Re: making maxima 4-20X or more faster sometimes, update
Subject: Re: making maxima 4-20X or more faster sometimes, update
From: Richard Fateman
Date: Tue, 28 Oct 2003 08:56:01 -0800
Earlier I reported on hacks to the simplifier;
I don't know if I also mentioned that I tried
running the xmaxima test suite with the speedups
loaded... it plows through most of rtest1-3 without
problems, though it gets different forms in some cases
because my simplifier no longer does "expand" commands
the same way. It crashes in rtest4 in the partition command.
This is with ?ret2list:true; so no hash tables are
left around outside the simpplus function.
With ?ret2list: false; it crashes much earlier.
Also, there are a number of other patches that I considered
worth experimenting with in that newsimp file, some of
which may be of interest, and some of which may be
bad ideas or buggy. There is a patch to nformat which allows you to use
common lisp COMPLEX numbers. e.g.
(c1) ... load newsimp...
(c2) ?complex(3,4);
(d2) 4 %i + 3
This works transparently for the most part since I
make %i = ?complex(0,1), and numbers work fine,
e.g. for d2+d2.
For other things, like realpart(d2+a) it doesn't work.
And in fact these other changes may be why the rtest
stuff may fail.
RJF
>