[CM] cm,clm and sbcl

Bill Schottstaedt bil@ccrma.Stanford.EDU
Thu, 16 May 2002 05:24:38 -0700

> Out of interest, roughly what performance increase is expected from
> :just-lisp to the FFI version?

It's been a while since I ran any timing tests, but a factor of 30
would not surprise me.  At Rick's urging, a long time ago, I wrote
a version of some of the mus.lisp code that actually got within
a factor of 3 or 4 of C speed, but it was incredibly ugly code --
basically every sub-expression had an explicit type declaration!
But maybe one really ugly pile of pure-lisp code would be better
than what I have now...  There would still need to be a C side
for the real-time controls (hmmm...maybe not...need to think about