compatibility): Likewise.
4846 * doc/invoke.texi (C++ Language Options): Likewise.
4847 (C++ Dialect Options): Likewise.
4849 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4852 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4853 _mm512_mask_mullox_epi64): New intrinsics.
4855 2018-05-03 Tom de Vries <tom@codesourcery.com>
4858 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4859 dump files): Add offload-tree.
4861 2018-05-03 Richard Biener <rguenther@suse.de>
4863 PR tree-optimization/85615
4864 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4865 to loops not nested in BBs loop father to avoid creating multi-entry
4868 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4870 PR tree-optimization/70291
4871 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4872 arguments. Change return type to tree. Emit libcall as a new
4873 statement rather than replacing existing one when inplace_p is true.
4874 (expand_complex_multiplication_components): New function.
4875 (expand_complex_multiplication): Expand floating-point complex
4876 multiplication using the above.
4877 (expand_complex_division): Rename inner_type parameter to type.
4878 Update expand_complex_libcall call-site.
4879 (expand_complex_operations_1): Update expand_complex_multiplication
4880 and expand_complex_division call-sites.
4882 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4885 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4886 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4887 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4888 the highest significant bit of the shift count mask is clear. In
4889 check whether and[sq]i3 is needed verify that all significant bits
4890 of the shift count other than the highest are set.
4892 2018-05-02 Tom de Vries <tom@codesourcery.com>
4895 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4896 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4897 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4898 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4899 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4900 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4901 __builtin_goacc_parlevel_size.
4903 2018-05-02 Richard Biener <rguenther@suse.de>
4905 PR tree-optimization/85597
4906 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4907 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4910 2018-05-02 Tom de Vries <tom@codesourcery.com>
4913 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4914 dump files): Add ltrans-tree.
4916 2018-05-02 Tom de Vries <tom@codesourcery.com>
4919 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4920 dump files): Add wpa-ipa.
4922 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4924 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4925 powerpc*-*-linux*paired* target.
4926 * config/rs6000/750cl.h: Delete.
4927 * config/rs6000/paired.h: Delete.
4928 * config/rs6000/paired.md: Delete.
4929 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4931 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4932 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4933 comment. Remove paired float support.
4934 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4935 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4937 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4938 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4940 * config/rs6000/rs6000.c: Remove paired float support.
4941 (paired_expand_vector_init, paired_expand_vector_move,
4942 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4943 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4944 paired_expand_builtin, paired_expand_predicate_builtin,
4945 paired_init_builtins): Delete.
4946 * config/rs6000/rs6000.h: Remove paired float support.
4947 * config/rs6000/rs6000.md: Remove paired float support.
4948 (move_from_CR_ov_bit): Delete.
4949 * config/rs6000/rs6000.opt (mpaired): Delete.
4950 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4951 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4953 2018-05-02 Richard Biener <rguenther@suse.de>
4956 * gimplify.c (gimplify_save_expr): When in SSA form allow
4957 SAVE_EXPRs to compute to SSA vars.
4959 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4962 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4963 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4964 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4965 clobber operands[2], instead use a new pseudo. Formatting fixes.
4967 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4969 PR tree-optimization/85586
4970 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4971 exit early for statements in the same group if the accesses are
4974 2018-05-02 Tom de Vries <tom@codesourcery.com>
4977 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4980 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4982 PR tree-optimization/85143
4983 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4985 2018-05-01 Tom de Vries <tom@codesourcery.com>
4988 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4989 not found" error message.
4991 2018-05-01 Tom de Vries <tom@codesourcery.com>
4994 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4995 * vec.c (test_ordered_remove_if): New function.
4996 (vec_c_tests): Call test_ordered_remove_if.
4997 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4998 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4999 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5000 VEC_ORDERED_REMOVE_IF.
5002 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5004 PR tree-optimization/82665
5005 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5006 pointer subtraction where arguments come from a memchr call.
5008 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5010 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5011 --push-state --as-needed and --pop-state instead of --as-needed and
5012 --no-as-needed if ld supports it.
5013 * configure: Regenerated.
5016 * doc/install.texi2html: Replace _002d with - and _002a with * in
5017 generated html files using sed.
5019 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5022 * gcc-rich-location.c (blank_line_before_p): New function.
5023 (use_new_line): New function.
5024 (gcc_rich_location::add_fixit_insert_formatted): New function.
5025 * gcc-rich-location.h
5026 (gcc_rich_location::add_fixit_insert_formatted): New function.
5028 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5030 * selftest.c (assert_streq): Rename "expected" and "actual" to
5031 "val1" and "val2". Extend NULL-handling to cover both inputs
5032 symmetrically, while still requiring both to be non-NULL for a pass.
5033 * selftest.h (assert_streq): Rename "expected" and "actual" to
5035 (ASSERT_EQ): Likewise.
5036 (ASSERT_EQ_AT): Likewise.
5037 (ASSERT_KNOWN_EQ): Likewise.
5038 (ASSERT_KNOWN_EQ_AT): Likewise.
5039 (ASSERT_NE): Likewise.
5040 (ASSERT_MAYBE_NE): Likewise.
5041 (ASSERT_MAYBE_NE_AT): Likewise.
5042 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5043 the assertion to pass.
5044 (ASSERT_STREQ_AT): Likewise.
5046 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5048 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5049 interaction with -pie.
5051 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5053 * selftest.h: Fix alphabetization of per-source-file selftest
5056 2018-04-30 Jason Merrill <jason@redhat.com>
5058 PR c++/61982 - dead stores to destroyed objects.
5059 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5062 2018-04-30 Jason Merrill <jason@redhat.com>
5064 * tree.c (build_clobber): New.
5065 * tree.h: Declare it.
5066 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5068 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5070 * diagnostic-show-locus.c (layout::layout): Update for
5071 location_get_source_line returning a char_span.
5072 (struct char_span): Move to input.h.
5073 (struct correction): Update for fields in char_span becoming
5075 (struct source_line): Update for location_get_source_line
5076 returning a char_span.
5077 (layout::print_line): Likewise.
5078 * edit-context.c (edited_file::print_content): Likewise.
5079 (edited_file::print_diff_hunk): Likewise.
5080 (edited_file::print_run_of_changed_lines): Likewise.
5081 (edited_file::get_num_lines): Likewise.
5082 (edited_line::edited_line): Likewise.
5083 * final.c (asm_show_source): Likewise.
5084 * input.c (location_get_source_line): Convert return type
5085 from const char * to char_span, losing the final "line_len"
5087 (dump_location_info): Update for the above.
5088 (get_substring_ranges_for_loc): Likewise. Use a char_span
5089 when handling the literal within the line.
5090 (test_reading_source_line): Update for location_get_source_line
5091 returning a char_span.
5092 * input.h (class char_span): Move here from
5093 diagnostic-show-locus.c, converting from a struct to a class.
5094 Make data members private.
5095 (char_span::operator bool): New.
5096 (char_span::length): New.
5097 (char_span::get_buffer): New.
5098 (char_span::operator[]): New.
5099 (char_span::subspan): Make const.
5100 (char_span::xstrdup): New.
5101 (location_get_source_line): Convert return type from const char *
5102 to char_span, losing the final "line_size" param.
5104 2018-04-30 Jan Hubicka <jh@suse.cz>
5106 * lto-wrapper.c (ltrans_priorities): New static var.
5107 (cmp_priority): New.
5108 (run_gcc): Read priorities and if doing parallel build order
5109 the Makefile by them.
5111 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5113 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5115 2018-04-30 Richard Biener <rguenther@suse.de>
5117 * tree-cfg.c (verify_address): Remove base argument, add
5118 flag whether to check TREE_ADDRESSABLE and do that.
5119 (verify_expr): Remove.
5120 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5121 (verify_gimple_assign_single): Likewise.
5122 (verify_gimple_switch): Likewise.
5123 (verify_expr_location_1): Dereference tp once. Add (disabled)
5124 piece from verify_expr.
5125 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5127 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5129 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5131 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5133 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5134 (small_data_pattern): Likewise.
5135 (arc_rewrite_small_data): Likewise.
5136 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5137 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5138 (get_symbol_alignment): New function.
5139 (legitimate_small_data_address_p): Likewise.
5140 (legitimate_scaled_address): Update, call
5141 legitimate_small_data_address_p.
5142 (output_sdata): New static variable.
5143 (arc_print_operand): Update how we handle small data operands.
5144 (arc_print_operand_address): Likewise.
5145 (arc_legitimate_address_p): Update, use
5146 legitimate_small_data_address_p.
5147 (arc_rewrite_small_data_p): Remove.
5148 (arc_rewrite_small_data_1): Likewise.
5149 (arc_rewrite_small_data): Likewise.
5150 (small_data_pattern): Likewise.
5151 (compact_sda_memory_operand): Update to use
5152 legitimate_small_data_address_p and get_symbol_alignment.
5153 (prepare_move_operands): Don't rewite sdata pattern.
5154 (prepare_extend_operands): Remove.
5155 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5157 (zero_extendqisi2): Likewise.
5158 (zero_extendhisi2): Likewise.
5159 (extendqihi2): Likewise.
5160 (extendqisi2): Likewise.
5161 (extendhisi2): Likewise.
5165 * config/arc/constraints.md (Usd): Change it to memory constraint.
5167 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5169 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5170 as source of std instructions.
5171 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5172 allow 6-bit constants as source for store instructions.
5173 (movdi_insn): Update instruction pattern to allow 6-bit constants
5174 as source for store instructions.
5176 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5178 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5180 2018-04-30 Nathan Sidwell <nathan@acm.org>
5181 Sandra Loosemore <sandra@codesourcery.com>
5183 * dumpfile.c (dump_open): Allow '-' for stdout.
5184 * doc/invoke.texi (Developer Options): Document dump filename
5185 determination early. Document stdin/stdout selection.
5187 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5189 Microblaze Target: PIC data text relative
5191 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5192 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5194 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5195 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5196 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5197 New addressing mode for data-text relative position indepenedent code.
5198 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5199 'ADDRESS_SYMBOLIC_TXT_REL'.
5200 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5201 (microblaze_legitimate_pic_operand): Exclude function calls from
5202 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5203 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5205 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5206 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5207 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5208 for 'address + offset'.
5209 (microblaze_expand_prologue): Add new function prologue call for
5211 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5212 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5213 table in case of TARGET_PIC_DATA_TEXT_REL.
5214 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5215 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5216 Add new macros 'UNSPEC_TEXT',
5217 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5218 + exclude function calls from 'UNSPEC_PLT' in case of data text
5220 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5221 new target hook for generating address diff vector tables in case of
5223 * doc/tm.texi : Regenerate.
5224 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5225 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5226 of addr diff vector generation.
5227 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5228 target hook definition.
5229 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5230 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5231 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5232 option for data text relative.
5234 2018-04-30 Richard Biener <rguenther@suse.de>
5236 * tree-chrec.h (evolution_function_is_constant_p): Remove
5238 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5240 2018-04-30 Richard Biener <rguenther@suse.de>
5243 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5245 2018-04-30 Richard Biener <rguenther@suse.de>
5247 PR tree-optimization/28364
5248 PR tree-optimization/85275
5249 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5250 copying first exit test.
5252 2018-04-28 Mark Wielaard <mark@klomp.org>
5254 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5256 (dwarf_AT): Handle DW_AT_addr_base.
5257 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5259 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5262 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5263 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5264 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5265 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5267 2018-04-28 Richard Biener <rguenther@suse.de>
5269 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5270 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5271 to reflect use. Only add interesting stmts.
5273 2018-04-27 Martin Jambor <mjambor@suse.cz>
5276 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5277 the jump function allows for passing through aggregate values.
5279 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5281 * input.h (in_system_header_at): Convert from macro to inline
5283 (from_macro_expansion_at): Likewise.
5284 (from_macro_definition_at): Likewise.
5286 2018-04-27 Jeff Law <law@redhat.com>
5288 * config.gcc: Mark tile* targets as deprecated/obsolete.
5290 2018-04-27 Richard Biener <rguenther@suse.de>
5292 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5295 2018-04-27 Richard Biener <rguenther@suse.de>
5297 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5299 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5301 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5302 with Yd constraint. Set "preferred_for_speed" attribute from
5303 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5305 (*movdi_internal): Ditto.
5306 (movti_interunit splitters): Remove
5307 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5308 (movdi_interunit splitters): Ditto.
5309 * config/i386/constraints.md (Ye): Remove.
5310 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5312 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5315 * config/aarch64/constraints.md (Usg): Limit to 31.
5318 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5320 PR tree-optimization/85529
5321 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5322 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5323 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5324 zero extension or masking of the MSB bit.
5325 (optimize_range_tests): Add FIRST_BB argument, pass it through
5326 to optimize_range_tests_var_bound.
5327 (maybe_optimize_range_tests, reassociate_bb): Adjust
5328 optimize_range_tests callers.
5330 2018-04-26 Richard Biener <rguenther@suse.de>
5331 Jakub Jelinek <jakub@redhat.com>
5333 * cgraph.h (symbol_table): Just declare debug method here.
5334 * symtab.c (symbol_table::debug): Define.
5336 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5338 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5340 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5342 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5343 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5344 (*movdi_internal): Substitute Yi and Yj constraint with x
5345 and Ym and Yn constraint with y constraint. Update "isa"
5346 attribute and set "preferred_for_speed" attribute from
5347 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5348 (*movsi_internal): Ditto.
5349 (*movdf_internal): Ditto.
5350 (*movsf_internal): Ditto.
5351 (*zero_extendsidi2): Ditto.
5352 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5353 (sse2_loadld): Ditto.
5354 (*vec_extract<ssevecmodelower>_0): Ditto.
5355 (*vec_extractv4si_0_zext_sse4): Ditto.
5356 (vec_concatv2di): Ditto.
5357 (*vec_dup<mode>): Ditto.
5358 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5359 * config/i386/constraints.md (Yi): Remove.
5364 2018-04-26 Nathan Sidwell <nathan@acm.org>
5366 * dumpfile.c (dump_open): New.
5367 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5368 (dump_finish): Detect stdio/stderr by value not name.
5370 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5372 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5374 2018-04-26 Tom de Vries <tom@codesourcery.com>
5377 * config/nvptx/nvptx.c (verify_neutering_jumps)
5378 (verify_neutering_labels): New function
5379 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5381 2018-04-26 Tom de Vries <tom@codesourcery.com>
5384 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5385 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5388 2018-04-26 Richard Biener <rguenther@suse.de>
5389 Tom de Vries <tom@codesourcery.com>
5392 * lto-streamer-out.c (output_function): Fixup loops if required to match
5393 discovery done in the reader.
5395 2018-04-26 Richard Biener <rguenther@suse.de>
5397 PR tree-optimization/85116
5398 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5399 have a loop exit from the single latch predecessor. Remove
5400 case of header with just condition.
5401 (ch_base::copy_headers): Exclude infinite loops from any
5403 (pass_ch::execute): Record exits.
5405 2018-04-26 Richard Biener <rguenther@suse.de>
5407 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5408 prologue cost vector and pass it to vect_get_load_cost.
5409 (vect_get_peeling_costs_all_drs): Likewise.
5410 (vect_peeling_hash_get_lowest_cost): Likewise.
5411 (vect_enhance_data_refs_alignment): Likewise.
5413 2018-04-26 Richard Biener <rguenther@suse.de>
5416 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5417 checking of integer<->pointer conversions.
5418 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5419 sign-/zero-extending pointer types.
5420 (expand_omp_for_static_chunk): Likewise.
5422 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5423 Jean Lee <xiaoyur347@gmail.com>
5425 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5426 (TARGET_ASAN_SHADOW_OFFSET): Define.
5427 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5428 true for -fsanitize=address.
5430 2018-04-25 Mark Wielaard <mark@klomp.org>
5432 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5435 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5437 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5438 than "alu", remove explicit "memory" and "imm_disp" attributes.
5439 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5442 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5443 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5444 gen_lowpart_no_emit.
5446 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5449 * config/i386/i386.c (ix86_expand_builtin): Change memory
5450 operand to XI, extend p0 to Pmode.
5451 * config/i386/i386.md: Change unspec volatile and operand
5452 1 mode to XI, change operand 0 mode to P.
5454 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5456 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5457 GET_MODE_MASK before any checking.
5458 (nds32_can_use_bset_p): Likewise.
5459 (nds32_can_use_btgl_p): Likewise.
5461 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5463 * config/nds32/nds32-doubleword.md: New define_split pattern for
5464 illegal register number.
5466 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5468 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5470 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5472 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5474 2018-04-25 Richard Biener <rguenther@suse.de>
5476 * lto-streamer.h (LTO_major_version): Bump to 8.
5478 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5480 * BASE-VER: Set to 9.0.0.
5482 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5484 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5485 in __abskf2 and __powikf2.
5487 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5490 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5491 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5492 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5493 Use the above on operand 2. Reindent.
5494 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5496 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5499 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5501 * config/i386/i386.opt (mcet): Removed.
5502 * doc/install.texi: Remove -mcet documentation.
5503 * doc/invoke.texi: Likewise.
5505 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5508 * doc/install.texi: Remove -mcet from bootstrap-cet.
5510 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5513 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5514 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5518 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5519 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5520 containing a CONST_VECTOR.
5522 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5524 * doc/install.texi: Update newlib dependency for nvptx.
5526 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5529 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5530 instead of INTVAL when shifting x left.
5532 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5534 PR tree-optimization/85478
5535 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5536 vect_grouped_store_supported for single element vectors.
5538 2018-04-24 Richard Biener <rguenther@suse.de>
5541 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5542 load cost increase to the case of non-constant step.
5544 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5547 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5548 destination if any_malformed_asm.
5550 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5553 * expr.c (store_field): In the bitfield case, if the value comes from
5554 a function call and is returned in registers by means of a PARALLEL,
5555 do not change the mode of the temporary unless BLKmode and VOIDmode.
5557 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5559 PR rtl-optimization/85423
5560 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5561 dependencies to debug insns when the previous insn is non-debug.
5563 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5565 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5566 enums into a single definition.
5567 (fls): Fix predicates and printing.
5570 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5572 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5573 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5574 and short u6 immediate.
5575 (check_if_valid_sleep_operand): Remove.
5576 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5578 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5580 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5581 flag_always_save_lp condition.
5582 * config/nds32/nds32.opt (malways-save-lp): New option.
5584 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5586 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5587 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5588 * config/nds32/nds32.h
5589 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5590 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5592 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5594 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5595 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5597 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5598 Chung-Ju Wu <jasonwucj@gmail.com>
5600 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5602 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5603 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5605 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5607 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5609 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5611 * config/nds32/nds32-protos.h (nds32_data_alignment,
5612 nds32_local_alignment): Declare.
5613 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5614 nds32_local_alignment): New functions.
5615 (TARGET_CONSTANT_ALIGNMENT): Define.
5616 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5618 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5620 * config/nds32/nds32.c
5621 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5622 (TARGET_MODES_TIEABLE_P): Likewise.
5624 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5626 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5629 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5630 Chung-Ju Wu <jasonwucj@gmail.com>
5632 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5633 for unaligned access.
5634 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5636 * config/nds32/nds32-intrinsic.md: Likewise.
5637 * config/nds32/nds32_intrinsic.h: Likewise.
5638 * config/nds32/nds32.h (nds32_builtins): Likewise.
5639 * config/nds32/nds32.opt (munaligned-access): New option.
5640 * config/nds32/nds32.c (nds32_asm_file_start): Display
5641 flag_unaligned_access status.
5643 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5645 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5646 -mno-relax is present.
5647 * config/riscv/linux.h (LINK_SPEC): Ditto.
5649 2018-04-20 Martin Sebor <msebor@redhat.com>
5652 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5654 (gimple_fold_builtin_stxcpy_chk): Same.
5655 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5657 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5660 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5661 __powikf2 when long double is IEEE 128-bit.
5663 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5665 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5666 step to make sure stack always aligned.
5668 2018-04-20 Carl Love <cel@us.ibm.com>
5671 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5672 size check for arg0.
5674 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5675 Tom de Vries <tom@codesourcery.com>
5678 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5679 Emit insns for calls too.
5680 (nvptx_find_par): Always look for worker-level predecessor insn.
5681 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5683 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5684 (nvptx_process_pars): Propagate frames for calls.
5686 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5689 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5691 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5692 (ix86_handle_option): Don't handle OPT_mibt.
5693 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5695 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5697 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5698 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5699 (ix86_target_macros): Define __CET__ with flag_cf_protection
5700 for -fcf-protection.
5701 * config/i386/i386.c (isa2_opts): Remove -mibt.
5702 * config/i386/i386.h (TARGET_IBT): Removed.
5703 (TARGET_IBT_P): Likewise.
5704 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5705 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5706 * config/i386/i386.opt (mcet): Update help message.
5709 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5710 -mcet as an alias for -mshstk.
5712 2018-04-20 Richard Biener <rguenther@suse.de>
5715 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5716 complexity by forcing a single use of the multiply operand.
5718 2018-04-20 Martin Jambor <mjambor@suse.cz>
5721 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5722 recursion dependency to only apply to non-clones.
5724 2018-04-20 Martin Jambor <mjambor@suse.cz>
5727 * ipa-cp.c (create_specialized_node): Check that clones of
5728 self-recursive edges exist during IPA-CP.
5730 2018-04-19 Toon Moene <toon@moene.org>
5732 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5735 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5737 PR tree-optimization/85467
5738 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5739 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5740 VECTOR_CST element to type.
5742 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5745 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5746 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5747 (builtin_longjmp): Likewise.
5748 (save_stack_nonlocal): New pattern.
5749 (restore_stack_nonlocal): Likewise.
5751 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5754 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5755 Replace ASM_OUTPUT_LABEL with fprintf.
5757 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5760 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5761 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5762 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5763 define __IBT__ and __SHSTK__ for -fcf-protection.
5764 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5766 (ix86_trampoline_init): Likewise.
5767 (x86_output_mi_thunk): Likewise.
5768 (ix86_notrack_prefixed_insn_p): Likewise.
5769 (ix86_option_override_internal): Don't disallow -fcf-protection.
5770 * config/i386/i386.md (rdssp<mode>): Also enable for
5772 (incssp<mode>): Likewise.
5773 (nop_endbr): Likewise.
5774 * config/i386/i386.opt (mcet): Change help message to built-in
5778 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5779 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5780 enable CET built-in functions.
5782 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5784 * common/config/i386/i386-common.c
5785 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5786 OPTION_MASK_ISA_MOVDIRI_UNSET,
5787 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5788 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5789 * config.gcc (movdirintrin.h): New header.
5790 * config/i386/cpuid.h (bit_MOVDIRI,
5791 bit_MOVDIR64B): New bits.
5792 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5794 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5795 (VOID, PVOID, PCVOID)): New function types.
5796 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5797 __builtin_ia32_directstoreu_u64,
5798 __builtin_ia32_movdir64b): New builtins.
5799 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5800 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5802 (ix86_valid_target_attribute_inner_p): Ditto.
5803 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5804 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5805 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5806 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5807 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5808 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5809 (movdiri<mode>, movdir64b_<mode>): New.
5810 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5811 * config/i386/immintrin.h: Include movdirintrin.h.
5812 * config/i386/movdirintrin.h: New file.
5813 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5815 2018-04-19 Richard Biener <rguenther@suse.de>
5818 * cfg.c (clear_bb_flags): When loop state says we have
5819 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5821 2018-04-19 Richard Biener <rguenther@suse.de>
5823 PR tree-optimization/84737
5824 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5825 copying restrict info.
5826 (vect_setup_realignment): Use it.
5827 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5828 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5829 the first DR to all generated stores.
5830 (vectorizable_load): Likewise for loads.
5832 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5834 PR tree-optimization/85446
5835 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5836 the integral and pointer types to have the same precision.
5838 * doc/install.texi: Document --disable-cet being the default and
5841 2018-04-18 Martin Liska <mliska@suse.cz>
5843 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5846 2018-04-18 Martin Liska <mliska@suse.cz>
5849 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5852 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5853 arguments if they are comparable.
5855 2018-04-18 Martin Liska <mliska@suse.cz>
5858 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5861 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5864 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5867 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5868 ENDBR after calling __morestack.
5870 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5873 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5874 by using gcc_base_ver to generate a gcc_driver_version, and use
5875 it when generating GCC_DRIVER_NAME.
5876 * configure: Regenerate.
5878 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5881 * config.gcc: Obsolete powerpc*-*-*spe*.
5883 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5886 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5887 (stabstr_D): Change type of unum from unsigned int to
5888 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5891 2018-04-17 Jim Wilson <jimw@sifive.com>
5894 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5895 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5896 Set arg_pointer_offset after using pretend_args_size.
5898 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5900 PR rtl-optimization/85431
5901 * dse.c (record_store): Ignore zero width stores.
5904 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5905 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5906 __builtin_stack_restore rather than after it.
5907 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5908 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5909 argument instead of virtual_dynamic_stack_rtx.
5911 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5913 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5915 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5916 Add note to error message to explain internal mapping of overloaded
5917 built-in function name to non-overloaded built-in function name.
5918 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5921 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5924 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5925 where the inputs overlap with the output.
5927 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5929 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5930 (=v, v) alternative and explicit "memory" attribute.
5931 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5932 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5934 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5935 "sselog1" type instead of "sselog".
5936 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5937 "sselog". Remove explicit "memory" attribute.
5938 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5939 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5941 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5942 "sselog1" type instead of "sselog". Remove explicit "memory"
5944 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5945 use "sselog1" type instead of "sselog". Remove explicit "memory"
5947 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5948 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5950 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5951 "sselog1" type instead of "sselog". Remove explicit "memory"
5953 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5954 use "sselog1" type instead of "sselog". Remove explicit "memory"
5958 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5961 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5964 2018-04-17 Martin Jambor <mjambor@suse.cz>
5967 * ipa-cp.c (create_specialized_node): Call
5968 expand_all_artificial_thunks if necessary.
5970 2018-04-17 Martin Liska <mliska@suse.cz>
5973 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5974 in message, remote space in between '_G' and '('.
5976 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5979 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5980 avx512f_vmcmp<mode>3<round_saeonly_name>,
5981 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5982 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5983 avx512f_rndscale<mode><round_saeonly_name>,
5984 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5985 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5986 Use %<iptr>2 instead of %2 for -masm=intel.
5987 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5988 avx512f_vcvttss2usi<round_saeonly_name>,
5989 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5991 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5992 avx512f_vcvttsd2usi<round_saeonly_name>,
5993 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5994 Use %q1 instead of %1 for -masm=intel.
5995 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5996 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5997 of %3 for -masm=intel.
5998 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6000 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6002 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6004 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6005 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6007 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6009 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6010 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6011 %g1 and one with %0 and %1.
6012 (avx512er_vmrcp28<mode><round_saeonly_name>,
6013 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6015 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6016 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6017 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6018 of %0 and %{%4%} for -masm=intel.
6019 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6020 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6021 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6022 order of %0 and %{%5%}%{z%} for -masm=intel.
6024 2018-04-17 Jan Hubicka <jh@suse.cz>
6027 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6029 2018-04-17 Martin Liska <mliska@suse.cz>
6032 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6033 for target_clone error message. Make default implementation
6034 clone to be a local declaration.
6035 (separate_attrs): Add new argument and check for an empty
6037 (expand_target_clones): Handle it.
6038 (ipa_target_clone): Make redirection just for target_clones
6041 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6042 Tom de Vries <tom@codesourcery.com>
6045 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6046 tiled basic blocks without omp continue statements.
6048 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6051 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6052 vec_extract expression as having side effects to make sure it gets
6055 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6058 * config/i386/i386.c (get_builtin_code_for_version): Check
6061 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6064 * gcc/config.gcc: Support "skylake".
6065 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6067 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6068 (processor_target_table): Add "skylake".
6069 (ix86_option_override_internal): Add "skylake".
6070 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6071 PROCESSOR_CANNONLAKE.
6072 (get_builtin_code_for_version): Fix priority for
6073 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6074 PROCESSOR_SKYLAKE-AVX512.
6075 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6076 (processor_type): Add PROCESSOR_SKYLAKE.
6078 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6079 Jason Merrill <jason@redhat.com>
6082 * convert.c (convert_to_integer_1): Use direct recursion for
6083 enumeral types and types with a precision less than the number
6084 of bits in their mode.
6086 2018-04-16 Julia Koval <julia.koval@intel.com>
6089 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6090 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6092 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6095 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6096 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6097 and -mno-direct-move.
6099 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6102 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6103 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6104 Ensure negative shifts result in {0}.
6106 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6108 PR rtl-optimization/79916
6109 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6110 regs (if any) to define how to gnerate SD moves when LRA is in
6113 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6115 PR rtl-optimization/85393
6116 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6117 * except.c (expand_dw2_landing_pad_for_region): Make static.
6118 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6119 a label and unconditional jump to old_bb, rather than
6120 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6123 PR rtl-optimization/85376
6124 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6125 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6126 instead of a specific value.
6128 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6129 Bin Cheng <bin.cheng@arm.com>
6131 PR tree-optimization/82965
6132 PR tree-optimization/83991
6133 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6134 by_profile_only parameter.
6135 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6136 information if the loop was predicted to iterate too many times.
6137 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6139 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6142 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6145 2018-04-13 Martin Liska <mliska@suse.cz>
6146 Jakub Jelinek <jakub@redhat.com>
6149 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6150 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6151 * builtins.c (expand_builtin_memory_copy_args): Use
6152 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6153 handle dest_addr == pc_rtx.
6155 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6158 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6159 asked to not generate direct moves.
6160 (fix_trunc<mode>si2_stfiwx): Similar.
6161 (fix_trunc<mode>si2_internal): Similar.
6163 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6166 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6167 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6168 lookup if dest in some wider mode is known to be const0_rtx and
6169 if so, record permanent equivalence for it to be ZERO_EXTEND of
6170 the narrower mode destination.
6172 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6174 * lto-streamer-out.c (output_function): Revert 259346.
6175 * omp-expand.c (expand_oacc_for): Likewise.
6177 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6179 PR rtl-optimization/85354
6180 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6181 * sel-sched.c (sel_global_init): ... here.
6183 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6186 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6187 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6188 mode for PE-COFF targets.
6189 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6190 (i386_pe_asm_lto_end): Likewise.
6191 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6192 (TARGET_ASM_LTO_END): Likewise.
6193 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6194 (i386_pe_asm_lto_start): New function.
6195 (i386_pe_asm_lto_end): Likewise.
6197 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6198 Richard Biener <rguenther@suse.de>
6201 * lto-streamer-out.c (output_function): Fix CFG loop state before
6203 * omp-expand.c (expand_oacc_for): Handle calls to internal
6204 functions like regular functions.
6206 2018-04-12 Richard Biener <rguenther@suse.de>
6209 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6210 for the early LTO debug to properly generate references to it
6211 during DIE emission. Do not re-use that for the skeleton for
6213 (dwarf2out_early_finish): Likewise.
6215 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6218 * config/i386/sse.md
6219 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6220 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6221 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6222 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6223 and output is a reg, avoid creating invalid lowpart subreg, but
6224 instead split into a 512-bit move. Don't split if not AVX512VL,
6225 input is xmm16+ reg and output is a mem.
6226 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6227 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6228 xmm16+ reg and output is a mem.
6230 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6232 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6233 also for flag_dwarf2_cfi_asm.
6235 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6237 PR rtl-optimization/85342
6238 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6239 a bool scalar var inside of the loop instead. Don't try to update
6240 recog_data.operand after failed apply_change_group.
6242 2018-04-12 Tom de Vries <tom@codesourcery.com>
6245 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6246 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6247 array with flexible array member as array without given dimension.
6248 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6249 argument for undefined param to true.
6251 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6254 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6255 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6256 from PowerPC section.
6257 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6258 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6259 help text that is too long.
6260 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6261 help text that is too long.
6262 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6263 help text that is too long.
6265 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6267 * config/alpha/alpha.md (stack_probe_internal): Rename
6268 from "probe_stack". Update all callers.
6270 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6272 PR rtl-optimization/84566
6273 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6274 sched_macro_fuse_insns.
6276 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6279 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6280 (compute_block_dependences): ... from here.
6282 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6284 PR tree-optimization/85331
6285 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6286 from int to HOST_WIDE_INT.
6288 2018-04-11 Martin Jambor <mjambor@suse.cz>
6291 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6292 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6293 not the same as the source val.
6294 (cgraph_edge_brings_value_p): New parameter.
6295 (gather_edges_for_value): Pass destination value to
6296 cgraph_edge_brings_value_p.
6297 (perhaps_add_new_callers): Likewise.
6298 (get_info_about_necessary_edges): Likewise and exclude values brought
6299 only by self-recursive edges.
6300 (create_specialized_node): Redirect only clones of self-calling edges.
6301 (+self_recursive_pass_through_p): New function.
6302 (find_more_scalar_values_for_callers_subset): Use it.
6303 (find_aggregate_values_for_callers_subset): Likewise.
6304 (known_aggs_to_agg_replacement_list): Removed.
6305 (decide_whether_version_node): Re-calculate known constants for all
6306 remaining context clones.
6308 2018-04-11 Richard Biener <rguenther@suse.de>
6311 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6312 from early DWARF output.
6313 (dwarf2out_early_finish): Output line info unconditionally into
6314 early DWARF and add reference to it.
6316 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6319 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6320 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6321 other than V2DFmode using iptr mode attribute.
6322 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6324 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6326 PR rtl-optimization/84659
6327 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6329 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6332 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6334 (output_loc_list): Pass address of a dummy size variable even in the
6335 locview handling loop.
6336 (index_location_lists): Add comment on why skip_loc_list_entry can't
6339 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6342 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6345 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6348 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6349 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6350 and -mstring-compare-inline-limit.
6352 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6355 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6356 for stack clash protection in a register whenever we need it to be in
6359 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6361 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6362 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6364 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6367 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6369 (mlong-double-): Ditto.
6370 * config/rs6000/sysv4.opt (msdata=): Ditto.
6371 (mtls-size=): Ditto.
6373 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6375 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6376 erroneous entries for
6377 "vector int vec_ldl (int, long int *)", and
6378 "vector unsigned int vec_ldl (int, unsigned long int *)".
6379 Add comments and entries for
6380 "vector bool char vec_ldl (int, bool char *)",
6381 "vector bool short vec_ldl (int, bool short *)",
6382 "vector bool int vec_ldl (int, bool int *)",
6383 "vector bool long long vec_ldl (int, bool long long *)",
6384 "vector pixel vec_ldl (int, pixel *)",
6385 "vector long long vec_ldl (int, long long *)",
6386 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6387 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6388 type tree bool_long_long_type_node and correct definition of
6389 bool_V2DI_type_node to make reference to this new type tree.
6390 (rs6000_mangle_type): Replace erroneous reference to
6391 bool_long_type_node with bool_long_long_type_node.
6392 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6393 comments to emphasize sign distinctions for char and int types and
6394 replace RS6000_BTI_bool_long constant with
6395 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6396 use of RS6000_BTI_pixel.
6397 (bool_long_type_node): Remove this macro definition.
6398 (bool_long_long_type_node): New macro definition
6400 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6402 PR rtl-optimization/85300
6403 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6404 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6405 simplify_unary_operation fails.
6407 2018-04-10 Martin Liska <mliska@suse.cz>
6409 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6410 cgraph_edge and ipa_ref.
6412 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6416 * config/i386/sse.md
6417 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6418 computation of the VEC_MERGE selector from mask.
6419 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6420 Fix decoding of the VEC_MERGE selector into mask.
6422 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6424 PR tree-optimization/85286
6425 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6427 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6429 * final.c (final_1): Set insn_last_address as well as
6430 insn_current_address.
6432 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6435 * explow.c (emit_stack_probe): Call validize_mem on memory location
6436 before passing it to gen_probe_stack. Create address operand and
6437 legitimize it for the probe_stack_address case.
6439 2018-04-09 Jan Hubicka <jh@suse.cz>
6442 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6443 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6444 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6446 (free_lang_data): Rebuild type inheritance graph.
6448 2018-04-09 Martin Sebor <msebor@redhat.com>
6450 * invoke.texi (-finline-small-functions): Mention other optimization
6452 (-findirect-inlining, -fpartial-inlining): Same.
6453 (-finline-functions-called-once): Same.
6454 (-freorder-blocks-and-partition): Same.
6456 2018-04-09 Jan Hubicka <jh@suse.cz>
6459 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6460 jumps; choose last target that matches the criteria (i.e.
6461 no partition changes for non-crossing jumps).
6462 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6463 support for redirecting crossing jumps to non-crossing.
6465 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6467 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6468 also for naked functions.
6470 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6472 * config/arc/arc.md (add_shift): New pattern.
6473 (add_shift2): Likewise.
6474 (sub_shift): Likewise.
6475 (sub_shift_cmp0_noout): Likewise.
6476 (compare_si_ashiftsi): Likewise.
6477 (xbfu_cmp0_noout): New combine pattern.
6478 (xbfu_cmp0"): Likewise.
6479 (movsi_set_cc_insn): Place the predicable variant first.
6480 (commutative_binary_cmp0_noout): Remove clobber.
6481 (commutative_binary_cmp0): New pattern.
6482 (noncommutative_binary_cmp0): Likewise.
6483 (noncommutative_binary_cmp0_noout): Likewise.
6484 (noncommutative_binary_comparison_result_used): Removed.
6485 (rsub_cmp0): New pattern.
6486 (rsub_cmp0_noout): Likewise.
6487 (extzvsi): Changed, keep only meaningful variants.
6488 (SQH, SEZ): New iterators.
6489 (SQH_postfix): New mode attribute.
6490 (SEZ_prefix): New code attribute.
6491 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6492 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6493 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6495 (noncommutative_operator): Check the availability of barrel
6498 2018-04-09 Richard Biener <rguenther@suse.de>
6500 PR tree-optimization/85284
6501 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6502 Only use the niter constraining form of simple_iv when the exit
6505 2018-04-09 Tom de Vries <tom@codesourcery.com>
6508 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6509 (define_expand "*memory_barrier"): New define_expand.
6510 (define_insn "memory_barrier"): New insn.
6512 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6514 PR rtl-optimization/80463
6515 PR rtl-optimization/83972
6516 PR rtl-optimization/83480
6518 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6519 correct producer for the insn.
6520 (tidy_control_flow): Fixup seqnos in case of debug insns.
6522 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6524 PR rtl-optimization/83913
6526 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6527 different sched-times when merging exprs.
6529 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6531 PR rtl-optimization/83962
6533 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6534 tidy_fallthru_edge and tidy_control_flow.
6536 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6538 PR rtl-optimization/83530
6540 * sel-sched.c (force_next_insn): New global variable.
6541 (remove_insn_for_debug): When force_next_insn is true, also leave only
6542 next insn in the ready list.
6543 (sel_sched_region): When the region wasn't scheduled, make another pass
6544 over it with force_next_insn set to 1.
6546 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6548 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6550 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6551 for interrupt control.
6552 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6553 functions for interrupt control.
6554 * config/nds32/nds32-intrinsic.md: Likewise.
6555 * config/nds32/nds32_intrinsic.h: Likewise.
6556 * config/nds32/nds32.h (nds32_builtins): Likewise.
6558 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6560 * config/nds32/nds32.c (nds32_init_machine_status,
6561 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6562 strict_aligned_p field.
6563 (nds32_expand_to_rtl_hook): New function.
6564 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6565 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6567 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6568 Chung-Ju Wu <jasonwucj@gmail.com>
6570 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6571 * config/nds32/nds32-n7.md: New file.
6572 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6573 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6575 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6576 * config/nds32/nds32.md (pipeline_model): Add n7.
6577 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6578 * config/nds32/pipelines.md: Include n7 settings.
6580 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6581 Chung-Ju Wu <jasonwucj@gmail.com>
6583 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6584 * config/nds32/nds32-e8.md: New file.
6585 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6586 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6588 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6589 * config/nds32/nds32.md (pipeline_model): Add e8.
6590 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6591 * config/nds32/pipelines.md: Include e8 settings.
6593 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6594 Chung-Ju Wu <jasonwucj@gmail.com>
6596 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6597 * config/nds32/nds32-n8.md: New file.
6598 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6599 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6601 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6602 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6603 * config/nds32/nds32.md (pipeline_model): Add n8.
6604 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6605 * config/nds32/pipelines.md: Include n8 settings.
6607 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6608 Chung-Ju Wu <jasonwucj@gmail.com>
6610 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6611 * config/nds32/nds32-n9-2r1w.md: New file.
6612 * config/nds32/nds32-n9-3r2w.md: New file.
6613 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6614 nds32_register_ports): New or modify for cpu n9.
6615 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6617 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6618 * config/nds32/nds32-utils.c: New file.
6619 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6620 TARGET_MUL_SLOW): Define.
6621 * config/nds32/nds32.md (pipeline_model): New attribute.
6622 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6623 New options that support cpu n9.
6624 * config/nds32/pipelines.md: Include n9 settings.
6625 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6627 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6629 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6630 information if necessary.
6631 (output_cond_branch_compare_zero): Likewise.
6632 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6633 (nds32_target_alignment): Refine for alignment.
6634 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6635 (FUNCTION_BOUNDARY): Modify.
6636 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6638 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6640 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6642 * config/nds32/constants.md (unspec_volatile_element): Add values for
6643 TLB operation and data prefetch.
6644 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6645 functions for TLB operation and data prefetch.
6646 * config/nds32/nds32-intrinsic.md: Likewise.
6647 * config/nds32/nds32_intrinsic.h: Likewise.
6648 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6649 (nds32_print_operand): Likewise.
6650 * config/nds32/nds32.h (nds32_builtins): Likewise.
6652 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6653 Andrew Pinski <pinsika@gcc.gnu.org>
6656 * match.pd: Use constant_boolean_node of correct type instead of
6657 boolean_true_node or boolean_false_node for simplifying
6658 pointer comparisons to zero.
6660 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6662 PR tree-optimization/80021
6663 * tree.c (verify_type_variant): Make error call in verify_variant_match
6664 translatable and remove final full stop.
6666 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6668 * config/nds32/constants.md (unspec_volatile_element): Add
6669 UNSPEC_VOLATILE_EH_RETURN.
6670 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6671 nds32_output_stack_pop): Support dwarf exception handling process.
6672 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6673 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6674 exception handling process.
6675 (nds32_compute_stack_frame): Likewise.
6676 (nds32_return_addr_rtx): Likewise.
6677 (nds32_initial_elimination_offset): Likewise.
6678 (nds32_expand_prologue): Likewise.
6679 (nds32_expand_epilogue): Likewise.
6680 (nds32_dynamic_chain_address): New function.
6681 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6683 (DYNAMIC_CHAIN_ADDRESS): Define.
6684 (EH_RETURN_DATA_REGNO): Define.
6685 (EH_RETURN_STACKADJ_RTX): Define.
6686 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6687 patterns for dwarf exception handling.
6689 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6691 * config/nds32/nds32.h: Clean up obsolete macros.
6693 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6695 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6696 Add enum values for particular instructions.
6697 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6698 particular intrinsic functions.
6699 * config/nds32/nds32-intrinsic.md: Likewise.
6700 * config/nds32/nds32_intrinsic.h: Likewise.
6701 * config/nds32/nds32.h (nds32_builtins): Likewise.
6702 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6703 (btst, ave): New patterns for particular instructions.
6705 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6707 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6708 Add enum values for atomic load/store and memory sync.
6709 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6711 * config/nds32/nds32-intrinsic.md: Likewise.
6712 * config/nds32/nds32_intrinsic.h: Likewise.
6713 * config/nds32/nds32.h (nds32_builtins): Likewise.
6715 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6717 PR tree-optimization/85257
6718 * fold-const.c (native_encode_vector): If not all elts could fit
6719 and off is -1, return 0 rather than offset.
6720 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6721 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6722 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6723 adjust buffer in native_interpret_expr call.
6725 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6727 * config/nds32/constants.md (unspec_volatile_element): Add cache
6728 control enum values.
6729 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6730 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6731 * config/nds32/nds32.c (nds32_cctl_names): New.
6732 (nds32_print_operand): Handle cache control register names.
6733 * config/nds32/nds32.h (nds32_builtins): New enum values.
6734 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6736 * config/nds32/nds32.md (type): Add mmu.
6737 * config/nds32/pipelines.md (simple_insn): Add mmu.
6739 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6741 * config/nds32/nds32.md (type): Remove call.
6742 * config/nds32/pipelines.md (simple_insn): Likewise.
6744 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6746 * config/nds32/constants.md (unspec_volatile_element): Add
6747 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6748 UNSPEC_VOLATILE_FMFCFG.
6749 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6750 description for fmfcfg and fmfcsr.
6751 (bdesc_1arg): Add fmtcsr.
6752 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6753 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6754 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6755 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6756 unspec_fmfcfg): New patterns.
6757 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6758 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6759 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6760 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6761 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6762 __nds32__fmfcfg): Define.
6764 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6766 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6767 intrinsic register names.
6768 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6769 intrinsic register enum values and macros.
6771 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6773 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6774 for load/store addressing form.
6775 (nds32_print_operand_address): Likewise.
6777 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6780 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6781 based on LABEL_REF. Remove useless assertion.
6782 (pic_address_needs_scratch): Fix formatting.
6783 (sparc_legitimize_pic_address): Minor tweaks.
6784 (sparc_delegitimize_address): Adjust assertion accordingly.
6785 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6786 into symbolic_operand.
6787 (movsi_high_pic_label_ref): Likewise.
6788 (movsi_lo_sum_pic_label_ref): Likewise.
6789 (movdi_pic_label_ref): Likewise.
6790 (movdi_high_pic_label_ref): Likewise.
6791 (movdi_lo_sum_pic_label_ref): Likewise.
6793 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6795 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6796 custom LIB_SPEC setup.
6798 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6799 Kito Cheng <kito.cheng@gmail.com>
6801 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6802 * config/riscv/freebsd.h: New.
6804 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6806 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6807 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6810 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6811 Kito Cheng <kito.cheng@gmail.com>
6813 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6814 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6815 * config/nds32/nds32-protos.h (nds32_output_call,
6816 nds32_output_return): Declare.
6817 * config/nds32/nds32.md: Refine all the call and return patterns.
6819 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6822 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6823 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6825 PR rtl-optimization/84872
6826 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6827 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6830 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6832 * expr.c (copy_blkmode_to_reg): Revert 254862.
6833 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6835 2018-04-06 Richard Biener <rguenther@suse.de>
6838 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6839 after seeing a component reference with an adjacent field. Treat
6840 refs to arrays at struct end of external decls similar to
6841 refs to unconstrained commons.
6843 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6846 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6847 look through SAVE_EXPRs with non-side-effects argument. Adjust
6849 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6852 2018-04-06 Richard Biener <rguenther@suse.de>
6855 * alias.c (find_base_term): New wrapper around find_base_term
6856 unwinding CSELIB_VAL_PTR changes.
6857 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6860 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6862 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6864 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6865 constant definitions.
6866 ("nop"): lr 0,0 -> nopr r0
6867 ("nop_lr0", "nop_lr1"): New insn definitions.
6869 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6871 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6872 NDS32_V3PUSH_AVAILABLE_P macro.
6874 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6875 Chung-Ju Wu <jasonwucj@gmail.com>
6877 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6878 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6879 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6880 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6881 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6882 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6883 * config/nds32/constraints.md: New constraints and checking for hard
6884 float configuration.
6885 * config/nds32/iterators.md: New mode iterator and attribute for hard
6886 float configuration.
6887 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6889 * config/nds32/nds32-fpu.md: New file.
6890 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6891 deal with hard float code generation.
6892 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6894 (abi_type, float_reg_number): New enum type.
6895 * config/nds32/nds32-predicates.c: New predicates for hard float.
6896 * config/nds32/nds32-protos.h: Declare functions for hard float.
6897 * config/nds32/nds32.c: Implementation for hard float configuration.
6898 * config/nds32/nds32.h: Definitions for hard float configuration.
6899 * config/nds32/nds32.md: Include hard float machine description and
6900 modify patterns for hard float configuration.
6901 * config/nds32/nds32.opt: New options for hard float configuration.
6902 * config/nds32/predicates.md: New predicates for hard float
6905 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6907 * common/config/nds32/nds32-common.c
6908 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6910 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6913 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6914 CONSTRUCTOR_ELT (ctor, ...)->value.
6916 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6919 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6921 2018-04-05 Tom de Vries <tom@codesourcery.com>
6924 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6927 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6928 Kito Cheng <kito.cheng@gmail.com>
6930 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6931 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6932 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6933 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6935 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6936 Kito Cheng <kito.cheng@gmail.com>
6938 * config/nds32/constraints.md (Ufe): New memory constraint.
6939 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6940 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6941 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6943 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6944 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6946 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6948 * config/nds32/nds32.md: Use optimize_size in the condition for
6949 alu-shift instructions.
6951 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6953 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6955 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6957 * config/nds32/nds32.md (negsi2): Refine pattern.
6959 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6960 Chung-Ju Wu <jasonwucj@gmail.com>
6962 * config/nds32/iterators.md (shift_rotate): New code iterator.
6963 (shift): New code attribute.
6964 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6965 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6966 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6967 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6968 bit-wise operations.
6969 (andsi3, *andsi3): Ditto.
6970 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6971 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6972 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6973 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6974 nds32_ior_operand, nds32_xor_operand): New predicates.
6976 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6978 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6979 (addsi3, subsi3): ... this.
6981 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6983 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6985 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6987 * config/nds32/nds32.md: Adjust indention.
6989 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6991 * config/nds32/nds32.md (feature): New attribute.
6993 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6995 * config/nds32/nds32.md (subtype): New attribute.
6997 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7000 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7001 expansion to perform a bitwise AND of the argument followed by a
7002 boolean negation of the result.
7004 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7006 PR rtl-optimization/84878
7007 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7008 the basic block. Assert the use reference is not artificial and that
7009 it has an associated insn.
7011 2018-04-04 Michael Matz <matz@suse.de>
7013 * builtins.c (compute_objsize): Pass correct operand
7014 to array_at_struct_end_p.
7016 2018-04-04 Richard Biener <rguenther@suse.de>
7019 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7020 from contexts for DINFO_LEVEL_TERSE and below.
7022 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7024 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7025 resiter_operand condition.
7026 * config/nds32/nds32.md (*move<mode>): Ditto.
7028 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7029 Monk Chiang <sh.chiang04@gmail.com>
7031 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7033 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7035 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7037 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7038 Kito Cheng <kito.cheng@gmail.com>
7040 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7041 nds32_cond_code_str, output_cond_branch,
7042 output_cond_branch_compare_zero, nds32_expand_cbranch,
7043 nds32_expand_cstore, nds32_expand_movcc,
7044 nds32_output_cbranchsi4_equality_zero,
7045 nds32_output_cbranchsi4_equality_reg,
7046 nds32_output_cbranchsi4_equality_reg_or_const_int,
7047 nds32_output_cbranchsi4_greater_less_zero: New functions.
7048 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7049 nds32_expand_cstore, nds32_expand_movcc,
7050 nds32_output_cbranchsi4_equality_zero,
7051 nds32_output_cbranchsi4_equality_reg,
7052 nds32_output_cbranchsi4_equality_reg_or_const_int,
7053 nds32_output_cbranchsi4_greater_less_zero): Declare.
7054 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7055 nds32_rimm11s_operand): New predicates.
7056 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7057 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7060 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7062 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7063 * config/nds32/nds32.md: Ditto.
7064 * config/nds32/pipelines.md: Ditto.
7066 2018-04-04 Richard Biener <rguenther@suse.de>
7068 PR tree-optimization/85168
7069 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7070 propagating abnormals.
7072 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7074 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7076 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7077 Kito Cheng <kito.cheng@gmail.com>
7079 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7080 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7081 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7082 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7083 * config/nds32/nds32.md (sibcall_internal): New.
7084 (sibcall_register): Remove.
7085 (sibcall_immediate): Remove.
7086 (sibcall_value_internal): New.
7087 (sibcall_value_register): Remove.
7088 (sibcall_value_immediate): Remove.
7089 * config/nds32/predicates.md (nds32_general_register_operand): New.
7090 (nds32_call_address_operand): New.
7092 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7094 PR rtl-optimization/85167
7095 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7096 bb_defs if *split_p, instead preinitialize it to NULL.
7098 PR tree-optimization/85156
7099 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7100 evaluating the argument multiple times.
7102 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7104 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7106 (_mm_cvtpd_ps): Likewise.
7107 (_mm_cvttpd_epi32): Likewise.
7108 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7109 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7110 vector, pixel, and bool following altivec.h include.
7112 2018-04-03 Martin Sebor <msebor@redhat.com>
7114 * doc/extend.texi (Common Function Attributes): Clarify.
7115 (const attribute): Likewise.
7116 (pure attribute): Likewise.
7118 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7121 * config/i386/i386.c (ix86_expand_vector_set): Use
7122 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7124 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7126 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7127 instructions when changing rounding bits to preserve precision bits
7128 in the x87 control word.
7130 2018-04-03 Martin Liska <mliska@suse.cz>
7132 PR tree-optimization/82491
7133 * rtl.h (strip_offset_and_add): Replace += suboffset with
7134 poly_uint64 () + suboffset.
7136 2018-03-29 Martin Liska <mliska@suse.cz>
7137 Martin Jambor <mjambor@suse.cz>
7140 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7141 param_type is not an integral or pointer type.
7143 2018-04-03 Richard Biener <rguenther@suse.de>
7145 * sese.h (recompute_all_dominators): Remove.
7147 2018-04-02 Martin Sebor <msebor@redhat.com>
7149 * doc/invoke.texi (-Wrestrict): Fix typos.
7151 2018-04-02 Jim Wilson <jimw@sifive.com>
7153 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7154 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7155 (<optab>di3, <optab>si3_extend): Likewise.
7156 (<optab>si3_mask, <optab>si3_mask_1): New.
7157 (<optab>di3_mask, <optab>di3_mask_1): New.
7158 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7159 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7160 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7162 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7164 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7167 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7169 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7170 (nds32_canonicalize_comparison): New function.
7172 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7173 Kito Cheng <kito.cheng@gmail.com>
7174 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7176 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7177 * config/nds32/constants.md (unspec_volatile_element): Add
7178 UNSPEC_VOLATILE_RELAX_GROUP.
7179 * config/nds32/nds32-relax-opt.c: New file.
7180 * config/nds32/nds32-predicates.c
7181 (nds32_symbol_load_store_p): New function.
7182 * config/nds32/nds32-protos.h
7183 (nds32_symbol_load_store_p): Declare function.
7184 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7185 * config/nds32/nds32.c
7186 (nds32_register_pass): New function to register pass.
7187 (nds32_register_passes): New function to register passes.
7188 * config/nds32/nds32.md (relax_group): New pattern.
7189 * config/nds32/nds32.opt (mrelax-hint): New option.
7190 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7192 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7194 * config/nds32/t-nds32: Modify files dependency.
7196 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7198 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7199 (PROFILE_HOOK): Define its implementation.
7201 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7203 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7204 type and 32-bit size.
7206 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7209 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7210 (V_128_256): New mode iterator.
7211 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7212 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7213 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7215 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7216 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7218 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7221 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7222 NaN inputs correctly.
7224 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7227 * config/rs6000/vsx.md (??r): New mode attribute.
7228 (*vsx_mov<mode>_64bit): Use it.
7229 (*vsx_mov<mode>_32bit): Likewise.
7231 2018-03-30 Martin Sebor <msebor@redhat.com>
7233 PR tree-optimization/84818
7234 * builtins.c (check_access): Use warning_n.
7236 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7239 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7241 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7244 2018-03-30 Julia Koval <julia.koval@intel.com>
7247 * x86-tune.def (movx, partial_reg_dependency): Enable for
7250 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7253 * lra-constraints.c (process_alt_operands): Move setting
7254 this_alternative_matches below.
7256 2018-03-29 Martin Liska <mliska@suse.cz>
7259 * doc/invoke.texi: Document how LTO works with debug info.
7260 Describe auto-load support of binutils. Mention 'x86-64'
7261 as valid option value of -march option.
7263 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7265 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7268 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7269 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7270 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7273 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7276 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7277 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7278 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7279 for RS6000_BTM_POWERPC64.
7280 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7281 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7282 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7287 2018-03-28 Carl Love <cel@us.ibm.com>
7290 2017-09-27 Carl Love <cel@us.ibm.com>
7292 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7293 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7294 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7297 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7299 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7300 instead of __vector bool.
7301 (_mm_max_pu8): Likewise.
7302 (_mm_min_pi16): Likewise.
7304 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7307 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7308 (DIVWEUO): Likewise.
7310 (DIVDEUO): Likewise.
7311 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7312 DIVWEUO and DIVDEUO.
7313 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7314 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7315 (div_extend): Likewise.
7316 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7318 (__builtin_divweuo): Likewise.
7319 (__builtin_divdeo): Likewise.
7320 (__builtin_divdeuo): Likewise.
7322 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7325 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7326 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7328 PR tree-optimization/82004
7329 * gimple-match-head.c (optimize_pow_to_exp): New function.
7330 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7331 Don't fold to exp if optimize_pow_to_exp is false.
7333 2018-03-28 Martin Liska <mliska@suse.cz>
7336 * calls.c (initialize_argument_information): Fix trailing space.
7337 * common.opt: Fix typo and provide better explanation for
7338 -fsanitize-coverage option.
7339 * config/i386/i386.opt: Fix typo.
7341 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7342 Martin Liska <mliska@suse.cz>
7345 * gimplify.c (asan_poison_variable): Don't do the check for
7346 gimplify_omp_ctxp here.
7347 (gimplify_decl_expr): Do it here.
7348 (gimplify_target_expr): Likewise.
7350 2018-03-28 Martin Liska <mliska@suse.cz>
7353 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7354 chkp_type_bounds_count if MPX is not enabled.
7356 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7358 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7360 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7363 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7364 function to create the function decl for complex long double
7365 multiply and divide for -mabi=ieeelongdouble.
7366 (init_float128_ieee): Call it.
7368 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7371 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7372 -fcf-protection=branch -mibt.
7373 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7375 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7378 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7379 arm_word_relocations.
7381 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7384 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7385 extern array declarations.
7387 2018-03-27 Richard Biener <rguenther@suse.de>
7390 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7391 explicit single_use checks.
7393 2018-03-27 Richard Biener <rguenther@suse.de>
7395 PR tree-optimization/85082
7396 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7399 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7401 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7402 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7403 Turn on fasynchronous-unwind-tables and funwind-tables.
7405 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7408 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7409 (*bmi_blsr_<mode>_ccz): Ditto.
7411 2018-03-26 Tom de Vries <tom@codesourcery.com>
7413 PR tree-optimization/85063
7414 * omp-general.c (offloading_function_p): New function. Factor out
7416 * omp-offload.c (pass_omp_target_link::gate): ... here.
7417 * omp-general.h (offloading_function_p): Declare.
7418 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7419 with attribute omp declare target for offloading functions.
7421 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7423 PR tree-optimization/84005
7424 * tree-data-ref.h (get_base_for_alignment): Declare.
7425 * tree-data-ref.c (get_base_for_alignment_1): New function.
7426 (get_base_for_alignment): Likewise.
7427 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7428 get_base_for_alignment to find a suitable base object, instead
7429 of always using drb->base_address.
7431 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7434 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7435 known size by default.
7437 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7440 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7441 and non BLKmode operands.
7443 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7446 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7447 Clean up attributes.
7449 2018-03-23 Richard Biener <rguenther@suse.de>
7452 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7453 we are going to emit early debug for LTO.
7455 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7458 * function.c (match_asm_constraints_1): Don't optimize if input
7459 doesn't satisfy general_operand predicate for output's mode.
7462 * alias.c (write_dependence_p): Don't require for x_canonicalized
7463 non-VOIDmode if x has VOIDmode.
7466 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7467 just don't try to optimize it rather than assert it never happens.
7469 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7471 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7472 macro expansions for definition of ST_INTERNAL_<mode> and
7473 LD_INTERNAL_<mode> builtins.
7474 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7476 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7478 (altivec_expand_st_builtin): Likewise.
7479 (altivec_expand_builtin): Remove calls to deleted functions.
7480 (rs6000_address_for_altivec): Delete this function.
7481 * config/rs6000/vector.md: Remove expands for
7482 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7484 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7487 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7488 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7489 re-computing once computed.
7490 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7491 (arm_init_machine_status): Initialize
7492 machine->static_chain_stack_bytes.
7494 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7497 * doc/extend.texi: Add four new prototypes for vec_ld.
7498 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7499 definitions for more logical presentation.
7500 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7501 entries for V1TI variants of __builtin_altivec_ld builtin.
7502 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7503 handling of V1TI variant of LVX icode pattern.
7504 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7505 (rs6000_gimple_fold_builtin): Likewise.
7506 (altivec_init_builtins): Add code to define
7507 __builtin_altivec_lvx_v1ti function.
7509 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7512 * function.c (match_asm_constraints_1): Don't do the optimization
7513 if input isn't a REG, SUBREG, MEM or constant.
7515 2018-03-22 Tom de Vries <tom@codesourcery.com>
7517 PR tree-optimization/84956
7518 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7519 bb_has_abnormal_pred.
7521 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7524 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7525 DECL_INITIAL (decl) to decl at the end.
7526 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7529 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7531 * doc/extend.texi (__builtin_tgmath): Document when complex
7532 integer types are treated as _Complex _Float64.
7534 2018-03-21 Tom de Vries <tom@codesourcery.com>
7536 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7538 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7540 PR tree-optimization/84960
7541 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7542 if it is ENTRY block, move them into single succ of ENTRY in that case.
7544 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7546 PR tree-optimization/84811
7547 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7548 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7549 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7550 (known_subrange_p): Update accordingly. Cast each value involved
7551 in the size comparison, rather than casting the result of the
7554 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7556 PR tree-optimization/84982
7557 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7558 by flipping the least significant bit rather than all bits from
7559 bitpos to bitpos + bitsize - 1.
7561 2018-03-21 Nathan Sidwell <nathan@acm.org>
7563 * doc/extend.texi (Deprecated Features): Remove mention of
7564 long-deleted deprecations.
7566 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7569 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7570 * configure: Regenerate.
7572 2018-03-21 Tom de Vries <tom@codesourcery.com>
7574 PR tree-optimization/83126
7575 * tree-parloops.c (num_phis): New function.
7576 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7578 2018-03-21 Nathan Sidwell <nathan@acm.org>
7580 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7581 mention anon-struct/union members and trailing attributes.
7583 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7585 PR tree-optimization/84969
7586 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7587 builtin memset partitions if they set different rhs values.
7589 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7591 PR rtl-optimization/84989
7592 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7593 VEC_DUPLICATE with scalar result mode.
7595 2018-03-21 Martin Liska <mliska@suse.cz>
7598 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7599 not intended return statement.
7601 2018-03-21 Martin Liska <mliska@suse.cz>
7604 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7605 (chkp_find_bound_slots_1): Limit number of iterations.
7607 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7610 * Minor grammar fixes for x86 options.
7612 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7615 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7616 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7618 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7621 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7622 (altivec_lvx_<mode>_1op): Likewise.
7623 (altivec_stvx_<mode>_2op): Likewise.
7624 (altivec_stvx_<mode>_1op): Likewise.
7625 (altivec_lvx_<VM2:mode>): New define_expand.
7626 (altivec_stvx_<VM2:mode>): Likewise.
7627 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7628 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7629 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7630 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7631 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7632 (rs6000_gen_lvx): Likewise.
7633 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7634 (altivec_expand_stv_builtin): Likewise.
7635 (altivec_expand_builtin): Likewise.
7636 * config/rs6000/vector.md: Likewise.
7638 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7641 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7644 2018-03-20 Richard Biener <rguenther@suse.de>
7647 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7648 sign-conversions as zero, fall back to standard scalar_stmt
7651 2018-03-20 Martin Liska <mliska@suse.cz>
7654 * predict.c (rebuild_frequencies): Handle case when we have
7655 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7657 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7660 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7661 flag_section_anchors.
7662 * varasm.c (use_blocks_for_decl_p): Remove hack for
7663 dw2_force_const_mem.
7666 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7668 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7669 be created, use lowpart_subreg of operands[0] rather than operands[0]
7671 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7672 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7673 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7674 and n constraint instead of aarch64_shift_imm_di and Usd.
7675 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7676 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7678 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7681 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7682 to favor GPR over NEON registers.
7683 (<shift>di3_neon): Likewise.
7685 2018-03-20 Tom de Vries <tom@codesourcery.com>
7688 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7689 (nvptx_process_pars): Emit bar.sync asap and alap.
7691 2018-03-20 Tom de Vries <tom@codesourcery.com>
7694 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7695 seen_label if seen_label is already set.
7697 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7700 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7701 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7702 Use 1U instead of 1. Formatting fixes.
7705 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7706 instead of TREE_TYPE (s1) for the return value.
7708 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7710 PR tree-optimization/84946
7711 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7712 bitsize + bitsize in poly_uint64 rather than poly_int64.
7715 * dwarf2asm.c: Include fold-const.c.
7716 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7717 of decl rather than decl itself.
7719 PR rtl-optimization/84643
7720 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7722 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7725 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7726 calling assemble_variable.
7728 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7731 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7732 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7734 2018-03-19 Jim Wilson <jimw@sifive.com>
7737 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7738 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7739 (riscv_first_stack_step): Likewise.
7740 (riscv_option_override): Use STACK_BOUNDARY instead of
7742 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7744 (BIGGEST_ALIGNMENT): Set to 128.
7745 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7746 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7749 2018-03-19 Richard Biener <rguenther@suse.de>
7751 PR tree-optimization/84933
7752 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7753 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7755 2018-03-19 Richard Biener <rguenther@suse.de>
7757 PR tree-optimization/84859
7758 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7759 (cond_if_else_store_replacement): Perform sinking operation on
7760 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7761 Generalize what a BB with a single eligible store is.
7763 2018-03-19 Richard Biener <rguenther@suse.de>
7765 PR tree-optimization/84929
7766 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7767 chrec_is_positive against non-chrec arg.
7769 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7772 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7774 2018-03-18 Martin Liska <mliska@suse.cz>
7776 PR rtl-optimization/84635
7777 * regrename.c (build_def_use): Use matches_mode only when
7780 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7782 PR tree-optimization/84913
7783 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7784 vectorize chains of COND_EXPRs.
7786 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7788 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7790 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7792 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7794 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7796 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7798 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7799 Kito Cheng <kito.cheng@gmail.com>
7801 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7802 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7803 (nds32_adjust_reg_alloc_order): New function.
7804 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7806 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7808 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7809 nds32_print_operand, nds32_print_operand_address): Use
7810 HOST_WIDE_INT_PRINT_DEC instead.
7812 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7814 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7816 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7819 * config/i386/i386.c (initial_ix86_tune_features,
7820 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7822 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7823 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7824 rather than 1u << ix86_tune. Formatting fix.
7825 (ix86_option_override_internal): Change ix86_arch_mask from
7826 unsigned int to unsigned HOST_WIDE_INT, initialize to
7827 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7828 (ix86_function_specific_restore): Likewise.
7830 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7833 * postreload.c (reload_combine_recognize_pattern): Perform
7834 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7835 truncate_int_for_mode the result for the destination's mode.
7838 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7839 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7841 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7844 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7847 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7848 regno_allocno_class_array and sorted_pseudos.
7849 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7850 insns where regno is used.
7852 2018-03-16 Martin Liska <mliska@suse.cz>
7855 * multiple_target.c (create_dispatcher_calls): Redirect
7856 reference in the symbol table.
7858 2018-03-16 Martin Liska <mliska@suse.cz>
7861 * multiple_target.c (create_dispatcher_calls): Redirect also
7864 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7868 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7869 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7870 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7872 2018-03-16 Julia Koval <julia.koval@intel.com>
7874 * doc/invoke.texi (Skylake Server): Add CLWB.
7875 Cannonlake): Remove CLWB.
7877 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7879 PR tree-optimization/84841
7880 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7882 (FLOAT_ONE_CONST_TYPE): Define.
7883 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7884 (sort_by_operand_rank): Put entries with higher constant_type last
7885 rather than first to match comments.
7887 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7889 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7892 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7895 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7896 check and use address of target always.
7898 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7901 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7902 (indirect_thunk_bnd_needed): Likewise.
7903 (indirect_thunks_used): Likewise.
7904 (indirect_thunks_bnd_used): Likewise.
7905 (indirect_return_needed): New.
7906 (indirect_return_bnd_needed): Likewise.
7907 (output_indirect_thunk_function): Add a bool argument for
7909 (output_indirect_thunk_function): Don't generate alias for
7910 function return thunk.
7911 (ix86_code_end): Call output_indirect_thunk_function to generate
7912 function return thunks.
7913 (ix86_output_function_return): Set indirect_return_bnd_needed
7914 and indirect_return_needed instead of indirect_thunk_bnd_needed
7915 and indirect_thunk_needed.
7917 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7919 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7920 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7921 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7923 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7924 Paul Hua <paul.hua.gm@gmail.com>
7927 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7929 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7931 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7932 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7935 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7938 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7939 instead of GET_MODE_SIZE when comparing Units.
7941 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7944 * varasm.c (hash_section): Return an unchangeble hash value
7945 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7946 Return !aarch64_can_use_per_function_literal_pools_p ().
7948 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7951 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7952 rather than address of cmode as last argument to prepare_cmp_insn.
7954 2018-03-15 Julia Koval <julia.koval@intel.com>
7956 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7957 F_AVX512VNNI, F_AVX512BITALG): New.
7959 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7962 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7963 insn for floating-point loads and stores.
7965 2018-03-14 Carl Love <cel@us.ibm.com>
7967 * config/rs6000/rs6000-c.c: Add macro definitions for
7968 ALTIVEC_BUILTIN_VEC_PERMXOR.
7969 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7970 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7971 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7972 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7974 * config/doc/extend.texi: Add prototypes for vec_permxor.
7976 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7979 * diagnostic-show-locus.c (class layout_point): Convert m_line
7980 from int to linenum_type.
7981 (line_span::comparator): Use linenum "compare" function when
7982 comparing line numbers.
7983 (test_line_span): New function.
7984 (layout_range::contains_point): Convert param "row" from int to
7986 (layout_range::intersects_line_p): Likewise.
7987 (layout::will_show_line_p): Likewise.
7988 (layout::print_source_line): Likewise.
7989 (layout::should_print_annotation_line_p): Likewise.
7990 (layout::print_annotation_line): Likewise.
7991 (layout::print_leading_fixits): Likewise.
7992 (layout::annotation_line_showed_range_p): Likewise.
7993 (struct line_corrections): Likewise for field m_row.
7994 (line_corrections::line_corrections): Likewise for param "row".
7995 (layout::print_trailing_fixits): Likewise.
7996 (layout::get_state_at_point): Likewise.
7997 (layout::get_x_bound_for_row): Likewise.
7998 (layout::print_line): Likewise.
7999 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8000 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8001 * input.c (selftest::test_linenum_comparisons): New function.
8002 (selftest::input_c_tests): Call it.
8003 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8004 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8005 * selftest.h (ASSERT_GT): New macro.
8006 (ASSERT_GT_AT): New macro.
8007 (ASSERT_LT): New macro.
8008 (ASSERT_LT_AT): New macro.
8010 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8012 PR rtl-optimization/84780
8013 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8015 2018-03-14 Martin Liska <mliska@suse.cz>
8017 * tree.c (record_node_allocation_statistics): Use
8018 get_stats_node_kind.
8019 (get_stats_node_kind): New function extracted from
8020 record_node_allocation_statistics.
8021 (free_node): Use get_stats_node_kind.
8023 2018-03-14 Richard Biener <rguenther@suse.de>
8025 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8026 that the value-set of ANTIC_IN doesn't grow.
8029 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8031 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8032 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8034 2018-03-14 Julia Koval <julia.koval@intel.com>
8036 * config.gcc (icelake-client, icelake-server): New.
8038 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8039 (initial_ix86_arch_features): Ditto.
8040 (PTA_SKYLAKE): Add SGX.
8041 (PTA_ICELAKE): Remove.
8042 (PTA_ICELAKE_CLIENT): New.
8043 (PTA_ICELAKE_SERVER): New.
8044 (ix86_option_override_internal): Split up icelake on icelake client and
8046 (get_builtin_code_for_version): Ditto.
8047 (fold_builtin_cpu): Ditto.
8048 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8049 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8050 * config/i386/i386.h (processor_type): Ditto.
8051 * doc/invoke.texi: Ditto.
8053 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8056 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8057 INTEGER_CST offset, add it together with bitpos / 8 and
8058 sign extend based on POINTER_SIZE.
8062 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8065 * config/i386/constraints.md (Yc): New register constraint.
8066 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8067 Use Yc constraint for alternative 2 of operand 0. Remove
8068 preferred_for_speed attribute.
8070 2018-03-14 Richard Biener <rguenther@suse.de>
8072 PR tree-optimization/84830
8073 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8074 with the old one to avoid oscillations.
8076 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8079 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8081 (assign_by_spills): Return a flag of reload assignment failure.
8082 Do not process the reload assignment failures. Do not spill other
8083 reload pseudos if they has the same reg class. Update n if
8085 (lra_assign): Add a return arg. Set up from the result of
8086 assign_by_spills call.
8087 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8088 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8089 usage_insns if it is not NULL.
8090 (spill_hard_reg_in_range): New function.
8091 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8092 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8093 function prototypes.
8094 (lra_assign): Change prototype.
8095 * lra.c (lra): Add code to deal with fails by splitting hard reg
8098 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8100 * config/riscv/riscv.opt (mrelax): New option.
8101 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8102 "norelax" when riscv_mrelax is disabled.
8103 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8105 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8108 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8109 reassociation for int modes.
8111 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8113 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8114 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8116 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8117 * config/aarch64/aarch64-sve.md
8118 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8119 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8120 (*extend<mode><Vwide>2): Rename to...
8121 (aarch64_sve_extend<mode><Vwide>2): ...this.
8122 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8123 renaming the old pattern to...
8124 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8126 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8127 define_expand, renaming the old pattern to...
8128 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8129 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8130 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8131 account when deciding which SVE instruction the optab should use.
8132 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8134 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8136 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8137 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8138 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8139 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8140 (tlsdesc_small_<mode>): Turn a define_expand and use
8141 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8142 (tlsdesc_small_advsimd_<mode>): ...this.
8143 (tlsdesc_small_sve_<mode>): New pattern.
8145 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8147 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8148 (UNSPEC_UMUL_HIGHPART): New constants.
8149 (MUL_HIGHPART): New int iteraor.
8150 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8151 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8153 (*<su>mul<mode>3_highpart): New define_insn.
8155 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8158 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8161 2018-03-13 Martin Liska <mliska@suse.cz>
8164 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8166 (sem_item_optimizer::~sem_item_optimizer): Release it.
8167 (sem_item_optimizer::merge_classes): Register variable aliases.
8168 (sem_item_optimizer::fixup_pt_set): New function.
8169 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8170 * ipa-icf.h: Declare new variables and functions.
8172 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8175 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8176 integer_pow2p@2 and test integer_pow2p in condition.
8177 (A < 0 ? C : 0): Similarly for @1.
8180 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8181 characters starting at p contain '\0' character, don't look beyond
8185 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8186 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8189 * reg-stack.c (change_stack): Change update_end var from int to
8190 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8191 also call set_block_for_insn on the newly added insns and rescan.
8194 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8195 on the last operand.
8198 * tree.c (stabilize_reference_1): Return save_expr (e) for
8199 STATEMENT_LIST even if it doesn't have side-effects.
8201 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8203 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8205 2018-03-12 Renlin Li <renlin.li@arm.com>
8207 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8208 aarch64_output_scalar_simd_mov_immediate.
8210 2018-03-12 Martin Sebor <msebor@redhat.com>
8212 PR tree-optimization/83456
8213 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8214 for perfectly overlapping calls to memcpy.
8215 (gimple_fold_builtin_memory_chk): Same.
8216 (gimple_fold_builtin_strcpy): Handle no-warning.
8217 (gimple_fold_builtin_stxcpy_chk): Same.
8218 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8220 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8222 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8223 parameter. Use it for SFmode.
8224 (rs6000_function_arg_advance_1): Adjust.
8225 (rs6000_function_arg): Adjust.
8226 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8228 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8230 PR rtl-optimization/84169
8231 PR rtl-optimization/84780
8232 * combine.c (can_combine_p): Check for a 2-insn combination whether
8233 the destination register is used between the two insns, too.
8235 2018-03-12 Richard Biener <rguenther@suse.de>
8237 PR tree-optimization/84803
8238 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8239 for refs DR analysis didn't process.
8241 2018-03-12 Richard Biener <rguenther@suse.de>
8243 PR tree-optimization/84777
8244 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8245 force-vectorize loops ignore whether we are optimizing for size.
8247 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8249 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8250 (TARGET_MD_ASM_ADJUST): Define.
8252 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8253 Kito Cheng <kito.cheng@gmail.com>
8254 Chung-Ju Wu <jasonwucj@gmail.com>
8256 * config/nds32/nds32.c (nds32_compute_stack_frame,
8257 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8258 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8259 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8260 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8261 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8262 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8263 * config/nds32/nds32.md (prologue, epilogue): Use macro
8264 NDS32_V3PUSH_AVAILABLE_P to do checking.
8266 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8269 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8270 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8271 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8272 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8273 addition of most attributes on !orig_type_die or the attribute not
8274 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8276 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8277 Chung-Ju Wu <jasonwucj@gmail.com>
8279 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8281 * config/nds32/nds32.opt (mvh): New option.
8283 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8284 Chung-Ju Wu <jasonwucj@gmail.com>
8286 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8288 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8289 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8292 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8293 Chung-Ju Wu <jasonwucj@gmail.com>
8295 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8297 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8298 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8300 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8301 Kito Cheng <kito.cheng@gmail.com>
8302 Chung-Ju Wu <jasonwucj@gmail.com>
8304 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8305 UNSPEC_FFMISM and UNSPEC_FLMISM.
8306 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8307 for ffb, ffmism and flmism.
8308 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8309 (unspec_ffmism): Ditto.
8310 (unspec_flmism): Ditto.
8311 (nds32_expand_builtin_impl): Check if string extension is available.
8312 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8313 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8315 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8318 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8321 * lra-assigns.c (assign_by_spills): Return a flag of reload
8322 assignment failure. Do not process the reload assignment
8323 failures. Do not spill other reload pseudos if they has the same
8325 (lra_assign): Add a return arg. Set up from the result of
8326 assign_by_spills call.
8327 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8328 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8329 usage_insns if it is not NULL.
8330 (spill_hard_reg_in_range): New function.
8331 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8332 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8333 function prototypes.
8334 (lra_assign): Change prototype.
8335 * lra.c (lra): Add code to deal with fails by splitting hard reg
8338 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8341 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8343 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8346 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8347 (dw_val_node): Add val_symbolic_view.
8348 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8349 (symview_upper_bound): New.
8350 (new_line_info_table): Initialize symviews_since_reset.
8351 (dwarf2out_source_line): Count symviews_since_reset and set
8352 symview_upper_bound.
8353 (dw_val_equal_p): Handle symview.
8354 (add_AT_symview): New.
8355 (print_dw_val): Handle symview.
8356 (attr_checksum, attr_checksum_ordered): Likewise.
8357 (same_dw_val_p, size_of_die): Likewise.
8358 (value_format, output_die): Likewise.
8359 (add_high_low_attributes): Use add_AT_symview for entry_view.
8360 (dwarf2out_finish): Reset symview_upper_bound, clear
8363 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8366 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8367 Add strict argument and use it.
8368 (rs6000_split_multireg_move): Update for new strict argument.
8369 (mem_operand_gpr): Disallow all non-offsettable addresses.
8370 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8372 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8375 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8376 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8377 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8380 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8381 decl, use remap_type if we want to use the type.
8383 2018-03-09 Martin Sebor <msebor@redhat.com>
8385 PR tree-optimization/84526
8386 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8388 (builtin_access::generic_overlap): Be prepared to handle non-array
8391 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8393 PR rtl-optimization/84682
8394 * lra-constraints.c (process_address_1): Check is_address flag
8395 for address constraints.
8396 (process_alt_operands): Likewise.
8397 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8398 preprocess_constraints.
8399 * recog.h (preprocess_constraints): Add oploc parameter.
8401 * recog.c (preprocess_constraints): Test address_operand for
8402 CT_ADDRESS constraints.
8404 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8407 * lra-assigns.c (assign_by_spills): Return a flag of reload
8408 assignment failure. Do not process the reload assignment
8409 failures. Do not spill other reload pseudos if they has the same
8411 (lra_assign): Add a return arg. Set up from the result of
8412 assign_by_spills call.
8413 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8414 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8415 usage_insns if it is not NULL.
8416 (spill_hard_reg_in_range): New function.
8417 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8418 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8419 function prototypes.
8420 (lra_assign): Change prototype.
8421 * lra.c (lra): Add code to deal with fails by splitting hard reg
8424 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8427 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8428 Accept complain bool parameter. Only emit errors if it is true.
8429 (arm_parse_cpu_option_name): Likewise.
8430 (arm_target_thumb_only): Adjust callers of the above.
8431 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8432 prototype to take a default true bool parameter.
8433 (arm_parse_arch_option_name): Likewise.
8435 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8436 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8440 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8441 * configure: Regenerate.
8442 * configure.ac ("linker --version-script option"): New.
8443 ("linker soname option"): New.
8445 2018-03-09 Richard Biener <rguenther@suse.de>
8447 PR tree-optimization/84775
8448 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8449 immediate uses of predicate stmts and mark them modified.
8452 PR tree-optimization/84178
8453 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8455 (version_loop_for_if_conversion): Delay update_ssa call.
8456 (tree_if_conversion): Delay update_ssa until after predicate
8459 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8462 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8463 when the function accesses prior frames.
8465 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8468 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8469 gen_llsym, otherwise call maybe_gen_llsym.
8472 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8473 has ',' character inside of it.
8475 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8478 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8479 as clobbering CC_REGNUM.
8481 2018-03-08 Richard Biener <rguenther@suse.de>
8484 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8485 (follow_copies_to_constant): Do not follow SSA names registered
8488 2018-03-08 Richard Biener <rguenther@suse.de>
8490 PR tree-optimization/84178
8491 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8493 (version_loop_for_if_conversion): Delay update_ssa call.
8494 (tree_if_conversion): Delay update_ssa until after predicate
8497 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8499 PR tree-optimization/84178
8500 * tree-if-conv.c (release_bb_predicate): Remove the
8501 the assertion that the stmts have NULL use_ops.
8502 Discard the statements, asserting that they haven't
8503 yet been added to a BB.
8505 2018-03-08 Richard Biener <rguenther@suse.de>
8507 PR tree-optimization/84746
8508 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8509 (phi_translate): Pass in destination ANTIC_OUT set.
8510 (phi_translate_1): Likewise. For a simplified result lookup
8511 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8512 (phi_translate_set): Adjust.
8513 (do_pre_regular_insertion): Likewise.
8514 (do_pre_partial_partial_insertion): Likewise.
8516 2018-03-08 Martin Liska <mliska@suse.cz>
8518 PR gcov-profile/84735
8519 * doc/gcov.texi: Document usage of profile files.
8520 * gcov-io.h: Document changes in the format.
8522 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8526 * dwarf2out.c (struct dw_line_info_table): Update comments for
8528 (FORCE_RESET_NEXT_VIEW): New.
8529 (FORCE_RESETTING_VIEW_P): New.
8530 (RESETTING_VIEW_P): Check for -1 too.
8531 (ZERO_VIEW_P): Likewise.
8532 (new_line_info_table): Force-reset next view.
8533 (dwarf2out_begin_function): Likewise.
8534 (dwarf2out_source_line): Simplify zero_view_p initialization.
8535 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8536 view directly. Omit view when omitting .loc at line 0.
8538 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8540 PR tree-optimization/84740
8541 * tree-switch-conversion.c (process_switch): Call build_constructors
8542 only if info.phi_count is non-zero.
8544 PR tree-optimization/84739
8545 * tree-tailcall.c (find_tail_calls): Check call arguments against
8546 DECL_ARGUMENTS (current_function_decl) rather than
8547 DECL_ARGUMENTS (func) when checking for tail recursion.
8549 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8551 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8552 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8553 Volker Reichelt's entry and add entries for people that perform
8554 GCC fuzzy testing and report numerous bugs.
8556 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8559 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8560 readonly data in sdata, if that is disabled.
8561 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8562 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8563 -mreadonly-in-sdata option.
8565 2018-03-07 Martin Sebor <msebor@redhat.com>
8567 PR tree-optimization/84468
8568 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8569 basic block when looking for nul assignment.
8571 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8574 * except.h (output_function_exception_table): Adjust prototype.
8575 * except.c (output_function_exception_table): Remove FNNAME parameter
8576 and add SECTION parameter. Ouput one part of the table at a time.
8577 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8578 the first part of the exception table and emit unwind directives.
8579 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8580 (i386_pe_seh_cold_init): Likewise.
8581 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8582 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8583 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8584 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8585 * config/i386/winnt.c: Include except.h.
8586 (struct seh_frame_state): Add reg_offset, after_prologue and
8587 in_cold_section fields.
8588 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8589 (i386_pe_seh_cold_init): New function.
8590 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8591 to seh->in_cold_section.
8592 (seh_emit_push): Record the offset of the push.
8593 (seh_emit_save): Record the offet of the save.
8594 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8595 Test seh->after_prologue to disregard the epilogue.
8596 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8597 (i386_pe_end_cold_function): New function.
8599 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8602 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8603 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8606 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8608 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8609 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8612 * multiple_target.c: Include tree-inline.h and intl.h.
8613 (expand_target_clones): Diagnose and fail if node->definition and
8614 !tree_versionable_function_p (node->decl).
8616 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8618 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8620 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8621 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8622 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8624 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8627 * combine.c (try_combine): Use reg_or_subregno instead of handling
8628 just paradoxical SUBREGs and REGs.
8630 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8632 * config/arc/arc.c (arc_finalize_pic): Remove function.
8633 (arc_must_save_register): We use single base PIC register, remove
8634 checks to save/restore the PIC register.
8635 (arc_expand_prologue): Likewise.
8636 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8638 (arc_verify_short): Likewise.
8639 (arc_attr_type): Likewise.
8640 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8641 (walk_stores): Likewise.
8642 (arc_address_cost): Make it static.
8643 (arc_verify_short): Likewise.
8644 (branch_dest): Likewise.
8645 (arc_attr_type): Likewise.
8646 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8647 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8648 (arc_final_prescan_insn): Remove inserting the nops due to
8649 hardware hazards. It is done in reorg step.
8650 (insn_length_variant_t): Remove.
8651 (insn_length_parameters_t): Likewise.
8652 (arc_insn_length_parameters): Likewise.
8653 (arc_get_insn_variants): Likewise.
8654 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8656 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8659 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8662 PR tree-optimization/84687
8663 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8665 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8667 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8669 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8670 Rename to ppc_speculation_barrier.
8671 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8672 __builtin_ppc_speculation_barrier.
8674 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8677 * combine.c (combine_simplify_rtx): Don't try to simplify if
8678 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8681 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8683 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8684 to 32 bytes when compiling for POWER9.
8686 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8689 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8690 regparm >= 3 with no arg reg available also for calls with
8691 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8694 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8696 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8698 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8701 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8703 2018-03-05 Richard Biener <rguenther@suse.de>
8705 PR tree-optimization/84486
8706 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8707 When inserting a __builtin_assume_aligned call set the LHS
8708 SSA name alignment info accordingly.
8710 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8712 PR tree-optimization/84114
8713 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8714 Avoid reassociation of FLOAT_MODE addition.
8716 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8718 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8719 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8720 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8721 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8722 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8723 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8724 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8726 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8727 __builtin_ia32_wbinvd): New builtins.
8728 (SPECIAL_ARGS2): New.
8729 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8730 (SPECIAL_ARGS2): New.
8731 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8732 (ix86_valid_target_attribute_inner_p): Ditto.
8733 (ix86_init_mmx_sse_builtins): Add special_args2.
8734 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8735 TARGET_WBNOINVD_P): New.
8736 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8737 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8738 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8739 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8740 * config/i386/pconfigintrin.h: New file.
8741 * config/i386/wbnoinvdintrin.h: Ditto.
8742 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8744 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8746 2018-03-05 Richard Biener <rguenther@suse.de>
8748 PR tree-optimization/84670
8749 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8751 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8752 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8753 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8754 doesn't grow if all successors (recursively) were visited at least
8757 2018-03-05 Richard Biener <rguenther@suse.de>
8759 PR tree-optimization/84650
8760 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8761 if executed in the loop pipeline.
8763 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8765 * doc/configfiles.texi (Configuration Files): Move info about
8766 conditionalizing $target-protos.h to...
8767 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8768 differs from $target-protos.h.
8770 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8771 Chung-Ju Wu <jasonwucj@gmail.com>
8773 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8774 * config/nds32/nds32-multiple.md (setmemsi): Define.
8775 * config/nds32/nds32-memory-manipulation.c
8776 (nds32_gen_dup_4_byte_to_word_value): New.
8777 (emit_setmem_word_loop): New.
8778 (emit_setmem_byte_loop): New.
8779 (nds32_expand_setmem_loop): New.
8780 (nds32_expand_setmem_loop_v3m): New.
8781 (nds32_expand_setmem_unroll): New.
8782 (nds32_expand_setmem): New.
8784 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8785 Chung-Ju Wu <jasonwucj@gmail.com>
8787 * config/nds32/nds32-memory-manipulation.c
8788 (nds32_emit_load_store): New.
8789 (nds32_emit_post_inc_load_store): New.
8790 (nds32_emit_mem_move): New.
8791 (nds32_emit_mem_move_block): New.
8792 (nds32_expand_movmemsi_loop_unknown_size): New.
8793 (nds32_expand_movmemsi_loop_known_size): New.
8794 (nds32_expand_movmemsi_loop): New.
8795 (nds32_expand_movmemsi_unroll): New.
8796 (nds32_expand_movmemqi): Rename ...
8797 (nds32_expand_movmemsi): ... to this.
8798 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8799 (movmemsi): ... to this.
8800 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8801 (nds32_expand_movmemsi): ... to this.
8803 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8804 Monk Chiang <sh.chiang04@gmail.com>
8805 Chung-Ju Wu <jasonwucj@gmail.com>
8807 * config/nds32/nds32-protos.h
8808 (nds32_expand_load_multiple): New arguments.
8809 (nds32_expand_store_multiple): Ditto.
8810 (nds32_valid_multiple_load_store): Rename ...
8811 (nds32_valid_multiple_load_store_p): ... to this.
8812 * config/nds32/nds32-memory-manipulation.c
8813 (nds32_expand_load_multiple): Refine implementation.
8814 (nds32_expand_store_multiple): Ditto.
8815 * config/nds32/nds32-multiple.md
8816 (load_multiple): Update nds32_expand_load_multiple interface.
8817 (store_multiple): Update nds32_expand_store_multiple interface.
8818 * config/nds32/nds32-predicates.c
8819 (nds32_valid_multiple_load_store): Rename ...
8820 (nds32_valid_multiple_load_store_p): ... to this and refine
8822 * config/nds32/predicates.md
8823 (nds32_load_multiple_and_update_address_operation): New predicate.
8824 (nds32_store_multiple_and_update_address_operation): New predicate.
8826 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8827 Chung-Ju Wu <jasonwucj@gmail.com>
8829 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8830 (combo): New attribute.
8831 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8833 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8835 * config/nds32/nds32.opt: Change -mcmodel= default value.
8837 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8838 Monk Chiang <sh.chiang04@gmail.com>
8839 Chung-Ju Wu <jasonwucj@gmail.com>
8841 * config/nds32/constants.md (unspec_element): New enum.
8842 * config/nds32/constraints.md (Umw): New constraint.
8843 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8844 * config/nds32/nds32-intrinsic.md: Likewise.
8845 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8846 (nds32_valid_smw_lwm_base_p): New.
8847 (nds32_output_smw_single_word): New.
8848 (nds32_output_lmw_single_word): New.
8849 (nds32_expand_unaligned_load): New.
8850 (nds32_expand_unaligned_store): New.
8851 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8852 (nds32_output_smw_single_word): Declare.
8853 (nds32_output_lmw_single_word): Declare.
8854 (nds32_expand_unaligned_load): Declare.
8855 (nds32_expand_unaligned_store): Declare.
8856 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8857 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8858 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8859 NDS32_BUILTIN_UASTORE_DW.
8860 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8863 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8864 Kito Cheng <kito.cheng@gmail.com>
8865 Chung-Ju Wu <jasonwucj@gmail.com>
8867 * config/nds32/nds32-intrinsic.c
8868 (nds32_expand_builtin_null_ftype_reg): Delete.
8869 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8870 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8871 (nds32_read_argument): New.
8872 (nds32_legitimize_target): Ditto.
8873 (nds32_legitimize_argument): Ditto.
8874 (nds32_check_constant_argument): Ditto.
8875 (nds32_expand_unop_builtin): Ditto.
8876 (nds32_expand_unopimm_builtin): Ditto.
8877 (nds32_expand_binop_builtin): Ditto.
8878 (nds32_builtin_decl_impl): Ditto.
8879 (builtin_description): Ditto.
8880 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8881 (nds32_init_builtins_impl): Ditto.
8882 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8883 (nds32_builtin_decl): New.
8884 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8885 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8887 2018-03-02 Jeff Law <law@redhat.com>
8889 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8890 (redundant_insn, fill_simple_delay_slots): Likewise.
8891 (fill_slots_from_thread): Likewise.
8892 * resource.c (mark_referenced_resources): Likewise.
8893 (mark_set_resources, find_dead_or_set_registers): Likewise.
8895 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8897 * substring-locations.h (format_warning_va): Formatting fix for
8899 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8901 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8902 * substring-locations.c: Include intl.h.
8903 (format_warning_va): Turned into small wrapper around
8904 format_warning_n_va, renamed to ...
8905 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8906 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8908 (format_warning_at_substring_n): New function.
8909 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8910 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8911 format_warning_at_substring with just a shorter name instead of
8912 const function pointer.
8913 (fmtwarn_n): New function.
8914 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8915 appropriate, get rid of all the fmtstr temporaries, move conditionals
8916 with G_() wrapped string literals directly into fmtwarn arguments,
8917 cast dir.len to (int), formatting fixes.
8919 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8921 * doc/invoke.texi: Remove "Cilk Plus" references.
8923 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8924 Richard Biener <rguenther@suse.de>
8927 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8928 for error or warning attributes if CALL_FROM_THUNK_P is set.
8931 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8934 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8935 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8938 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8939 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8941 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8942 instead of -1U in last predictors element's probability member.
8944 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8947 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8948 arguments if they are comparable.
8950 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8952 PR tree-optimization/84634
8953 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8954 masks and masked_loop_p with a single loop_masks, making sure it's
8955 null for bb vectorization.
8957 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8959 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8960 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8961 loop->force_vectorize to check whether there is no alias.
8963 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8966 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8968 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8970 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8971 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8972 instead of a loop around prev_real_insn.
8973 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8977 * config/i386/i386.c (ix86_print_operand): Use conditional
8978 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8981 2018-03-02 Richard Biener <rguenther@suse.de>
8983 PR tree-optimization/84427
8984 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8985 (bitmap_set_subtract_values): Rewrite to handle multiple
8988 (prune_clobbered_mems): Likewise.
8989 (phi_translate): Take edge instead of pred/phiblock.
8990 (phi_translate_1): Likewise.
8991 (phi_translate_set): Likewise. Insert all translated
8992 exprs for a value into the set, keeping possibly multiple
8993 expressions per value.
8994 (compute_antic_aux): Adjust for phi_translate changes.
8995 When intersecting union the expressions and prune those
8996 not in the final value set, keeping possibly multiple
8997 expressions per value. Do not use value-insertion
8998 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8999 all expressions. Add verification that the value-sets
9000 only shrink during iteration.
9001 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9002 (do_pre_regular_insertion): Likewise.
9003 (do_pre_partial_partial_insertion): Likewise.
9005 2018-03-02 Richard Biener <rguenther@suse.de>
9008 * config/darwin.c (saved_debug_info_level): New static global.
9009 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9010 (darwin_asm_lto_end): Restore debug info generation settings.
9012 2018-03-01 Martin Liska <mliska@suse.cz>
9015 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9018 2018-03-01 Richard Biener <rguenther@suse.de>
9021 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9022 type annotation with LTO when debug was disabled at compile-time.
9024 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9026 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9028 (mips_final_postscan_insn): Likewise.
9030 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9032 PR rtl-optimization/84528
9033 * alias.c (init_alias_target): Add commentary.
9034 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9035 a unique base value if the frame pointer is not eliminated
9036 to the stack pointer.
9038 2018-03-01 Tom de Vries <tom@codesourcery.com>
9040 PR rtl-optimization/83327
9041 * lra-int.h (hard_regs_spilled_into): Declare.
9042 * lra.c (hard_regs_spilled_into): Define.
9043 (init_reg_info): Init hard_regs_spilled_into.
9044 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9045 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9046 (process_bb_lives): Handle hard_regs_spilled_into.
9047 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9048 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9050 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9052 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9053 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9054 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9055 * config/rs6000/aix72.h: New file.
9057 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9059 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9060 instead of warning_at with conditional singular and plural messages
9064 * stor-layout.c (update_alignment_for_field): For
9065 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9066 && !DECL_PACKED (field), do the alignment update, just use
9067 only desired_align instead of MAX (type_align, desired_align)
9069 (place_field): Don't do known_align < desired_align handling
9070 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9071 is non-NULL, instead do it after rli->prev_field handling and
9072 only if not within a bitfield word. For DECL_PACKED (field)
9073 use type_align of BITS_PER_UNIT.
9075 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9077 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9078 superfluous parentheses and trailing spaces.
9080 2018-02-28 Richard Biener <rguenther@suse.de>
9082 PR tree-optimization/84584
9083 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9084 SCoPs with fake exit edge.
9086 2018-02-28 Martin Liska <mliska@suse.cz>
9089 * timevar.c (timer::print): Fix format to properly print 100%
9092 2018-02-28 Richard Biener <rguenther@suse.de>
9095 * genmatch.c (capture_info::walk_match): Do not mark
9096 captured expressions without operands as expr_p given
9097 they act more like predicates and should be subject to
9098 "lost tail" side-effect preserving.
9100 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9102 PR rtl-optimization/81611
9103 * auto-inc-dec.c (attempt_change): Move dead note from
9104 mem_insn if it's the next use of regno
9105 (find_address): Take address use of reg holding
9106 non-incremented value. Add parm to limit search to the named
9108 (merge_in_block): Attempt to use a mem insn that is the next
9109 use of the original regno.
9111 2018-02-27 Martin Sebor <msebor@redhat.com>
9114 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9115 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9117 2018-02-27 Martin Sebor <msebor@redhat.com>
9119 PR translation/84207
9120 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9121 n argument to unsigned HOST_WIDE_INT.
9122 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9123 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9124 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9125 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9127 2018-02-27 Richard Biener <rguenther@suse.de>
9129 PR tree-optimization/84512
9130 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9131 Do not use the estimate returned from record_stmt_cost for
9132 the scalar iteration cost but sum properly using add_stmt_cost.
9134 2018-02-27 Richard Biener <rguenther@suse.de>
9136 PR tree-optimization/84466
9137 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9138 Adjust last change to less strictly validate use operands.
9140 2018-02-27 Martin Liska <mliska@suse.cz>
9142 PR gcov-profile/84548
9143 * gcov.c (process_file): Allow partial overlap and consider it
9144 also as group functions.
9145 (output_lines): Properly calculate range of lines for a group.
9147 2018-02-27 Martin Liska <mliska@suse.cz>
9149 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9150 'ggc' suffixes. Change first column width.
9151 (timer::print): Fix formatting of the column.
9153 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9155 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9156 preserve inline entry blocks for the sake of debug inline
9157 entry point markers alone.
9158 (remove_unused_locals): Suggest in comments a better place to
9159 force the preservation of inline entry blocks that are
9160 otherwise unused, but do not preserve them.
9162 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9164 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9166 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9169 * config/i386/constraints.md (Bs): Replace
9170 ix86_indirect_branch_register with
9171 TARGET_INDIRECT_BRANCH_REGISTER.
9173 * config/i386/i386.md (indirect_jump): Likewise.
9174 (tablejump): Likewise.
9175 (*sibcall_memory): Likewise.
9176 (*sibcall_value_memory): Likewise.
9177 Peepholes of indirect call and jump via memory: Likewise.
9178 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9179 (*sibcall_value_GOT_32): Likewise.
9180 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9181 (GOT_memory_operand): Likewise.
9182 (call_insn_operand): Likewise.
9183 (sibcall_insn_operand): Likewise.
9184 (GOT32_symbol_operand): Likewise.
9185 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9187 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9189 PR rtl-optimization/83496
9190 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9191 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9192 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9193 redundant insn, if any.
9194 (relax_delay_slots): Likewise.
9195 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9197 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9199 PR tree-optimization/83965
9200 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9201 that grouped statements are part of a reduction chain. Return
9202 true if the statement is not marked as a reduction itself but
9204 (vect_recog_dot_prod_pattern): Don't check whether the statement
9205 is part of a group here.
9206 (vect_recog_sad_pattern): Likewise.
9207 (vect_recog_widen_sum_pattern): Likewise.
9209 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9212 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9214 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9217 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9219 (ix86_output_indirect_function_return): New prototype.
9220 (ix86_split_simple_return_pop_internal): Likewise.
9221 * config/i386/i386.c (indirect_return_via_cx): New.
9222 (indirect_return_via_cx_bnd): Likewise.
9223 (indirect_thunk_name): Handle return va CX_REG.
9224 (output_indirect_thunk_function): Create alias for
9225 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9226 (ix86_output_indirect_jmp): Remove the bool argument.
9227 (ix86_output_indirect_function_return): New function.
9228 (ix86_split_simple_return_pop_internal): Likewise.
9229 * config/i386/i386.md (*indirect_jump): Don't pass false
9230 to ix86_output_indirect_jmp.
9231 (*tablejump_1): Likewise.
9232 (simple_return_pop_internal): Change it to define_insn_and_split.
9233 Call ix86_split_simple_return_pop_internal to split it for
9235 (simple_return_indirect_internal): Call
9236 ix86_output_indirect_function_return instead of
9237 ix86_output_indirect_jmp.
9239 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9242 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9243 memset and value initialization afterwards.
9245 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9247 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9249 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9252 * common/config/aarch64/aarch64-common.c
9253 (aarch_option_optimization_table[]): Switch
9254 off fomit-frame-pointer
9256 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9257 Chung-Ju Wu <jasonwucj@gmail.com>
9259 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9261 (store_multiple): Ditto.
9263 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9265 * config.gcc: Add --with-cpu support for nds32 target.
9266 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9267 * config/nds32/nds32.opt: Add -mcpu= option.
9269 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9271 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9272 isel=yes): Warn for these deprecated options.
9274 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9276 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9277 ISA_2_5_MASKS_EMBEDDED.
9279 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9281 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9282 p->max as pointers rather than using iterative_hash_expr.
9284 2018-02-23 Carl Love <cel@us.ibm.com>
9286 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9287 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9289 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9290 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9291 P8V_BUILTIN_VEC_VUNSIGNED2.
9293 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9296 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9297 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9299 * lra-constraints.c (curr_insn_transform): Set up
9300 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9302 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9303 * lra-eliminations.c (spill_pseudos): Ditto.
9304 (process_insn_for_elimination): Ditto.
9305 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9306 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9307 LRA_NON_CLOBBERED_ALT.
9309 2018-02-22 Martin Sebor <msebor@redhat.com>
9311 PR tree-optimization/84480
9312 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9313 to maybe_diag_stxncpy_trunc. Call it.
9314 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9315 from gimple_fold_builtin_strcpy. Print inlining stack.
9316 (handle_builtin_stxncpy): Print inlining stack.
9317 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9319 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9322 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9323 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9324 and -fcheck-pointer-bounds are used together.
9325 (indirect_thunk_prefix): New enum.
9326 (indirect_thunk_need_prefix): New function.
9327 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9328 "_nt" instead of "_bnd" for NOTRACK prefix.
9329 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9330 (output_indirect_thunk_function): Likewise.
9332 (ix86_code_end): Update output_indirect_thunk_function calls.
9333 (ix86_output_indirect_branch_via_reg): Replace
9334 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9335 (ix86_output_indirect_branch_via_push): Likewise.
9336 (ix86_output_function_return): Likewise.
9337 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9338 incompatible with -fcf-protection=branch and
9339 -fcheck-pointer-bounds.
9341 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9344 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9345 Change gcc_assert call to output_operand_lossage.
9347 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9349 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9351 2018-02-22 DJ Delorie <dj@redhat.com>
9352 Sebastian Perta <sebastian.perta@renesas.com>
9353 Oleg Endo <olegendo@gcc.gnu.org>
9355 * config/rx/rx.c (rx_rtx_costs): New function.
9356 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9358 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9360 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9362 2018-02-22 Martin Liska <mliska@suse.cz>
9365 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9366 Add "native" as a possible value.
9368 2018-02-22 Martin Liska <mliska@suse.cz>
9371 * config/i386/i386.c (ix86_option_override_internal):
9372 Add "native" as a possible value for -march and -mtune.
9374 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9377 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9378 to all type variants.
9380 PR tree-optimization/84503
9381 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9382 width as info->bitpos + info->bitsize - start.
9383 (merged_store_group::merge_overlapping): Simplify width computation.
9384 (check_no_overlap): New function.
9385 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9386 start + width and last_order of the group, fail if check_no_overlap
9388 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9389 to group if check_no_overlap fails.
9391 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9393 * config/rs6000/altivec.md: Delete contraint arguments to
9394 define_expand, define_split, and define_peephole2, and in
9395 define_insn_and_split if always unused.
9396 * config/rs6000/darwin.md: Ditto.
9397 * config/rs6000/dfp.md: Ditto.
9398 * config/rs6000/rs6000.md: Ditto.
9399 * config/rs6000/sync.md: Ditto.
9400 * config/rs6000/vector.md: Ditto.
9401 * config/rs6000/vsx.md: Ditto.
9403 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9405 * config/rs6000/altivec.md: Write output control strings as braced
9406 blocks instead of double-quoted strings.
9407 * config/rs6000/darwin.md: Ditto.
9408 * config/rs6000/rs6000.md: Ditto.
9409 * config/rs6000/vector.md: Ditto.
9410 * config/rs6000/vsx.md: Ditto.
9412 2018-02-21 Jason Merrill <jason@redhat.com>
9414 PR c++/84314 - ICE with templates and fastcall attribute.
9415 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9417 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9419 * ipa-cp.c (determine_versionability): Fix comment typos.
9421 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9424 * ipa-cp.c (determine_versionability): Do not version functions caling
9427 2018-02-21 Martin Liska <mliska@suse.cz>
9430 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9431 Add "native" as a possible value.
9432 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9433 the macro when native cpu detection is available.
9435 2018-02-21 Martin Liska <mliska@suse.cz>
9438 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9439 Add "native" as a possible value.
9440 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9441 when native cpu detection is available.
9443 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9444 Martin Sebor <msebor@redhat.com>
9446 PR tree-optimization/84478
9447 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9449 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9450 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9451 support which is conservatively correct, for 2 only stay conservative
9452 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9453 argument to the 2 argument get_range_strlen, adjust 6 arg
9454 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9456 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9457 (gimple_fold_builtin_strlen): Pass true as last argument to
9460 2018-02-20 Martin Sebor <msebor@redhat.com>
9463 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9464 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9465 (builtin_memref::builtin_memref): Factor out parts into
9466 set_base_and_offset and call it.
9468 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9471 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9472 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9473 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9474 search at the associated MODE_INT.
9476 2018-02-20 Jeff Law <law@redhat.com>
9479 PR tree-optimization/81592
9481 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9482 for range data rather than using global data.
9483 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9484 range data rather than using global data.
9485 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9486 pass it to children as needed.
9487 (struct directive::fmtresult): Similarly.
9488 (struct directive::set_width): Similarly.
9489 (struct directive::set_precision): Similarly.
9490 (format_integer, format_directive, parse_directive): Similarly.
9491 (format_none): Accept unnamed vr_values parameter.
9492 (format_percent, format_floating, format_character): Similarly.
9493 (format_string, format_plain): Similarly.
9494 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9495 the EVRP range analyzer for range data rather than using global data.
9496 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9497 gimple-ssa-evrp-analyze.h
9498 (class sprintf_dom_walker): Add after_dom_children member function.
9499 Add evrp_range_analyzer member.
9500 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9501 range analyzer as needed.
9502 (sprintf_dom_walker::after_dom_children): New member function.
9503 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9505 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9506 (evrp_range_analyzer::pop_to_marker): Likewise.
9508 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9510 PR tree-optimization/84419
9511 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9512 with the required type if its current type is compatible but
9515 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9518 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9519 after vectorization.
9521 2018-02-20 Martin Liska <mliska@suse.cz>
9524 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9525 possible values if we don't have a hint.
9527 2018-02-20 Martin Liska <mliska@suse.cz>
9531 * final.c (shorten_branches): Build align_tab array with one
9533 * opts.c (finish_options): Add alignment option limit check.
9534 (MAX_CODE_ALIGN): Likewise.
9535 (MAX_CODE_ALIGN_VALUE): Likewise.
9536 * doc/invoke.texi: Document maximum allowed option value for
9537 all -falign-* options.
9539 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9542 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9543 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9544 * var-tracking.c (emit_note_insn_var_location): Remove all references
9545 to NOTE_INSN_CALL_ARG_LOCATION.
9546 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9547 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9548 Use copy_rtx_if_shared.
9549 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9550 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9551 (dwarf2out_var_location): Remove handling of
9552 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9554 * final.c (final_scan_insn): Remove all references to
9555 NOTE_INSN_CALL_ARG_LOCATION.
9556 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9557 before dumping final insns.
9558 * except.c (emit_note_eh_region_end): Remove all references to
9559 NOTE_INSN_CALL_ARG_LOCATION.
9560 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9561 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9562 * config/arc/arc.c (hwloop_optimize): Likewise.
9563 * config/arm/arm.c (create_fix_barrier): Likewise.
9564 * config/s390/s390.c (s390_chunkify_start): Likewise.
9565 * config/sh/sh.c (find_barrier): Likewise.
9566 * config/i386/i386.c (rest_of_insert_endbranch,
9567 ix86_seh_fixup_eh_fallthru): Likewise.
9568 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9569 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9570 * config/frv/frv.c (frv_function_prologue): Likewise.
9571 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9573 (note_outside_basic_block_p): Remove all references to
9574 NOTE_INSN_CALL_ARG_LOCATION.
9575 * gengtype.c (adjust_field_rtx_def): Likewise.
9576 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9578 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9579 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9582 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9585 PR tree-optimization/84452
9586 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9587 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9590 2018-02-19 Martin Liska <mliska@suse.cz>
9593 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9595 2018-02-19 Martin Liska <mliska@suse.cz>
9596 Richard Sandiford <richard.sandiford@linaro.org>
9598 PR tree-optimization/82491
9599 * gimple-fold.c (get_base_constructor): Make earlier bail out
9602 2018-02-19 Carl Love <cel@us.ibm.com>
9604 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9605 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9607 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9608 P8V_BUILTIN_VEC_NEG.
9610 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9612 * config/rl78/rl78.md (movdf): New define expand.
9614 2018-02-19 Martin Liska <mliska@suse.cz>
9617 * doc/invoke.texi: Fix typo.
9618 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9620 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9622 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9623 handle rs6000_single_float and rs6000_double_float specially for
9626 2018-02-16 Jeff Law <law@redhat.com>
9628 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9631 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9634 * ipa-inline.c (inline_small_functions): Fix a typo.
9636 2018-02-16 Nathan Sidwell <nathan@acm.org>
9638 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9640 2018-02-16 Carl Love <cel@us.ibm.com>
9642 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9643 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9644 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9645 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9646 expansion to P8V_BUILTIN_VEC_FLOAT2.
9648 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9650 PR rtl-optimization/70023
9651 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9652 src_regno into account.
9654 2018-02-16 Carl Love <cel@us.ibm.com>
9656 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9657 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9658 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9659 * config/rs6000/rs6000.c: Remove case statements for
9660 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9661 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9662 and P9V_BUILTIN_VEC_VINSERT4B.
9663 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9664 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9665 * config/rs6000/vsx.md:
9666 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9669 2018-02-16 Carl Love <cel@us.ibm.com>
9671 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9673 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9675 * config/rs6000/rs6000-c.c: Add the definitions for
9676 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9677 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9678 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9679 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9680 definition for insert4b and define insn *insert3b_internal.
9681 * doc/extend.texi: Add documentation for vec_extract4b.
9683 2018-02-16 Nathan Sidwell <nathan@acm.org>
9685 * doc/extend.texi (Backwards Compatibility): Mention friend
9686 injection. Note for-scope is deprecated.
9687 * doc/invoke.texi (-ffriend-injection): Deprecate.
9689 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9691 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9692 that moved to I2, also allow destinations that are a paradoxical
9693 subreg (instead of a normal reg).
9695 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9698 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9701 2018-02-16 Richard Biener <rguenther@suse.de>
9703 PR tree-optimization/84037
9704 PR tree-optimization/84016
9706 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9707 Adjust vec_construct for the fact we need additional higher latency
9708 128bit inserts for AVX256 and AVX512 vector builds.
9709 (ix86_add_stmt_cost): Scale vector construction cost for
9712 2018-02-16 Richard Biener <rguenther@suse.de>
9714 PR tree-optimization/84417
9715 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9716 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9717 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9719 2018-02-16 Martin Liska <mliska@suse.cz>
9722 * internal-fn.def (ASAN_CHECK): Set proper flags.
9723 (ASAN_MARK): Likewise.
9725 2018-02-16 Julia Koval <julia.koval@intel.com>
9727 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9728 from PTA_CANNONLAKE.
9730 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9733 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9734 Use ++iter rather than iter++ for std::list iterators.
9735 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9736 defer deleting them until all nodes in the forest are processed. Do
9737 free even leaf nodes. Change to_process into auto_vec.
9740 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9741 * vec.h (vec_default_construct): Use memset instead of placement new
9742 if BROKEN_VALUE_INITIALIZATION is defined.
9743 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9744 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9747 PR rtl-optimization/83723
9748 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9749 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9750 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9752 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9754 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9756 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9758 PR rtl-optimization/81443
9759 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9760 from inner REGs to paradoxical SUBREGs.
9762 2018-02-16 Richard Biener <rguenther@suse.de>
9764 PR tree-optimization/84399
9765 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9766 For operands we can analyze at their definition make sure we can
9767 analyze them at each use as well.
9769 2018-02-16 Richard Biener <rguenther@suse.de>
9771 PR tree-optimization/84190
9772 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9773 volatile accesses if the decl isn't volatile.
9775 2018-02-15 Jason Merrill <jason@redhat.com>
9777 PR c++/84314 - ICE with templates and fastcall attribute.
9778 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9779 TYPE_CANONICAL on an existing type.
9781 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9783 PR tree-optimization/84383
9784 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9785 dstoff nor call operand_equal_p if dstbase is NULL.
9787 PR tree-optimization/84334
9788 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9789 also a CONSTANT_CLASS_P, punt.
9791 2018-02-14 Jim Wilson <jimw@sifive.com>
9793 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9794 first SMALL_OPERAND check. New local min_second_step. Move assert
9795 to where locals are set. Add TARGET_RVC support.
9796 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9798 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9800 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9802 2018-02-14 Martin Sebor <msebor@redhat.com>
9804 PR tree-optimization/83698
9805 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9806 arrays constrain the offset range to their bounds.
9807 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9808 (builtin_access::overlap): Avoid setting the size of overlap if it's
9810 (maybe_diag_overlap): Also consider arrays when deciding what values
9811 of offsets to include in diagnostics.
9813 2018-02-14 Martin Sebor <msebor@redhat.com>
9816 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9817 that correspond to the kind of a declaration.
9819 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9822 * config/pa/pa.md: Load address of PIC label using the linkage table
9823 if the label is nonlocal.
9825 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9827 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9828 warning message if user requests -maltivec=be.
9829 * doc/invoke.texi: Document deprecation of -maltivec=be.
9831 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9834 * config/rs6000/rs6000-c.c: Update definitions for
9835 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9836 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9838 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9841 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9842 add _get_ssp intrinsics. Remove argument from
9843 __builtin_ia32_rdssp[d|q].
9844 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9845 * config/i386/i386-builtin.def: Remove argument from
9846 __builtin_ia32_rdssp[d|q].
9847 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9848 ix86_expand_special_args_builtin for _rdssp[d|q].
9849 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9850 Clear register before usage.
9851 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9852 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9854 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9856 PR tree-optimization/84357
9857 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9858 operand 1 of an ARRAY_REF too.
9860 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9863 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9864 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9866 (set_of_reg): New struct.
9867 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9868 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9869 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9871 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9872 Split into bitclr, bitset, bitinvert patterns if appropriate.
9873 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9874 use rx_fuse_in_memory_bitop.
9875 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9876 to named insn, correct maximum insn length.
9878 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9881 * machmode.def: Define a complex mode for PARTIAL_INT.
9882 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9884 * doc/rtl.texi: Document CSPImode.
9885 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9887 (msp430_hard_regno_nregs_with_padding): Likewise.
9889 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9892 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9894 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9896 PR rtl-optimization/84169
9897 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9898 we generated a parallel as new i3 and we split that to new i2 and i3
9899 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9900 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9901 those to i2, not i1. Partially rewrite this scan code.
9903 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9906 * stor-layout.c (place_field): For variable length fields, adjust
9907 offset_align afterwards not just based on the field's alignment,
9908 but also on the size.
9911 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9912 of exps and logs in the use_exp2 case.
9914 2018-02-13 Jeff Law <law@redhat.com>
9916 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9919 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9922 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9926 * common.opt (gas-loc-support, gas-locview-support): New.
9927 (ginline-points, ginternal-reset-location-views): New.
9928 * doc/invoke.texi: Document them. Use @itemx where intended.
9929 (gvariable-location-views): Adjust.
9930 * target.def (reset_location_view): New.
9931 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9932 (TARGET_RESET_LOCATION_VIEW): New.
9933 * doc/tm.texi: Rebuilt.
9934 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9935 (dwarf2out_default_as_locview_support): New.
9936 (output_asm_line_debug_info): Use option variables.
9937 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9938 (output_loc_list): Likewise.
9939 (add_high_low_attributes): Check option variables.
9940 Don't output entry view attribute in strict mode.
9941 (gen_inlined_subroutine_die): Check option variables.
9942 (dwarf2out_inline_entry): Likewise.
9943 (init_sections_and_labels): Likewise.
9944 (dwarf2out_early_finish): Likewise.
9945 (maybe_reset_location_view): New, from...
9946 (dwarf2out_var_location): ... here. Call it.
9947 * debug.h (dwarf2out_default_as_loc_support): Declare.
9948 (dwarf2out_default_as_locview_support): Declare.
9949 * hooks.c (hook_int_rtx_insn_0): New.
9950 * hooks.h (hook_int_rtx_insn_0): Declare.
9951 * toplev.c (process_options): Take -gas-loc-support and
9952 -gas-locview-support from dwarf2out. Enable
9953 -gvariable-location-views by default only with locview
9954 assembler support. Enable -ginternal-reset-location-views by
9955 default only if the target defines the corresponding hook.
9956 Enable -ginline-points by default if location views are
9957 enabled; force it disabled if statement frontiers are
9959 * tree-inline.c (expand_call_inline): Check option variables.
9960 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9962 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9964 PR tree-optimization/84321
9965 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9966 handling. Also check whether the anti-range contains any values
9967 that satisfy the mask; switch to a VR_RANGE if not.
9969 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9972 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9974 2018-02-13 Martin Jambor <mjambor@suse.cz>
9977 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9978 of call statements, also set location of a load to a temporary.
9980 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9982 * config/rl78/rl78.c (add_vector_labels): New function.
9983 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9984 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9985 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9986 which checks that no arguments are passed.
9987 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9988 * doc/extend.texi: Documentation for the new attribute.
9990 2018-02-13 Andreas Schwab <schwab@suse.de>
9992 * config/riscv/linux.h (CPP_SPEC): Define.
9994 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9997 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9998 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9999 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10000 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10001 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10002 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10003 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10004 temporarily for AES and PCLMUL builtins.
10006 PR tree-optimization/84339
10007 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10008 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10011 PR middle-end/84309
10012 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10013 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10014 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10016 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10018 * omp-simd-clone.h: New file.
10019 * omp-simd-clone.c: Include omp-simd-clone.h.
10020 (expand_simd_clones): No longer static.
10021 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10022 cgraph.h and omp-simd-clone.h.
10023 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10024 (vect_recog_widen_shift_pattern): Formatting fix.
10025 (vect_pattern_recog_1): Don't check optab for calls.
10028 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10029 operands[2] into a REG before using gen_lowpart on it.
10031 2018-02-12 Jeff Law <law@redhat.com>
10034 * config/sh/sh.c (find_barrier): Consider a sibling call
10037 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10038 successfully back substituting a reg.
10040 2018-02-12 Richard Biener <rguenther@suse.de>
10042 PR tree-optimization/84037
10043 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10044 parameter, move visited init to caller.
10045 (vect_slp_analyze_operations): Separate cost from validity
10046 check, initialize visited once for all instances.
10047 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10049 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10051 (vect_model_promotion_demotion_cost): Likewise.
10052 (vectorizable_bswap): Guard cost modeling with !slp_node
10053 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10055 (vectorizable_call): Likewise.
10056 (vectorizable_conversion): Likewise.
10057 (vectorizable_assignment): Likewise.
10058 (vectorizable_shift): Likewise.
10059 (vectorizable_operation): Likewise.
10060 (vectorizable_store): Likewise.
10061 (vectorizable_load): Likewise.
10062 (vectorizable_condition): Likewise.
10063 (vectorizable_comparison): Likewise.
10065 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10068 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10069 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10070 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10072 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10074 PR middle-end/83665
10075 * params.def (inline-min-speedup): Increase from 8 to 15.
10076 (max-inline-insns-auto): Decrease from 40 to 30.
10077 * ipa-split.c (consider_split): Add some buffer for function to
10078 be considered inlining candidate.
10079 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10082 2018-02-12 Richard Biener <rguenther@suse.de>
10084 PR tree-optimization/84037
10085 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10086 matched stmts if we cannot swap the non-matched ones.
10088 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10090 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10091 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10092 _mm_maskz_scalef_round_ss): New intrinsics.
10093 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10094 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10095 __builtin_ia32_scalefss_round): Remove.
10096 (__builtin_ia32_scalefsd_mask_round,
10097 __builtin_ia32_scalefss_mask_round): New intrinsics.
10098 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10099 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10100 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10101 "<round_constraint>")): Changed to ...
10102 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10103 "<round_scalar_constraint>")): ... this.
10104 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10105 %0, %1, %2<round_op3>}"): Changed to ...
10106 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10107 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10108 %2<round_scalar_mask_op3>}"): ... this.
10109 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10111 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10113 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10114 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10115 (_mm_maskz_sqrt_round_ss): New intrinsics.
10116 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10117 (__builtin_ia32_sqrtsd_mask_round)
10118 (__builtin_ia32_sqrtss_mask_round): New builtins.
10119 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10120 (__builtin_ia32_sqrtss_round): Remove.
10121 (__builtin_ia32_sqrtsd_mask_round)
10122 (__builtin_ia32_sqrtss_mask_round): New builtins.
10123 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10124 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10125 ((match_operand:VF_128 1 "vector_operand"
10126 "xBm,<round_constraint>")): Changed to ...
10127 ((match_operand:VF_128 1 "vector_operand"
10128 "xBm,<round_scalar_constraint>")): ... this.
10129 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10130 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10131 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10132 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10133 %<iptr>1<round_scalar_mask_op3>}): ... this.
10134 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10135 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10137 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10140 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10141 Cast vec_cmpeq result to correct type.
10142 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10143 Cast vec_cmpgt result to correct type.
10145 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10147 * final.c (final_scan_insn_1): Renamed from...
10148 (final_scan_insn): ... this. New wrapper, to recover
10149 seen from the outermost call in recursive ones.
10150 * config/sparc/sparc.c (output_return): Drop seen from call.
10151 (output_sibcall): Likewise.
10152 * config/visium/visium.c (output_branch): Likewise.
10154 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10156 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10159 2018-02-10 Alan Modra <amodra@gmail.com>
10162 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10163 Specify LR as an input.
10165 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10168 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10169 remove_member_access_dummy_vars): New functions.
10170 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10171 lower_omp_1, execute_lower_omp): Use them.
10173 PR rtl-optimization/84308
10174 * shrink-wrap.c (spread_components): Release todo vector.
10176 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10178 PR rtl-optimization/57193
10179 * ira-color.c (struct allocno_color_data): Add member
10180 conflict_allocno_hard_prefs.
10181 (update_conflict_allocno_hard_prefs): New.
10182 (bucket_allocno_compare_func): Add a preference based on
10183 conflict_allocno_hard_prefs.
10184 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10185 (color_allocnos): Remove a dead code. Initiate
10186 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10188 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10191 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10192 constraint from =wa to wa. Avoid a subreg on the output operand,
10193 instead use a pseudo and subreg it in a move.
10194 (p9_xxbrd_<mode>): Changed to ...
10195 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10196 (p9_xxbrd_v2df): New expander.
10197 (p9_xxbrw_<mode>): Changed to ...
10198 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10199 (p9_xxbrw_v4sf): New expander.
10201 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10203 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10205 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10208 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10209 multiply in 32-bit mode.
10210 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10211 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10214 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10216 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10217 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10218 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10219 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10221 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10224 * dwarf2out.c (is_trivial_indirect_ref): New function.
10225 (dwarf2out_late_global_decl): Do not generate a location
10226 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10227 and that are not defined in the current unit.
10229 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10231 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10232 instead of a libcall for UNORDERED.
10234 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10237 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10238 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10240 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10243 * config/s390/s390.c (s390_set_current_function): Invoke
10244 s390_indirect_branch_settings also if fndecl didn't change.
10246 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10248 * config/rs6000/rs6000.md (blockage): Set length to zero.
10250 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10252 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10254 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10257 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10258 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10262 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10263 PARALLEL incoming that failed vt_get_decl_and_offset check.
10265 PR middle-end/84237
10266 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10267 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10269 (get_variable_section): For decls in named .bss* sections pass true as
10270 second argument to bss_initializer_p.
10272 2018-02-09 Marek Polacek <polacek@redhat.com>
10273 Jakub Jelinek <jakub@redhat.com>
10276 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10277 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10278 Sync some changes from cxx_fold_indirect_ref.
10280 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10282 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10284 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10285 (BLOCK_INLINE_ENTRY_LABEL): New.
10286 (dwarf2out_var_location): Disregard inline entry markers.
10287 (inline_entry_data): New struct.
10288 (inline_entry_data_hasher): New hashtable type.
10289 (inline_entry_data_hasher::hash): New.
10290 (inline_entry_data_hasher::equal): New.
10291 (inline_entry_data_table): New variable.
10292 (add_high_low_attributes): Add DW_AT_entry_pc and
10293 DW_AT_GNU_entry_view attributes if a pending entry is found
10294 in inline_entry_data_table. Add old entry_pc attribute only
10295 if debug nonbinding markers are disabled.
10296 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10297 markers are enabled.
10298 (block_within_block_p, dwarf2out_inline_entry): New.
10299 (dwarf2out_finish): Check that no entries remained in
10300 inline_entry_data_table.
10301 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10302 (final_scan_insn, notice_source_line): Likewise.
10303 (rest_of_clean_state): Skip inline entry markers.
10304 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10306 * gimple.c (gimple_build_debug_inline_entry): New.
10307 * gimple.h (enum gimple_debug_subcode): Add
10308 GIMPLE_DEBUG_INLINE_ENTRY.
10309 (gimple_build_debug_inline_entry): Declare.
10310 (gimple_debug_inline_entry_p): New.
10311 (gimple_debug_nonbind_marker_p): Adjust.
10312 * insn-notes.def (INLINE_ENTRY): New.
10313 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10314 inline entry marker notes.
10315 (print_insn): Likewise.
10316 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10317 (INSN_DEBUG_MARKER_KIND): Likewise.
10318 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10319 * tree-inline.c (expand_call_inline): Build and insert
10320 debug_inline_entry stmt.
10321 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10322 inline entry blocks early, if nonbind markers are enabled.
10323 (dump_scope_block): Dump fragment info.
10324 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10325 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10326 (gimple_build_debug_inline_entry): New.
10327 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10328 Enable/disable inline entry points too.
10329 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10330 (DEBUG_INSN): Describe inline entry markers.
10332 * common.opt (gvariable-location-views): New.
10333 (gvariable-location-views=incompat5): New.
10334 * config.in: Rebuilt.
10335 * configure: Rebuilt.
10336 * configure.ac: Test assembler for view support.
10337 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10338 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10339 * dwarf2out.c (var_loc_view): New typedef.
10340 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10341 (dwarf2out_locviews_in_attribute): New.
10342 (dwarf2out_locviews_in_loclist): New.
10343 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10344 (enum dw_line_info_opcode): Add LI_adv_address.
10345 (struct dw_line_info_table): Add view.
10346 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10347 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10348 (zero_view_p): New variable.
10349 (ZERO_VIEW_P): New macro.
10350 (output_asm_line_debug_info): New.
10351 (struct var_loc_node): Add view.
10352 (add_AT_view_list, AT_loc_list): New.
10353 (add_var_loc_to_decl): Add view param. Test it against last.
10354 (new_loc_list): Add view params. Record them.
10355 (AT_loc_list_ptr): Handle loc and view lists.
10356 (view_list_to_loc_list_val_node): New.
10357 (print_dw_val): Handle dw_val_class_view_list.
10358 (size_of_die): Likewise.
10359 (value_format): Likewise.
10360 (loc_list_has_views): New.
10361 (gen_llsym): Set vl_symbol too.
10362 (maybe_gen_llsym, skip_loc_list_entry): New.
10363 (dwarf2out_maybe_output_loclist_view_pair): New.
10364 (output_loc_list): Output view list or entries too.
10365 (output_view_list_offset): New.
10366 (output_die): Handle dw_val_class_view_list.
10367 (output_dwarf_version): New.
10368 (output_compilation_unit_header): Use it.
10369 (output_skeleton_debug_sections): Likewise.
10370 (output_rnglists, output_line_info): Likewise.
10371 (output_pubnames, output_aranges): Update version comments.
10372 (output_one_line_info_table): Output view numbers in asm comments.
10373 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10374 Call maybe_gen_llsym.
10375 (loc_list_from_tree_1): Adjust.
10376 (add_AT_location_description): Create view list attribute if
10377 needed, check it's absent otherwise.
10378 (convert_cfa_to_fb_loc_list): Adjust.
10379 (maybe_emit_file): Call output_asm_line_debug_info for test.
10380 (dwarf2out_var_location): Reset views as needed. Precompute
10381 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10382 attribute. Set view.
10383 (new_line_info_table): Reset next view.
10384 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10385 (dwarf2out_source_line): Likewise. Output view resets and labels to
10386 the assembler, or select appropriate line info opcodes.
10387 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10388 (optimize_string_length): Catch it. Adjust.
10389 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10390 dw_val_class_view_list, and remove it if no longer needed.
10391 (hash_loc_list): Hash view numbers.
10392 (loc_list_hasher::equal): Compare them.
10393 (optimize_location_lists): Check whether a view list symbol is
10394 needed, and whether the locview attribute is present, and
10395 whether they match. Remove the locview attribute if no longer
10397 (index_location_lists): Call skip_loc_list_entry for test.
10398 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10399 Use output_dwarf_version.
10400 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10401 (struct dw_val_node): Add val_view_list.
10402 * final.c (SEEN_NEXT_VIEW): New.
10403 (set_next_view_needed): New.
10404 (clear_next_view_needed): New.
10405 (maybe_output_next_view): New.
10406 (final_start_function): Rename to...
10407 (final_start_function_1): ... this. Take pointer to FIRST,
10408 add SEEN parameter. Emit param bindings in the initial view.
10409 (final_start_function): Reintroduce SEEN-less interface.
10410 (final): Rename to...
10411 (final_1): ... this. Take SEEN parameter. Output final pending
10412 next view at the end.
10413 (final): Reintroduce seen-less interface.
10414 (final_scan_insn): Output pending next view before switching
10415 sections or ending a block. Mark the next view as needed when
10416 outputting variable locations. Notify debug backend of section
10417 changes, and of location view changes.
10418 (rest_of_handle_final): Adjust.
10419 * toplev.c (process_options): Autodetect value for debug variable
10420 location views option. Warn on incompat5 without -gdwarf-5.
10421 * doc/invoke.texi (gvariable-location-views): New.
10422 (gvariable-location-views=incompat5): New.
10423 (gno-variable-location-views): New.
10425 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10427 PR tree-optimization/84136
10428 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10429 that the result of find_edge is non-NULL.
10431 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10434 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10435 storing integer register in SImode. Fix cost of 256 and 512
10436 byte aligned SSE register store.
10438 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10440 * config/i386/i386.c (ix86_multiplication_cost): Fix
10441 multiplication cost for TARGET_AVX512DQ.
10443 2018-02-08 Marek Polacek <polacek@redhat.com>
10445 PR tree-optimization/84238
10446 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10449 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10451 PR tree-optimization/84265
10452 * tree-vect-stmts.c (vectorizable_store): Don't treat
10453 VMAT_CONTIGUOUS accesses as grouped.
10454 (vectorizable_load): Likewise.
10456 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10458 PR tree-optimization/81635
10459 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10460 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10461 (test_round_for_mask): New functions.
10462 (wide_int_cc_tests): Call test_round_for_mask.
10463 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10464 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10465 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10466 range returned by get_range_info.
10468 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10471 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10472 * symtab.c: Include builtins.h
10473 (symtab_node::output_to_lto_symbol_table_p): Move here
10474 from lto-streamer-out.c:output_symbol_p.
10475 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10476 (output_symbol_p): Move all logic to symtab.c
10477 (produce_symtab): Update.
10479 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10481 * config/s390/s390-opts.h (enum indirect_branch): Define.
10482 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10483 (s390_indirect_branch_via_thunk)
10484 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10485 (enum s390_indirect_branch_type): Define.
10486 * config/s390/s390.c (struct s390_frame_layout, struct
10487 machine_function): Remove.
10488 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10489 (indirect_branch_table_label_no, indirect_branch_table_name):
10491 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10492 (enum s390_indirect_branch_option): Define.
10493 (s390_return_addr_from_memory): New function.
10494 (s390_handle_string_attribute): New function.
10495 (s390_attribute_table): Add new attribute handler.
10496 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10497 (s390_indirect_branch_via_thunk): New function.
10498 (s390_indirect_branch_via_inline_thunk): New function.
10499 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10500 sibling call optimization for non z10 compiles.
10501 (s390_emit_call): Force indirect branch target to be a single
10502 register. Add r1 clobber for non-z10 compiles.
10503 (s390_emit_epilogue): Emit return jump via return_use expander.
10504 (s390_reorg): Handle JUMP_INSNs as execute targets.
10505 (s390_option_override_internal): Perform validity checks for the
10506 new command line options.
10507 (s390_indirect_branch_attrvalue): New function.
10508 (s390_indirect_branch_settings): New function.
10509 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10510 (s390_output_indirect_thunk_function): New function.
10511 (s390_code_end): Implement target hook.
10512 (s390_case_values_threshold): Implement target hook.
10513 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10515 * config/s390/s390.h (struct s390_frame_layout)
10516 (struct machine_function): Move here from s390.c.
10517 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10518 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10519 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10520 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10521 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10522 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10523 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10524 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10525 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10526 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10527 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10528 (mnemonic attribute): Add values which aren't recognized
10530 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10531 pattern for branch conversion. Fix mnemonic attribute.
10532 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10533 indirect branch via thunk if requested.
10534 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10535 ("*indirect_jump"): Disable for branch conversion using out of
10537 ("indirect_jump_via_thunk<mode>_z10")
10538 ("indirect_jump_via_thunk<mode>")
10539 ("indirect_jump_via_inlinethunk<mode>_z10")
10540 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10541 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10542 ("casesi_jump_via_inlinethunk<mode>_z10")
10543 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10544 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10545 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10546 ("*indirect2_jump"): Disable for branch conversion.
10547 ("casesi_jump"): Turn into expander and expand patterns for branch
10549 ("return_use"): New expander.
10550 ("*return"): Emit return via thunk and rename it to ...
10551 ("*return<mode>"): ... this one.
10552 * config/s390/s390.opt: Add new options and and enum for the
10555 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10557 * lra-constraints.c (match_reload): Unconditionally use
10558 gen_lowpart_SUBREG, rather than selecting between that
10559 and equivalent gen_rtx_SUBREG code.
10561 2018-02-08 Richard Biener <rguenther@suse.de>
10563 PR tree-optimization/84233
10564 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10565 changed flag instead of boguously re-using phi_inserted.
10567 2018-02-08 Martin Jambor <mjambor@suse.cz>
10569 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10570 static local variables.
10572 2018-02-08 Richard Biener <rguenther@suse.de>
10574 PR tree-optimization/84278
10575 * tree-vect-stmts.c (vectorizable_store): When looking for
10576 smaller vector types to perform grouped strided loads/stores
10577 make sure the mode is supported by the target.
10578 (vectorizable_load): Likewise.
10580 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10582 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10583 Increase LDP/STP opportunities by adding adjacent callee-saves.
10585 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10587 PR rtl-optimization/84068
10588 PR rtl-optimization/83459
10589 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10591 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10593 PR tree-optimization/84224
10594 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10595 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10596 non-zero arguments.
10598 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10601 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10602 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10603 count, remove one USE.
10605 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10607 * doc/install.texi (Configuration): Document the
10608 --with-long-double-format={ibm,ieee} PowerPC configuration
10612 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10613 Convert from define_expand to be define_insn_and_split. Rework
10614 float/double/_Float128 conversions to QI/HI/SImode to work with
10615 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10616 conversions to QI/HImode types did a store and then a load to
10617 truncate the value. For conversions to VSX registers, don't split
10618 the insn, instead emit the code directly. Use the code iterator
10619 any_fix to combine signed and unsigned conversions.
10620 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10621 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10622 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10623 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10624 (fix_<mode>di2_hw): Likewise.
10625 (fixuns_<mode>di2_hw): Likewise.
10626 (fix_<mode>si2_hw): Likewise.
10627 (fixuns_<mode>si2_hw): Likewise.
10628 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10629 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10630 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10631 fix<uns>_trunc<SFDF:mode>si2_p8.
10632 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10633 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10634 (fix<uns>_<mode>_mem): Likewise.
10635 (fctiw<u>z_<mode>_mem): Likewise.
10636 (fix<uns>_<mode>_mem): Likewise.
10637 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10638 the register allocator from doing a direct move to the GPRs to do
10639 a store, and instead use the ISA 3.0 store byte/half-word from
10640 vector register instruction. For IEEE 128-bit floating point,
10641 also optimize stores of 32-bit ints.
10642 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10644 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10646 * genextract.c (push_pathstr_operand): New function to support
10648 (walk_rtx): Call push_pathstr_operand.
10649 (print_path): Support [a-zA-Z].
10651 2018-02-07 Richard Biener <rguenther@suse.de>
10653 PR tree-optimization/84037
10654 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10655 (cse_and_gimplify_to_preheader): Declare.
10656 (vect_get_place_in_interleaving_chain): Likewise.
10657 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10659 (_loop_vec_info::~_loop_vec_info): Delete it.
10660 (cse_and_gimplify_to_preheader): New function.
10661 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10662 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10663 (vectorizable_load): Likewise. For grouped stores always base
10664 the IV on the first element.
10665 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10666 condition before gimplifying.
10668 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10670 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10671 *DIV_EXPR and *MOD_EXPR.
10673 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10676 * config/i386/i386.c (ix86_option_override_internal): Mask out
10677 the CF_SET bit when checking -fcf-protection.
10679 2018-02-07 Tom de Vries <tom@codesourcery.com>
10682 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10685 2018-02-07 Richard Biener <rguenther@suse.de>
10687 PR tree-optimization/84204
10688 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10691 PR tree-optimization/84205
10692 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10693 special-case isl_ast_op_zdiv_r.
10695 PR tree-optimization/84223
10696 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10697 Only add conditions from within the region.
10698 (gather_bbs::after_dom_children): Adjust.
10700 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10703 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10704 * config/avr/avr.md: Only post-reload split REG-REG moves if
10705 either register is GENERAL_REG_P.
10707 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10709 PR tree-optimization/84235
10710 * tree-ssa-scopedtables.c
10711 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10712 if the subtraction is performed in floating point type where NaNs are
10713 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10714 build 1. Formatting fix.
10716 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10719 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10720 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10721 and skip it regardless of bb boundaries. Use CALL_P macro,
10722 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10723 unnecessarily, formatting fix.
10725 2018-02-06 Michael Collison <michael.collison@arm.com>
10727 * config/arm/thumb2.md:
10728 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10729 (*thumb_mov_notscc): Ditto.
10731 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10734 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10735 unsigned_fix, not "s".
10737 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10739 * configure.ac (gcc_fn_eh_frame_ro): New function.
10740 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10741 correct .eh_frame permissions.
10742 * configure: Regenerate.
10744 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10746 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10747 irrelevant options.
10749 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10751 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10752 Display warning message for -mno-speculate-indirect-jumps.
10754 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10756 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10758 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10760 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10762 PR tree-optimization/84225
10763 * tree-eh.c (find_trapping_overflow): Only call
10764 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10766 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10769 * config/i386/i386.c: Reimplement the check of possible options
10770 -mibt/-mshstk conbination. Change error messages.
10771 * doc/invoke.texi: Fix a typo: remove extra '='.
10773 2018-02-06 Marek Polacek <polacek@redhat.com>
10775 PR tree-optimization/84228
10776 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10778 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10781 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10782 emitted arch directives.
10783 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10784 __ARM_FEATURE_COPROC before changing architectures.
10786 2018-02-06 Richard Biener <rguenther@suse.de>
10788 * config/i386/i386.c (print_reg): Fix typo.
10789 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10791 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10793 * configure: Regenerate.
10795 2018-02-05 Martin Sebor <msebor@redhat.com>
10797 PR tree-optimization/83369
10798 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10801 2018-02-05 Martin Liska <mliska@suse.cz>
10803 * doc/invoke.texi: Cherry-pick upstream r323995.
10805 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10807 * ira.c (ira_init_register_move_cost): Adjust comment.
10809 2018-02-05 Martin Liska <mliska@suse.cz>
10811 PR gcov-profile/84137
10812 * doc/gcov.texi: Fix typo in documentation.
10814 2018-02-05 Martin Liska <mliska@suse.cz>
10816 PR gcov-profile/83879
10817 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10818 using dlopen functionality.
10820 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10822 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10823 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10824 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10825 _mm_maskz_range_round_ss): New intrinsics.
10826 (__builtin_ia32_rangesd128_round)
10827 (__builtin_ia32_rangess128_round): Remove.
10828 (__builtin_ia32_rangesd128_mask_round,
10829 __builtin_ia32_rangess128_mask_round): New builtins.
10830 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10831 __builtin_ia32_rangess128_round): Remove.
10832 (__builtin_ia32_rangesd128_mask_round,
10833 __builtin_ia32_rangess128_mask_round): New builtins.
10834 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10835 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10836 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10837 "<round_saeonly_constraint>")): Changed to ...
10838 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10839 "<round_saeonly_scalar_constraint>")): ... this.
10840 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10841 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10842 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10843 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10844 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10846 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10848 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10850 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10851 Remove all values except native, 8540 and 8548.
10853 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10855 * config/i386/i386.c (ix86_output_function_return): Pass
10856 INVALID_REGNUM, instead of -1, as invalid register number to
10857 indirect_thunk_name and output_indirect_thunk.
10859 2018-02-02 Julia Koval <julia.koval@intel.com>
10861 * config.gcc: Add -march=icelake.
10862 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10863 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10864 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10865 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10866 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10867 (processor_target_table): Add icelake.
10868 (ix86_option_override_internal): Handle new PTAs.
10869 (get_builtin_code_for_version): Handle icelake.
10870 (M_INTEL_COREI7_ICELAKE): New.
10871 (fold_builtin_cpu): Handle icelake.
10872 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10873 * doc/invoke.texi: Add -march=icelake.
10875 2018-02-02 Julia Koval <julia.koval@intel.com>
10877 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10878 to wide_int_bitmask.
10879 * wide-int-bitmask.h: New.
10881 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10884 * config/i386/i386.md: Replace Pmode with word_mode in
10885 builtin_setjmp_setup and builtin_longjmp to support x32.
10887 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10891 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10893 (rs6000_supported_cpu_names): New static variable.
10894 (linux_cpu_translation_table): Likewise.
10895 (elf_platform) <cpu>: Define new static variable and use it.
10896 Translate kernel AT_PLATFORM name to canonical name if needed.
10897 Error if platform name is unknown.
10899 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10902 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10904 2018-02-01 Jeff Law <law@redhat.com>
10907 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10908 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10909 the scratch if RELEASE_VIA_POP is false.
10910 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10911 If we have to save a temporary register, decrement SIZE appropriately.
10912 Pass new arguments to release_scratch_register_on_entry.
10913 (ix86_adjust_stack_and_probe): Likewise.
10914 (ix86_emit_probe_stack_range): Pass new arguments to
10915 release_scratch_register_on_entry.
10917 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10919 PR rtl-optimization/84157
10920 * combine.c (change_zero_ext): Use REG_P predicate in
10921 front of HARD_REGISTER_P predicate.
10923 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10925 * config/avr/avr.c (avr_option_override): Move disabling of
10926 -fdelete-null-pointer-checks to...
10927 * common/config/avr/avr-common.c (avr_option_optimization_table):
10930 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10932 PR tree-optimization/81635
10933 * tree-data-ref.c (split_constant_offset_1): For types that
10934 wrap on overflow, try to use range info to prove that wrapping
10937 2018-02-01 Renlin Li <renlin.li@arm.com>
10940 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10941 TAILCALL_ADDR_REGS.
10942 (aarch64_register_move_cost): Likewise.
10943 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10944 TAILCALL_ADDR_REGS.
10945 (REG_CLASS_NAMES): Likewise.
10946 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10947 TAILCALL_ADDR_REGS. Remove IP registers.
10948 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10950 2018-02-01 Richard Biener <rguenther@suse.de>
10952 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10953 for specifying RPO order and allow NULL for that.
10954 * domwalk.c (dom_walker::dom_walker): Likewise.
10955 (dom_walker::walk): Handle NULL RPO order.
10956 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10958 (rewrite_update_dom_walker): Likewise.
10959 (mark_def_dom_walker): Likewise.
10961 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10963 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10964 (aarch64_maybe_expand_sve_subreg_move): Declare.
10965 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10966 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10968 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10969 that are semantically a reverse operation.
10970 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10971 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10972 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10974 (aarch64_can_change_mode_class): For big-endian, forbid changes
10975 between two SVE modes if they have different element sizes.
10977 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10979 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10980 the TImode handling for big-endian targets.
10982 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10984 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10985 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10987 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10988 Remove BSWAP handing for big-endian targets and use the form of
10989 LD1RQ appropariate for the mode.
10991 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10993 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10994 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10995 duplicated element.
10997 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11000 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11001 check for operands that need to go through aarch64_sve_reload_be.
11003 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11005 PR tree-optimization/81661
11006 PR tree-optimization/84117
11007 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11008 * tree-eh.c: Include gimplify.h.
11009 (find_trapping_overflow, replace_trapping_overflow,
11010 rewrite_to_non_trapping_overflow): New functions.
11011 * tree-vect-loop.c: Include tree-eh.h.
11012 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11013 * tree-data-ref.c: Include tree-eh.h.
11014 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11016 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11018 PR rtl-optimization/84123
11019 * combine.c (change_zero_ext): Check if hard register satisfies
11020 can_change_dest_mode before calling gen_lowpart_SUBREG.
11022 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11025 * ira.c (ira_init_register_move_cost): Remove assert.
11027 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11029 PR rtl-optimization/84071
11030 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11031 * doc/tm.texi: Regenerate.
11033 2018-01-31 Richard Biener <rguenther@suse.de>
11035 PR tree-optimization/84132
11036 * tree-data-ref.c (analyze_miv_subscript): Properly
11037 check whether evolution_function_is_affine_multivariate_p
11038 before calling gcd_of_steps_may_divide_p.
11040 2018-01-31 Julia Koval <julia.koval@intel.com>
11043 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11044 * config/i386/i386.md (rdpid_rex64) New.
11045 (rdpid): Make 32bit only.
11047 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11050 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11051 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11053 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11056 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11058 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11060 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11062 PR rtl-optimization/84071
11063 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11064 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11066 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11068 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11069 (arc_attribute_table): Add 'aux' attribute.
11070 (arc_in_small_data_p): Consider aux like variables.
11071 (arc_is_aux_reg_p): New function.
11072 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11073 (arc_get_aux_arg): New function.
11074 (prepare_move_operands): Handle aux-register access.
11075 (arc_handle_aux_attribute): New function.
11076 * doc/extend.texi (ARC Variable attributes): Add subsection.
11078 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11080 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11081 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11082 (arc_attribute_table): Add 'uncached' attribute.
11083 (arc_print_operand): Print '.di' flag for uncached memory
11085 (arc_in_small_data_p): Do not consider for small data the uncached
11087 (arc_is_uncached_mem_p): New function.
11088 * config/arc/predicates.md (compact_store_memory_operand): Check
11089 for uncached memory accesses.
11090 (nonvol_nonimm_operand): Likewise.
11091 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11093 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11096 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11097 falign-loops=): Add Optimization flag.
11099 2018-01-30 Jeff Law <law@redhat.com>
11102 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11103 INT_REGISTERS_SAVED. Check it prior to calling
11104 get_scratch_register_on_entry.
11105 (ix86_adjust_stack_and_probe): Similarly.
11106 (ix86_emit_probe_stack_range): Similarly.
11107 (ix86_expand_prologue): Corresponding changes.
11109 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11112 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11113 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11115 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11118 * lra-constraints.c (curr_insn_transform): Process AND in the
11121 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11123 PR rtl-optimization/83986
11124 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11125 dependence against last_pending_memory_flush in addition to
11126 pending_jump_insns.
11128 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11130 PR tree-optimization/81611
11131 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11134 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11137 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11140 2018-01-30 Richard Biener <rguenther@suse.de>
11141 Jakub Jelinek <jakub@redhat.com>
11143 PR tree-optimization/84111
11144 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11145 inner loops added during recursion, as they don't have up-to-date
11148 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11151 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11152 (can_inline_edge_by_limits_p): ... here.
11153 (can_early_inline_edge_p, check_callers,
11154 update_caller_keys, update_callee_keys, recursive_inlining,
11155 add_new_edges_to_heap, speculation_useful_p,
11156 inline_small_functions,
11157 inline_small_functions, flatten_function,
11158 inline_to_all_callers_1): Update.
11160 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11162 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11165 2018-01-30 Richard Biener <rguenther@suse.de>
11167 PR tree-optimization/83008
11168 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11169 invariant and constant vector uses in stmts when they need
11170 more than one stmt.
11172 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11175 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11176 * configure: Regenerate.
11178 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11180 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11182 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11183 Use gen_rtx_REG rather than gen_lowpart.
11185 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11187 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11188 rather than 0 when creating partial subregs.
11190 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11192 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11195 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11198 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11199 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11200 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11201 flags. This restores the settings used before the 2017-07-24.
11202 Turning off pre increment/decrement/modify allows IVOPTS to
11203 optimize DF/SF loops where the index is an int.
11205 2018-01-29 Richard Biener <rguenther@suse.de>
11206 Kelvin Nilsen <kelvin@gcc.gnu.org>
11209 * tree-vect-stmts.c (vectorizable_call): Don't call
11210 targetm.vectorize_builtin_md_vectorized_function if callee is
11213 2018-01-22 Carl Love <cel@us.ibm.com>
11215 * doc/extend.tex: Fix typo in second arg in
11216 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11218 2018-01-29 Richard Biener <rguenther@suse.de>
11220 PR tree-optimization/84086
11221 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11222 (flush_ssaname_freelist): When SSA names were released reset
11223 the SCEV hash table.
11225 2018-01-29 Richard Biener <rguenther@suse.de>
11227 PR tree-optimization/84057
11228 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11229 removed paths when removing edges.
11231 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11233 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11234 -mfunction-return=@var{choice}.
11236 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11238 PR diagnostic/84034
11239 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11240 Handle CR like TAB.
11241 (layout::print_source_line): Likewise.
11242 (test_get_line_width_without_trailing_whitespace): Add test cases.
11244 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11246 PR middle-end/84040
11247 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11250 2018-01-26 Jim Wilson <jimw@sifive.com>
11252 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11254 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11257 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11259 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11260 and CMP + SUB-immediate -> SUBS.
11262 2018-01-26 Martin Sebor <msebor@redhat.com>
11264 PR tree-optimization/83896
11265 * tree-ssa-strlen.c (get_string_len): Rename...
11266 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11267 Avoid assuming length is constant.
11268 (handle_char_store): Use HOST_WIDE_INT for string length.
11270 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11273 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11274 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11276 2018-01-26 Richard Biener <rguenther@suse.de>
11278 PR rtl-optimization/84003
11279 * dse.c (record_store): Only record redundant stores when
11280 the earlier store aliases at least all accesses the later one does.
11282 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11284 PR rtl-optimization/83985
11285 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11286 REG_CFA_RESTORE insns.
11287 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11288 REG_CFA_RESTORE insns here.
11291 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11292 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11294 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11296 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11297 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11298 (arc_init): Likewise.
11299 (arc_override_options): Likewise.
11300 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11302 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11304 * config/arc/arc.h (TARGET_DBNZ): Define.
11305 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11306 properly set the tune attribute.
11307 (dbnz): Use TARGET_DBNZ guard.
11308 * config/arc/arc.opt (mtune): Add core3 option.
11310 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11312 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11313 recognize new pic like addresses.
11314 (arc_delegitimize_address): Clean up.
11316 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11318 * config/arc/arc-arches.def: Option mrf16 valid for all
11320 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11321 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11322 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11323 * config/arc/arc-tables.opt: Regenerate.
11324 * config/arc/arc.c (arc_conditional_register_usage): Handle
11325 reduced register file case.
11326 (arc_file_start): Set must have build attributes.
11327 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11328 mrf16 option value.
11329 * config/arc/arc.opt (mrf16): Add new option.
11330 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11331 * config/arc/genmultilib.awk: Handle new mrf16 option.
11332 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11333 * config/arc/t-multilib: Regenerate.
11334 * doc/invoke.texi (ARC Options): Document mrf16 option.
11336 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11338 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11339 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11340 (arc_attribute_table): Add 'secure_call' attribute.
11341 (arc_print_operand): Print secure call operand.
11342 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11344 (arc_is_secure_call_p): New function. * config/arc/arc.md
11345 (call_i): Add support for sjli instruction.
11346 (call_value_i): Likewise.
11347 * config/arc/constraints.md (Csc): New constraint.
11349 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11350 John Eric Martin <John.Martin@emmicro-us.com>
11352 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11353 * config/arc/arc.c (_arc_jli_section): New struct.
11354 (arc_jli_section): New type.
11355 (rc_jli_sections): New static variable.
11356 (arc_handle_jli_attribute): New function.
11357 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11358 (arc_file_end): New function.
11359 (TARGET_ASM_FILE_END): Define.
11360 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11361 (arc_add_jli_section): New function.
11362 (jli_call_scan): Likewise.
11363 (arc_reorg): Call jli_call_scan.
11364 (arc_output_addsi): Remove 'S' from printing asm operand.
11365 (arc_is_jli_call_p): New function.
11366 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11368 (movhi_insn): Likewise.
11369 (movsi_insn): Likewise.
11370 (movsi_set_cc_insn): Likewise.
11371 (loadqi_update): Likewise.
11372 (load_zeroextendqisi_update): Likewise.
11373 (load_signextendqisi_update): Likewise.
11374 (loadhi_update): Likewise.
11375 (load_zeroextendhisi_update): Likewise.
11376 (load_signextendhisi_update): Likewise.
11377 (loadsi_update): Likewise.
11378 (loadsf_update): Likewise.
11379 (movsicc_insn): Likewise.
11380 (bset_insn): Likewise.
11381 (bxor_insn): Likewise.
11382 (bclr_insn): Likewise.
11383 (bmsk_insn): Likewise.
11384 (bicsi3_insn): Likewise.
11385 (cmpsi_cc_c_insn): Likewise.
11386 (movsi_ne): Likewise.
11387 (movsi_cond_exec): Likewise.
11388 (clrsbsi2): Likewise.
11389 (norm_f): Likewise.
11399 (call_i): Remove 'S' asm letter, add jli instruction.
11400 (call_value_i): Likewise.
11401 * config/arc/arc.op (mjli-always): New option.
11402 * config/arc/constraints.md (Cji): New constraint.
11403 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11405 (subsf3_fpx): Likewise.
11406 (mulsf3_fpx): Likewise.
11407 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11409 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11410 function attrbutes.
11411 * doc/invoke.texi (ARC): Document mjli-always option.
11413 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11415 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11416 avoid addition with 0 and use incw and decw where possible.
11418 2018-01-26 Richard Biener <rguenther@suse.de>
11420 PR tree-optimization/81082
11421 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11422 association if it requires casting to unsigned.
11423 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11424 from fold_plusminus_mult_expr to catch important cases late when
11425 range info is available.
11427 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11429 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11430 * configure.ac (hidden_linkonce): New test.
11431 * configure: Regenerate.
11432 * config.in: Regenerate.
11434 2018-01-26 Julia Koval <julia.koval@intel.com>
11436 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11437 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11438 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11439 _mm_mask_bitshuffle_epi64_mask): Fix type.
11440 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11441 USI_FTYPE_V4DI_V4DI_USI): Remove.
11442 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11443 __builtin_ia32_vpshufbitqmb256_mask,
11444 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11445 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11446 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11448 2018-01-26 Alan Modra <amodra@gmail.com>
11451 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11452 UNSPEC_VBPERMQ. Sort other unspecs.
11454 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11456 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11458 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11460 PR middle-end/83055
11461 * predict.c (drop_profile): Do not push/pop cfun; update also
11463 (handle_missing_profiles): Fix logic looking for zero profiles.
11465 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11467 PR middle-end/83977
11468 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11469 on functions with #pragma omp declare simd or functions with simd
11471 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11472 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11473 Remove trailing \n from warning_at calls.
11475 2018-01-25 Tom de Vries <tom@codesourcery.com>
11478 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11479 for neutered workers.
11481 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11484 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11485 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11487 2018-01-24 Jeff Law <law@redhat.com>
11490 * i386.c (get_probe_interval): Move to earlier point.
11491 (ix86_compute_frame_layout): If -fstack-clash-protection and
11492 the frame is larger than the probe interval, then use pushes
11493 to save registers rather than reg->mem moves.
11494 (ix86_expand_prologue): Remove conditional for int_registers_saved
11497 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11500 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11501 min/max for never referenced object.
11503 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11505 PR middle-end/83977
11506 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11508 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11509 attributes from DECL_ATTRIBUTES (decl) without affecting
11510 DECL_ATTRIBUTES (current_function_decl).
11511 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11512 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11514 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11516 PR tree-optimization/83979
11517 * fold-const.c (fold_comparison): Use constant_boolean_node
11518 instead of boolean_{true,false}_node.
11520 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11522 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11525 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11527 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11528 Simplify the clause that sets the length attribute.
11529 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11530 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11531 clause that sets the length attribute.
11532 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11534 2018-01-24 Tom de Vries <tom@codesourcery.com>
11537 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11538 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11539 Add strict parameter.
11540 (prevent_branch_around_nothing): Insert dummy insn between branch to
11541 label and label with no ptx insn inbetween.
11542 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11544 2018-01-24 Tom de Vries <tom@codesourcery.com>
11547 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11548 for neutered threads in warp.
11549 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11551 2018-01-24 Richard Biener <rguenther@suse.de>
11553 PR tree-optimization/83176
11554 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11557 2018-01-24 Richard Biener <rguenther@suse.de>
11559 PR tree-optimization/82819
11560 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11561 code generating pluses that are no-ops in the target precision.
11563 2018-01-24 Richard Biener <rguenther@suse.de>
11565 PR middle-end/84000
11566 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11568 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11570 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11571 to merge probabilities.
11572 * predict.c (probably_never_executed): Also mark as cold functions
11573 with global 0 profile and guessed local profile.
11574 * profile-count.c (profile_probability::combine_with_count): New
11576 * profile-count.h (profile_probability::operator*,
11577 profile_probability::operator*=, profile_probability::operator/,
11578 profile_probability::operator/=): Reduce precision to adjusted
11579 and set value to guessed on contradictory divisions.
11580 (profile_probability::combine_with_freq): Remove.
11581 (profile_probability::combine_wiht_count): Declare.
11582 (profile_count::force_nonzero):: Set to adjusted.
11583 (profile_count::probability_in):: Set quality to adjusted.
11584 * tree-ssa-tail-merge.c (replace_block_by): Use
11585 combine_with_count.
11587 2018-01-23 Andrew Waterman <andrew@sifive.com>
11588 Jim Wilson <jimw@sifive.com>
11590 * config/riscv/riscv.c (riscv_stack_boundary): New.
11591 (riscv_option_override): Set riscv_stack_boundary. Handle
11592 riscv_preferred_stack_boundary_arg.
11593 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11594 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11595 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11596 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11597 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11598 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11600 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11603 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11604 of struct ix86_frame.
11605 (ix86_expand_epilogue): Likewise. Add a local variable for
11606 the reg_save_offset field in struct ix86_frame.
11608 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11610 PR tree-optimization/82604
11611 * tree-loop-distribution.c (enum partition_kind): New enum item
11612 PKIND_PARTIAL_MEMSET.
11613 (partition_builtin_p): Support above new enum item.
11614 (generate_code_for_partition): Ditto.
11615 (compute_access_range): Differentiate cases that equality can be
11616 proven at all loops, the innermost loops or no loops.
11617 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11618 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11619 (finalize_partitions, distribute_loop): Don't fuse partition of
11620 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11621 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11622 parloop is enabled.
11624 2018-01-23 Martin Liska <mliska@suse.cz>
11626 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11627 order to ignore the predictor.
11628 (PRED_POLYMORPHIC_CALL): Likewise.
11629 (PRED_RECURSIVE_CALL): Likewise.
11631 2018-01-23 Martin Liska <mliska@suse.cz>
11633 * tree-profile.c (tree_profiling): Print function header to
11634 aware reader which function we are working on.
11635 * value-prof.c (gimple_find_values_to_profile): Do not print
11636 not interesting value histograms.
11638 2018-01-23 Martin Liska <mliska@suse.cz>
11640 * profile-count.h (enum profile_quality): Add
11641 profile_uninitialized as the first value. Do not number values
11642 as they are zero based.
11643 (profile_count::verify): Update sanity check.
11644 (profile_probability::verify): Likewise.
11646 2018-01-23 Nathan Sidwell <nathan@acm.org>
11648 * doc/invoke.texi (ffor-scope): Deprecate.
11650 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11652 PR tree-optimization/83510
11653 * domwalk.c (set_all_edges_as_executable): New function.
11654 (dom_walker::dom_walker): Convert bool param
11655 "skip_unreachable_blocks" to enum reachability. Move setup of
11656 edge flags to set_all_edges_as_executable and only do it when
11657 reachability is REACHABLE_BLOCKS.
11658 * domwalk.h (enum dom_walker::reachability): New enum.
11659 (dom_walker::dom_walker): Convert bool param
11660 "skip_unreachable_blocks" to enum reachability.
11661 (set_all_edges_as_executable): New decl.
11662 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11663 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11665 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11666 but converting true to REACHABLE_BLOCKS.
11667 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11669 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11670 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11671 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11673 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11674 if check_all_array_refs will be called.
11676 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11678 * tree.c (selftest::test_location_wrappers): Add more test
11681 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11683 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11684 (selftest::test_bit_in_range): Likewise.
11686 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11689 * doc/sourcebuild.texi (vect_float): Say that the selector
11690 only describes the situation when -funsafe-math-optimizations is on.
11691 (vect_float_strict): Document.
11693 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11695 PR tree-optimization/83965
11696 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11697 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11698 instead of checking only for a reduction.
11699 (vect_recog_widen_sum_pattern): Likewise.
11701 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11703 * predict.c (probably_never_executed): Only use precise profile info.
11704 (compute_function_frequency): Skip after inlining hack since we now
11705 have quality checking.
11707 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11709 * profile-count.h (profile_probability::very_unlikely,
11710 profile_probability::unlikely, profile_probability::even): Set
11711 precision to guessed.
11713 2018-01-23 Richard Biener <rguenther@suse.de>
11715 PR tree-optimization/83963
11716 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11717 Properly terminate dominator walk when crossing the exit edge not
11718 when visiting its source block.
11720 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11723 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11724 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11726 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11728 PR tree-optimization/83957
11729 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11730 semicolon after for body surrounded by braces.
11732 PR tree-optimization/83081
11733 * profile-count.h (profile_probability::split): New method.
11734 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11735 Use profile_probability::split.
11736 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11737 when splitting a single conditional jump into 2.
11739 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11741 PR tree-optimization/69452
11742 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11745 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11747 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11748 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11749 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11751 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11753 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11755 * config/rl78/rl78.md (movdi): New define_expand.
11756 * config/rl78/rl78.c (rl78_split_movdi): New function.
11758 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11761 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11763 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11764 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11765 128-bit to produce an UNSPEC move to get the double word with the
11766 signbit and then a shift directly to do signbit.
11767 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11768 implementation with a new version that just does either a direct
11769 move or a regular move. Move memory interface to separate insns.
11770 Move insns so they are next to the expander.
11771 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11772 with signbit move. Split big and little endian case.
11773 (signbit<mode>2_dm_mem_le): Likewise.
11774 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11775 (signbit<mode>2_dm2): Likewise.
11777 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11779 * config/rl78/rl78.md (anddi3): New define_expand.
11781 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11783 * config/rl78/rl78.md (umindi3): New define_expand.
11785 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11787 * config/rl78/rl78.md (smindi3): New define_expand.
11789 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11791 * config/rl78/rl78.md (smaxdi3): New define_expand.
11793 2018-01-22 Carl Love <cel@us.ibm.com>
11795 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11796 LVX_V1TI): Add macro expansion.
11797 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11798 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11799 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11800 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11801 Change check to determine if the instruction is a byte reversing
11802 entry. Fix typo in comment.
11803 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11804 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11805 Add def_builtin calls for new builtins.
11806 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11807 Add define_insn expansion.
11809 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11811 * config/rl78/rl78.md (umaxdi3): New define_expand.
11813 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11815 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11816 for non-QImode registers.
11818 2018-01-22 Richard Biener <rguenther@suse.de>
11820 PR tree-optimization/83963
11821 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11822 including the loop exit block.
11823 (scop_detection::merge_sese): Likewise.
11824 (scop_detection::add_scop): Do it here instead.
11826 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11828 * doc/sourcebuild.texi (arm_softfloat): Document.
11830 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11833 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11834 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11835 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11837 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11838 David Edelsohn <dje.gcc@gmail.com>
11841 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11842 Change "crset eq" to "crset 2".
11843 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11844 (*call_indirect_aix<mode>_nospec): Likewise.
11845 (*call_value_indirect_aix<mode>_nospec): Likewise.
11846 (*call_indirect_elfv2<mode>_nospec): Likewise.
11847 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11848 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11849 change assembly output from . to $.
11850 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11851 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11852 (*tablejump<mode>_internal1_nospec): Likewise.
11854 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11857 * config/sh/sh_optimize_sett_clrt.cc:
11858 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11860 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11862 PR tree-optimization/83940
11863 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11864 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11865 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11866 use it to pass back the definition type.
11867 (vect_check_store_rhs): Likewise rhs_dt_out.
11868 (vect_build_gather_load_calls): Add a mask_dt argument and use
11869 it instead of a call to vect_is_simple_use.
11870 (vectorizable_store): Update calls to vect_check_load_store_mask
11871 and vect_check_store_rhs. Use the dt returned by the latter instead
11872 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11873 instead of calls to vect_is_simple_use. Pass the scalar rather
11874 than the vector operand to vect_is_simple_use when handling
11875 second and subsequent copies of an rhs value.
11876 (vectorizable_load): Update calls to vect_check_load_store_mask
11877 and vect_build_gather_load_calls. Use the cached mask_dt and
11878 gs_info.offset_dt instead of calls to vect_is_simple_use.
11880 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11882 PR middle-end/83945
11883 * tree-emutls.c: Include gimplify.h.
11884 (lower_emutls_2): New function.
11885 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11886 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11887 it before further processing.
11890 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11891 UINTVAL (trueop1) instead of INTVAL (op1).
11893 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11897 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11898 INCOMING_FRAME_SP_OFFSET if not defined.
11899 (scan_trace): Add ENTRY argument. If true and
11900 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11901 emit a note to adjust the CFA offset.
11902 (create_cfi_notes): Adjust scan_trace callers.
11903 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11904 INCOMING_FRAME_SP_OFFSET in the CIE.
11905 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11906 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11908 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11909 * doc/tm.texi: Regenerated.
11911 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11913 PR rtl-optimization/83147
11914 * lra-constraints.c (remove_inheritance_pseudos): Use
11915 lra_substitute_pseudo_within_insn.
11917 2018-01-19 Tom de Vries <tom@codesourcery.com>
11918 Cesar Philippidis <cesar@codesourcery.com>
11921 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11923 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11926 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11927 spaces for function labels.
11929 2018-01-19 Martin Liska <mliska@suse.cz>
11931 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11932 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11933 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11934 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11935 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11936 (PRED_CONST_RETURN): Change from 69 to 65.
11937 (PRED_NULL_RETURN): Change from 91 to 71.
11938 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11939 (PRED_LOOP_GUARD): Change from 66 to 73.
11941 2018-01-19 Martin Liska <mliska@suse.cz>
11943 * predict.c (predict_insn_def): Add new assert.
11944 (struct branch_predictor): Change type to signed integer.
11945 (test_prediction_value_range): Amend test to cover
11946 PROB_UNINITIALIZED.
11947 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11948 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11949 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11950 (PRED_LOOP_IV_COMPARE): Likewise.
11951 * predict.h (PROB_UNINITIALIZED): Define new constant.
11953 2018-01-19 Martin Liska <mliska@suse.cz>
11955 * predict.c (dump_prediction): Add new format for
11956 analyze_brprob.py script which is enabled with -details
11958 * profile-count.h (precise_p): New function.
11960 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11962 PR tree-optimization/83922
11963 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11964 there are no statements that need masking.
11965 (vect_active_double_reduction_p): New function.
11966 (vect_analyze_loop_operations): Use it when handling phis that
11967 are not in the loop header.
11969 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11971 PR tree-optimization/83914
11972 * tree-vect-loop.c (vectorizable_induction): Don't convert
11973 init_expr or apply the peeling adjustment for inductions
11974 that are nested within the vectorized loop.
11976 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11978 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11981 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11985 * function.h (gimplify_parameters): Add gimple_seq * argument.
11986 * function.c: Include gimple.h and options.h.
11987 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11988 for the added local temporaries if needed.
11989 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11990 if there are any parameter cleanups, wrap whole body into a
11991 try/finally with the cleanups.
11993 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11996 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11997 Use GET_MODE_CLASS for scalar floating point.
11999 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12003 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12004 Fix call of call_cgraph_insertion_hooks.
12006 2018-01-18 Martin Sebor <msebor@redhat.com>
12008 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12010 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12013 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12016 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12019 * common.opt: (-ffile-prefix-map): New option.
12020 * opts.c (common_handle_option): Defer it.
12021 * opts-global.c (handle_common_deferred_options): Handle it.
12022 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12023 * file-prefix-map.h: New file.
12024 (remap_debug_filename, add_debug_prefix_map): ...here.
12025 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12026 * final.c (debug_prefix_map, add_debug_prefix_map
12027 remap_debug_filename): Move to...
12028 * file-prefix-map.c: New file.
12029 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12030 generalize, get rid of alloca(), use strrchr() instead of strchr().
12031 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12032 Implement in terms of add_prefix_map().
12033 (remap_macro_filename, remap_debug_filename): Implement in term of
12035 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12036 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12037 * dbxout.c: Include file-prefix-map.h.
12038 * varasm.c: Likewise.
12039 * vmsdbgout.c: Likewise.
12040 * xcoffout.c: Likewise.
12041 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12042 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12043 * doc/invoke.texi (-ffile-prefix-map): Document.
12044 (-fdebug-prefix-map): Update description.
12046 2018-01-18 Martin Liska <mliska@suse.cz>
12048 * config/i386/i386.c (indirect_thunk_name): Document that also
12050 (output_indirect_thunk): Document why both instructions
12051 (pause and lfence) are generated.
12053 2018-01-18 Richard Biener <rguenther@suse.de>
12055 PR tree-optimization/83887
12056 * graphite-scop-detection.c
12057 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12058 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12059 (scop_detection::merge_sese): Re-implement with a flood-fill
12060 algorithm that properly finds a SESE region if it exists.
12062 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12065 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12066 pointer_diff optimizations use view_convert instead of convert.
12068 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12070 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12071 Generate different code for -mno-speculate-indirect-jumps.
12072 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12073 (*call_indirect_aix<mode>): Disable for
12074 -mno-speculate-indirect-jumps.
12075 (*call_indirect_aix<mode>_nospec): New define_insn.
12076 (*call_value_indirect_aix<mode>): Disable for
12077 -mno-speculate-indirect-jumps.
12078 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12079 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12080 -mno-speculate-indirect-jumps.
12081 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12083 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12085 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12086 long double type, set the flags for noting the default long double
12087 type, even if we don't pass or return a long double type.
12089 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12092 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12095 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12097 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12098 support for merge[hl].
12099 (fold_mergehl_helper): New helper function.
12100 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12101 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12102 (altivec_vmrglw_direct): Add xxmrglw insn.
12104 2018-01-17 Andrew Waterman <andrew@sifive.com>
12106 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12107 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12109 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12112 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12113 call the lto_location_cache before reading the
12114 DECL_SOURCE_LOCATION of the types.
12116 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12117 Richard Sandiford <richard.sandiford@linaro.org>
12119 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12120 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12121 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12122 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12123 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12125 * config/aarch64/constraints.md (aarch64_movti_operand):
12127 * config/aarch64/predicates.md (Uti): Add new constraint.
12129 2018-01-17 Carl Love <cel@us.ibm.com>
12131 * config/rs6000/vsx.md (define_expand xl_len_r,
12132 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12133 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12135 (define_expand, define_insn): Move the shift left from the
12136 define_insn to the define_expand for lxvl and stxvl instructions.
12137 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12138 and XL_LEN_R definitions to PURE.
12140 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12142 * config/i386/i386.c (indirect_thunk_name): Declare regno
12143 as unsigned int. Compare regno with INVALID_REGNUM.
12144 (output_indirect_thunk): Ditto.
12145 (output_indirect_thunk_function): Ditto.
12146 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12147 in the call to output_indirect_thunk_function.
12149 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12151 PR middle-end/83884
12152 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12153 rather than the size of inner_type to determine the stack slot size
12154 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12156 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12159 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12162 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12164 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12165 endian Linux systems to optionally enable multilibs for selecting
12166 the long double type if the user configured an explicit type.
12167 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12168 have no long double multilibs if not defined.
12169 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12170 warn if the user used -mabi={ieee,ibm}longdouble and we built
12171 multilibs for long double.
12172 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12173 appropriate multilib option.
12174 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12176 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12177 for building long double multilibs.
12178 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12180 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12182 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12185 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12187 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12190 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12193 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12196 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12198 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12199 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12201 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12203 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12204 different rtl trees depending on TARGET_64BIT.
12205 (rs6000_gen_lvx): Likewise.
12207 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12209 * config/visium/visium.md (nop): Tweak comment.
12210 (hazard_nop): Likewise.
12212 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12214 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12215 -mspeculate-indirect-jumps.
12216 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12217 for -mno-speculate-indirect-jumps.
12218 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12219 (*call_value_indirect_elfv2<mode>): Disable for
12220 -mno-speculate-indirect-jumps.
12221 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12222 (indirect_jump): Emit different RTL for
12223 -mno-speculate-indirect-jumps.
12224 (*indirect_jump<mode>): Disable for
12225 -mno-speculate-indirect-jumps.
12226 (*indirect_jump<mode>_nospec): New define_insn.
12227 (tablejump): Emit different RTL for
12228 -mno-speculate-indirect-jumps.
12229 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12230 (tablejumpsi_nospec): New define_expand.
12231 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12232 (tablejumpdi_nospec): New define_expand.
12233 (*tablejump<mode>_internal1): Disable for
12234 -mno-speculate-indirect-jumps.
12235 (*tablejump<mode>_internal1_nospec): New define_insn.
12236 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12239 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12241 * caller-save.c (insert_save): Drop unnecessary parameter. All
12244 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12245 Richard Biener <rguenth@suse.de>
12248 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12249 return early, inline manually is_gimple_sizepos. Make sure if we
12250 call gimplify_expr we don't end up with a gimple constant.
12251 * tree.c (variably_modified_type_p): Don't return true for
12252 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12253 * gimplify.h (is_gimple_sizepos): Remove.
12255 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12257 PR tree-optimization/83857
12258 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12259 vectorizable_live_operation for pure SLP statements.
12260 (vectorizable_live_operation): Handle PHIs.
12262 2018-01-16 Richard Biener <rguenther@suse.de>
12264 PR tree-optimization/83867
12265 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12266 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12268 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12271 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12272 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12273 If off is not INTEGER_CST, issue a may not be aligned warning
12274 rather than isn't aligned. Use isn%'t rather than isn't.
12275 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12277 <case MULT_EXPR>: Improve the case when bottom and one of the
12278 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12279 operand, in that case check if the other operand is multiple of
12280 bottom divided by the INTEGER_CST operand.
12282 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12285 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12286 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12287 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12288 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12289 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12290 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12291 (pa_function_arg_size): New function.
12293 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12295 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12296 in a separate statement.
12298 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12300 PR tree-optimization/83847
12301 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12302 group gathers and scatters.
12304 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12306 PR rtl-optimization/86620
12307 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12309 PR rtl-optimization/83213
12310 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12311 to last if both are JUMP_INSNs.
12313 PR tree-optimization/83843
12314 * gimple-ssa-store-merging.c
12315 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12316 store_immediate_info for bswap/nop orig_stores.
12318 2018-01-15 Andrew Waterman <andrew@sifive.com>
12320 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12322 <UDIV>: Increase cost if !TARGET_DIV.
12324 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12326 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12327 (define_attr "cr_logical_3op"): New.
12328 (cceq_ior_compare): Adjust.
12329 (cceq_ior_compare_complement): Adjust.
12330 (*cceq_rev_compare): Adjust.
12331 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12332 (is_cracked_insn): Adjust.
12333 (insn_must_be_first_in_group): Adjust.
12334 * config/rs6000/40x.md: Adjust.
12335 * config/rs6000/440.md: Adjust.
12336 * config/rs6000/476.md: Adjust.
12337 * config/rs6000/601.md: Adjust.
12338 * config/rs6000/603.md: Adjust.
12339 * config/rs6000/6xx.md: Adjust.
12340 * config/rs6000/7450.md: Adjust.
12341 * config/rs6000/7xx.md: Adjust.
12342 * config/rs6000/8540.md: Adjust.
12343 * config/rs6000/cell.md: Adjust.
12344 * config/rs6000/e300c2c3.md: Adjust.
12345 * config/rs6000/e500mc.md: Adjust.
12346 * config/rs6000/e500mc64.md: Adjust.
12347 * config/rs6000/e5500.md: Adjust.
12348 * config/rs6000/e6500.md: Adjust.
12349 * config/rs6000/mpc.md: Adjust.
12350 * config/rs6000/power4.md: Adjust.
12351 * config/rs6000/power5.md: Adjust.
12352 * config/rs6000/power6.md: Adjust.
12353 * config/rs6000/power7.md: Adjust.
12354 * config/rs6000/power8.md: Adjust.
12355 * config/rs6000/power9.md: Adjust.
12356 * config/rs6000/rs64.md: Adjust.
12357 * config/rs6000/titan.md: Adjust.
12359 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12361 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12362 ix86_indirect_branch_register logic.
12364 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12366 * config/i386/constraints.md (Bs): Update
12367 ix86_indirect_branch_register check. Don't check
12368 ix86_indirect_branch_register with GOT_memory_operand.
12370 * config/i386/predicates.md (GOT_memory_operand): Don't check
12371 ix86_indirect_branch_register here.
12372 (GOT32_symbol_operand): Likewise.
12374 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12376 * config/i386/predicates.md (constant_call_address_operand):
12377 Rewrite ix86_indirect_branch_register logic.
12378 (sibcall_insn_operand): Likewise.
12380 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12382 * config/i386/constraints.md (Bs): Replace
12383 ix86_indirect_branch_thunk_register with
12384 ix86_indirect_branch_register.
12386 * config/i386/i386.md (indirect_jump): Likewise.
12387 (tablejump): Likewise.
12388 (*sibcall_memory): Likewise.
12389 (*sibcall_value_memory): Likewise.
12390 Peepholes of indirect call and jump via memory: Likewise.
12391 * config/i386/i386.opt: Likewise.
12392 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12393 (GOT_memory_operand): Likewise.
12394 (call_insn_operand): Likewise.
12395 (sibcall_insn_operand): Likewise.
12396 (GOT32_symbol_operand): Likewise.
12398 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12400 PR middle-end/83837
12401 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12402 type rather than type addr's type points to.
12403 (expand_omp_atomic_mutex): Likewise.
12404 (expand_omp_atomic): Likewise.
12406 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12409 * config/i386/i386.c (output_indirect_thunk_function): Use
12410 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12411 for __x86_return_thunk.
12413 2018-01-15 Richard Biener <rguenther@suse.de>
12415 PR middle-end/83850
12416 * expmed.c (extract_bit_field_1): Fix typo.
12418 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12421 * config/arm/iterators.md (VF): New mode iterator.
12422 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12423 Remove integer-related logic from pattern.
12424 (neon_vabd<mode>_3): Likewise.
12426 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12428 PR middle-end/82694
12429 * common.opt (fstrict-overflow): No longer an alias.
12430 (fwrapv-pointer): New option.
12431 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12432 also for pointer types based on flag_wrapv_pointer.
12433 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12434 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12435 opts->x_flag_wrapv got set.
12436 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12437 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12438 POINTER_TYPE_OVERFLOW_UNDEFINED.
12439 * match.pd: Likewise in address comparison pattern.
12440 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12442 2018-01-15 Richard Biener <rguenther@suse.de>
12445 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12446 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12447 Reset type names to their identifier if their TYPE_DECL doesn't
12448 have linkage (and thus is used for ODR and devirt).
12449 (save_debug_info_for_decl): Remove.
12450 (save_debug_info_for_type): Likewise.
12451 (add_tree_to_fld_list): Adjust.
12452 * tree-pretty-print.c (dump_generic_node): Make dumping of
12453 type names more robust.
12455 2018-01-15 Richard Biener <rguenther@suse.de>
12457 * BASE-VER: Bump to 8.0.1.
12459 2018-01-14 Martin Sebor <msebor@redhat.com>
12462 * builtins.c (check_access): Avoid warning when the no-warning bit
12465 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12467 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12468 * ira-color (allocno_hard_regs_compare): Likewise.
12470 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12473 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12474 Use .pushsection/.popsection.
12476 2018-01-14 Martin Sebor <msebor@redhat.com>
12479 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12481 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12483 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12484 entry from extra_headers.
12485 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12486 extra_headers, make the list bitwise identical to the i?86-*-* one.
12488 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12490 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12491 -mcmodel=large with -mindirect-branch=thunk,
12492 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12493 -mfunction-return=thunk-extern.
12494 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12495 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12496 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12498 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12500 * config/i386/i386.c (print_reg): Print the name of the full
12501 integer register without '%'.
12502 (ix86_print_operand): Handle 'V'.
12503 * doc/extend.texi: Document 'V' modifier.
12505 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12507 * config/i386/constraints.md (Bs): Disallow memory operand for
12508 -mindirect-branch-register.
12510 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12511 (GOT_memory_operand): Likewise.
12512 (call_insn_operand): Likewise.
12513 (sibcall_insn_operand): Likewise.
12514 (GOT32_symbol_operand): Likewise.
12515 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12516 for -mindirect-branch-register.
12517 (tablejump): Likewise.
12518 (*sibcall_memory): Likewise.
12519 (*sibcall_value_memory): Likewise.
12520 Disallow peepholes of indirect call and jump via memory for
12521 -mindirect-branch-register.
12522 (*call_pop): Replace m with Bw.
12523 (*call_value_pop): Likewise.
12524 (*sibcall_pop_memory): Replace m with Bs.
12525 * config/i386/i386.opt (mindirect-branch-register): New option.
12526 * doc/invoke.texi: Document -mindirect-branch-register option.
12528 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12530 * config/i386/i386-protos.h (ix86_output_function_return): New.
12531 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12532 set function_return_type.
12533 (indirect_thunk_name): Add ret_p to indicate thunk for function
12535 (output_indirect_thunk_function): Pass false to
12536 indirect_thunk_name.
12537 (ix86_output_indirect_branch_via_reg): Likewise.
12538 (ix86_output_indirect_branch_via_push): Likewise.
12539 (output_indirect_thunk_function): Create alias for function
12540 return thunk if regno < 0.
12541 (ix86_output_function_return): New function.
12542 (ix86_handle_fndecl_attribute): Handle function_return.
12543 (ix86_attribute_table): Add function_return.
12544 * config/i386/i386.h (machine_function): Add
12545 function_return_type.
12546 * config/i386/i386.md (simple_return_internal): Use
12547 ix86_output_function_return.
12548 (simple_return_internal_long): Likewise.
12549 * config/i386/i386.opt (mfunction-return=): New option.
12550 (indirect_branch): Mention -mfunction-return=.
12551 * doc/extend.texi: Document function_return function attribute.
12552 * doc/invoke.texi: Document -mfunction-return= option.
12554 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12556 * config/i386/i386-opts.h (indirect_branch): New.
12557 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12558 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12559 with local indirect jump when converting indirect call and jump.
12560 (ix86_set_indirect_branch_type): New.
12561 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12562 (indirectlabelno): New.
12563 (indirect_thunk_needed): Likewise.
12564 (indirect_thunk_bnd_needed): Likewise.
12565 (indirect_thunks_used): Likewise.
12566 (indirect_thunks_bnd_used): Likewise.
12567 (INDIRECT_LABEL): Likewise.
12568 (indirect_thunk_name): Likewise.
12569 (output_indirect_thunk): Likewise.
12570 (output_indirect_thunk_function): Likewise.
12571 (ix86_output_indirect_branch_via_reg): Likewise.
12572 (ix86_output_indirect_branch_via_push): Likewise.
12573 (ix86_output_indirect_branch): Likewise.
12574 (ix86_output_indirect_jmp): Likewise.
12575 (ix86_code_end): Call output_indirect_thunk_function if needed.
12576 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12578 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12579 (ix86_attribute_table): Add indirect_branch.
12580 * config/i386/i386.h (machine_function): Add indirect_branch_type
12581 and has_local_indirect_jump.
12582 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12584 (tablejump): Likewise.
12585 (*indirect_jump): Use ix86_output_indirect_jmp.
12586 (*tablejump_1): Likewise.
12587 (simple_return_indirect_internal): Likewise.
12588 * config/i386/i386.opt (mindirect-branch=): New option.
12589 (indirect_branch): New.
12592 (thunk-inline): Likewise.
12593 (thunk-extern): Likewise.
12594 * doc/extend.texi: Document indirect_branch function attribute.
12595 * doc/invoke.texi: Document -mindirect-branch= option.
12597 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12600 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12602 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12604 * ipa-inline.c (want_inline_small_function_p): Return false if
12605 inlining has already failed with CIF_FINAL_ERROR.
12606 (update_caller_keys): Call want_inline_small_function_p before
12608 (update_callee_keys): Likewise.
12610 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12612 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12614 (rs6000_quadword_masked_address_p): Likewise.
12615 (quad_aligned_load_p): Likewise.
12616 (quad_aligned_store_p): Likewise.
12617 (const_load_sequence_p): Add comment to describe the outer-most loop.
12618 (mimic_memory_attributes_and_flags): New function.
12619 (rs6000_gen_stvx): Likewise.
12620 (replace_swapped_aligned_store): Likewise.
12621 (rs6000_gen_lvx): Likewise.
12622 (replace_swapped_aligned_load): Likewise.
12623 (replace_swapped_load_constant): Capitalize argument name in
12624 comment describing this function.
12625 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12626 and stores that access quad-word aligned addresses and replace
12627 with stvx or lvx instructions when appropriate.
12628 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12629 New function prototype.
12630 (rs6000_quadword_masked_address_p): Likewise.
12631 (rs6000_gen_lvx): Likewise.
12632 (rs6000_gen_stvx): Likewise.
12633 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12634 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12635 when memory address is aligned.
12636 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12637 this split to select lvx instruction when memory address is aligned.
12638 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12639 instruction when memory address is aligned.
12640 (*vsx_le_perm_load_v16qi): Likewise.
12641 (four unnamed splitters): Modify to select the stvx instruction
12642 when memory is aligned.
12644 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12646 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12647 which appears in the queue multiple times.
12649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12650 Alan Hayward <alan.hayward@arm.com>
12651 David Sherwood <david.sherwood@arm.com>
12653 * tree-vectorizer.h (vec_lower_bound): New structure.
12654 (_loop_vec_info): Add check_nonzero and lower_bounds.
12655 (LOOP_VINFO_CHECK_NONZERO): New macro.
12656 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12657 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12658 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12659 fields. Make seg_len the distance travelled, not including the
12661 (dr_direction_indicator): Declare.
12662 (dr_zero_step_indicator): Likewise.
12663 (dr_known_forward_stride_p): Likewise.
12664 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12666 (runtime_alias_check_p): Allow runtime alias checks with
12668 (operator ==): Compare access_size and align.
12669 (prune_runtime_alias_test_list): Rework for new distinction between
12670 the access_size and seg_len.
12671 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12673 (get_segment_min_max): New function.
12674 (create_intersect_range_checks): Use it.
12675 (dr_step_indicator): New function.
12676 (dr_direction_indicator): Likewise.
12677 (dr_zero_step_indicator): Likewise.
12678 (dr_known_forward_stride_p): Likewise.
12679 * tree-loop-distribution.c (data_ref_segment_size): Return
12680 DR_STEP * (niters - 1).
12681 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12683 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12684 (vect_preserves_scalar_order_p): New function, split out from...
12685 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12686 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12687 (vect_vfa_access_size): New function.
12688 (vect_vfa_align): Likewise.
12689 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12690 (dump_lower_bound): New function.
12691 (vect_check_lower_bound): Likewise.
12692 (vect_small_gap_p): Likewise.
12693 (vectorizable_with_step_bound_p): Likewise.
12694 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12695 depencies if the vectorization factor is 1. Convert the checks
12696 for nonzero steps into checks on the bounds of DR_STEP. Try using
12697 a bunds check for variable steps if the minimum required step is
12698 relatively small. Update calls to the dr_with_seg_len
12699 constructor and to vect_compile_time_alias.
12700 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12702 (vect_loop_versioning): Call it.
12703 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12705 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12708 Alan Hayward <alan.hayward@arm.com>
12709 David Sherwood <david.sherwood@arm.com>
12711 * doc/sourcebuild.texi (vect_scatter_store): Document.
12712 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12714 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12716 * genopinit.c (main): Add supports_vec_scatter_store and
12717 supports_vec_scatter_store_cached to target_optabs.
12718 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12719 IFN_MASK_SCATTER_STORE.
12720 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12722 * internal-fn.h (internal_store_fn_p): Declare.
12723 (internal_fn_stored_value_index): Likewise.
12724 * internal-fn.c (scatter_store_direct): New macro.
12725 (expand_scatter_store_optab_fn): New function.
12726 (direct_scatter_store_optab_supported_p): New macro.
12727 (internal_store_fn_p): New function.
12728 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12729 IFN_MASK_SCATTER_STORE.
12730 (internal_fn_mask_index): Likewise.
12731 (internal_fn_stored_value_index): New function.
12732 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12733 for scatter stores.
12734 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12735 * optabs-query.c (supports_vec_scatter_store_p): New function.
12736 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12737 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12738 true for scatter stores.
12739 (vect_gather_scatter_fn_p): Handle scatter stores too.
12740 (vect_check_gather_scatter): Consider using scatter stores if
12741 supports_vec_scatter_store_p.
12742 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12743 scatter stores too.
12744 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12745 internal_fn_stored_value_index.
12746 (check_load_store_masking): Handle scatter stores too.
12747 (vect_get_store_rhs): Make public.
12748 (vectorizable_call): Use internal_store_fn_p.
12749 (vectorizable_store): Handle scatter store internal functions.
12750 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12751 when deciding whether the end of the group has been reached.
12752 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12753 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12754 (mask_scatter_store<mode>): New insns.
12756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12757 Alan Hayward <alan.hayward@arm.com>
12758 David Sherwood <david.sherwood@arm.com>
12760 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12761 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12762 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12764 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12765 Use vect_truncate_gather_scatter_offset if we can't treat the
12766 operation as a normal gather load or scatter store.
12767 (get_group_load_store_type): Take the gather_scatter_info
12768 as argument. Try using a gather load or scatter store for
12769 single-element groups.
12770 (get_load_store_type): Update calls to get_group_load_store_type
12771 and vect_use_strided_gather_scatters_p.
12773 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12774 Alan Hayward <alan.hayward@arm.com>
12775 David Sherwood <david.sherwood@arm.com>
12777 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12778 optional tree argument.
12779 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12781 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12782 but continue to use the current value as a fallback.
12783 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12784 to compare the updates.
12785 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12786 (get_load_store_type): Use it when handling a strided access.
12787 (vect_get_strided_load_store_ops): New function.
12788 (vect_get_data_ptr_increment): Likewise.
12789 (vectorizable_load): Handle strided gather loads. Always pass
12790 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12793 Alan Hayward <alan.hayward@arm.com>
12794 David Sherwood <david.sherwood@arm.com>
12796 * doc/md.texi (gather_load@var{m}): Document.
12797 (mask_gather_load@var{m}): Likewise.
12798 * genopinit.c (main): Add supports_vec_gather_load and
12799 supports_vec_gather_load_cached to target_optabs.
12800 * optabs-tree.c (init_tree_optimization_optabs): Use
12801 ggc_cleared_alloc to allocate target_optabs.
12802 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12803 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12805 * internal-fn.h (internal_load_fn_p): Declare.
12806 (internal_gather_scatter_fn_p): Likewise.
12807 (internal_fn_mask_index): Likewise.
12808 (internal_gather_scatter_fn_supported_p): Likewise.
12809 * internal-fn.c (gather_load_direct): New macro.
12810 (expand_gather_load_optab_fn): New function.
12811 (direct_gather_load_optab_supported_p): New macro.
12812 (direct_internal_fn_optab): New function.
12813 (internal_load_fn_p): Likewise.
12814 (internal_gather_scatter_fn_p): Likewise.
12815 (internal_fn_mask_index): Likewise.
12816 (internal_gather_scatter_fn_supported_p): Likewise.
12817 * optabs-query.c (supports_at_least_one_mode_p): New function.
12818 (supports_vec_gather_load_p): Likewise.
12819 * optabs-query.h (supports_vec_gather_load_p): Declare.
12820 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12821 and memory_type field.
12822 (NUM_PATTERNS): Bump to 15.
12823 * tree-vect-data-refs.c: Include internal-fn.h.
12824 (vect_gather_scatter_fn_p): New function.
12825 (vect_describe_gather_scatter_call): Likewise.
12826 (vect_check_gather_scatter): Try using internal functions for
12827 gather loads. Recognize existing calls to a gather load function.
12828 (vect_analyze_data_refs): Consider using gather loads if
12829 supports_vec_gather_load_p.
12830 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12831 (vect_get_gather_scatter_offset_type): Likewise.
12832 (vect_convert_mask_for_vectype): Likewise.
12833 (vect_add_conversion_to_patterm): Likewise.
12834 (vect_try_gather_scatter_pattern): Likewise.
12835 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12836 (vect_vect_recog_func_ptrs): Add it.
12837 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12838 internal_fn_mask_index and internal_gather_scatter_fn_p.
12839 (check_load_store_masking): Take the gather_scatter_info as an
12840 argument and handle gather loads.
12841 (vect_get_gather_scatter_ops): New function.
12842 (vectorizable_call): Check internal_load_fn_p.
12843 (vectorizable_load): Likewise. Handle gather load internal
12845 (vectorizable_store): Update call to check_load_store_masking.
12846 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12847 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12848 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12849 (aarch64_gather_scale_operand_d): New predicates.
12850 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12851 (mask_gather_load<mode>): New insns.
12853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12854 Alan Hayward <alan.hayward@arm.com>
12855 David Sherwood <david.sherwood@arm.com>
12857 * optabs.def (fold_left_plus_optab): New optab.
12858 * doc/md.texi (fold_left_plus_@var{m}): Document.
12859 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12860 * internal-fn.c (fold_left_direct): Define.
12861 (expand_fold_left_optab_fn): Likewise.
12862 (direct_fold_left_optab_supported_p): Likewise.
12863 * fold-const-call.c (fold_const_fold_left): New function.
12864 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12865 * tree-parloops.c (valid_reduction_p): New function.
12866 (gather_scalar_reductions): Use it.
12867 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12868 (vect_finish_replace_stmt): Declare.
12869 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12870 (needs_fold_left_reduction_p): New function, split out from...
12871 (vect_is_simple_reduction): ...here. Accept reductions that
12872 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12873 (vect_force_simple_reduction): Also store the reduction type in
12874 the assignment's STMT_VINFO_REDUC_TYPE.
12875 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12876 (merge_with_identity): New function.
12877 (vect_expand_fold_left): Likewise.
12878 (vectorize_fold_left_reduction): Likewise.
12879 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12880 scalar phi in place for it. Check for target support and reject
12881 cases that would reassociate the operation. Defer the transform
12882 phase to vectorize_fold_left_reduction.
12883 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12884 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12885 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12887 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12889 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12890 call to ifc_temp_var.
12892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12893 Alan Hayward <alan.hayward@arm.com>
12894 David Sherwood <david.sherwood@arm.com>
12896 * target.def (legitimize_address_displacement): Take the original
12897 offset as a poly_int.
12898 * targhooks.h (default_legitimize_address_displacement): Update
12900 * targhooks.c (default_legitimize_address_displacement): Likewise.
12901 * doc/tm.texi: Regenerate.
12902 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12903 as an argument, moving assert of ad->disp == ad->disp_term to...
12904 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12905 Try calling targetm.legitimize_address_displacement before expanding
12906 the address rather than afterwards, and adjust for the new interface.
12907 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12908 Match the new hook interface. Handle SVE addresses.
12909 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12910 new hook interface.
12912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12914 * Makefile.in (OBJS): Add early-remat.o.
12915 * target.def (select_early_remat_modes): New hook.
12916 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12917 * doc/tm.texi: Regenerate.
12918 * targhooks.h (default_select_early_remat_modes): Declare.
12919 * targhooks.c (default_select_early_remat_modes): New function.
12920 * timevar.def (TV_EARLY_REMAT): New timevar.
12921 * passes.def (pass_early_remat): New pass.
12922 * tree-pass.h (make_pass_early_remat): Declare.
12923 * early-remat.c: New file.
12924 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12926 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12929 Alan Hayward <alan.hayward@arm.com>
12930 David Sherwood <david.sherwood@arm.com>
12932 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12933 vfm1 with a bound_epilog parameter.
12934 (vect_do_peeling): Update calls accordingly, and move the prologue
12935 call earlier in the function. Treat the base bound_epilog as 0 for
12936 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12937 this base when peeling for gaps.
12938 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12939 with fully-masked loops.
12940 (vect_estimate_min_profitable_iters): Handle the single peeled
12941 iteration in that case.
12943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12944 Alan Hayward <alan.hayward@arm.com>
12945 David Sherwood <david.sherwood@arm.com>
12947 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12948 single-element interleaving even if the size is not a power of 2.
12949 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12950 accesses for single-element interleaving if the group size is
12953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12954 Alan Hayward <alan.hayward@arm.com>
12955 David Sherwood <david.sherwood@arm.com>
12957 * doc/md.texi (fold_extract_last_@var{m}): Document.
12958 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12959 * optabs.def (fold_extract_last_optab): New optab.
12960 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12961 * internal-fn.c (fold_extract_direct): New macro.
12962 (expand_fold_extract_optab_fn): Likewise.
12963 (direct_fold_extract_optab_supported_p): Likewise.
12964 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12965 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12966 EXTRACT_LAST_REDUCTION.
12967 (get_initial_def_for_reduction): Do not create an initial vector
12968 for EXTRACT_LAST_REDUCTION reductions.
12969 (vectorizable_reduction): Leave the scalar phi in place for
12970 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12971 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12972 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12973 transform phase to vectorizable_condition.
12974 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12976 (vect_finish_stmt_generation): ...here.
12977 (vect_finish_replace_stmt): New function.
12978 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12979 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12981 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12984 Alan Hayward <alan.hayward@arm.com>
12985 David Sherwood <david.sherwood@arm.com>
12987 * doc/md.texi (extract_last_@var{m}): Document.
12988 * optabs.def (extract_last_optab): New optab.
12989 * internal-fn.def (EXTRACT_LAST): New internal function.
12990 * internal-fn.c (cond_unary_direct): New macro.
12991 (expand_cond_unary_optab_fn): Likewise.
12992 (direct_cond_unary_optab_supported_p): Likewise.
12993 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12994 loops using EXTRACT_LAST.
12995 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12996 (extract_last_<mode>): ...this optab.
12997 (vec_extract<mode><Vel>): Update accordingly.
12999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13000 Alan Hayward <alan.hayward@arm.com>
13001 David Sherwood <david.sherwood@arm.com>
13003 * target.def (empty_mask_is_expensive): New hook.
13004 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13005 * doc/tm.texi: Regenerate.
13006 * targhooks.h (default_empty_mask_is_expensive): Declare.
13007 * targhooks.c (default_empty_mask_is_expensive): New function.
13008 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13009 if the target says that empty masks are expensive.
13010 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13012 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13015 Alan Hayward <alan.hayward@arm.com>
13016 David Sherwood <david.sherwood@arm.com>
13018 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13019 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13020 (vect_use_loop_mask_for_alignment_p): New function.
13021 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13022 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13023 niters_skip argument. Make sure that the first niters_skip elements
13024 of the first iteration are inactive.
13025 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13026 Update call to vect_set_loop_masks_directly.
13027 (get_misalign_in_elems): New function, split out from...
13028 (vect_gen_prolog_loop_niters): ...here.
13029 (vect_update_init_of_dr): Take a code argument that specifies whether
13030 the adjustment should be added or subtracted.
13031 (vect_update_init_of_drs): Likewise.
13032 (vect_prepare_for_masked_peels): New function.
13033 (vect_do_peeling): Skip prologue peeling if we're using a mask
13034 instead. Update call to vect_update_inits_of_drs.
13035 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13037 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13038 alignment. Do not include the number of peeled iterations in
13039 the minimum threshold in that case.
13040 (vectorizable_induction): Adjust the start value down by
13041 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13042 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13043 Take the number of skipped iterations into account when calculating
13045 * tree-vect-stmts.c (vect_gen_while_not): New function.
13047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13048 Alan Hayward <alan.hayward@arm.com>
13049 David Sherwood <david.sherwood@arm.com>
13051 * doc/sourcebuild.texi (vect_fully_masked): Document.
13052 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13053 default value to 0.
13054 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13056 (vect_analyze_loop_2): ...here. Don't check the vectorization
13057 factor against the number of loop iterations if the loop is
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13065 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13066 (dump_groups): Update accordingly.
13067 (iv_use::mem_type): New member variable.
13068 (address_p): New function.
13069 (record_use): Add a mem_type argument and initialize the new
13071 (record_group_use): Add a mem_type argument. Use address_p.
13072 Remove obsolete null checks of base_object. Update call to record_use.
13073 (find_interesting_uses_op): Update call to record_group_use.
13074 (find_interesting_uses_cond): Likewise.
13075 (find_interesting_uses_address): Likewise.
13076 (get_mem_type_for_internal_fn): New function.
13077 (find_address_like_use): Likewise.
13078 (find_interesting_uses_stmt): Try find_address_like_use before
13079 calling find_interesting_uses_op.
13080 (addr_offset_valid_p): Use the iv mem_type field as the type
13081 of the addressed memory.
13082 (add_autoinc_candidates): Likewise.
13083 (get_address_cost): Likewise.
13084 (split_small_address_groups_p): Use address_p.
13085 (split_address_groups): Likewise.
13086 (add_iv_candidate_for_use): Likewise.
13087 (autoinc_possible_for_pair): Likewise.
13088 (rewrite_groups): Likewise.
13089 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13090 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13091 (get_alias_ptr_type_for_ptr_address): New function.
13092 (rewrite_use_address): Rewrite address uses in calls that were
13093 identified by find_address_like_use.
13095 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13096 Alan Hayward <alan.hayward@arm.com>
13097 David Sherwood <david.sherwood@arm.com>
13099 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13101 * gimple-expr.h (is_gimple_addressable: Likewise.
13102 * gimple-expr.c (is_gimple_address): Likewise.
13103 * internal-fn.c (expand_call_mem_ref): New function.
13104 (expand_mask_load_optab_fn): Use it.
13105 (expand_mask_store_optab_fn): Likewise.
13107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13108 Alan Hayward <alan.hayward@arm.com>
13109 David Sherwood <david.sherwood@arm.com>
13111 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13112 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13113 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13114 (cond_umax@var{mode}): Document.
13115 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13116 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13117 (cond_umin_optab, cond_umax_optab): New optabs.
13118 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13119 (COND_IOR, COND_XOR): New internal functions.
13120 * internal-fn.h (get_conditional_internal_fn): Declare.
13121 * internal-fn.c (cond_binary_direct): New macro.
13122 (expand_cond_binary_optab_fn): Likewise.
13123 (direct_cond_binary_optab_supported_p): Likewise.
13124 (get_conditional_internal_fn): New function.
13125 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13126 Cope with reduction statements that are vectorized as calls rather
13128 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13129 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13130 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13131 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13132 (UNSPEC_COND_EOR): New unspecs.
13133 (optab): Add mappings for them.
13134 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13135 (sve_int_op, sve_fp_op): New int attributes.
13137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13138 Alan Hayward <alan.hayward@arm.com>
13139 David Sherwood <david.sherwood@arm.com>
13141 * optabs.def (while_ult_optab): New optab.
13142 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13143 * internal-fn.def (WHILE_ULT): New internal function.
13144 * internal-fn.h (direct_internal_fn_supported_p): New override
13145 that takes two types as argument.
13146 * internal-fn.c (while_direct): New macro.
13147 (expand_while_optab_fn): New function.
13148 (convert_optab_supported_p): Likewise.
13149 (direct_while_optab_supported_p): New macro.
13150 * wide-int.h (wi::udiv_ceil): New function.
13151 * tree-vectorizer.h (rgroup_masks): New structure.
13152 (vec_loop_masks): New typedef.
13153 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13154 and fully_masked_p.
13155 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13156 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13157 (vect_max_vf): New function.
13158 (slpeel_make_loop_iterate_ntimes): Delete.
13159 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13160 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13161 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13162 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13163 internal-fn.h, stor-layout.h and optabs-query.h.
13164 (vect_set_loop_mask): New function.
13165 (add_preheader_seq): Likewise.
13166 (add_header_seq): Likewise.
13167 (interleave_supported_p): Likewise.
13168 (vect_maybe_permute_loop_masks): Likewise.
13169 (vect_set_loop_masks_directly): Likewise.
13170 (vect_set_loop_condition_masked): Likewise.
13171 (vect_set_loop_condition_unmasked): New function, split out from
13172 slpeel_make_loop_iterate_ntimes.
13173 (slpeel_make_loop_iterate_ntimes): Rename to..
13174 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13175 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13176 (vect_do_peeling): Update call accordingly.
13177 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13179 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13180 mask_compare_type, can_fully_mask_p and fully_masked_p.
13181 (release_vec_loop_masks): New function.
13182 (_loop_vec_info): Use it to free the loop masks.
13183 (can_produce_all_loop_masks_p): New function.
13184 (vect_get_max_nscalars_per_iter): Likewise.
13185 (vect_verify_full_masking): Likewise.
13186 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13187 retries, and free the mask rgroups before retrying. Check loop-wide
13188 reasons for disallowing fully-masked loops. Make the final decision
13189 about whether use a fully-masked loop or not.
13190 (vect_estimate_min_profitable_iters): Do not assume that peeling
13191 for the number of iterations will be needed for fully-masked loops.
13192 (vectorizable_reduction): Disable fully-masked loops.
13193 (vectorizable_live_operation): Likewise.
13194 (vect_halve_mask_nunits): New function.
13195 (vect_double_mask_nunits): Likewise.
13196 (vect_record_loop_mask): Likewise.
13197 (vect_get_loop_mask): Likewise.
13198 (vect_transform_loop): Handle the case in which the final loop
13199 iteration might handle a partial vector. Call vect_set_loop_condition
13200 instead of slpeel_make_loop_iterate_ntimes.
13201 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13202 (check_load_store_masking): New function.
13203 (prepare_load_store_mask): Likewise.
13204 (vectorizable_store): Handle fully-masked loops.
13205 (vectorizable_load): Likewise.
13206 (supportable_widening_operation): Use vect_halve_mask_nunits for
13208 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13209 (vect_gen_while): New function.
13210 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13211 (aarch64_uqdec<mode>): New insn.
13213 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13214 Alan Hayward <alan.hayward@arm.com>
13215 David Sherwood <david.sherwood@arm.com>
13217 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13218 (reduc_xor_scal_optab): New optabs.
13219 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13220 (reduc_xor_scal_@var{m}): Document.
13221 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13222 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13223 internal functions.
13224 * fold-const-call.c (fold_const_call): Handle them.
13225 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13226 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13227 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13228 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13229 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13230 (UNSPEC_XORV): New unspecs.
13231 (optab): Add entries for them.
13232 (BITWISEV): New int iterator.
13233 (bit_reduc_op): New int attributes.
13235 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13236 Alan Hayward <alan.hayward@arm.com>
13237 David Sherwood <david.sherwood@arm.com>
13239 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13240 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13241 * optabs.def (vec_shl_insert_optab): New optab.
13242 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13243 (duplicate_and_interleave): Likewise.
13244 * tree-vect-loop.c: Include internal-fn.h.
13245 (neutral_op_for_slp_reduction): New function, split out from
13246 get_initial_defs_for_reduction.
13247 (get_initial_def_for_reduction): Handle option 2 for variable-length
13248 vectors by loading the neutral value into a vector and then shifting
13249 the initial value into element 0.
13250 (get_initial_defs_for_reduction): Replace the code argument with
13251 the neutral value calculated by neutral_op_for_slp_reduction.
13252 Use gimple_build_vector for constant-length vectors.
13253 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13254 but the first group_size elements have a neutral value.
13255 Use duplicate_and_interleave otherwise.
13256 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13257 Update call to get_initial_defs_for_reduction. Handle SLP
13258 reductions for variable-length vectors by creating one vector
13259 result for each scalar result, with the elements associated
13260 with other scalar results stubbed out with the neutral value.
13261 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13262 Require IFN_VEC_SHL_INSERT for double reductions on
13263 variable-length vectors, or SLP reductions that have
13264 a neutral value. Require can_duplicate_and_interleave_p
13265 support for variable-length unchained SLP reductions if there
13266 is no neutral value, such as for MIN/MAX reductions. Also require
13267 the number of vector elements to be a multiple of the number of
13268 SLP statements when doing variable-length unchained SLP reductions.
13269 Update call to vect_create_epilog_for_reduction.
13270 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13271 and remove initial values.
13272 (duplicate_and_interleave): Make public.
13273 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13274 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13277 Alan Hayward <alan.hayward@arm.com>
13278 David Sherwood <david.sherwood@arm.com>
13280 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13281 (can_duplicate_and_interleave_p): New function.
13282 (vect_get_and_check_slp_defs): Take the vector of statements
13283 rather than just the current one. Remove excess parentheses.
13284 Restriction rejectinon of vect_constant_def and vect_external_def
13285 for variable-length vectors to boolean types, or types for which
13286 can_duplicate_and_interleave_p is false.
13287 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13288 (duplicate_and_interleave): New function.
13289 (vect_get_constant_vectors): Use gimple_build_vector for
13290 constant-length vectors and suitable variable-length constant
13291 vectors. Use duplicate_and_interleave for other variable-length
13292 vectors. Don't defer the update when inserting new statements.
13294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13295 Alan Hayward <alan.hayward@arm.com>
13296 David Sherwood <david.sherwood@arm.com>
13298 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13299 min_profitable_iters doesn't go negative.
13301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13302 Alan Hayward <alan.hayward@arm.com>
13303 David Sherwood <david.sherwood@arm.com>
13305 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13306 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13307 * optabs.def (vec_mask_load_lanes_optab): New optab.
13308 (vec_mask_store_lanes_optab): Likewise.
13309 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13310 (MASK_STORE_LANES): Likewise.
13311 * internal-fn.c (mask_load_lanes_direct): New macro.
13312 (mask_store_lanes_direct): Likewise.
13313 (expand_mask_load_optab_fn): Handle masked operations.
13314 (expand_mask_load_lanes_optab_fn): New macro.
13315 (expand_mask_store_optab_fn): Handle masked operations.
13316 (expand_mask_store_lanes_optab_fn): New macro.
13317 (direct_mask_load_lanes_optab_supported_p): Likewise.
13318 (direct_mask_store_lanes_optab_supported_p): Likewise.
13319 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13321 (vect_load_lanes_supported): Likewise.
13322 * tree-vect-data-refs.c (strip_conversion): New function.
13323 (can_group_stmts_p): Likewise.
13324 (vect_analyze_data_ref_accesses): Use it instead of checking
13325 for a pair of assignments.
13326 (vect_store_lanes_supported): Take a masked_p parameter.
13327 (vect_load_lanes_supported): Likewise.
13328 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13329 vect_store_lanes_supported and vect_load_lanes_supported.
13330 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13331 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13332 parameter. Don't allow gaps for masked accesses.
13333 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13334 and vect_load_lanes_supported.
13335 (get_load_store_type): Take a masked_p parameter and update
13336 call to get_group_load_store_type.
13337 (vectorizable_store): Update call to get_load_store_type.
13338 Handle IFN_MASK_STORE_LANES.
13339 (vectorizable_load): Update call to get_load_store_type.
13340 Handle IFN_MASK_LOAD_LANES.
13342 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13343 Alan Hayward <alan.hayward@arm.com>
13344 David Sherwood <david.sherwood@arm.com>
13346 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13348 * config/aarch64/aarch64-protos.h
13349 (aarch64_sve_struct_memory_operand_p): Declare.
13350 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13351 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13352 (VPRED, vpred): Handle SVE structure modes.
13353 * config/aarch64/constraints.md (Utx): New constraint.
13354 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13355 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13356 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13357 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13358 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13359 structure modes. Split into pieces after RA.
13360 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13361 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13363 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13364 SVE structure modes.
13365 (aarch64_classify_address): Likewise.
13366 (sizetochar): Move earlier in file.
13367 (aarch64_print_operand): Handle SVE register lists.
13368 (aarch64_array_mode): New function.
13369 (aarch64_sve_struct_memory_operand_p): Likewise.
13370 (TARGET_ARRAY_MODE): Redefine.
13372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13373 Alan Hayward <alan.hayward@arm.com>
13374 David Sherwood <david.sherwood@arm.com>
13376 * target.def (array_mode): New target hook.
13377 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13378 * doc/tm.texi: Regenerate.
13379 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13380 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13381 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13382 targetm.array_mode.
13383 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13387 Alan Hayward <alan.hayward@arm.com>
13388 David Sherwood <david.sherwood@arm.com>
13390 * fold-const.c (fold_binary_loc): Check the argument types
13391 rather than the result type when testing for a vector operation.
13393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13395 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13396 * doc/tm.texi: Regenerate.
13398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13399 Alan Hayward <alan.hayward@arm.com>
13400 David Sherwood <david.sherwood@arm.com>
13402 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13403 (sve): Document new AArch64 extension.
13404 * doc/md.texi (w): Extend the description of the AArch64
13405 constraint to include SVE vectors.
13406 (Upl, Upa): Document new AArch64 predicate constraints.
13407 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13409 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13410 (msve-vector-bits=): New option.
13411 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13412 SVE when these are disabled.
13413 (sve): New extension.
13414 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13415 modes. Adjust their number of units based on aarch64_sve_vg.
13416 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13417 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13418 aarch64_addr_query_type.
13419 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13420 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13421 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13422 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13423 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13424 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13425 (aarch64_simd_imm_zero_p): Delete.
13426 (aarch64_check_zero_based_sve_index_immediate): Declare.
13427 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13428 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13429 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13430 (aarch64_sve_float_mul_immediate_p): Likewise.
13431 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13432 rather than an rtx.
13433 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13434 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13435 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13436 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13437 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13438 (aarch64_regmode_natural_size): Likewise.
13439 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13440 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13442 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13443 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13444 for VG and the SVE predicate registers.
13445 (V_ALIASES): Add a "z"-prefixed alias.
13446 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13447 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13448 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13449 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13450 (REG_CLASS_NAMES): Add entries for them.
13451 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13452 and the predicate registers.
13453 (aarch64_sve_vg): Declare.
13454 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13455 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13456 (REGMODE_NATURAL_SIZE): Define.
13457 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13459 * config/aarch64/aarch64.c: Include cfgrtl.h.
13460 (simd_immediate_info): Add a constructor for series vectors,
13461 and an associated step field.
13462 (aarch64_sve_vg): New variable.
13463 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13464 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13465 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13466 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13467 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13468 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13469 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13470 (aarch64_get_mask_mode): New functions.
13471 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13472 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13473 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13474 predicate modes and predicate registers. Explicitly restrict
13475 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13476 to store a vector mode if it is recognized by
13477 aarch64_classify_vector_mode.
13478 (aarch64_regmode_natural_size): New function.
13479 (aarch64_hard_regno_caller_save_mode): Return the original mode
13481 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13482 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13483 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13484 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13486 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13487 does not overlap dest if the function is frame-related. Handle
13489 (aarch64_split_add_offset): New function.
13490 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13491 them aarch64_add_offset.
13492 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13493 and update call to aarch64_sub_sp.
13494 (aarch64_add_cfa_expression): New function.
13495 (aarch64_expand_prologue): Pass extra temporary registers to the
13496 functions above. Handle the case in which we need to emit new
13497 DW_CFA_expressions for registers that were originally saved
13498 relative to the stack pointer, but now have to be expressed
13499 relative to the frame pointer.
13500 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13502 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13503 IP0 and IP1 values for SVE frames.
13504 (aarch64_expand_vec_series): New function.
13505 (aarch64_expand_sve_widened_duplicate): Likewise.
13506 (aarch64_expand_sve_const_vector): Likewise.
13507 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13508 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13509 into the register, rather than emitting a SET directly.
13510 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13511 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13512 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13513 (offset_9bit_signed_scaled_p): New functions.
13514 (aarch64_replicate_bitmask_imm): New function.
13515 (aarch64_bitmask_imm): Use it.
13516 (aarch64_cannot_force_const_mem): Reject expressions involving
13517 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13518 (aarch64_classify_index): Handle SVE indices, by requiring
13519 a plain register index with a scale that matches the element size.
13520 (aarch64_classify_address): Handle SVE addresses. Assert that
13521 the mode of the address is VOIDmode or an integer mode.
13522 Update call to aarch64_classify_symbol.
13523 (aarch64_classify_symbolic_expression): Update call to
13524 aarch64_classify_symbol.
13525 (aarch64_const_vec_all_in_range_p): New function.
13526 (aarch64_print_vector_float_operand): Likewise.
13527 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13528 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13529 and the FP immediates 1.0 and 0.5.
13530 (aarch64_print_address_internal): Handle SVE addresses.
13531 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13532 (aarch64_regno_regclass): Handle predicate registers.
13533 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13535 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13536 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13537 (aarch64_convert_sve_vector_bits): New function.
13538 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13539 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13540 rather than an rtx.
13541 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13542 Handle SVE vector and predicate modes. Accept VL-based constants
13543 that need only one temporary register, and VL offsets that require
13544 no temporary registers.
13545 (aarch64_conditional_register_usage): Mark the predicate registers
13546 as fixed if SVE isn't available.
13547 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13548 Return true for SVE vector and predicate modes.
13549 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13550 rather than an unsigned int. Handle SVE modes.
13551 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13553 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13555 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13556 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13557 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13558 (aarch64_sve_float_mul_immediate_p): New functions.
13559 (aarch64_sve_valid_immediate): New function.
13560 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13561 Explicitly reject structure modes. Check for INDEX constants.
13562 Handle PTRUE and PFALSE constants.
13563 (aarch64_check_zero_based_sve_index_immediate): New function.
13564 (aarch64_simd_imm_zero_p): Delete.
13565 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13566 vector modes. Accept constants in the range of CNT[BHWD].
13567 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13568 ask for an Advanced SIMD mode.
13569 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13570 (aarch64_simd_vector_alignment): Handle SVE predicates.
13571 (aarch64_vectorize_preferred_vector_alignment): New function.
13572 (aarch64_simd_vector_alignment_reachable): Use it instead of
13574 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13575 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13577 (MAX_VECT_LEN): Delete.
13578 (expand_vec_perm_d): Add a vec_flags field.
13579 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13580 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13581 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13583 (aarch64_evpc_rev): Rename to...
13584 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13585 (aarch64_evpc_rev_global): New function.
13586 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13587 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13589 (aarch64_evpc_sve_tbl): New function.
13590 (aarch64_expand_vec_perm_const_1): Update after rename of
13591 aarch64_evpc_rev. Handle SVE permutes too, trying
13592 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13593 than aarch64_evpc_tbl.
13594 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13595 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13596 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13597 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13598 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13599 (aarch64_expand_sve_vcond): New functions.
13600 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13601 of aarch64_vector_mode_p.
13602 (aarch64_dwarf_poly_indeterminate_value): New function.
13603 (aarch64_compute_pressure_classes): Likewise.
13604 (aarch64_can_change_mode_class): Likewise.
13605 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13606 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13607 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13608 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13609 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13610 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13611 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13612 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13614 (Dn, Dl, Dr): Accept const as well as const_vector.
13615 (Dz): Likewise. Compare against CONST0_RTX.
13616 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13617 of "vector" where appropriate.
13618 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13619 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13620 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13621 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13622 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13623 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13624 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13625 (v_int_equiv): Extend to SVE modes.
13626 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13628 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13629 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13630 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13631 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13632 (SVE_COND_FP_CMP): New int iterators.
13633 (perm_hilo): Handle the new unpack unspecs.
13634 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13636 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13637 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13638 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13639 (aarch64_equality_operator, aarch64_constant_vector_operand)
13640 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13641 (aarch64_sve_nonimmediate_operand): Likewise.
13642 (aarch64_sve_general_operand): Likewise.
13643 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13644 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13645 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13646 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13647 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13648 (aarch64_sve_float_arith_immediate): Likewise.
13649 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13650 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13651 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13652 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13653 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13654 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13655 (aarch64_sve_float_arith_operand): Likewise.
13656 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13657 (aarch64_sve_float_mul_operand): Likewise.
13658 (aarch64_sve_vec_perm_operand): Likewise.
13659 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13660 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13661 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13662 as well as const_vector.
13663 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13664 in file. Use CONST0_RTX and CONSTM1_RTX.
13665 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13666 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13667 Use aarch64_simd_imm_zero.
13668 * config/aarch64/aarch64-sve.md: New file.
13669 * config/aarch64/aarch64.md: Include it.
13670 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13671 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13672 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13673 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13674 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13675 (sve): New attribute.
13676 (enabled): Disable instructions with the sve attribute unless
13678 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13679 aarch64_expand_mov_immediate.
13680 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13681 CNT[BHSD] immediates.
13682 (movti): Split CONST_POLY_INT moves into two halves.
13683 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13684 Split additions that need a temporary here if the destination
13685 is the stack pointer.
13686 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13687 (*add<mode>3_poly_1): New instruction.
13688 (set_clobber_cc): New expander.
13690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13692 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13693 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13694 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13695 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13696 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13697 Change innermode from fixed_mode_size to machine_mode.
13698 (simplify_subreg): Update call accordingly. Handle a constant-sized
13699 subreg of a variable-length CONST_VECTOR.
13701 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13702 Alan Hayward <alan.hayward@arm.com>
13703 David Sherwood <david.sherwood@arm.com>
13705 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13706 (add_offset_to_base): New function, split out from...
13707 (create_mem_ref): ...here. When handling a scale other than 1,
13708 check first whether the address is valid without the offset.
13709 Add it into the base if so, leaving the index and scale as-is.
13711 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13714 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13715 fold_for_warn before checking if arg2 is INTEGER_CST.
13717 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13719 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13720 (store_multiple_operation): Delete.
13721 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13722 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13723 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13724 guarded by TARGET_STRING.
13725 (rs6000_output_load_multiple): Delete.
13726 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13727 OPTION_MASK_STRING / TARGET_STRING handling.
13728 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13729 (const rs6000_opt_masks) <"string">: Change mask to 0.
13730 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13731 (MASK_STRING): Delete.
13732 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13734 (load_multiple): Delete.
13741 (store_multiple): Delete.
13748 (movmemsi_8reg): Delete.
13749 (corresponding unnamed define_insn): Delete.
13750 (movmemsi_6reg): Delete.
13751 (corresponding unnamed define_insn): Delete.
13752 (movmemsi_4reg): Delete.
13753 (corresponding unnamed define_insn): Delete.
13754 (movmemsi_2reg): Delete.
13755 (corresponding unnamed define_insn): Delete.
13756 (movmemsi_1reg): Delete.
13757 (corresponding unnamed define_insn): Delete.
13758 * config/rs6000/rs6000.opt (mno-string): New.
13759 (mstring): Replace by deprecation warning stub.
13760 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13762 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13764 * regrename.c (regrename_do_replace): If replacing the same
13765 reg multiple times, try to reuse last created gen_raw_REG.
13768 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13769 main to workaround a bug in GDB.
13771 2018-01-12 Tom de Vries <tom@codesourcery.com>
13774 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13776 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13778 PR rtl-optimization/80481
13779 * ira-color.c (get_cap_member): New function.
13780 (allocnos_conflict_by_live_ranges_p): Use it.
13781 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13782 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13784 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13787 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13788 (*saddl_se_1): Ditto.
13789 (*ssubsi_1): Ditto.
13790 (*ssubl_se_1): Ditto.
13792 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13794 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13795 rather than wi::to_widest for DR_INITs.
13796 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13797 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13798 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13800 (vect_analyze_group_access_1): Note that here.
13802 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13804 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13805 polynomial type sizes.
13807 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13809 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13810 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13811 (gimple_add_tmp_var): Likewise.
13813 2018-01-12 Martin Liska <mliska@suse.cz>
13815 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13816 (gimple_alloc_sizes): Likewise.
13817 (dump_gimple_statistics): Use PRIu64 in printf format.
13818 * gimple.h: Change uint64_t to int.
13820 2018-01-12 Martin Liska <mliska@suse.cz>
13822 * tree-core.h: Use uint64_t instead of int.
13823 * tree.c (tree_node_counts): Likewise.
13824 (tree_node_sizes): Likewise.
13825 (dump_tree_statistics): Use PRIu64 in printf format.
13827 2018-01-12 Martin Liska <mliska@suse.cz>
13829 * Makefile.in: As qsort_chk is implemented in vec.c, add
13830 vec.o to linkage of gencfn-macros.
13831 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13832 passing the info to record_node_allocation_statistics.
13833 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13835 * ggc-common.c (struct ggc_usage): Add operator== and use
13836 it in operator< and compare function.
13837 * mem-stats.h (struct mem_usage): Likewise.
13838 * vec.c (struct vec_usage): Remove operator< and compare
13839 function. Can be simply inherited.
13841 2018-01-12 Martin Jambor <mjambor@suse.cz>
13844 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13845 * tree-ssa-math-opts.c: Include domwalk.h.
13846 (convert_mult_to_fma_1): New function.
13847 (fma_transformation_info): New type.
13848 (fma_deferring_state): Likewise.
13849 (cancel_fma_deferring): New function.
13850 (result_of_phi): Likewise.
13851 (last_fma_candidate_feeds_initial_phi): Likewise.
13852 (convert_mult_to_fma): Added deferring logic, split actual
13853 transformation to convert_mult_to_fma_1.
13854 (math_opts_dom_walker): New type.
13855 (math_opts_dom_walker::after_dom_children): New method, body moved
13856 here from pass_optimize_widening_mul::execute, added deferring logic
13858 (pass_optimize_widening_mul::execute): Moved most of code to
13859 math_opts_dom_walker::after_dom_children.
13860 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13861 * config/i386/i386.c (ix86_option_override_internal): Added
13862 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13864 2018-01-12 Richard Biener <rguenther@suse.de>
13867 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13868 inline instance vars.
13870 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13873 * config/rx/rx.c (rx_is_restricted_memory_address):
13874 Handle SUBREG case.
13876 2018-01-12 Richard Biener <rguenther@suse.de>
13878 PR tree-optimization/80846
13879 * target.def (split_reduction): New target hook.
13880 * targhooks.c (default_split_reduction): New function.
13881 * targhooks.h (default_split_reduction): Declare.
13882 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13883 target requests first reduce vectors by combining low and high
13885 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13886 (get_vectype_for_scalar_type_and_size): Export.
13887 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13888 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13889 * doc/tm.texi: Regenerate.
13890 * config/i386/i386.c (ix86_split_reduction): Implement
13891 TARGET_VECTORIZE_SPLIT_REDUCTION.
13893 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13896 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13897 in PIC mode except for TARGET_VXWORKS_RTP.
13898 * config/sparc/sparc.c: Include cfgrtl.h.
13899 (TARGET_INIT_PIC_REG): Define.
13900 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13901 (sparc_pic_register_p): New predicate.
13902 (sparc_legitimate_address_p): Use it.
13903 (sparc_legitimize_pic_address): Likewise.
13904 (sparc_delegitimize_address): Likewise.
13905 (sparc_mode_dependent_address_p): Likewise.
13906 (gen_load_pcrel_sym): Remove 4th parameter.
13907 (load_got_register): Adjust call to above. Remove obsolete stuff.
13908 (sparc_expand_prologue): Do not call load_got_register here.
13909 (sparc_flat_expand_prologue): Likewise.
13910 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13911 (sparc_use_pseudo_pic_reg): New function.
13912 (sparc_init_pic_reg): Likewise.
13913 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13914 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13916 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13918 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13919 Add item for branch_cost.
13921 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13923 PR rtl-optimization/83565
13924 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13925 not extend the result to a larger mode for rotate operations.
13926 (num_sign_bit_copies1): Likewise.
13928 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13931 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13933 Use values-Xc.o for -pedantic.
13934 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13936 2018-01-12 Martin Liska <mliska@suse.cz>
13939 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13941 (possible_polymorphic_call_targets): Use it.
13942 (ipa_devirt): Likewise.
13944 2018-01-12 Martin Liska <mliska@suse.cz>
13946 * profile-count.h (enum profile_quality): Use 0 as invalid
13947 enum value of profile_quality.
13949 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13951 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13952 -mext-string options.
13954 2018-01-12 Richard Biener <rguenther@suse.de>
13956 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13957 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13958 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13960 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13962 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13964 * configure.ac (--with-long-double-format): Add support for the
13965 configuration option to change the default long double format on
13967 * config.gcc (powerpc*-linux*-*): Likewise.
13968 * configure: Regenerate.
13969 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13970 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13971 used without modification.
13973 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13975 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13976 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13977 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13978 MISC_BUILTIN_SPEC_BARRIER.
13979 (rs6000_init_builtins): Likewise.
13980 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13982 (speculation_barrier): New define_insn.
13983 * doc/extend.texi: Document __builtin_speculation_barrier.
13985 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13988 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13989 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13990 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13992 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13993 integral modes instead of "ss" and "sd".
13994 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13995 vectors with 32-bit and 64-bit elements.
13996 (vecdupssescalarmodesuffix): New mode attribute.
13997 (vec_dup<mode>): Use it.
13999 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14002 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14003 frame if argument is passed on stack.
14005 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14008 * ree.c (combine_reaching_defs): Optimize also
14009 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14010 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14012 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14014 PR middle-end/83189
14015 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14017 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14019 PR middle-end/83718
14020 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14021 after they are computed.
14023 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14025 PR tree-optimization/83695
14026 * gimple-loop-linterchange.cc
14027 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14028 reset cached scev information after interchange.
14029 (pass_linterchange::execute): Remove call to scev_reset_htab.
14031 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14033 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14034 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14035 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14036 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14037 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14038 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14039 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14040 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14041 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14042 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14043 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14044 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14045 (V_lane_reg): Likewise.
14046 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14048 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14049 (vfmal_lane_low<mode>_intrinsic,
14050 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14051 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14052 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14053 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14054 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14055 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14057 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14059 * config/arm/arm-cpus.in (fp16fml): New feature.
14060 (ALL_SIMD): Add fp16fml.
14061 (armv8.2-a): Add fp16fml as an option.
14062 (armv8.3-a): Likewise.
14063 (armv8.4-a): Add fp16fml as part of fp16.
14064 * config/arm/arm.h (TARGET_FP16FML): Define.
14065 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14067 * config/arm/arm-modes.def (V2HF): Define.
14068 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14069 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14070 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14071 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14072 vfmsl_low, vfmsl_high): New set of builtins.
14073 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14074 (vfml_op): New code attribute.
14075 (VFMLHALVES): New int iterator.
14076 (VFML, VFMLSEL): New mode attributes.
14077 (V_reg): Define mapping for V2HF.
14078 (V_hi, V_lo): New mode attributes.
14079 (VF_constraint): Likewise.
14080 (vfml_half, vfml_half_selector): New int attributes.
14081 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14083 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14084 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14086 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14087 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14088 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14089 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14091 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14092 Document new effective target and option set.
14094 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14096 * config/arm/arm-cpus.in (armv8_4): New feature.
14097 (ARMv8_4a): New fgroup.
14098 (armv8.4-a): New arch.
14099 * config/arm/arm-tables.opt: Regenerate.
14100 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14101 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14102 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14103 Add matching rules for -march=armv8.4-a and extensions.
14104 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14106 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14109 * config/rx/rx.md (BW): New mode attribute.
14110 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14112 2018-01-11 Richard Biener <rguenther@suse.de>
14114 PR tree-optimization/83435
14115 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14116 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14117 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14119 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14120 Alan Hayward <alan.hayward@arm.com>
14121 David Sherwood <david.sherwood@arm.com>
14123 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14125 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14126 (aarch64_print_address_internal): Use it to check for a zero offset.
14128 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14129 Alan Hayward <alan.hayward@arm.com>
14130 David Sherwood <david.sherwood@arm.com>
14132 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14133 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14134 Return a poly_int64 rather than a HOST_WIDE_INT.
14135 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14136 rather than a HOST_WIDE_INT.
14137 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14138 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14139 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14140 final_offset from HOST_WIDE_INT to poly_int64.
14141 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14142 to_constant when getting the number of units in an Advanced SIMD
14144 (aarch64_builtin_vectorized_function): Check for a constant number
14146 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14148 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14149 attribute instead of GET_MODE_NUNITS.
14150 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14151 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14152 GET_MODE_SIZE for fixed-size registers.
14153 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14154 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14155 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14156 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14157 (aarch64_print_operand, aarch64_print_address_internal)
14158 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14159 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14160 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14161 Handle polynomial GET_MODE_SIZE.
14162 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14163 wider than SImode without modification.
14164 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14165 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14166 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14167 passing and returning SVE modes.
14168 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14169 rather than GEN_INT.
14170 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14171 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14172 (aarch64_allocate_and_probe_stack_space): Likewise.
14173 (aarch64_layout_frame): Cope with polynomial offsets.
14174 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14175 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14176 polynomial offsets.
14177 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14178 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14179 poly_int64 rather than a HOST_WIDE_INT.
14180 (aarch64_get_separate_components, aarch64_process_components)
14181 (aarch64_expand_prologue, aarch64_expand_epilogue)
14182 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14183 (aarch64_anchor_offset): New function, split out from...
14184 (aarch64_legitimize_address): ...here.
14185 (aarch64_builtin_vectorization_cost): Handle polynomial
14186 TYPE_VECTOR_SUBPARTS.
14187 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14189 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14190 number of elements from the PARALLEL rather than the mode.
14191 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14192 rather than GET_MODE_BITSIZE.
14193 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14194 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14195 (aarch64_expand_vec_perm_const_1): Handle polynomial
14196 d->perm.length () and d->perm elements.
14197 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14198 Apply to_constant to d->perm elements.
14199 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14200 polynomial CONST_VECTOR_NUNITS.
14201 (aarch64_move_pointer): Take amount as a poly_int64 rather
14203 (aarch64_progress_pointer): Avoid temporary variable.
14204 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14205 the mode attribute instead of GET_MODE.
14207 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14208 Alan Hayward <alan.hayward@arm.com>
14209 David Sherwood <david.sherwood@arm.com>
14211 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14212 x exists before using it.
14213 (aarch64_add_constant_internal): Rename to...
14214 (aarch64_add_offset_1): ...this. Replace regnum with separate
14215 src and dest rtxes. Handle the case in which they're different,
14216 including when the offset is zero. Replace scratchreg with an rtx.
14217 Use 2 additions if there is no spare register into which we can
14218 move a 16-bit constant.
14219 (aarch64_add_constant): Delete.
14220 (aarch64_add_offset): Replace reg with separate src and dest
14221 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14222 Use aarch64_add_offset_1.
14223 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14224 an rtx rather than an int. Take the delta as a poly_int64
14225 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14226 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14227 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14228 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14229 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14230 and aarch64_add_sp.
14231 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14232 aarch64_add_constant.
14234 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14236 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14237 Use scalar_float_mode.
14239 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14241 * config/aarch64/aarch64-simd.md
14242 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14243 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14244 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14245 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14246 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14247 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14248 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14249 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14250 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14251 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14253 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14256 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14257 targ_options->x_arm_arch_string is non NULL.
14259 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14261 * config/aarch64/aarch64.h
14262 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14264 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14267 * expmed.c (emit_store_flag_force): Swap if const op0
14268 and change VOIDmode to mode of op0.
14270 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14272 PR rtl-optimization/83761
14273 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14274 than bytes to mode_for_size.
14276 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14278 PR middle-end/83189
14279 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14280 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14283 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14285 PR middle-end/83575
14286 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14287 when in layout mode.
14288 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14289 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14292 2018-01-10 Michael Collison <michael.collison@arm.com>
14294 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14295 * config/aarch64/aarch64-option-extension.def: Add
14296 AARCH64_OPT_EXTENSION of 'fp16fml'.
14297 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14298 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14299 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14300 * config/aarch64/constraints.md (Ui7): New constraint.
14301 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14302 (VFMLA_SEL_W): Ditto.
14305 (VFMLA16_LOW): New int iterator.
14306 (VFMLA16_HIGH): Ditto.
14307 (UNSPEC_FMLAL): New unspec.
14308 (UNSPEC_FMLSL): Ditto.
14309 (UNSPEC_FMLAL2): Ditto.
14310 (UNSPEC_FMLSL2): Ditto.
14311 (f16mac): New code attribute.
14312 * config/aarch64/aarch64-simd-builtins.def
14313 (aarch64_fmlal_lowv2sf): Ditto.
14314 (aarch64_fmlsl_lowv2sf): Ditto.
14315 (aarch64_fmlalq_lowv4sf): Ditto.
14316 (aarch64_fmlslq_lowv4sf): Ditto.
14317 (aarch64_fmlal_highv2sf): Ditto.
14318 (aarch64_fmlsl_highv2sf): Ditto.
14319 (aarch64_fmlalq_highv4sf): Ditto.
14320 (aarch64_fmlslq_highv4sf): Ditto.
14321 (aarch64_fmlal_lane_lowv2sf): Ditto.
14322 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14323 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14324 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14325 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14326 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14327 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14328 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14329 (aarch64_fmlal_lane_highv2sf): Ditto.
14330 (aarch64_fmlsl_lane_highv2sf): Ditto.
14331 (aarch64_fmlal_laneq_highv2sf): Ditto.
14332 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14333 (aarch64_fmlalq_lane_highv4sf): Ditto.
14334 (aarch64_fmlsl_lane_highv4sf): Ditto.
14335 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14336 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14337 * config/aarch64/aarch64-simd.md:
14338 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14339 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14340 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14341 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14342 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14343 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14344 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14345 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14346 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14347 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14348 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14349 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14350 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14351 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14352 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14353 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14354 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14355 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14356 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14357 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14358 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14359 (vfmlsl_low_u32): Ditto.
14360 (vfmlalq_low_u32): Ditto.
14361 (vfmlslq_low_u32): Ditto.
14362 (vfmlal_high_u32): Ditto.
14363 (vfmlsl_high_u32): Ditto.
14364 (vfmlalq_high_u32): Ditto.
14365 (vfmlslq_high_u32): Ditto.
14366 (vfmlal_lane_low_u32): Ditto.
14367 (vfmlsl_lane_low_u32): Ditto.
14368 (vfmlal_laneq_low_u32): Ditto.
14369 (vfmlsl_laneq_low_u32): Ditto.
14370 (vfmlalq_lane_low_u32): Ditto.
14371 (vfmlslq_lane_low_u32): Ditto.
14372 (vfmlalq_laneq_low_u32): Ditto.
14373 (vfmlslq_laneq_low_u32): Ditto.
14374 (vfmlal_lane_high_u32): Ditto.
14375 (vfmlsl_lane_high_u32): Ditto.
14376 (vfmlal_laneq_high_u32): Ditto.
14377 (vfmlsl_laneq_high_u32): Ditto.
14378 (vfmlalq_lane_high_u32): Ditto.
14379 (vfmlslq_lane_high_u32): Ditto.
14380 (vfmlalq_laneq_high_u32): Ditto.
14381 (vfmlslq_laneq_high_u32): Ditto.
14382 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14383 (AARCH64_FL_FOR_ARCH8_4): New.
14384 (AARCH64_ISA_F16FML): New ISA flag.
14385 (TARGET_F16FML): New feature flag for fp16fml.
14386 (doc/invoke.texi): Document new fp16fml option.
14388 2018-01-10 Michael Collison <michael.collison@arm.com>
14390 * config/aarch64/aarch64-builtins.c:
14391 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14392 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14393 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14394 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14395 (AARCH64_ISA_SHA3): New ISA flag.
14396 (TARGET_SHA3): New feature flag for sha3.
14397 * config/aarch64/iterators.md (sha512_op): New int attribute.
14398 (CRYPTO_SHA512): New int iterator.
14399 (UNSPEC_SHA512H): New unspec.
14400 (UNSPEC_SHA512H2): Ditto.
14401 (UNSPEC_SHA512SU0): Ditto.
14402 (UNSPEC_SHA512SU1): Ditto.
14403 * config/aarch64/aarch64-simd-builtins.def
14404 (aarch64_crypto_sha512hqv2di): New builtin.
14405 (aarch64_crypto_sha512h2qv2di): Ditto.
14406 (aarch64_crypto_sha512su0qv2di): Ditto.
14407 (aarch64_crypto_sha512su1qv2di): Ditto.
14408 (aarch64_eor3qv8hi): Ditto.
14409 (aarch64_rax1qv2di): Ditto.
14410 (aarch64_xarqv2di): Ditto.
14411 (aarch64_bcaxqv8hi): Ditto.
14412 * config/aarch64/aarch64-simd.md:
14413 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14414 (aarch64_crypto_sha512su0qv2di): Ditto.
14415 (aarch64_crypto_sha512su1qv2di): Ditto.
14416 (aarch64_eor3qv8hi): Ditto.
14417 (aarch64_rax1qv2di): Ditto.
14418 (aarch64_xarqv2di): Ditto.
14419 (aarch64_bcaxqv8hi): Ditto.
14420 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14421 (vsha512h2q_u64): Ditto.
14422 (vsha512su0q_u64): Ditto.
14423 (vsha512su1q_u64): Ditto.
14424 (veor3q_u16): Ditto.
14425 (vrax1q_u64): Ditto.
14426 (vxarq_u64): Ditto.
14427 (vbcaxq_u16): Ditto.
14428 * config/arm/types.md (crypto_sha512): New type attribute.
14429 (crypto_sha3): Ditto.
14430 (doc/invoke.texi): Document new sha3 option.
14432 2018-01-10 Michael Collison <michael.collison@arm.com>
14434 * config/aarch64/aarch64-builtins.c:
14435 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14436 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14437 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14438 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14439 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14440 (AARCH64_ISA_SM4): New ISA flag.
14441 (TARGET_SM4): New feature flag for sm4.
14442 * config/aarch64/aarch64-simd-builtins.def
14443 (aarch64_sm3ss1qv4si): Ditto.
14444 (aarch64_sm3tt1aq4si): Ditto.
14445 (aarch64_sm3tt1bq4si): Ditto.
14446 (aarch64_sm3tt2aq4si): Ditto.
14447 (aarch64_sm3tt2bq4si): Ditto.
14448 (aarch64_sm3partw1qv4si): Ditto.
14449 (aarch64_sm3partw2qv4si): Ditto.
14450 (aarch64_sm4eqv4si): Ditto.
14451 (aarch64_sm4ekeyqv4si): Ditto.
14452 * config/aarch64/aarch64-simd.md:
14453 (aarch64_sm3ss1qv4si): Ditto.
14454 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14455 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14456 (aarch64_sm4eqv4si): Ditto.
14457 (aarch64_sm4ekeyqv4si): Ditto.
14458 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14459 (sm3part_op): Ditto.
14460 (CRYPTO_SM3TT): Ditto.
14461 (CRYPTO_SM3PART): Ditto.
14462 (UNSPEC_SM3SS1): New unspec.
14463 (UNSPEC_SM3TT1A): Ditto.
14464 (UNSPEC_SM3TT1B): Ditto.
14465 (UNSPEC_SM3TT2A): Ditto.
14466 (UNSPEC_SM3TT2B): Ditto.
14467 (UNSPEC_SM3PARTW1): Ditto.
14468 (UNSPEC_SM3PARTW2): Ditto.
14469 (UNSPEC_SM4E): Ditto.
14470 (UNSPEC_SM4EKEY): Ditto.
14471 * config/aarch64/constraints.md (Ui2): New constraint.
14472 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14473 * config/arm/types.md (crypto_sm3): New type attribute.
14474 (crypto_sm4): Ditto.
14475 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14476 (vsm3tt1aq_u32): Ditto.
14477 (vsm3tt1bq_u32): Ditto.
14478 (vsm3tt2aq_u32): Ditto.
14479 (vsm3tt2bq_u32): Ditto.
14480 (vsm3partw1q_u32): Ditto.
14481 (vsm3partw2q_u32): Ditto.
14482 (vsm4eq_u32): Ditto.
14483 (vsm4ekeyq_u32): Ditto.
14484 (doc/invoke.texi): Document new sm4 option.
14486 2018-01-10 Michael Collison <michael.collison@arm.com>
14488 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14489 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14490 (AARCH64_FL_FOR_ARCH8_4): New.
14491 (AARCH64_FL_V8_4): New flag.
14492 (doc/invoke.texi): Document new armv8.4-a option.
14494 2018-01-10 Michael Collison <michael.collison@arm.com>
14496 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14497 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14498 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14499 * config/aarch64/aarch64-option-extension.def: Add
14500 AARCH64_OPT_EXTENSION of 'sha2'.
14501 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14502 (crypto): Disable sha2 and aes if crypto disabled.
14503 (crypto): Enable aes and sha2 if enabled.
14504 (simd): Disable sha2 and aes if simd disabled.
14505 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14507 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14508 (TARGET_SHA2): New feature flag for sha2.
14509 (TARGET_AES): New feature flag for aes.
14510 * config/aarch64/aarch64-simd.md:
14511 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14512 conditional on TARGET_AES.
14513 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14514 (aarch64_crypto_sha1hsi): Make pattern conditional
14516 (aarch64_crypto_sha1hv4si): Ditto.
14517 (aarch64_be_crypto_sha1hv4si): Ditto.
14518 (aarch64_crypto_sha1su1v4si): Ditto.
14519 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14520 (aarch64_crypto_sha1su0v4si): Ditto.
14521 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14522 (aarch64_crypto_sha256su0v4si): Ditto.
14523 (aarch64_crypto_sha256su1v4si): Ditto.
14524 (doc/invoke.texi): Document new aes and sha2 options.
14526 2018-01-10 Martin Sebor <msebor@redhat.com>
14528 PR tree-optimization/83781
14529 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14532 2018-01-11 Martin Sebor <msebor@gmail.com>
14533 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14535 PR tree-optimization/83501
14536 PR tree-optimization/81703
14538 * tree-ssa-strlen.c (get_string_cst): Rename...
14539 (get_string_len): ...to this. Handle global constants.
14540 (handle_char_store): Adjust.
14542 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14543 Jim Wilson <jimw@sifive.com>
14545 * config/riscv/riscv-protos.h (riscv_output_return): New.
14546 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14547 (riscv_attribute_table, riscv_output_return),
14548 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14549 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14550 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14552 (riscv_expand_prologue): Add early return for naked function.
14553 (riscv_expand_epilogue): Likewise.
14554 (riscv_function_ok_for_sibcall): Return false for naked function.
14555 (riscv_set_current_function): New.
14556 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14557 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14558 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14559 * doc/extend.texi (RISC-V Function Attributes): New.
14561 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14563 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14564 check for 128-bit long double before checking TCmode.
14565 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14566 128-bit long doubles before checking TFmode or TCmode.
14567 (FLOAT128_IBM_P): Likewise.
14569 2018-01-10 Martin Sebor <msebor@redhat.com>
14571 PR tree-optimization/83671
14572 * builtins.c (c_strlen): Unconditionally return zero for the empty
14574 Use -Warray-bounds for warnings.
14575 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14576 for non-constant array indices with COMPONENT_REF, arrays of
14577 arrays, and pointers to arrays.
14578 (gimple_fold_builtin_strlen): Determine and set length range for
14579 non-constant character arrays.
14581 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14583 PR middle-end/81897
14584 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14587 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14589 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14591 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14594 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14595 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14596 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14597 indexed_or_indirect_operand predicate.
14598 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14599 (*vsx_le_perm_load_v8hi): Likewise.
14600 (*vsx_le_perm_load_v16qi): Likewise.
14601 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14602 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14603 (*vsx_le_perm_store_v8hi): Likewise.
14604 (*vsx_le_perm_store_v16qi): Likewise.
14605 (eight unnamed splitters): Likewise.
14607 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14609 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14610 * config/rs6000/emmintrin.h: Likewise.
14611 * config/rs6000/mmintrin.h: Likewise.
14612 * config/rs6000/xmmintrin.h: Likewise.
14614 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14617 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14619 * tree.c (tree_nop_conversion): Return true for location wrapper
14621 (maybe_wrap_with_location): New function.
14622 (selftest::check_strip_nops): New function.
14623 (selftest::test_location_wrappers): New function.
14624 (selftest::tree_c_tests): Call it.
14625 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14626 (maybe_wrap_with_location): New decl.
14627 (EXPR_LOCATION_WRAPPER_P): New macro.
14628 (location_wrapper_p): New inline function.
14629 (tree_strip_any_location_wrapper): New inline function.
14631 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14634 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14635 stack_realign_offset for the largest alignment of stack slot
14637 (ix86_find_max_used_stack_alignment): New function.
14638 (ix86_finalize_stack_frame_flags): Use it. Set
14639 max_used_stack_alignment if we don't realign stack.
14640 * config/i386/i386.h (machine_function): Add
14641 max_used_stack_alignment.
14643 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14645 * config/arm/arm.opt (-mbranch-cost): New option.
14646 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14649 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14652 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14653 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14655 2018-01-10 Richard Biener <rguenther@suse.de>
14658 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14659 early out so it also covers the case where we have a non-NULL
14662 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14664 PR tree-optimization/83753
14665 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14666 for non-strided grouped accesses if the number of elements is 1.
14668 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14671 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14672 * i386.h (TARGET_USE_GATHER): Define.
14673 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14675 2018-01-10 Martin Liska <mliska@suse.cz>
14678 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14679 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14681 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14682 CLEANUP_NO_PARTITIONING is not set.
14684 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14686 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14687 for vectors, as a partial revert of r254296.
14688 * rtl.h (const_vec_p): Delete.
14689 (const_vec_duplicate_p): Don't test for vector CONSTs.
14690 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14691 * expmed.c (make_tree): Likewise.
14694 * common.md (E, F): Use CONSTANT_P instead of checking for
14696 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14697 checking for CONST_VECTOR.
14699 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14701 PR middle-end/83575
14702 * predict.c (force_edge_cold): Handle in more sane way edges
14703 with no prediction.
14705 2018-01-09 Carl Love <cel@us.ibm.com>
14707 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14709 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14710 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14711 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14712 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14713 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14714 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14715 * config/rs6000/rs6000-protos.h: Add extern defition for
14716 rs6000_generate_float2_double_code.
14717 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14719 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14720 (float2_v2df): Add define_expand.
14722 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14725 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14726 op_mode in the force_to_mode call.
14728 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14730 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14731 instead of checking each element individually.
14732 (aarch64_evpc_uzp): Likewise.
14733 (aarch64_evpc_zip): Likewise.
14734 (aarch64_evpc_ext): Likewise.
14735 (aarch64_evpc_rev): Likewise.
14736 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14737 instead of checking each element individually. Return true without
14739 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14740 whether all selected elements come from the same input, instead of
14741 checking each element individually. Remove calls to gen_rtx_REG,
14742 start_sequence and end_sequence and instead assert that no rtl is
14745 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14747 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14748 order of HIGH and CONST checks.
14750 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14752 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14753 if the destination isn't an SSA_NAME.
14755 2018-01-09 Richard Biener <rguenther@suse.de>
14757 PR tree-optimization/83668
14758 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14760 (canonicalize_loop_form): ... here, renamed from ...
14761 (canonicalize_loop_closed_ssa_form): ... this and amended to
14762 swap successor edges for loop exit blocks to make us use
14763 the RPO order we need for initial schedule generation.
14765 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14767 PR tree-optimization/64811
14768 * match.pd: When optimizing comparisons with Inf, avoid
14769 introducing or losing exceptions from comparisons with NaN.
14771 2018-01-09 Martin Liska <mliska@suse.cz>
14774 * asan.c (shadow_mem_size): Add gcc_assert.
14776 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14778 Don't save registers in main().
14781 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14782 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14783 * config/avr/avr.c (avr_set_current_function): Don't error if
14784 naked, OS_task or OS_main are specified at the same time.
14785 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14787 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14789 * common/config/avr/avr-common.c (avr_option_optimization_table):
14790 Switch on -mmain-is-OS_task for optimizing compilations.
14792 2018-01-09 Richard Biener <rguenther@suse.de>
14794 PR tree-optimization/83572
14795 * graphite.c: Include cfganal.h.
14796 (graphite_transform_loops): Connect infinite loops to exit
14797 and remove fake edges at the end.
14799 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14801 * ipa-inline.c (edge_badness): Revert accidental checkin.
14803 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14806 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14807 symbols; not inline clones.
14809 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14812 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14813 hard registers. Formatting fixes.
14815 PR preprocessor/83722
14816 * gcc.c (try_generate_repro): Pass
14817 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14818 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14821 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14822 Kito Cheng <kito.cheng@gmail.com>
14824 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14825 (riscv_leaf_function_p): Delete.
14826 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14828 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14830 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14832 (do_ifelse): New function.
14833 (do_isel): New function.
14834 (do_sub3): New function.
14835 (do_add3): New function.
14836 (do_load_mask_compare): New function.
14837 (do_overlap_load_compare): New function.
14838 (expand_compare_loop): New function.
14839 (expand_block_compare): Call expand_compare_loop() when appropriate.
14840 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14841 option description.
14842 (-mblock-compare-inline-loop-limit): New option.
14844 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14847 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14848 Reverse order of second and third operands in first alternative.
14849 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14850 of first and second elements in UNSPEC_VPERMR vector.
14851 (altivec_expand_vec_perm_le): Likewise.
14853 2018-01-08 Jeff Law <law@redhat.com>
14855 PR rtl-optimizatin/81308
14856 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14857 (process_switch): If group_case_labels makes a change, then set
14859 (pass_convert_switch::execute): If a switch is converted, then
14860 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14862 PR rtl-optimization/81308
14863 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14866 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14868 PR target/83663 - Revert r255946
14869 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14870 generation for cases where splatting a value is not useful.
14871 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14872 across a vec_duplicate and a paradoxical subreg forming a vector
14873 mode to a vec_concat.
14875 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14877 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14878 -march=armv8.3-a variants.
14879 * config/arm/t-multilib: Likewise.
14880 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14882 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14884 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14886 (cceq_ior_compare_complement): Give it a name so I can use it, and
14887 change boolean_or_operator predicate to boolean_operator so it can
14888 be used to generate a crand.
14889 (eqne): New code iterator.
14890 (bd/bd_neg): New code_attrs.
14891 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14892 a single define_insn.
14893 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14894 decrement (bdnzt/bdnzf/bdzt/bdzf).
14895 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14896 with the new names of the branch decrement patterns, and added the
14897 names of the branch decrement conditional patterns.
14899 2018-01-08 Richard Biener <rguenther@suse.de>
14901 PR tree-optimization/83563
14902 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14905 2018-01-08 Richard Biener <rguenther@suse.de>
14907 PR middle-end/83713
14908 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14910 2018-01-08 Richard Biener <rguenther@suse.de>
14912 PR tree-optimization/83685
14913 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14914 references to abnormals.
14916 2018-01-08 Richard Biener <rguenther@suse.de>
14919 * dwarf2out.c (output_indirect_strings): Handle empty
14920 skeleton_debug_str_hash.
14921 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14923 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14925 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14926 (emit_store_direct): Likewise.
14927 (arc_trampoline_adjust_address): Likewise.
14928 (arc_asm_trampoline_template): New function.
14929 (arc_initialize_trampoline): Use asm_trampoline_template.
14930 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14931 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14932 * config/arc/arc.md (flush_icache): Delete pattern.
14934 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14936 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14937 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14940 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14943 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14944 by not USED_FOR_TARGET.
14945 (make_pass_resolve_sw_modes): Likewise.
14947 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14949 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14952 2018-01-08 Richard Biener <rguenther@suse.de>
14954 PR middle-end/83580
14955 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14957 2018-01-08 Richard Biener <rguenther@suse.de>
14959 PR middle-end/83517
14960 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14962 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14964 PR middle-end/81897
14965 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14966 basic blocks with a small number of successors.
14967 (convert_control_dep_chain_into_preds): Improve handling of
14969 (dump_predicates): Split apart into...
14970 (dump_pred_chain): ...here...
14971 (dump_pred_info): ...and here.
14972 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14973 (can_chain_union_be_invalidated_p): Improve check for invalidation
14975 (uninit_uses_cannot_happen): Avoid unnecessary if
14976 convert_control_dep_chain_into_preds yielded nothing.
14978 2018-01-06 Martin Sebor <msebor@redhat.com>
14980 PR tree-optimization/83640
14981 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14982 subtracting negative offset from size.
14983 (builtin_access::overlap): Adjust offset bounds of the access to fall
14984 within the size of the object if possible.
14986 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14988 PR rtl-optimization/83699
14989 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14990 extract_bit_field_as_subreg to cases in which the extracted
14991 value is also a vector.
14993 * lra-constraints.c (process_alt_operands): Test for the equivalence
14994 substitutions when detecting a possible reload cycle.
14996 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14999 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15000 by default if flag_selective_schedling{,2}. Formatting fixes.
15002 PR rtl-optimization/83682
15003 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15004 if it has non-VECTOR_MODE element mode.
15005 (vec_duplicate_p): Likewise.
15007 PR middle-end/83694
15008 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15009 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15011 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15014 * config/i386/i386-builtin.def
15015 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15016 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15017 Require also OPTION_MASK_ISA_AVX512F in addition to
15018 OPTION_MASK_ISA_GFNI.
15019 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15020 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15021 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15022 to OPTION_MASK_ISA_GFNI.
15023 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15024 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15025 OPTION_MASK_ISA_AVX512BW.
15026 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15027 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15028 addition to OPTION_MASK_ISA_GFNI.
15029 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15030 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15031 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15032 to OPTION_MASK_ISA_GFNI.
15033 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15034 a requirement for all ISAs rather than any of them with a few
15036 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15038 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15039 bitmasks to be enabled with 3 exceptions, instead of requiring any
15040 enabled ISA with lots of exceptions.
15041 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15042 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15043 Change avx512bw in isa attribute to avx512f.
15044 * config/i386/sgxintrin.h: Add license boilerplate.
15045 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15046 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15047 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15048 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15050 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15051 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15052 temporarily sse2 rather than sse if not enabled already.
15055 * config/i386/sse.md (VI248_VLBW): Rename to ...
15056 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15057 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15058 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15059 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15060 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15061 mode iterator instead of VI248_VLBW.
15063 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15065 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15066 (record_modified): Skip clobbers; add debug output.
15067 (param_change_prob): Use sreal frequencies.
15069 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15071 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15072 punt for user-aligned variables.
15074 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15076 * tree-chrec.c (chrec_contains_symbols): Return true for
15079 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15082 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15083 of (x|y) == x for BICS pattern.
15085 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15087 PR tree-optimization/83605
15088 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15089 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15092 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15094 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15095 * config/epiphany/rtems.h: New file.
15097 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15098 Uros Bizjak <ubizjak@gmail.com>
15101 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15102 QIreg_operand instead of register_operand predicate.
15103 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15104 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15105 comments instead of -fmitigate[-_]rop.
15107 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15110 * cgraphunit.c (symbol_table::compile): Switch to text_section
15111 before calling assembly_start debug hook.
15112 * run-rtl-passes.c (run_rtl_passes): Likewise.
15115 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15117 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15118 range_int_cst_p rather than !symbolic_range_p before calling
15119 extract_range_from_multiplicative_op_1.
15121 2018-01-04 Jeff Law <law@redhat.com>
15123 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15124 redundant test in assertion.
15126 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15128 * doc/rtl.texi: Document machine_mode wrapper classes.
15130 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15132 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15133 using tree_to_uhwi.
15135 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15137 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15138 the VEC_PERM_EXPR fold to fail.
15140 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15143 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15144 to switched_sections.
15146 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15149 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15150 test for d.testing.
15152 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15155 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15156 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15158 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15161 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15162 is BLKmode and bitpos not zero or mode change is needed.
15164 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15167 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15170 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15173 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15174 instead of MULT rtx. Update all corresponding splitters.
15175 (*saddl_se): Ditto.
15176 (*ssub<modesuffix>): Ditto.
15177 (*ssubl_se): Ditto.
15178 (*cmp_sadd_di): Update split patterns.
15179 (*cmp_sadd_si): Ditto.
15180 (*cmp_sadd_sidi): Ditto.
15181 (*cmp_ssub_di): Ditto.
15182 (*cmp_ssub_si): Ditto.
15183 (*cmp_ssub_sidi): Ditto.
15184 * config/alpha/predicates.md (const23_operand): New predicate.
15185 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15186 Look for ASHIFT, not MULT inner operand.
15187 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15189 2018-01-04 Martin Liska <mliska@suse.cz>
15191 PR gcov-profile/83669
15192 * gcov.c (output_intermediate_file): Add version to intermediate
15194 * doc/gcov.texi: Document new field 'version' in intermediate
15195 file format. Fix location of '-k' option of gcov command.
15197 2018-01-04 Martin Liska <mliska@suse.cz>
15200 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15202 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15204 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15206 2018-01-03 Martin Sebor <msebor@redhat.com>
15208 PR tree-optimization/83655
15209 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15210 checking calls with invalid arguments.
15212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15214 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15215 (vectorizable_mask_load_store): Delete.
15216 (vectorizable_call): Return false for masked loads and stores.
15217 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15218 instead of gimple_assign_rhs1.
15219 (vectorizable_load): Handle IFN_MASK_LOAD.
15220 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15224 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15226 (vectorizable_mask_load_store): ...here.
15227 (vectorizable_load): ...and here.
15229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15231 * tree-vect-stmts.c (vect_build_all_ones_mask)
15232 (vect_build_zero_merge_argument): New functions, split out from...
15233 (vectorizable_load): ...here.
15235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15237 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15239 (vectorizable_mask_load_store): ...here.
15240 (vectorizable_store): ...and here.
15242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15244 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15246 (vectorizable_mask_load_store): ...here.
15248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15250 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15251 (vect_model_store_cost): Take a vec_load_store_type instead of a
15253 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15254 (vect_model_store_cost): Take a vec_load_store_type instead of a
15256 (vectorizable_mask_load_store): Update accordingly.
15257 (vectorizable_store): Likewise.
15258 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15262 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15263 IFN_MASK_LOAD calls here rather than...
15264 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15267 Alan Hayward <alan.hayward@arm.com>
15268 David Sherwood <david.sherwood@arm.com>
15270 * expmed.c (extract_bit_field_1): For vector extracts,
15271 fall back to extract_bit_field_as_subreg if vec_extract
15274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15275 Alan Hayward <alan.hayward@arm.com>
15276 David Sherwood <david.sherwood@arm.com>
15278 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15279 they are variable or constant sized.
15280 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15281 slots for constant-sized data.
15283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15284 Alan Hayward <alan.hayward@arm.com>
15285 David Sherwood <david.sherwood@arm.com>
15287 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15288 handling COND_EXPRs with boolean comparisons, try to find a better
15289 basis for the mask type than the boolean itself.
15291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15293 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15294 is calculated and how it can be overridden.
15295 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15296 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15298 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15302 Alan Hayward <alan.hayward@arm.com>
15303 David Sherwood <david.sherwood@arm.com>
15305 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15306 Remove the mode argument.
15307 (aarch64_simd_valid_immediate): Remove the mode and inverse
15309 * config/aarch64/iterators.md (bitsize): New iterator.
15310 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15311 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15312 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15313 aarch64_simd_valid_immediate.
15314 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15315 (aarch64_reg_or_bic_imm): Likewise.
15316 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15317 with an insn_type enum and msl with a modifier_type enum.
15318 Replace element_width with a scalar_mode. Change the shift
15319 to unsigned int. Add constructors for scalar_float_mode and
15320 scalar_int_mode elements.
15321 (aarch64_vect_float_const_representable_p): Delete.
15322 (aarch64_can_const_movi_rtx_p)
15323 (aarch64_simd_scalar_immediate_valid_for_move)
15324 (aarch64_simd_make_constant): Update call to
15325 aarch64_simd_valid_immediate.
15326 (aarch64_advsimd_valid_immediate_hs): New function.
15327 (aarch64_advsimd_valid_immediate): Likewise.
15328 (aarch64_simd_valid_immediate): Remove mode and inverse
15329 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15330 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15331 and aarch64_float_const_representable_p on the result.
15332 (aarch64_output_simd_mov_immediate): Remove mode argument.
15333 Update call to aarch64_simd_valid_immediate and use of
15334 simd_immediate_info.
15335 (aarch64_output_scalar_simd_mov_immediate): Update call
15338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15339 Alan Hayward <alan.hayward@arm.com>
15340 David Sherwood <david.sherwood@arm.com>
15342 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15343 (mode_nunits): Likewise CONST_MODE_NUNITS.
15344 * machmode.def (ADJUST_NUNITS): Document.
15345 * genmodes.c (mode_data::need_nunits_adj): New field.
15346 (blank_mode): Update accordingly.
15347 (adj_nunits): New variable.
15348 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15350 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15351 listed in adj_nunits.
15352 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15353 listed in adj_nunits. Don't emit case statements for such modes.
15354 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15355 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15356 nothing if adj_nunits is nonnull.
15357 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15358 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15359 (emit_mode_fbit): Update use of print_maybe_const_decl.
15360 (emit_move_size): Likewise. Treat the array as non-const
15362 (emit_mode_adjustments): Handle adj_nunits.
15364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15366 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15367 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15368 (VECTOR_MODES): Use it.
15369 (make_vector_modes): Take the prefix as an argument.
15371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15372 Alan Hayward <alan.hayward@arm.com>
15373 David Sherwood <david.sherwood@arm.com>
15375 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15376 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15377 for MODE_VECTOR_BOOL.
15378 * machmode.def (VECTOR_BOOL_MODE): Document.
15379 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15380 (make_vector_bool_mode): New function.
15381 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15383 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15384 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15386 * stor-layout.c (int_mode_for_mode): Likewise.
15387 * tree.c (build_vector_type_for_mode): Likewise.
15388 * varasm.c (output_constant_pool_2): Likewise.
15389 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15390 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15391 for MODE_VECTOR_BOOL.
15392 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15393 of mode class checks.
15394 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15395 instead of a list of mode class checks.
15396 (expand_vector_scalar_condition): Likewise.
15397 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15400 Alan Hayward <alan.hayward@arm.com>
15401 David Sherwood <david.sherwood@arm.com>
15403 * machmode.h (mode_size): Change from unsigned short to
15405 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15406 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15407 or if measurement_type is not polynomial.
15408 (fixed_size_mode::includes_p): Check for constant-sized modes.
15409 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15410 return a poly_uint16 rather than an unsigned short.
15411 (emit_mode_size): Change the type of mode_size from unsigned short
15412 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15413 (emit_mode_adjustments): Cope with polynomial vector sizes.
15414 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15416 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15418 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15419 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15420 * caller-save.c (setup_save_areas): Likewise.
15421 (replace_reg_with_saved_mem): Likewise.
15422 * calls.c (emit_library_call_value_1): Likewise.
15423 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15424 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15425 (gen_lowpart_for_combine): Likewise.
15426 * convert.c (convert_to_integer_1): Likewise.
15427 * cse.c (equiv_constant, cse_insn): Likewise.
15428 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15429 (cselib_subst_to_values): Likewise.
15430 * dce.c (word_dce_process_block): Likewise.
15431 * df-problems.c (df_word_lr_mark_ref): Likewise.
15432 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15433 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15434 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15435 (rtl_for_decl_location): Likewise.
15436 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15437 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15438 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15439 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15440 (expand_expr_real_1): Likewise.
15441 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15442 (pad_below): Likewise.
15443 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15444 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15445 * ira.c (get_subreg_tracking_sizes): Likewise.
15446 * ira-build.c (ira_create_allocno_objects): Likewise.
15447 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15448 (ira_sort_regnos_for_alter_reg): Likewise.
15449 * ira-costs.c (record_operand_costs): Likewise.
15450 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15451 (resolve_simple_move): Likewise.
15452 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15453 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15454 (lra_constraints): Likewise.
15455 (CONST_POOL_OK_P): Reject variable-sized modes.
15456 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15457 (add_pseudo_to_slot, lra_spill): Likewise.
15458 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15459 * optabs-query.c (get_best_extraction_insn): Likewise.
15460 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15461 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15462 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15463 * recog.c (offsettable_address_addr_space_p): Likewise.
15464 * regcprop.c (maybe_mode_change): Likewise.
15465 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15466 * regrename.c (build_def_use): Likewise.
15467 * regstat.c (dump_reg_info): Likewise.
15468 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15469 (find_reloads, find_reloads_subreg_address): Likewise.
15470 * reload1.c (eliminate_regs_1): Likewise.
15471 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15472 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15473 (simplify_binary_operation_1, simplify_subreg): Likewise.
15474 * targhooks.c (default_function_arg_padding): Likewise.
15475 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15476 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15477 (verify_gimple_assign_ternary): Likewise.
15478 * tree-inline.c (estimate_move_cost): Likewise.
15479 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15480 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15481 (get_address_cost_ainc): Likewise.
15482 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15483 (vect_supportable_dr_alignment): Likewise.
15484 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15485 (vectorizable_reduction): Likewise.
15486 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15487 (vectorizable_operation, vectorizable_load): Likewise.
15488 * tree.c (build_same_sized_truth_vector_type): Likewise.
15489 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15490 * var-tracking.c (emit_note_insn_var_location): Likewise.
15491 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15492 (ADDR_VEC_ALIGN): Likewise.
15494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15495 Alan Hayward <alan.hayward@arm.com>
15496 David Sherwood <david.sherwood@arm.com>
15498 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15500 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15501 or if measurement_type is polynomial.
15502 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15503 * combine.c (make_extraction): Likewise.
15504 * dse.c (find_shift_sequence): Likewise.
15505 * dwarf2out.c (mem_loc_descriptor): Likewise.
15506 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15507 (extract_bit_field, extract_low_bits): Likewise.
15508 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15509 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15510 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15511 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15512 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15513 * reload.c (find_reloads): Likewise.
15514 * reload1.c (alter_reg): Likewise.
15515 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15516 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15517 * tree-if-conv.c (predicate_mem_writes): Likewise.
15518 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15519 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15520 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15521 * valtrack.c (dead_debug_insert_temp): Likewise.
15522 * varasm.c (mergeable_constant_section): Likewise.
15523 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15526 Alan Hayward <alan.hayward@arm.com>
15527 David Sherwood <david.sherwood@arm.com>
15529 * expr.c (expand_assignment): Cope with polynomial mode sizes
15530 when assigning to a CONCAT.
15532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15533 Alan Hayward <alan.hayward@arm.com>
15534 David Sherwood <david.sherwood@arm.com>
15536 * machmode.h (mode_precision): Change from unsigned short to
15538 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15540 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15541 or if measurement_type is not polynomial.
15542 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15543 in which the mode is already known to be a scalar_int_mode.
15544 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15545 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15547 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15548 for GET_MODE_PRECISION.
15549 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15550 for GET_MODE_PRECISION.
15551 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15553 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15554 (expand_field_assignment, make_extraction): Likewise.
15555 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15556 (get_last_value): Likewise.
15557 * convert.c (convert_to_integer_1): Likewise.
15558 * cse.c (cse_insn): Likewise.
15559 * expr.c (expand_expr_real_1): Likewise.
15560 * lra-constraints.c (simplify_operand_subreg): Likewise.
15561 * optabs-query.c (can_atomic_load_p): Likewise.
15562 * optabs.c (expand_atomic_load): Likewise.
15563 (expand_atomic_store): Likewise.
15564 * ree.c (combine_reaching_defs): Likewise.
15565 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15566 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15567 * tree.h (type_has_mode_precision_p): Likewise.
15568 * ubsan.c (instrument_si_overflow): Likewise.
15570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15571 Alan Hayward <alan.hayward@arm.com>
15572 David Sherwood <david.sherwood@arm.com>
15574 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15575 polynomial numbers of units.
15576 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15577 (valid_vector_subparts_p): New function.
15578 (build_vector_type): Remove temporary shim and take the number
15579 of units as a poly_uint64 rather than an int.
15580 (build_opaque_vector_type): Take the number of units as a
15581 poly_uint64 rather than an int.
15582 * tree.c (build_vector_from_ctor): Handle polynomial
15583 TYPE_VECTOR_SUBPARTS.
15584 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15585 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15586 (build_vector_from_val): If the number of units is variable,
15587 use build_vec_duplicate_cst for constant operands and
15588 VEC_DUPLICATE_EXPR otherwise.
15589 (make_vector_type): Remove temporary is_constant ().
15590 (build_vector_type, build_opaque_vector_type): Take the number of
15591 units as a poly_uint64 rather than an int.
15592 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15594 * cfgexpand.c (expand_debug_expr): Likewise.
15595 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15596 (store_constructor, expand_expr_real_1): Likewise.
15597 (const_scalar_mask_from_tree): Likewise.
15598 * fold-const-call.c (fold_const_reduction): Likewise.
15599 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15600 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15601 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15602 (fold_relational_const): Likewise.
15603 (native_interpret_vector): Likewise. Change the size from an
15604 int to an unsigned int.
15605 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15606 TYPE_VECTOR_SUBPARTS.
15607 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15608 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15609 duplicating a non-constant operand into a variable-length vector.
15610 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15611 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15612 * ipa-icf.c (sem_variable::equals): Likewise.
15613 * match.pd: Likewise.
15614 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15615 * print-tree.c (print_node): Likewise.
15616 * stor-layout.c (layout_type): Likewise.
15617 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15618 * tree-cfg.c (verify_gimple_comparison): Likewise.
15619 (verify_gimple_assign_binary): Likewise.
15620 (verify_gimple_assign_ternary): Likewise.
15621 (verify_gimple_assign_single): Likewise.
15622 * tree-pretty-print.c (dump_generic_node): Likewise.
15623 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15624 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15625 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15626 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15627 (vect_shift_permute_load_chain): Likewise.
15628 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15629 (expand_vector_condition, optimize_vector_constructor): Likewise.
15630 (lower_vec_perm, get_compute_type): Likewise.
15631 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15632 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15633 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15634 (vect_recog_mask_conversion_pattern): Likewise.
15635 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15636 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15637 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15638 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15639 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15640 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15641 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15642 (supportable_widening_operation): Likewise.
15643 (supportable_narrowing_operation): Likewise.
15644 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15646 * varasm.c (output_constant): Likewise.
15648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15649 Alan Hayward <alan.hayward@arm.com>
15650 David Sherwood <david.sherwood@arm.com>
15652 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15653 so that both the length == 3 and length != 3 cases set up their
15654 own permute vectors. Add comments explaining why we know the
15655 number of elements is constant.
15656 (vect_permute_load_chain): Likewise.
15658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15659 Alan Hayward <alan.hayward@arm.com>
15660 David Sherwood <david.sherwood@arm.com>
15662 * machmode.h (mode_nunits): Change from unsigned char to
15664 (ONLY_FIXED_SIZE_MODES): New macro.
15665 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15666 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15667 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15669 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15670 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15671 or if measurement_type is not polynomial.
15672 * genmodes.c (ZERO_COEFFS): New macro.
15673 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15675 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15676 Use ZERO_COEFFS when emitting initializers.
15677 * data-streamer.h (bp_pack_poly_value): New function.
15678 (bp_unpack_poly_value): Likewise.
15679 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15680 for GET_MODE_NUNITS.
15681 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15682 for GET_MODE_NUNITS.
15683 * tree.c (make_vector_type): Remove temporary shim and make
15684 the real function take the number of units as a poly_uint64
15685 rather than an int.
15686 (build_vector_type_for_mode): Handle polynomial nunits.
15687 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15688 * emit-rtl.c (const_vec_series_p_1): Likewise.
15689 (gen_rtx_CONST_VECTOR): Likewise.
15690 * fold-const.c (test_vec_duplicate_folding): Likewise.
15691 * genrecog.c (validate_pattern): Likewise.
15692 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15693 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15694 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15695 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15696 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15697 * rtlanal.c (subreg_get_info): Likewise.
15698 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15699 (vect_grouped_load_supported): Likewise.
15700 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15701 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15702 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15703 (simplify_const_unary_operation, simplify_binary_operation_1)
15704 (simplify_const_binary_operation, simplify_ternary_operation)
15705 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15706 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15707 instead of CONST_VECTOR_NUNITS.
15708 * varasm.c (output_constant_pool_2): Likewise.
15709 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15710 explicit-encoded elements in the XVEC for variable-length vectors.
15712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15714 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15717 Alan Hayward <alan.hayward@arm.com>
15718 David Sherwood <david.sherwood@arm.com>
15720 * coretypes.h (fixed_size_mode): Declare.
15721 (fixed_size_mode_pod): New typedef.
15722 * builtins.h (target_builtins::x_apply_args_mode)
15723 (target_builtins::x_apply_result_mode): Change type to
15724 fixed_size_mode_pod.
15725 * builtins.c (apply_args_size, apply_result_size, result_vector)
15726 (expand_builtin_apply_args_1, expand_builtin_apply)
15727 (expand_builtin_return): Update accordingly.
15729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15731 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15732 * cselib.c (cselib_hash_rtx): Likewise.
15733 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15734 CONST_VECTOR encoding.
15736 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15737 Jeff Law <law@redhat.com>
15740 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15741 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15742 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15743 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15746 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15747 explicitly probe *sp in a noreturn function if there were any callee
15748 register saves or frame pointer is needed.
15750 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15753 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15754 BLKmode for ternary, binary or unary expressions.
15757 * var-tracking.c (delete_vta_debug_insn): New inline function.
15758 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15759 insns from get_insns () to NULL instead of each bb separately.
15760 Use delete_vta_debug_insn. No longer static.
15761 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15762 delete_vta_debug_insns callers.
15763 * rtl.h (delete_vta_debug_insns): Declare.
15764 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15765 instead of variable_tracking_main.
15767 2018-01-03 Martin Sebor <msebor@redhat.com>
15769 PR tree-optimization/83603
15770 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15771 arguments past the endof the argument list in functions declared
15772 without a prototype.
15773 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15774 Avoid checking when arguments are null.
15776 2018-01-03 Martin Sebor <msebor@redhat.com>
15779 * doc/extend.texi (attribute const): Fix a typo.
15780 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15781 issuing -Wsuggest-attribute for void functions.
15783 2018-01-03 Martin Sebor <msebor@redhat.com>
15785 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15786 offset_int::from instead of wide_int::to_shwi.
15787 (maybe_diag_overlap): Remove assertion.
15788 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15789 * gimple-ssa-sprintf.c (format_directive): Same.
15790 (parse_directive): Same.
15791 (sprintf_dom_walker::compute_format_length): Same.
15792 (try_substitute_return_value): Same.
15794 2018-01-03 Jeff Law <law@redhat.com>
15796 PR middle-end/83654
15797 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15798 non-constant residual for zero at runtime and avoid probing in
15799 that case. Reorganize code for trailing problem to mirror handling
15802 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15804 PR tree-optimization/83501
15805 * tree-ssa-strlen.c (get_string_cst): New.
15806 (handle_char_store): Call get_string_cst.
15808 2018-01-03 Martin Liska <mliska@suse.cz>
15810 PR tree-optimization/83593
15811 * tree-ssa-strlen.c: Include tree-cfg.h.
15812 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15813 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15814 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15816 (strlen_dom_walker::before_dom_children): Call
15817 gimple_purge_dead_eh_edges. Dump tranformation with details
15819 (strlen_dom_walker::before_dom_children): Update call by adding
15820 new argument cleanup_eh.
15821 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15823 2018-01-03 Martin Liska <mliska@suse.cz>
15826 * cif-code.def (VARIADIC_THUNK): New enum value.
15827 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15830 2018-01-03 Jan Beulich <jbeulich@suse.com>
15832 * sse.md (mov<mode>_internal): Tighten condition for when to use
15833 vmovdqu<ssescalarsize> for TI and OI modes.
15835 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15837 Update copyright years.
15839 2018-01-03 Martin Liska <mliska@suse.cz>
15842 * ipa-visibility.c (function_and_variable_visibility): Skip
15843 functions with noipa attribure.
15845 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15847 * gcc.c (process_command): Update copyright notice dates.
15848 * gcov-dump.c (print_version): Ditto.
15849 * gcov.c (print_version): Ditto.
15850 * gcov-tool.c (print_version): Ditto.
15851 * gengtype.c (create_file): Ditto.
15852 * doc/cpp.texi: Bump @copying's copyright year.
15853 * doc/cppinternals.texi: Ditto.
15854 * doc/gcc.texi: Ditto.
15855 * doc/gccint.texi: Ditto.
15856 * doc/gcov.texi: Ditto.
15857 * doc/install.texi: Ditto.
15858 * doc/invoke.texi: Ditto.
15860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15862 * vector-builder.h (vector_builder::m_full_nelts): Change from
15863 unsigned int to poly_uint64.
15864 (vector_builder::full_nelts): Update prototype accordingly.
15865 (vector_builder::new_vector): Likewise.
15866 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15867 (vector_builder::operator ==): Likewise.
15868 (vector_builder::finalize): Likewise.
15869 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15870 Take the number of elements as a poly_uint64 rather than an
15872 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15873 from unsigned int to poly_uint64.
15874 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15875 (vec_perm_indices::new_vector): Likewise.
15876 (vec_perm_indices::length): Likewise.
15877 (vec_perm_indices::nelts_per_input): Likewise.
15878 (vec_perm_indices::input_nelts): Likewise.
15879 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15880 number of elements per input as a poly_uint64 rather than an
15881 unsigned int. Use the original encoding for variable-length
15882 vectors, rather than clamping each individual element.
15883 For the second and subsequent elements in each pattern,
15884 clamp the step and base before clamping their sum.
15885 (vec_perm_indices::series_p): Handle polynomial element counts.
15886 (vec_perm_indices::all_in_range_p): Likewise.
15887 (vec_perm_indices_to_tree): Likewise.
15888 (vec_perm_indices_to_rtx): Likewise.
15889 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15890 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15891 (tree_vector_builder::new_binary_operation): Handle polynomial
15892 element counts. Return false if we need to know the number
15893 of elements at compile time.
15894 * fold-const.c (fold_vec_perm): Punt if the number of elements
15895 isn't known at compile time.
15897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15899 * vec-perm-indices.h (vec_perm_builder): Change element type
15900 from HOST_WIDE_INT to poly_int64.
15901 (vec_perm_indices::element_type): Update accordingly.
15902 (vec_perm_indices::clamp): Handle polynomial element_types.
15903 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15904 (vec_perm_indices::all_in_range_p): Likewise.
15905 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15907 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15908 polynomial vec_perm_indices element types.
15909 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15910 * fold-const.c (fold_vec_perm): Likewise.
15911 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15912 * tree-vect-generic.c (lower_vec_perm): Likewise.
15913 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15914 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15915 element type to HOST_WIDE_INT.
15917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15918 Alan Hayward <alan.hayward@arm.com>
15919 David Sherwood <david.sherwood@arm.com>
15921 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15922 rather than an int. Use plus_constant.
15923 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15924 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15927 Alan Hayward <alan.hayward@arm.com>
15928 David Sherwood <david.sherwood@arm.com>
15930 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15931 a HOST_WIDE_INT to a poly_int64.
15933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15934 Alan Hayward <alan.hayward@arm.com>
15935 David Sherwood <david.sherwood@arm.com>
15937 * calls.c (load_register_parameters): Cope with polynomial
15938 mode sizes. Require a constant size for BLKmode parameters
15939 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15940 forces a parameter to be padded at the lsb end in order to
15941 fill a complete number of words, require the parameter size
15942 to be ordered wrt UNITS_PER_WORD.
15944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15945 Alan Hayward <alan.hayward@arm.com>
15946 David Sherwood <david.sherwood@arm.com>
15948 * reload1.c (spill_stack_slot_width): Change element type
15949 from unsigned int to poly_uint64_pod.
15950 (alter_reg): Treat mode sizes as polynomial.
15952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15953 Alan Hayward <alan.hayward@arm.com>
15954 David Sherwood <david.sherwood@arm.com>
15956 * reload.c (complex_word_subreg_p): New function.
15957 (reload_inner_reg_of_subreg, push_reload): Use it.
15959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15960 Alan Hayward <alan.hayward@arm.com>
15961 David Sherwood <david.sherwood@arm.com>
15963 * lra-constraints.c (process_alt_operands): Reject matched
15964 operands whose sizes aren't ordered.
15965 (match_reload): Refer to this check here.
15967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15968 Alan Hayward <alan.hayward@arm.com>
15969 David Sherwood <david.sherwood@arm.com>
15971 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15972 that the mode size is in the set {1, 2, 4, 8, 16}.
15974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15975 Alan Hayward <alan.hayward@arm.com>
15976 David Sherwood <david.sherwood@arm.com>
15978 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15979 Use plus_constant instead of gen_rtx_PLUS.
15981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15982 Alan Hayward <alan.hayward@arm.com>
15983 David Sherwood <david.sherwood@arm.com>
15985 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15986 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15987 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15988 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15989 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15990 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15991 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15992 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15993 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15994 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15996 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15997 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15998 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15999 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16000 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16001 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16002 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16003 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16004 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16005 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16007 * expr.c (emit_move_resolve_push): Treat the input and result
16008 of PUSH_ROUNDING as a poly_int64.
16009 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16010 (emit_push_insn): Likewise.
16011 * lra-eliminations.c (mark_not_eliminable): Likewise.
16012 * recog.c (push_operand): Likewise.
16013 * reload1.c (elimination_effects): Likewise.
16014 * rtlanal.c (nonzero_bits1): Likewise.
16015 * calls.c (store_one_arg): Likewise. Require the padding to be
16016 known at compile time.
16018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16019 Alan Hayward <alan.hayward@arm.com>
16020 David Sherwood <david.sherwood@arm.com>
16022 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16023 Use plus_constant instead of gen_rtx_PLUS.
16025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16026 Alan Hayward <alan.hayward@arm.com>
16027 David Sherwood <david.sherwood@arm.com>
16029 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16030 rather than an int.
16032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16033 Alan Hayward <alan.hayward@arm.com>
16034 David Sherwood <david.sherwood@arm.com>
16036 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16037 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16038 via stack temporaries. Treat the mode size as polynomial too.
16040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16041 Alan Hayward <alan.hayward@arm.com>
16042 David Sherwood <david.sherwood@arm.com>
16044 * expr.c (expand_expr_real_2): When handling conversions involving
16045 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16046 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16047 as a poly_uint64 too.
16049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16050 Alan Hayward <alan.hayward@arm.com>
16051 David Sherwood <david.sherwood@arm.com>
16053 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16056 Alan Hayward <alan.hayward@arm.com>
16057 David Sherwood <david.sherwood@arm.com>
16059 * combine.c (can_change_dest_mode): Handle polynomial
16060 REGMODE_NATURAL_SIZE.
16061 * expmed.c (store_bit_field_1): Likewise.
16062 * expr.c (store_constructor): Likewise.
16063 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16064 and polynomial REGMODE_NATURAL_SIZE.
16065 (gen_lowpart_common): Likewise.
16066 * reginfo.c (record_subregs_of_mode): Likewise.
16067 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16074 numbers of elements.
16076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16077 Alan Hayward <alan.hayward@arm.com>
16078 David Sherwood <david.sherwood@arm.com>
16080 * match.pd: Cope with polynomial numbers of vector elements.
16082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16083 Alan Hayward <alan.hayward@arm.com>
16084 David Sherwood <david.sherwood@arm.com>
16086 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16087 in a POINTER_PLUS_EXPR.
16089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16090 Alan Hayward <alan.hayward@arm.com>
16091 David Sherwood <david.sherwood@arm.com>
16093 * omp-simd-clone.c (simd_clone_subparts): New function.
16094 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16095 (ipa_simd_modify_function_body): Likewise.
16097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16098 Alan Hayward <alan.hayward@arm.com>
16099 David Sherwood <david.sherwood@arm.com>
16101 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16102 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16103 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16104 (expand_vector_condition, vector_element): Likewise.
16105 (subparts_gt): New function.
16106 (get_compute_type): Use subparts_gt.
16107 (count_type_subparts): Delete.
16108 (expand_vector_operations_1): Use subparts_gt instead of
16109 count_type_subparts.
16111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16112 Alan Hayward <alan.hayward@arm.com>
16113 David Sherwood <david.sherwood@arm.com>
16115 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16116 (vect_compile_time_alias): ...this new function. Do the calculation
16117 on poly_ints rather than trees.
16118 (vect_prune_runtime_alias_test_list): Update call accordingly.
16120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16121 Alan Hayward <alan.hayward@arm.com>
16122 David Sherwood <david.sherwood@arm.com>
16124 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16126 (vect_schedule_slp_instance): Likewise.
16128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16129 Alan Hayward <alan.hayward@arm.com>
16130 David Sherwood <david.sherwood@arm.com>
16132 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16133 constant and extern definitions for variable-length vectors.
16134 (vect_get_constant_vectors): Note that the number of units
16135 is known to be constant.
16137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16138 Alan Hayward <alan.hayward@arm.com>
16139 David Sherwood <david.sherwood@arm.com>
16141 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16142 of units as polynomial. Choose between WIDE and NARROW based
16145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16146 Alan Hayward <alan.hayward@arm.com>
16147 David Sherwood <david.sherwood@arm.com>
16149 * tree-vect-stmts.c (simd_clone_subparts): New function.
16150 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16153 Alan Hayward <alan.hayward@arm.com>
16154 David Sherwood <david.sherwood@arm.com>
16156 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16157 vectors as polynomial. Use build_index_vector for
16158 IFN_GOMP_SIMD_LANE.
16160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16161 Alan Hayward <alan.hayward@arm.com>
16162 David Sherwood <david.sherwood@arm.com>
16164 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16165 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16166 for variable-length vectors.
16167 (vectorizable_mask_load_store): Treat the number of units as
16168 polynomial, asserting that it is constant if the condition has
16169 already been enforced.
16170 (vectorizable_store, vectorizable_load): Likewise.
16172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16173 Alan Hayward <alan.hayward@arm.com>
16174 David Sherwood <david.sherwood@arm.com>
16176 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16177 of units as polynomial. Punt if we can't tell at compile time
16178 which vector contains the final result.
16180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16181 Alan Hayward <alan.hayward@arm.com>
16182 David Sherwood <david.sherwood@arm.com>
16184 * tree-vect-loop.c (vectorizable_induction): Treat the number
16185 of units as polynomial. Punt on SLP inductions. Use an integer
16186 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16187 cast of such a series for variable-length floating-point
16190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16191 Alan Hayward <alan.hayward@arm.com>
16192 David Sherwood <david.sherwood@arm.com>
16194 * tree.h (build_index_vector): Declare.
16195 * tree.c (build_index_vector): New function.
16196 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16197 of units as polynomial, forcibly converting it to a constant if
16198 vectorizable_reduction has already enforced the condition.
16199 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16200 to create a {1,2,3,...} vector.
16201 (vectorizable_reduction): Treat the number of units as polynomial.
16202 Choose vectype_in based on the largest scalar element size rather
16203 than the smallest number of units. Enforce the restrictions
16206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16207 Alan Hayward <alan.hayward@arm.com>
16208 David Sherwood <david.sherwood@arm.com>
16210 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16211 number of units as polynomial.
16213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16214 Alan Hayward <alan.hayward@arm.com>
16215 David Sherwood <david.sherwood@arm.com>
16217 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16218 * target.def (autovectorize_vector_sizes): Return the vector sizes
16219 by pointer, using vector_sizes rather than a bitmask.
16220 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16221 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16222 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16224 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16225 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16226 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16227 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16228 * omp-general.c (omp_max_vf): Likewise.
16229 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16230 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16231 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16232 * tree-vect-slp.c (vect_slp_bb): Likewise.
16233 * doc/tm.texi: Regenerate.
16234 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16236 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16237 the vector size as a poly_uint64 rather than an unsigned int.
16238 (current_vector_size): Change from an unsigned int to a poly_uint64.
16239 (get_vectype_for_scalar_type): Update accordingly.
16240 * tree.h (build_truth_vector_type): Take the size and number of
16241 units as a poly_uint64 rather than an unsigned int.
16242 (build_vector_type): Add a temporary overload that takes
16243 the number of units as a poly_uint64 rather than an unsigned int.
16244 * tree.c (make_vector_type): Likewise.
16245 (build_truth_vector_type): Take the number of units as a poly_uint64
16246 rather than an unsigned int.
16248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16249 Alan Hayward <alan.hayward@arm.com>
16250 David Sherwood <david.sherwood@arm.com>
16252 * target.def (get_mask_mode): Take the number of units and length
16253 as poly_uint64s rather than unsigned ints.
16254 * targhooks.h (default_get_mask_mode): Update accordingly.
16255 * targhooks.c (default_get_mask_mode): Likewise.
16256 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16257 * doc/tm.texi: Regenerate.
16259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16260 Alan Hayward <alan.hayward@arm.com>
16261 David Sherwood <david.sherwood@arm.com>
16263 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16264 * omp-general.c (omp_max_vf): Likewise.
16265 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16266 (expand_omp_simd): Handle polynomial safelen.
16267 * omp-low.c (omplow_simd_context): Add a default constructor.
16268 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16269 (lower_rec_simd_input_clauses): Update accordingly.
16270 (lower_rec_input_clauses): Likewise.
16272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16273 Alan Hayward <alan.hayward@arm.com>
16274 David Sherwood <david.sherwood@arm.com>
16276 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16277 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16278 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16279 (vect_analyze_slp_cost): Likewise.
16280 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16281 (vect_model_load_cost): Likewise.
16283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16284 Alan Hayward <alan.hayward@arm.com>
16285 David Sherwood <david.sherwood@arm.com>
16287 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16288 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16289 from an unsigned int * to a poly_uint64_pod *.
16290 (calculate_unrolling_factor): New function.
16291 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16294 Alan Hayward <alan.hayward@arm.com>
16295 David Sherwood <david.sherwood@arm.com>
16297 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16298 from an unsigned int to a poly_uint64.
16299 (_loop_vec_info::slp_unrolling_factor): Likewise.
16300 (_loop_vec_info::vectorization_factor): Change from an int
16302 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16303 (vect_get_num_vectors): New function.
16304 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16305 (vect_get_num_copies): Use vect_get_num_vectors.
16306 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16307 to an unsigned int *.
16308 (vect_analyze_data_refs): Change min_vf from an int * to a
16310 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16311 than an unsigned HOST_WIDE_INT.
16312 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16313 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16314 to an unsigned int *.
16315 (vect_analyze_data_ref_dependences): Likewise.
16316 (vect_compute_data_ref_alignment): Handle polynomial vf.
16317 (vect_enhance_data_refs_alignment): Likewise.
16318 (vect_prune_runtime_alias_test_list): Likewise.
16319 (vect_shift_permute_load_chain): Likewise.
16320 (vect_supportable_dr_alignment): Likewise.
16321 (dependence_distance_ge_vf): Take the vectorization factor as a
16322 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16323 (vect_analyze_data_refs): Change min_vf from an int * to a
16325 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16326 vfm1 as a poly_uint64 rather than an int. Make the same change
16327 for the returned bound_scalar.
16328 (vect_gen_vector_loop_niters): Handle polynomial vf.
16329 (vect_do_peeling): Likewise. Update call to
16330 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16331 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16333 * tree-vect-loop.c (vect_determine_vectorization_factor)
16334 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16335 (vect_get_known_peeling_cost): Likewise.
16336 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16337 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16338 (vect_transform_loop): Likewise. Use the lowest possible VF when
16339 updating the upper bounds of the loop.
16340 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16341 rather than an int.
16342 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16343 polynomial unroll factors.
16344 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16345 (vect_make_slp_decision): Likewise.
16346 (vect_supported_load_permutation_p): Likewise, and polynomial
16348 (vect_analyze_slp_cost): Handle polynomial vf.
16349 (vect_slp_analyze_node_operations): Likewise.
16350 (vect_slp_analyze_bb_1): Likewise.
16351 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16352 than an unsigned HOST_WIDE_INT.
16353 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16354 (vectorizable_load): Handle polynomial vf.
16355 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16357 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16360 Alan Hayward <alan.hayward@arm.com>
16361 David Sherwood <david.sherwood@arm.com>
16363 * match.pd: Handle bit operations involving three constants
16364 and try to fold one pair.
16366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16368 * tree-vect-loop-manip.c: Include gimple-fold.h.
16369 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16370 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16371 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16372 Add a path that uses a step of VF instead of 1, but disable it
16374 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16375 and niters_no_overflow parameters. Update calls to
16376 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16377 Create a new SSA name if the latter choses to use a ste other
16378 than zero, and return it via niters_vector_mult_vf_var.
16379 * tree-vect-loop.c (vect_transform_loop): Update calls to
16380 vect_do_peeling, vect_gen_vector_loop_niters and
16381 slpeel_make_loop_iterate_ntimes.
16382 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16383 (vect_gen_vector_loop_niters): Update declarations after above changes.
16385 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16387 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16388 128-bit round to integer instructions.
16389 (ceil<mode>2): Likewise.
16390 (btrunc<mode>2): Likewise.
16391 (round<mode>2): Likewise.
16393 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16395 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16396 unaligned VSX load/store on P8/P9.
16397 (expand_block_clear): Allow the use of unaligned VSX
16398 load/store on P8/P9.
16400 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16402 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16404 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16405 swap associated with both a load and a store.
16407 2018-01-02 Andrew Waterman <andrew@sifive.com>
16409 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16410 * config/riscv/riscv.md (clear_cache): Use it.
16412 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16414 * web.c: Remove out-of-date comment.
16416 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16418 * expr.c (fixup_args_size_notes): Check that any existing
16419 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16420 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16421 (emit_single_push_insn): ...here.
16423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16425 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16426 (const_vector_encoded_nelts): New function.
16427 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16428 (const_vector_int_elt, const_vector_elt): Declare.
16429 * emit-rtl.c (const_vector_int_elt_1): New function.
16430 (const_vector_elt): Likewise.
16431 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16432 of CONST_VECTOR_ELT.
16434 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16436 * expr.c: Include rtx-vector-builder.h.
16437 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16438 directly on the tree encoding.
16439 (const_vector_from_tree): Likewise.
16440 * optabs.c: Include rtx-vector-builder.h.
16441 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16442 sequence of "u" values.
16443 * vec-perm-indices.c: Include rtx-vector-builder.h.
16444 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16445 directly on the vec_perm_indices encoding.
16447 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16449 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16450 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16451 * rtx-vector-builder.h: New file.
16452 * rtx-vector-builder.c: Likewise.
16453 * rtl.h (rtx_def::u2): Add a const_vector field.
16454 (CONST_VECTOR_NPATTERNS): New macro.
16455 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16456 (CONST_VECTOR_DUPLICATE_P): Likewise.
16457 (CONST_VECTOR_STEPPED_P): Likewise.
16458 (CONST_VECTOR_ENCODED_ELT): Likewise.
16459 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16460 (unwrap_const_vec_duplicate): Likewise.
16461 (const_vec_series_p): Check for a non-duplicated vector encoding.
16462 Say that the function only returns true for integer vectors.
16463 * emit-rtl.c: Include rtx-vector-builder.h.
16464 (gen_const_vec_duplicate_1): Delete.
16465 (gen_const_vector): Call gen_const_vec_duplicate instead of
16466 gen_const_vec_duplicate_1.
16467 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16468 (gen_const_vec_duplicate): Use rtx_vector_builder.
16469 (gen_const_vec_series): Likewise.
16470 (gen_rtx_CONST_VECTOR): Likewise.
16471 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16472 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16473 Build a new vector rather than modifying a CONST_VECTOR in-place.
16474 (handle_special_swappables): Update call accordingly.
16475 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16476 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16477 Build a new vector rather than modifying a CONST_VECTOR in-place.
16478 (handle_special_swappables): Update call accordingly.
16480 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16482 * simplify-rtx.c (simplify_const_binary_operation): Use
16483 CONST_VECTOR_ELT instead of XVECEXP.
16485 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16487 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16488 the selector elements to be different from the data elements
16489 if the selector is a VECTOR_CST.
16490 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16491 ssizetype for the selector.
16493 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16495 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16496 before testing each element individually.
16497 * tree-vect-generic.c (lower_vec_perm): Likewise.
16499 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16501 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16502 * selftest-run-tests.c (selftest::run_tests): Call it.
16503 * vector-builder.h (vector_builder::operator ==): New function.
16504 (vector_builder::operator !=): Likewise.
16505 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16506 (vec_perm_indices::all_from_input_p): New function.
16507 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16508 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16509 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16510 instead of reading the VECTOR_CST directly. Detect whether both
16511 vector inputs are the same before constructing the vec_perm_indices,
16512 and update the number of inputs argument accordingly. Use the
16513 utility functions added above. Only construct sel2 if we need to.
16515 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16517 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16518 the broadcast of the low byte.
16519 (expand_mult_highpart): Use an explicit encoding for the permutes.
16520 * optabs-query.c (can_mult_highpart_p): Likewise.
16521 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16522 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16523 (vectorizable_bswap): Likewise.
16524 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16525 explicit encoding for the power-of-2 permutes.
16526 (vect_permute_store_chain): Likewise.
16527 (vect_grouped_load_supported): Likewise.
16528 (vect_permute_load_chain): Likewise.
16530 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16532 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16533 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16534 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16535 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16536 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16537 (vect_gen_perm_mask_any): Likewise.
16539 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16541 * int-vector-builder.h: New file.
16542 * vec-perm-indices.h: Include int-vector-builder.h.
16543 (vec_perm_indices): Redefine as an int_vector_builder.
16544 (auto_vec_perm_indices): Delete.
16545 (vec_perm_builder): Redefine as a stand-alone class.
16546 (vec_perm_indices::vec_perm_indices): New function.
16547 (vec_perm_indices::clamp): Likewise.
16548 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16549 (vec_perm_indices::new_vector): New function.
16550 (vec_perm_indices::new_expanded_vector): Update for new
16551 vec_perm_indices class.
16552 (vec_perm_indices::rotate_inputs): New function.
16553 (vec_perm_indices::all_in_range_p): Operate directly on the
16554 encoded form, without computing elided elements.
16555 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16556 encoding. Update for new vec_perm_indices class.
16557 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16558 the given vec_perm_builder.
16559 (expand_vec_perm_var): Update vec_perm_builder constructor.
16560 (expand_mult_highpart): Use vec_perm_builder instead of
16561 auto_vec_perm_indices.
16562 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16563 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16564 or double series encoding as appropriate.
16565 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16566 vec_perm_indices instead of auto_vec_perm_indices.
16567 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16568 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16569 (vect_permute_store_chain): Likewise.
16570 (vect_grouped_load_supported): Likewise.
16571 (vect_permute_load_chain): Likewise.
16572 (vect_shift_permute_load_chain): Likewise.
16573 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16574 (vect_transform_slp_perm_load): Likewise.
16575 (vect_schedule_slp_instance): Likewise.
16576 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16577 (vectorizable_mask_load_store): Likewise.
16578 (vectorizable_bswap): Likewise.
16579 (vectorizable_store): Likewise.
16580 (vectorizable_load): Likewise.
16581 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16582 vec_perm_indices instead of auto_vec_perm_indices. Use
16583 tree_to_vec_perm_builder to read the vector from a tree.
16584 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16585 vec_perm_builder instead of a vec_perm_indices.
16586 (have_whole_vector_shift): Use vec_perm_builder and
16587 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16588 truncation to calc_vec_perm_mask_for_shift.
16589 (vect_create_epilog_for_reduction): Likewise.
16590 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16591 from auto_vec_perm_indices to vec_perm_indices.
16592 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16593 instead of changing individual elements.
16594 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16595 the vector in d.perm.
16596 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16597 from auto_vec_perm_indices to vec_perm_indices.
16598 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16599 instead of changing individual elements.
16600 (arm_vectorize_vec_perm_const): Use new_vector to install
16601 the vector in d.perm.
16602 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16603 Update vec_perm_builder constructor.
16604 (rs6000_expand_interleave): Likewise.
16605 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16606 (rs6000_expand_interleave): Likewise.
16608 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16610 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16611 to qimode could truncate the indices.
16612 * optabs.c (expand_vec_perm_var): Likewise.
16614 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16616 * Makefile.in (OBJS): Add vec-perm-indices.o.
16617 * vec-perm-indices.h: New file.
16618 * vec-perm-indices.c: Likewise.
16619 * target.h (vec_perm_indices): Replace with a forward class
16621 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16622 * optabs.h: Include vec-perm-indices.h.
16623 (expand_vec_perm): Delete.
16624 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16625 (expand_vec_perm_const): Declare.
16626 * target.def (vec_perm_const_ok): Replace with...
16627 (vec_perm_const): ...this new hook.
16628 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16629 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16630 * doc/tm.texi: Regenerate.
16631 * optabs.def (vec_perm_const): Delete.
16632 * doc/md.texi (vec_perm_const): Likewise.
16633 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16634 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16635 expand_vec_perm for constant permutation vectors. Assert that
16636 the mode of variable permutation vectors is the integer equivalent
16637 of the mode that is being permuted.
16638 * optabs-query.h (selector_fits_mode_p): Declare.
16639 * optabs-query.c: Include vec-perm-indices.h.
16640 (selector_fits_mode_p): New function.
16641 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16642 is defined, instead of checking whether the vec_perm_const_optab
16643 exists. Use targetm.vectorize.vec_perm_const instead of
16644 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16645 fit in the vector mode before using a variable permute.
16646 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16647 vec_perm_indices instead of an rtx.
16648 (expand_vec_perm): Replace with...
16649 (expand_vec_perm_const): ...this new function. Take the selector
16650 as a vec_perm_indices rather than an rtx. Also take the mode of
16651 the selector. Update call to shift_amt_for_vec_perm_mask.
16652 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16653 Use vec_perm_indices::new_expanded_vector to expand the original
16654 selector into bytes. Check whether the indices fit in the vector
16655 mode before using a variable permute.
16656 (expand_vec_perm_var): Make global.
16657 (expand_mult_highpart): Use expand_vec_perm_const.
16658 * fold-const.c: Includes vec-perm-indices.h.
16659 * tree-ssa-forwprop.c: Likewise.
16660 * tree-vect-data-refs.c: Likewise.
16661 * tree-vect-generic.c: Likewise.
16662 * tree-vect-loop.c: Likewise.
16663 * tree-vect-slp.c: Likewise.
16664 * tree-vect-stmts.c: Likewise.
16665 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16667 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16668 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16669 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16670 (aarch64_vectorize_vec_perm_const): ...this new function.
16671 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16672 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16673 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16674 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16675 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16676 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16677 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16679 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16680 check for NEON modes.
16681 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16682 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16683 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16684 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16686 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16687 the old VEC_PERM_CONST conditions.
16688 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16689 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16690 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16691 (ia64_vectorize_vec_perm_const_ok): Merge into...
16692 (ia64_vectorize_vec_perm_const): ...this new function.
16693 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16694 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16695 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16696 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16697 * config/mips/mips.c (mips_expand_vec_perm_const)
16698 (mips_vectorize_vec_perm_const_ok): Merge into...
16699 (mips_vectorize_vec_perm_const): ...this new function.
16700 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16701 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16702 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16703 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16704 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16705 (rs6000_expand_vec_perm_const): Delete.
16706 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16708 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16709 (altivec_expand_vec_perm_const_le): Take each operand individually.
16710 Operate on constant selectors rather than rtxes.
16711 (altivec_expand_vec_perm_const): Likewise. Update call to
16712 altivec_expand_vec_perm_const_le.
16713 (rs6000_expand_vec_perm_const): Delete.
16714 (rs6000_vectorize_vec_perm_const_ok): Delete.
16715 (rs6000_vectorize_vec_perm_const): New function.
16716 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16717 an element count and rtx array.
16718 (rs6000_expand_extract_even): Update call accordingly.
16719 (rs6000_expand_interleave): Likewise.
16720 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16721 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16722 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16723 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16724 (rs6000_expand_vec_perm_const): Delete.
16725 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16726 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16727 (altivec_expand_vec_perm_const_le): Take each operand individually.
16728 Operate on constant selectors rather than rtxes.
16729 (altivec_expand_vec_perm_const): Likewise. Update call to
16730 altivec_expand_vec_perm_const_le.
16731 (rs6000_expand_vec_perm_const): Delete.
16732 (rs6000_vectorize_vec_perm_const_ok): Delete.
16733 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16734 reference to the SPE evmerge intructions.
16735 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16736 an element count and rtx array.
16737 (rs6000_expand_extract_even): Update call accordingly.
16738 (rs6000_expand_interleave): Likewise.
16739 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16740 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16742 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16744 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16746 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16747 vector mode and that that mode matches the mode of the data
16749 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16750 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16751 directly using expand_vec_perm_1 when forcing selectors into
16753 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16755 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16757 * optabs-query.h (can_vec_perm_p): Delete.
16758 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16759 * optabs-query.c (can_vec_perm_p): Split into...
16760 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16761 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16762 particular selector is valid.
16763 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16764 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16765 (vect_grouped_load_supported): Likewise.
16766 (vect_shift_permute_load_chain): Likewise.
16767 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16768 (vect_transform_slp_perm_load): Likewise.
16769 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16770 (vectorizable_bswap): Likewise.
16771 (vect_gen_perm_mask_checked): Likewise.
16772 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16773 implementations of variable permutation vectors into account
16774 when deciding which selector to use.
16775 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16776 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16777 with a false third argument.
16778 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16779 to test whether the constant selector is valid and can_vec_perm_var_p
16780 to test whether a variable selector is valid.
16782 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16784 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16785 * optabs-query.c (can_vec_perm_p): Likewise.
16786 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16787 instead of vec_perm_indices.
16788 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16789 (vect_gen_perm_mask_checked): Likewise,
16790 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16791 (vect_gen_perm_mask_checked): Likewise,
16793 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16795 * optabs-query.h (qimode_for_vec_perm): Declare.
16796 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16797 (qimode_for_vec_perm): ...this new function.
16798 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16800 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16802 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16803 does not have a conditional at the top.
16805 2018-01-02 Richard Biener <rguenther@suse.de>
16807 * ipa-inline.c (big_speedup_p): Fix expression.
16809 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16812 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16815 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16819 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16820 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16821 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16822 cond_taken_branch_cost 3->4.
16824 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16826 PR tree-optimization/83581
16827 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16828 TODO_cleanup_cfg if any changes have been made.
16830 PR middle-end/83608
16831 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16832 convert_modes if target mode has the right side, but different mode
16835 PR middle-end/83609
16836 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16837 last argument when extracting from CONCAT. If either from_real or
16838 from_imag is NULL, use expansion through memory. If result is not
16839 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16840 the parts directly to inner mode, if even that fails, use expansion
16843 PR middle-end/83623
16844 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16845 check for bswap in mode rather than HImode and use that in expand_unop
16848 Copyright (C) 2018 Free Software Foundation, Inc.
16850 Copying and distribution of this file, with or without modification,
16851 are permitted in any medium without royalty provided the copyright
16852 notice and this notice are preserved.