[mythtv-users] State of VA-API

Mark Kendall mark.kendall at gmail.com
Tue Nov 16 10:18:07 UTC 2010


On 16 Nov 2010, at 17:13, dargllun <dargllun at googlemail.com> wrote:

> On 15.11.2010 19:47, Richard Morton wrote:
>> As I understand it, a lot of the coding to support VAAPI has been put
>> in place but it isnt ready for primetime and it is disabled; whether
>> this is a build option or something else I dont know.
>> 
>> I guess if you want to discuss this bleeding edge feature you should
>> ask on the dev list if they are looking for testers; if the feature is
>> still being developed then I am sure they wont want lots of people
>> saying this and that isnt working - which in almost all cases they
>> will be fully aware of.
> 
> Actually I did ask on the dev mailing list, but this list seems to be rather dead, there's next to no new messages, let alone an answer to my question wrt to VA-API.
> 
>> Essentially; the devs will ask or make it known when a feature is
>> ready for beta-testers. Supporting VAAPI isnt going to be a trivial
>> piece of work.
> 
> I was afraid of that... Guess I'll stick with KMS and OSS drivers for my HD3200, works great for SD content and for H.264, well, the CPU must do 

VAAPI never made it into 0.24 largely because it wasn't stable. The VDPAU backend worked reasonably well, with a few quirks, but if you're on nvidia hardware you should just use vdpau directly - it's feature 'full' and stable.

ATI/AMD support is a bit of a joke. You are reliant on a single binary blob released to a single developer/organisation that is not feature complete (no deinterlacing) and horribly buggy.

Intel hardware faired slightly better but was still problematic enough that I wouldn't dream of enabling it by default until it has improved significantly.

Add to that the current situation of largely relying on a third party library that is a moving feast and you might appreciate why I'm not exactly straining at the leash to start working on it again.

Fwiw, I'm quite sure some of the problems were self inflicted by my own code and the way in which the rest of the mythtv codebase works - as well as my desire to integrate vaapi into the existing OpenGL code - but it really shouldn't be that painful!

Regards
Mark


More information about the mythtv-users mailing list