-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
james@email-addr-hidden-dot-dat.net escribió:
> On Mon, 29 May, 2006 at 04:49PM -0400, Dan Easley spake thus:
>> Slashdot recently mentioned the following FFT via GPU library:
>>
>> http://gamma.cs.unc.edu/GPUFFTW/index.html
>>
>> (slashdot article)
>> http://hardware.slashdot.org/hardware/06/05/29/1424213.shtml
>>
>> it has a custom license, which seems friendly enough, but i'd
>> prefer GPL.
>
> It would be nice if fftw switched to this when it was available.
>
> Or would it? Maybe that would cause too much heterogeneity of
> platform. Small changes from one system to another (ATI instead of
> nVidia, for example) could change performance drastically. You'd
> be unable to rely on an "average" level of performance. I might
> write something nifty with my nVidia card and it be unusable by
> some significant percentage of people because they just can't run
> it that fast.
>
> I dunno. What does everyone think? It's nice, but should we use
> it?
>
> James
>
>
>
My thoughts exactly... Given the great degree of variation in terms of
raw horsepower in graphics cards, especially the ones from the two
major vendors (ATi and nVidia), say you wrote something with a 7900,
how would a GF 4 440 MX, or an R9200 play that back?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org
iD8DBQFEe507XM+XOp70dwoRAv2/AKCJCfBG21Nb7HTX0tfanJiOmy/WxwCfWVKz
wF3aS6Y44vhcSRXbklrosG8=
=BkB3
-----END PGP SIGNATURE-----
Received on Tue May 30 04:15:05 2006
This archive was generated by hypermail 2.1.8 : Tue May 30 2006 - 04:15:05 EEST