Any chance of support for AMD APP & nVidea CUDA acceleration?

Started by TrevorS, June 25, 2014, 12:02:16 AM

Previous topic - Next topic

TrevorS

I'm seeing AMD and nVidea GPU accelerated parallel processing being adopted by more and more commercial video tool developers.  It's really great to see Avidemux max out my four FX4300 cpu cores,  it would be even neater if it could cut loose the 1024 shaders on my HD7850 video card as well :)!

Any possibility?

Jan Gruuthuse

Currently 2.6 has also support for VDPAU, so that editing H.264 video is slighly faster.
source: http://www.avidemux.org/admWiki/doku.php?id=versions:main&s[]=vdpau#experimental

On cuda: probably this still stands:
Quote from: Agent_007 on March 12, 2012, 06:22:56 PM
It is good thing to have multiple video standards. Competition usually leads to better solutions in the long run.

But as is said, supporting hardware GPU encoding would require some major work in many fronts, so I doubt CUDE or similar support will happen.
Can't speak myself for the developer(s).

TrevorS

Thanks much for your input :)!  My impression is that VDPAU is perhaps Unix oriented, though I may be wrong on that!  I'm hoping this forum is at least occasionally perused  by the developers and they might look further into input such as mine.  As can be told from my post, my primary concern is with AMD GCN video cards (like mine), but logically, it would be great for nVidea Cuda cards to be supported as well :)!  I'd really like to enjoy the benefits of AMD HSA without requiring an APU (though that's what I have installed in my HTPC).  I'm hoping the developers take this seriously, from what I read, video card acceleration is very beneficial!

TrevorS

Just a follow up thought, "OpenCL" has been designed to provide clean access to video GPU acceleration (plus other computation devices).  I understand it works with both AMD and nVidea video cards.  That is to say, it's not necessary to get into the nitty-gritty of card/GPUs themselves in order to access their calculation capability.

http://en.wikipedia.org/wiki/OpenCL

Jan Gruuthuse

Might not be so simple as it looks at 1st hand:
QuoteAn implementation of OpenCL for a number of platforms is maintained as part of the Gallium Compute Project, which builds on the work of the Mesa project to support multiple platforms. An implementation by Intel for its Ivy Bridge hardware was released in 2013. This software, called "Beignet", is not based on Mesa/Gallium, which has attracted criticism from developers at AMD and Red Hat, as well as Michael Larabel of Phoronix.
source: OpenCL: Implementation