Re: Firmware and libva2 backports

[Sebastian Ramacher <sramacher@debian.org>]

On 2022-04-21 17:58:52, Nicholas D Steeves wrote:
> Hi Sebatian,
> 
> Thanks for the quick reply :-)
> 
> Sebastian Ramacher <sramacher@debian.org> writes:
> > On 2022-04-21 16:51:22, Nicholas D Steeves wrote:
> [snip]
> >> > Another thing that would be nice to have in backports is newer libva2 (and related packages libva-drm2, libva-wayland2, libva-x11-2). VAAPI hardware accelerated video decoding is not working in
> >> > latest Chromium based browsers like Chrome, Brave with libva2 from stable, libva2 from testing works but it pulls libc upgrade and creates FrankenDebian.
> >> >  
> >> 
> >> I've done this one in past releases.  In addition to libva2, it needs
> >> intel-vaapi-driver, and also libdrm.  Finally, I seem to remember that I
> >> experienced graphics corruption on my system unless I tightened the
> >> dependencies so that these three packages built against the versions in
> >> backports rather than in stable.
> >
> > Also don't forget to backport intel-gmmlib and intel-media-driver.
> >
> 
> Oh, those are new!  Is this a refactoring, or does the use of these
> packages depend on the firmware-nonfree binary blobs IIRC needed for
> hardware acceleration on non-ancient Intel GPUs?  I think that's in
> firmware-misc-nonfree, no?

intel-media-driver is the VAAPI driver for anything after Broadwell and
enables hardware accelerated video decoding. If you also want encoding,
then intel-media-driver-non-free is the better option. It contains some
non-free shaders/kernels that are required for encoding.

But yes, one will also require the binary blobs from the firmware
package … not only for video decoding/encoding, but also for a sensible
experience for OpenGL-based rendering.

Cheers
-- 
Sebastian Ramacher