PDA

View Full Version : A Unified GPGPU API In Gallium3D?


phoronix
02-06-2008, 02:00 PM
Phoronix: A Unified GPGPU API In Gallium3D?

Zack Rusin today has covered on his blog GPGPU and ultimately answers the question as to whether there will be a GPGPU API within the new Gallium3D architecture. GPGPU, or General Purpose computing on GPUs, has been a hot topic as of late with both NVIDIA and ATI/AMD having their own SDK/APIs and the latest graphics cards offering dozens of stream processors...

http://www.phoronix.com/vr.php?view=NjMyMg

nanonyme
01-01-2009, 01:05 PM
"AMD has decided to back OpenCL (and DirectX 11) instead of its now deprecated Close to Metal (aka Stream) framework" Wikipedia, OpenCL article (Their sources are http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543~127451,00.html and http://www.eweek.com/c/a/Desktops-and-Notebooks/AMD-Backing-OpenCL-and-Microsoft-DirectX-11/)
Just thought to mention since this change apparently happened after the Phoronix article was written.

bridgman
01-01-2009, 01:38 PM
Just a slight tweak to this; CTM was the lowest level of the Stream SDK, and was replaced by CAL. The Stream SDK isn't going away, we just expect to add OpenCL as another (and, over time, perhaps the most important) open API in the stack.

On the open source side, I expect that any OpenCL implementation would run over drm for sure, and the implementors would probably try to run over Gallium3D unless they ran into serious problems. OpenCL is arguably higher level than CAL or Gallium3D -- both CAL and Gallium3D are primarily there to expose the shaders of a modern GPU in a relatively hardware-independent way.

EDIT - I fixed the Wikipedia page; previously it didn't say exactly the same as the pages it referenced ;)

nanonyme
01-01-2009, 03:31 PM
EDIT - I fixed the Wikipedia page; previously it didn't say exactly the same as the pages it referenced ;)
Thanks. I was hoping to catch the attention of some AMD representative who could point out which parts I had possibly misunderstood. (Well, the fact that the misinterpretation was on Wikipedia's side was unexpected, I expected it would be on mine)