> In general, I think everybody would welcome if somebody who cares and
> knows about fp-intensive apps would spend some time on improving the
> math library (if you would like to do so, please start with the one
> that comes with glibc---it's a much better starting point than the old
> glibc).
I've downloaded a glibc beta, and maybe I even will take a night or two
hacking at it (even though I ought to finish my thesis, and tend to my
sick girlfriend, and write letters to rich aunts, and...) but I just
wonder:
Why does lack of optimization seem to hurt the Alpha in particular? As
I see it, it's the *algorithms* used, that for a large part should be
optimized, with hardware specific optimizations being considered when
algorithms already are optimal. Is this not the case?
Does anybody have any good references to literature for the algorithms
involved? I guess I can find this on my own, but if anything pops into
your head as you read this, feel free to drop me a note...
Will I be able to compile libm on an Intel Linux box? Is this
preferrable to using an Alpha-based OSF box? Those are basically my
choices (though I /could/ use Solaris :-) -- and I often ask myself
what I'm doing on this list. :-)
As I said, I don't really have time for this, so if anybody else wants
to take charge and coordinate everything, I'd be more than happy to
supply that person with patches etc if and when I actually do
something. Drop me a mail!
~kzm