On Fri, 2002-03-29 at 13:43, Daniel Lemire wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Thanks James for these numbers, but what do they mean?
Well, I wouldn't read too much into them.
> Here are my naive conclusions (naive on purpose):
>
> 1) Maxima should run on CMUCL... it seems so fast!!!
> The new results with GCL are fine, but you are still at
> half the speed!!!
There is a reason we've been working so hard to get Maxima to work
equally well with Lisps other than GCL. One reason is that CMUCL is
pretty fast. (There are others.)
> To someone who has never done symbolic programming,
> this appears surprising. I work mostly on signal processing
> applications... and if you can get something done twice as
> fast, it is worth a major paper...!!!
> Why are the numbers sooo different? And what can we tell
> from this one benchmark?
The given benchmark is particularly simple. I would not necessary expect
the speed disparity to be as great with all Maxima code. I think lisp
implementations differ in speed much more than, say, C compilers.
Nonetheless, the benchmark I forwarded behaved roughly the way I
expected.
> 2) Mathematica and Maple are soooo slow. The result is
> not surprising as far as Maple goes because I know it is
> really, really slow and programming Maple is typically a
> waste of time... but I'm surprised Mathematica does so badly???
It would be interesting to see benchmarks on some more complicated code.
--Jim