>>>>> "blake" == blake <[EMAIL PROTECTED]> writes:
blake> To the CMUCL team, nice job! I am using cmucl for my graphics art work
blake> and these are the results from my little test. It appears to be 4 times
blake> faster, out of the box, than 18e for my routines. That's just great!
blake> Thanks a lot!
blake> Blake
blake> ;; Under version 19a,for size-multiple=1, with interlocking circles as
blake> underlying grid
blake> ; Evaluation took:
blake> ; 10.8 seconds of real time
blake> ; 10.722656 seconds of user run time
blake> ; 0.048828 seconds of system run time
blake> ; 25,824,925,688 CPU cycles
blake> ; 78 page faults and
blake> ; 3,290,016 bytes consed.
blake> ;
blake> ;;Under version 18e, for size-multiple=1, with interlocking circles as
blake> underlying grid
blake> ; Evaluation took:
blake> ; 41.6 seconds of real time
blake> ; 38.51172 seconds of user run time
blake> ; 3.0625 seconds of system run time
blake> ; 99,512,323,048 CPU cycles
blake> ; [Run times include 12.21 seconds GC run time]
blake> ; 202 page faults and
blake> ; 753,545,048 bytes consed.
4 times faster, and 200 times less consing. Are you sure you compiled
both versions before testing? This is pretty amazing.
Can you tell us why it's so much faster? Or maybe what you're doing?
The only thing I can think of is CLOS that might make a big
difference. I wouldn't expect any other changes to make such a huge
performance gain.
Ray
