All you have to do is build a mechanism that distributes threads between interpreter processes then just spawn an interpreter for each core. For many applications, that's all you need.