Currently, we print a simple `lwu` for this node, which is not accurate becasue we do a `ld` and logic shift right the loaded 64-bit value for this node. We should change it into `load_narrow_klass_compact` like other CPU platforms.