To be sure our generic class writer isn't too much of a performance hit when we move it to trunk, I've run a profiling session compiling Maxima sources.

The outcome is here: http://paste.lisp.org/display/113517.

Conclusion: the old "lets store .cls files before we create a fasl out of them" is still the performance bottleneck (on Windows at least). So, the good thing is that the change in the generator doesn't show up in the list of hot spots (not to be confused with HotSpot).

Bye,

Erik.