Intel graphics and the future of XvMC
dcm at MIT.EDU
Fri May 11 21:14:00 PDT 2007
Recently there was a thread here discussing implementation plans for
XvMC for the Intel 965 chipsets. I'd like to take the discussion a bit
further: Can we do better than XvMC?
Based on my limited experience with XvMC, it suffers from a number of
significant drawbacks: It's not cross-platform, it doesn't play nice
with applications that use 3D acceleration (OpenGL), and it's hard to
add new features to the specification.
On the other hand, OpenGL has been around for a long time, has a large
and active development community, and has a good track record of
extensibility. Furthermore, OpenGL is the cross-platform way to access
the full compliment of features on modern graphics hardware (with the
notable exception of CUDA).
Wouldn't it make more sense for Intel to innovate here and come up with
a new OpenGL extension that would expose video acceleration features
such as iDCT and motion compensation? Applications like mythtv would be
able to really innovate given a full OpenGL code path.
I'm just going to throw that out there in the hopes of spawning some
interesting discussion. I'd love to help out with any effort that
develops. I'm really glad that Intel is taking an active role in free
graphics drivers. By the way, are the data sheets for Intel's hardware
P.S. I would be even _more_ grateful to Intel if someone took a look at
my G965 Mesa bug reports. :)
More information about the xorg