Thanks, that's an interesting paper. I have never heard of that paper, I'm thinking specifically of a VM that implemented the feature. Not sure if we are talking about the same algorithm, either, although it sounds similar.

Thinking about it, that sounds like inline threading

http://blog.mozilla.com/dmandelin/2008/08/27/inline-threadin...

The bodies of these opcodes are only single machine instruction or two themselves and are inlined. This changes the game.