Do you attribute most of the difference in performance to the OS change or to the implementation change, though? (I'll never use Windows, but I've been considering using Lispworks again on OS X for a few things.) I've always used LW with Emacs in the past, not the LW IDE. I use Aquamacs for almost everything now...and SLIME is great.

What Lisp libs are you using, if you don't mind my asking? I'm trying to figure out why CL-Memcached is slower than freaking PHP and Python for most things (!)

- dcphillips /AT/ runbox /DOT/ com

You call being slower by a (sometimes) 34 msec over 10000 network io requests as slow ? Most of the io stuff for php etc. is pure C. If you have /any/ non trivial application then the stability and performance of CL systems become apparent. If that is your problem, then try benchmarking string operations of CL vs. the rest of the gang.

We're guessing primarily Windows, but with no LW Linux version to compare with, it is a guess.

We're using cl-sql, uffi, md5, portableaserve (with net.html and minimal use of webactions).

We have our own caching system (for key mysql db data) running in the application.

Out of curiosity (as a Lispworks and SBCL user who has never seen this sort of performance difference between the two) is there anything in particular that you could attribute this to (eg. FFI, Streams)

It's hard to say, as we don't have LW on Linux to compare to. My guess is that Windows is a big contributor to the slowness.

Really? That is a lot. I noticed a factor of two when I moved from CLisp to SBCL, and I was under the impression that CLisp is generally the slowest implementation.