There's a limitation in the path discovery bits in the backward
threader that I ran into recently and I'd like to document it so we
don't lose track of it.
Basically we stop looking if we find a threadable path through a PHI,
without taking into account that there could be multiple
paths through a PHI that resolve the path. For example:
x_5 = PHI <10(4), 20(5), ...>
if (x_5 > 5)
I don't remember how we ended up skipping this, but it could existing
behavior from the old bits. It probably skipped multiple threads
through a PHI since the generic copier couldn't re-using existing
threading paths anyhow.
Documenting for later fixing.
---
gcc/tree-ssa-threadbackward.c | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/gcc/tree-ssa-threadbackward.c b/gcc/tree-ssa-threadbackward.c
index 28c7ef8c872..496b68e0a82 100644
--- a/gcc/tree-ssa-threadbackward.c
+++ b/gcc/tree-ssa-threadbackward.c
@@ -282,6 +282,13 @@ back_threader::resolve_phi (gphi *phi, bitmap interesting)
continue;
}
+ // FIXME: We currently stop looking if we find a threadable path
+ // through a PHI. This is pessimistic, as there can be multiple
+ // paths that can resolve the path. For example:
+ //
+ // x_5 = PHI <10(4), 20(5), ...>
+ // if (x_5 > 5)
+
tree arg = gimple_phi_arg_def (phi, i);
if (TREE_CODE (arg) == SSA_NAME)
{
--
2.31.1
