Tim Smith wrote:
> In article <giq1ds$nnm$1@xxxxxxxxxxxxxxxxxxxxxxxxx>,
> Hadron <hadronquark@xxxxxxxxx> wrote:
>> > What we've got now are several vendor-specific systems (CUDA from
>> > NVidia, and Stream Computing SDK from AMD, for example). You don't
>> > think having a portable, standard way to do GPGPU stuff is something
>> > we need!?
>>
>> No. I dont. At this time we need a shove of the existing standards. And
>> b "we" I mean the Linux side. Gods knows its having enough difficulty
>> keeping up and getting SW vendors on their side.
>
> What existing standards? As far as I know, there isn't an existing
> standard. If I want to do, say, a GPU-accelerated FFT, I have to decide
> if I'm going to target NVidia GPUs or ATI GPUs, then use either NVidia's
> CUDA software or AMD's Stream Computing SDK, respectively.
>
> With OpenCL, I'll be able to write my GPU-assisted FFT code in a way
> that doesn't depend on the particular GPU. How is that not a good and
> much needed thing?
>
Simply *because* it is a good thing Hadron Quark is against it.
Not only because it is "choice" (something Hadron *hates* like few other
things), but also because it is good for OpenGL. Can't have that, as MS
favours DirectX and wants to bury OpenGL, the sooner the better
--
Don't abandon hope: your Tom Mix decoder ring arrives tomorrow
|
|
