diff --git a/src/u32x4_.rs b/src/u32x4_.rs
index 7f0229d..dbdb48a 100644
--- a/src/u32x4_.rs
+++ b/src/u32x4_.rs
@@ -458,7 +458,10 @@ impl u32x4 {
         u64x2 { simd: i64x2_mul((self & i64x2::splat(0xffffffff)).simd, (rhs.simd & i64x2::splat(0xffffffff))).simd }
       } else if #[cfg(all(target_feature="neon",target_arch="aarch64"))] {
         unsafe {
-          u64x2 { neon: vmull_u32(vget_low_u32(self.neon), vget_low_u32(rhs.neon)) }
+          let a = vuzpq_u32(self.neon, self.neon);
+          let b = vuzpq_u32(rhs.neon, rhs.neon);
+
+          u64x2 { neon: vmull_u32(vget_low_u32(a), vget_low_u32(b)) }
         }
       } else {
         let a: [u32; 4] = cast(self);