MAXIMA Problem Suite
			   by Bill Schelter

We have collected a number of test problems for MAXIMA, and include
below some information on timings in various lisps.  The files we used
are in the maxima/doc directory of my standard distribution of MAXIMA.
These problems are in the files rtest*.mac.  Files rtest1.mac through
rtest13s.mac contain a total of 808 problems.

In the following reference to tests [n,m] mean tests rtest`n'.mac
through rtest`m'.mac inclusive.  The following times were obtained on
a Sun 280 (rascal.ics.utexas.edu).

I used MAXIMA version 4.71, AKCL 1.34, and LUCID 2.1.

TESTS         AKCL   LUCID    EXPLORER I      
[1,8]	     157.55  199.66        
[9,13s]      103.00  138.01        
[1,13s]      278.63  388.14    511.1

4 Users(cpu) 324.86  500.60
4 Users(tot)1478.64  2318.49  
There were no errors.

Tests were run with the output directed to a file.  The tests
represent a wide spectrum of problems for MAXIMA, and are the standard
verification suite, which I use after making changes to the underlying
lisp or to the MAXIMA files, to help determine correctness.  The files
consist of alternating problem, and answer, with the result of the
problem being compared against the answer.

In neither lisp did I use the floating point chip.  This is because the
saved image is often used on machines without such a chip.  Generally
the LUCID floating point calculations were single float and the AKCL
ones double float.   I don't believe however that the tests are
very floating point intensive.

The 4 Users line is the Cpu time for the same program but running in 4
processes simultaneously.  The AKCL probably degraded less from the
single user time (16% as opposed to 29% for Lucid), because a large
amount of the maxima image is shared between processes.


The complete list of files is 
 '("rtest1" "rtest1a" "rtest2" "rtest3" "rtest4" "rtest5"
   "rtest6" "rtest6a" "rtest6b" "rtest7" "rtest8" "rtest9"
   "rtest9a" "rtest10" "rtest11" "rtest12" "rtest13" "rtest13s")

It is still the experience of the author, that lisp machines do show
important gains on very large problems, much larger than those done
here.


They are included in /usr2/ftp/pub/maxima-xx.tar.Z on rascal.