Christian Neukirchen said:

> While we are at it, did you consider llvm?  llvm.cs.uiuc.edu

Honestly no. We were more concerned about evaluating how well C-- would
perform as the target for a dynamically type checked, object oriented
language. Also by actually compiling ruby to machine code we wanted to
break new ground, which we wouldn't have done with llvm[1].

Cheers,
Andreas

[1] http://llvm.cs.uiuc.edu/pubs/2004-Spring-AlexanderssonMSThesis.html