[Libav-user] Performance of Slice Based Processing

[Robert Beattie]

I am curious as to why the performance (throughput) does not increase
linearly with the number of slices. I am using a scale filter
(scale=1280:720:flags=bilinear) to convert from 1080 to 720p in a stand
alone app.

When I go from 1 to 2 slices and 2 to 4 slices I see a large increase in
throughput. However, going from 4 to 8 slices there is almost no change at
all. Can anyone enlighten me as to the reason for this? There are plenty of
cores available on the machines I tried it on.

Thanks
Rob
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://ffmpeg.org/pipermail/libav-user/attachments/20230629/2bb26596/attachment.htm>