>>>>> "Raymond" == Raymond Toy <toy@rtp.ericsson.se> writes:
>>>>> "James" == James Amundson <amundson@fnal.gov> writes:
James> You make some very good points. I would love to see the results of some
James> optimization experiments with Maxima. I think the optimization settings
James> all go back to (at least) Schelter's time.
James> Anyone willing to try some tests?
Raymond> I might try a few simple tests with CMUCL and maybe Clisp. I'd
Raymond> probably just zap any (safety 0) from the source code. The test would
Raymond> probably be the maxima test suite since it prints out the elapsed
Raymond> time.
Ok, I did the test using CMUCL (recent CVS version). For some reason
Clisp 2.30 crashes when loading bessel.lisp. Don't know why. This is
all on a Solaris box.
Timing:
; 12.81f0 seconds of real time
; 9.39f0 seconds of user run time
; 2.62f0 seconds of system run time
; [Run times include 0.61f0 seconds GC run time]
*** end of summary for tests-cmucl.log
Timing results with current maxima sources, removing the safety 0
declarations:
Timing:
; 12.8f0 seconds of real time
; 9.35f0 seconds of user run time
; 2.54f0 seconds of system run time
; [Run times include 0.63f0 seconds GC run time]
These times seem fairly typical, with the rea and user times not
varying by much more than 0.1-0.2 sec.
A grep shows that safety = 0 is only used in a few places:
o bessel.lisp, which I left in
o commac.lisp, which I commented out
Anyway, those are the results, FWIW.
Ray