[plt-scheme] On understanding performance

   I writing a small PROLOG implementation using DrScheme (MrEd) version 
350.

   I decided to see about improving performance using Marc Feeley's 
compilation to closures technique.  An initial benchmark (the naive 
Fibonacci predicate) shows little difference in speed.  When implementing 
Scheme-like languages in other environments, I am used to a significant 
improvement.  Any guesses about why I am seeing so little improvement 
here?

   I saw a hint that profiling tools exist somewhere.  Can anyone give me a 
little more information about how to use them?

   My implementation is currently leaning heavily on gensyming.  What 
performance/gc issues are involved with gensym?  In particular, are 
gensyms garbage collected?

   Thanks.

-Arthur