They pretty clearly layout that PARSEC is their benchmarking suite. You can see that here: http://parsec.cs.princeton.edu/ http://parsec.cs.princeton.edu/ Fee

They pretty clearly layout that PARSEC is their benchmarking suite. You can see that here: http://parsec.cs.princeton.edu/ http://parsec.cs.princeton.edu/ Feel free to bug Princeton if you need specifics. I mean, I don't really care about this program, but it's not that difficult to beat qemu. Their whole thing is that they focus on accuracy and cross-platform compatibility. Both those things make optimizations difficult-at-best. If you're only coding a VM for a certain translation (x86->ARM, for instance), you can optimize like crazy.