If there is a separately wrapped register restore on some path that
is dead (say, control goes into an endless loop after it), then we
cannot delete that restore because that would confuse the DWARF CFI
(if there is another path joining).
This happens with gcc.dg/torture/pr53168.c, for example.
2016-06-07 Segher Boessenkool <seg...@kernel.crashing.org>
* dce.c (delete_unmarked_insns): Don't delete instructions with
a REG_CFA_RESTORE note.
---
gcc/dce.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/gcc/dce.c b/gcc/dce.c
index ea3fb00..d510287 100644
--- a/gcc/dce.c
+++ b/gcc/dce.c
@@ -587,6 +587,15 @@ delete_unmarked_insns (void)
if (!dbg_cnt (dce))
continue;
+ if (crtl->shrink_wrapped_separate
+ && find_reg_note (insn, REG_CFA_RESTORE, NULL))
+ {
+ if (dump_file)
+ fprintf (dump_file, "DCE: NOT deleting insn %d, it's a "
+ "callee-save restore\n", INSN_UID (insn));
+ continue;
+ }
+
if (dump_file)
fprintf (dump_file, "DCE: Deleting insn %d\n", INSN_UID (insn));
--
1.9.3
