making maxima 4-20X or more faster sometimes

Here is a benchmark of macsyma/ maxima running
with my speedups and without.  This is running
in commercial macsyma, but the (single) lisp
file could be compiled/read into maxima too, I think.
I would be interested in benchmarks on various
other lisps.
The source file would probably
have to be examined by someone else
who has been following the maxima source.
The comments and bugs would have to be taken note of.

The lisp source file
is in http://www.cs.berkeley.edu/~fateman/temp/newsimp.cl

Here is a benchmark.  In each case I show 2 or 3 times.
The first time is with the original simplifier.
The second is with the new simplifier.
If there is a third, it is time shown by xmaxima 5.9.


First 2 times are on a pentium 4 2.6Ghz using Allegro CL;
3rd time is same machine with xmaxima.

(c1) showtime:all$

Time= 0 msecs
Time= 0 msecs
(c2) m1:sum(z[i],i,0,200)$

Time= 31 msecs
Time= 15 msecs
(c3) m2:sum(z[i],i,0,200)$

Time= 32 msecs
Time= 0 msecs
(c4) m1-m2;

Time= 15 msecs
Time= 0 msecs
xmaxima 15 msecs
(d4)                        0
(c5) m1+m2$

Time= 0 msecs
Time= 0 msecs

(c6) d5-m1-m2;

Time= 62 msecs
Time= 0 msecs
xmaxima 18 msecs
(d6)                        0
(c7) m10:m1+sum(z[i],i,201,500)$

Time= 78 msecs
Time= 16 msecs
(c8) m20:m2+sum(z[i],i,201,500)$

Time= 47 msecs
Time= 46 msecs (including 46 msecs in GC)
(c9) m10-m20;

Time= 172 msecs
Time= 16 msecs
xmaxima 62 msecs
(d9)                        0
(c10) m10+m20$

Time= 0 msecs
Time= 0 msecs
(c11) d10-m10-m20;

Time= 343 msecs
Time= 0 msecs
xmaxima 1200 msecs !! I tried this twice, got similar times
(d11)                        0