Re: Statistical profiling

Gerd Moellmann <gerd.moellmann@xxxxxxxxxxx> writes:

> I've hacked something this afternoon, as a feasibility test, that I'd
> like to share.  It's by no means complete or anything else, and it
> might not even work for others, but I think it's interesting, and
> maybe some kind soul takes it as a start for something...

Sorry, I wanted to attach an example, but forgot:

CL-USER ==> (sprof:start-profiling)
t
CL-USER ==> (dotimes (i 10000000) nil)
nil
CL-USER ==> (sprof:stop-profiling)
nil
CL-USER ==> (sprof:report)
   NR   COUNT    % %FUNCTION
   1     2358 38.0 (flet #:g30
                     eval::internal-apply-loop)
   2      612  9.9 eval::eval-stack-args
   3      560  9.0 eval::reference-this-var-to-keep-it-alive
   4      281  4.5 eval::maybe-trace-nodes
   5      280  4.5 eval::eval-stack-push
   6      221  3.6 eval::eval-stack-pop
   7      213  3.4 "DEFUN LEAF-VALUE"
   8      194  3.1 eval::leaf-value-lambda-var
   9      173  2.8 "DEFUN GET-INFO-VALUE"
  10      160  2.6 eval::internal-invoke