[VDPAU] [PATCH] try DRI2DriverDRI name on DRI2DriverVDPAU absence

[Aaron Plattner]

On 01/08/2014 09:33 PM, Eric Anholt wrote:
> Aaron Plattner <aplattner at nvidia.com> writes:
>
>> Thanks for the link, Rinat.
>>
>> It looks like from the discussion, that Eric (CC'd) is suggesting that
>> libvdpau should enumerate and load all available VDPAU drivers and let
>> each one determine, through vendor-specific mechanisms, whether it is
>> applicable for the requested X screen.  Eric, does that sound like what
>> you're suggesting?  That would be doable, but was not the approach that
>> was suggested when I asked the X.Org community how the X screen-to-VDPAU
>> driver mapping should be achieved when this was first being designed.
>>
>> Can you point me to where Mesa's DRI3 support determines which driver to
>> use for which X screen?
>
> dri3_common.c.  We should really rig up a mechanism based on either
> inspecting driver binaries to see what hw they support, or having a file
> along the side saying what hw the drivers support, so that you don't
> need all the pci ids in the build of the loader.  We just haven't gotten
> around to it.

So if I'm reading this right, the server opens a DRI device and passes 
the fd through to the client, and then the client uses that to match a 
driver based on a table of PCI IDs?  That might work as long as it can 
fall back to DRI2, since it needs to continue to work on older systems 
and on Solaris and FreeBSD.  It won't work on Tegra, though, because the 
GPU is not a PCI device.

I'm not going to have time to wire up the nvidia X driver to that path 
any time soon, so I can't sign up to write and test that path right now.

-- 
Aaron