On 23/06/16 21:13, Anton Khirnov wrote:
> Quoting Mark Thompson (2016-06-22 00:28:31)
>> Previously we would allocate a new one for every frame.  This instead
>> maintains an AVBufferPool of them to use as-needed.
>>
>> Also makes the maximum size of an output buffer adapt to the frame
>> size - the fixed upper bound was a bit too easy to hit when encoding
>> large pictures at high quality.
> 
> Looks ok. Does this have a measurable performance impact?

Not significant for normal streams; noticable but not large (a few fps) at high 
bitrates (with the buffers suitably bigger but still created/destroyed 
per-picture).

I wrote this when pursuing why Skylake was so much slower than Haswell: 
actually it was Intel forgetting to turn on some caching 
<https://cgit.freedesktop.org/vaapi/intel-driver/commit/?id=dd9a0fb7a885f79f6413df0bd1afd5556c919a03>,
 but I think this approach to buffer handling is nicer anyway.

Thanks,

- Mark

_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to