[FFmpeg-user] Copying YUV data take huge amount of memory and time

Roger Pack rogerdpack2 at gmail.com
Fri Jul 6 22:34:34 CEST 2012


> I am integrating hardware codecs in ffmpeg code.
> But while I am trying decode video data using function virtual to physical memory it takes lots of time and memory.
> The api used to copy data is ioremap.
> Can any one suggest how can I reduce this overhead so that it takes less amount of time and memory.

I know when the VLC guys integrated DXVA into ffmpeg, they mentioned
that it "still required a memory copy of frames" so...it may require
other measures, I'm not sure.


More information about the ffmpeg-user mailing list