I agree with Stavros -- speed is important in
writing papers that
(a) compare systems. Even if the systems are quite different,
they use the same clock. You can't say which is better,
Maple, Mathematica, C++, Maxima, MathCad, Matlab, Java,
Python, Ruby, Perl, Tcl, Common Lisp.
But you can say which is faster.
(b) study asymptotic computational complexity. If you have
a technique that is faster for multiplying polynomials with
20,000 terms or more, you can write a paper about it.
(In a (very useful!) paper about how theorists should write
experimental papers Dave Johnson points out (I think) that anything
that takes less than a few seconds isn't worth timing.
If my speedups for maxima are installed, (a) they MIGHT
break something; (b) most things will not run faster, and may
run a little slower and take more memory; (c) some (large)
computations may run hundreds of times faster.
RJF
Stavros Macrakis wrote:
>Speeding up Maxima would be nice. In fact, I have improved the speed of
>parts of Nset and of GCL's sort routine and am always aware of speed
>issues (probably too aware).
>
>However, I think it is FAR more important to improve Maxima's
>correctness, documentation, coverage, features, and usability. That
>is, we need to correct bugs (the bug list is long), write good documents
>(our current documents are incomplete and rough), add more mathematical
>functionality, improve features (e.g. plotting), and make it easier to
>use.
>
> -s
>
>_______________________________________________
>Maxima mailing list
>Maxima@www.math.utexas.edu
>http://www.math.utexas.edu/mailman/listinfo/maxima
>
>