[FFmpeg-devel] [PATCH 2/3] libavcodec: add NETINT Quadra HW decoders & encoders

Timo Rothenpieler timo at rothenpieler.org
Wed Jul 9 15:04:46 EEST 2025


On 09/07/2025 04:11, Desmond Liu wrote:
> Hi Ronald, thanks for your response.
> 
>> We previously had a similar debate regarding realmedia (search for "Codec
>> wrapper for librv11 and RMHD muxer/demuxer").
> 
> I couldn't find much in the way of debate when I searched for this except
> for a patch submission in Jan 2018. Was this debate in #ffmpeg-devel IRC?
> 
>> The debate comes down to this:
>> - it's clear that for the company, there's a huge advantage in patches
>> being upstream. For example, upstream does maintenance/upkeep for free.
>> - it's not clear whether there is any benefit to the community / project in
>> this patch being upstream. For example, how do we test this code in fate?
> 
> As far as I can tell, FATE doesn't test hwaccel codecs/filters. But
> we can donate some hardware for FATE testing if needed. We plan on
> continuous testing against master to ensure upstream FFmpeg changes
> don't break our specific changes. We may also have to donate some VMs to
> help with testing.
> 
>> The fact that something is for sale does not mean it's a general benefit to
>> the FFmpeg developer or user community.
> 
> There were some developers in #ffmpeg-devel IRC that thought we should upstream.
> 
> Our Netint Quadra hardware has been used by a large customer base including
> several large cloud service providers in production for high density video
> encoding and decoding.
> 
> There is a lot of interest in using FFmpeg with our ASIC hardware from regular
> developers. We've made our ASIC hardware available via Akamai cloud. You can
> provision a VM, then use our publicly available FFmpeg changes to work with our
> ASIC. Now regular developers can also easily build high density video encoding
> solutions with ffmpeg in the cloud.

I'm very much in favour of adding support. Assuming it is then also 
properly maintained and not just "dumped" on us.
Of course we can't reasonably add support for every single niche bit of 
hardware, but this seems mature and adopted enough to me to warrant it.


More information about the ffmpeg-devel mailing list