[FFmpeg-user] Multicore CPU run

[Dennis Mungai]

On Sat, 2 Feb 2019 at 19:09, Mahmood Naderan <mahmood.nt at gmail.com> wrote:

> >Note that hardware accelerated
> >decode is targeted for real time PLAYBACK without using any CPU resources,
> >and does not necessarily imply to be faster than a software based decode.
>
> If I got the point, you are saying that GPU version is somehow used to free
> CPU cores, so that CPU cores are ready for non-gpu task. However, we know
> that video processing is GPU friendly due to the nature of that.
> How that can be justified?
>
>
>
> Regards,
> Mahmood
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-user
>
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".



For DECODE: Yes. GPU-based hwaccel guarantees that no CPU resources will be
used for DECODE, where supported. See notes below.
Video processing is NOT necessarily "GPU friendly". It can however, be
significantly FASTER depending on the accelerator in use AND the encoder
implementation in place.
Also, others, such as VAAPI and Intel's QuickSync will also require a bit
more setup due to limited format support, etc. This implies that
performance will differ significantly on the content you're encoding and
the presets available to the encoder.

See the wiki for more details:
1. https://trac.ffmpeg.org/wiki/Hardware/QuickSync
2. https://trac.ffmpeg.org/wiki/Hardware/VAAPI
3. https://trac.ffmpeg.org/wiki/HWAccelIntro