[Numpy-discussion] Solaris Sparc build broken
Charles R Harris
charlesr.harris at gmail.com
Thu Nov 5 04:15:29 EST 2009
On Thu, Nov 5, 2009 at 1:39 AM, David Cournapeau <
david at ar.media.kyoto-u.ac.jp> wrote:
> Charles R Harris wrote:
> > Hmm, you mean use the architecture information along with the size?
> > How does that work for open solaris, which at some point will be
> > implementing quads on intel, but doesn't now?
> I would say let's focus on existing platforms which cause problem before
> worrying about non existent platforms :)
I don't think it is that far off. The quad precision isn't done in hardware,
not on SPARC or Intel, it is done in software and that software is already
present on Intel, just not turned on. Someone else's quick and dirty fix.
But folks need to port software from one platform to the other, so I expect
things will change.
> Setting the functions at runtime is too inconvenient to warrant it
> unless strictly necessary (slow, complicated, etc...).
Well, I don't agree that it is slow or complicated ;) Hey, if we needed code
that went zoom, zoom, zoom, we wouldn't be using python, now would we. We
would write in assembler like real programmers do.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the NumPy-Discussion