Re: PCL benchmark

[Chris Burdorf <burdorf@rand.org>]

Ok,

I got rid of the formats and compiled with safety set to 0 and
speed at 3.  I got the following times:


AKCL

20 cars

real time : 8.550 secs
run time  : 8.483 secs
----------------------

50 cars

real time : 28.133 secs
run time  : 26.300 secs


Allegro -----------

20 cars
cpu time (non-gc) 1984 msec user, 0 msec system
cpu time (gc)     0 msec user, 0 msec system
cpu time (total)  1984 msec user, 0 msec system
real time  2020 msec

50 cars
cpu time (non-gc) 5300 msec user, 67 msec system
cpu time (gc)     967 msec user, 0 msec system
cpu time (total) real time  6340 msec

ERNIE -----------

20 cars --- 1.0003
50 cars --- 2.686


I ran them on a sun 3/60.  I couldn't compile pcl under allegro with
safety at 0 and speed at 3, but I could compile the simulation.

It's clear that PCL is faster under some Common LISP's than others.
I did 50 cars, because the numbers with 20 were too low.  Note
with 50 cars, ERNIE is 10 times faster than AKCL PCL.  It's also
close to 2.5 times faster than ALLEGRO PCL.  


I can't tell you much about ERNIE, because it is funded by DOD and
I don't want to get tried for treason.  It's basically a reimplemtation of
FLAVORS.  It has practically everything Flavors has, but it was 
designed especially to be fast.  I double checked my ERNIE
code and it looks to do the same things as my PCL code.  You'll have to
take my word for it. I can't send you the code.  


I wouldn't say that comparing ERNIE and PCL is like comparing APPLES and
ORANGES, but rather TANGERINES and ORANGES.  ORANGES are bigger, but
they take longer to eat.  My main reason for sending out this comparison
of ERNIE and PCL was not to try and convert people to ERNIE, but to
try to  stir a dialog on efficiency.  Hopefully PCL will be fast for all of us:
rich or poor. 


Chris