Libva/VA-API integration with freerdp

[Miguel Amez]

Good morning list,

This is my first post to the list, so I don't know very well the rules that
you have about posting (if there is anyone).

I've contacted to Paul to know which was the best way of sending you my
purposal/suggestion, and he told me that sending to this list probably
would be the best one, so hands on it!

I've been working with Allwinner A20 and freerdp in a production
environment for 3 years with kernel 3.4+. Our main problem is the video
acceleration, that makes the experience a litle bit  weird, so that is why
we decided to test the newest cedrus driver with mainline kernel. Final
experience is more or less the same and troubles found are again related to
multimedia and video acceleration.

After taking a look at freerdp's git repository, I realized about one
commit that talks about the integration in their software of libva/VA-API
drivers/decoders, so I'm asking them about the posibility of the
integration of cedrus-libva with their software.

My suggestion: I consider that giving this integration functionality to
sunxi-cedrus could be great for the comunity as a different use case apart
from multimedia playing resources, that you are more focused on nowadays.

What do you thing? Would you give me support on this?

Thanks in advance,

[Miguel Amez]

Hi again list,

After asking to the freerdp list, this is the answer that they send me:

Hi Miguel,

no, currently there is no (direct) support for libva, but it may be

indirectly used with the FFMPEG AVC444 decoder backend.

Currently the biggest obstacle there is the custom YUV to RGB filter for

the AVC444 mode, which is done CPU only.

Accelerating older RGB modes will most likely not be possible though, as

they use custom codecs not really supported by hardware.

From my point of view libva support will only be half of the work as the

main problem currently is the lack of GPU accelerated filter/image

manipulation necessary after the video decoding is already complete.

But, of course, getting a speedup is always what we're looking for, so

maybe we can find some way.

I think they aren't bad news indeed, but we need to build a use case scenary with all this characteristics in order to figure out the way we could "do the miracle". Any idea/suggestion that you have on this?

Thanks for your help,