Write_U16_Swap leaves the upper 32 bits alone. Reimplementing this correctly in the JIT would require more than one instruction, so let's just call Write_U16_Swap instead, like Jit64 does.