The following avoids using TYPE_PRECISION on VECTOR_TYPE when
looking for bit-precision changes in vectorizable_assignment.
We didn't anticipate a stmt like
_21 = VIEW_CONVERT_EXPR<unsigned int>(vect__1.7_28);
and the following makes sure to handle that.
Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed.
* tree-vect-stmts.cc (vectorizable_assignment):
Properly handle non-integral operands when analyzing
conversions.
---
gcc/tree-vect-stmts.cc | 13 ++++++++-----
1 file changed, 8 insertions(+), 5 deletions(-)
diff --git a/gcc/tree-vect-stmts.cc b/gcc/tree-vect-stmts.cc
index e6649789540..01cb19ce933 100644
--- a/gcc/tree-vect-stmts.cc
+++ b/gcc/tree-vect-stmts.cc
@@ -5833,12 +5833,15 @@ vectorizable_assignment (vec_info *vinfo,
/* We do not handle bit-precision changes. */
if ((CONVERT_EXPR_CODE_P (code)
|| code == VIEW_CONVERT_EXPR)
- && INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
- && (!type_has_mode_precision_p (TREE_TYPE (scalar_dest))
- || !type_has_mode_precision_p (TREE_TYPE (op)))
+ && ((INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
+ && !type_has_mode_precision_p (TREE_TYPE (scalar_dest)))
+ || (INTEGRAL_TYPE_P (TREE_TYPE (op))
+ && !type_has_mode_precision_p (TREE_TYPE (op))))
/* But a conversion that does not change the bit-pattern is ok. */
- && !((TYPE_PRECISION (TREE_TYPE (scalar_dest))
- > TYPE_PRECISION (TREE_TYPE (op)))
+ && !(INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
+ && INTEGRAL_TYPE_P (TREE_TYPE (op))
+ && (TYPE_PRECISION (TREE_TYPE (scalar_dest))
+ > TYPE_PRECISION (TREE_TYPE (op)))
&& TYPE_UNSIGNED (TREE_TYPE (op))))
{
if (dump_enabled_p ())
--
2.35.3
