Here are the timing results:
(user::run)
Maxima 5.5 Fri Mar 30 11:08:06 CST 2001 (with enhancements by W. Schelter).
Licensed under the GNU Public License (see file COPYING)
(C1) showtime:all;
Evaluation took 0.00 seconds (0.00 elapsed)
(D1) ALL
(C2) (a+b+c+d+1)^35;
Evaluation took 0.00 seconds (0.00 elapsed)
35
(D2) (d + c + b + a + 1)
(C3) (ratsimp(d2),1)$
Evaluation took 7.07 seconds (7.34 elapsed)
(C4) (ratsimp(d2),1)$
Evaluation took 4.96 seconds (5.00 elapsed)
(C5) (ratsimp(d2),1)$
Evaluation took 3.27 seconds (3.45 elapsed)
(C6)
....
unfortunately, there are a few glitches.
ACL is unhappy about the redefinition of //
it complains about reference to the serror package in
sys-proclaim (seems I can #-allegro comment it out?)
and I just commented out a function in spgcd
which did some array store bogosity.
I haven't figured out the dumplisp/restart
sequence which for ACL6.0 is slightly different.
You dump only a dll or dxl and then when you
start up,
you specify this file. When I do, I'll send
an compile-allegro.lisp file out.
It looks like the GCL speed and the Allegro
speed on this problem are fairly close. Since
GCL is running on a slower machine, the difference
may be in GCL's favor. (I don't know if WFS
has some GCL-only speedups in use on this task.)
RJF