[Libva] How to detect the type of memory returned...

[Jean-Yves Avenard]

On 22 June 2014 20:16, Peter Frühberger <peter.fruehberger at gmail.com> wrote:

> On the other hand I benchmarked the OpenGL approach. This approach has
> won by more than factor 5 with arround 0.8ms per frame.

how did you time the OpenGL rendering of the NV12 frame. I would guess
that it's done in a different thread than the decoding thread.

I doubt you are measuring the same thing in both intances.
hence why you are seeing such a major difference.