[vlc-devel] [PATCH 3/8] picture_pool: use the internal condition/lock indirectly

[Steve Lhomme]

> For now the issue is with hardware decoders. We could have a picture 
> pool exactly the size of the surface pool and only wait for pictures 
> (and then attach the surface to it). If we have a picture available that 
> means we have a matching surface available, no need for a blocking call.
> The problem is to use the proper pool size for decoder_t->out_pool. It 
> may even be dynamic. The pool size required could be passed via the 
> format_update() callback.
> 
> That would mean, contrary to what we said, decoders cannot handle their 
> own picture pool. They have to rely on the owner to provide the 
> pictures. Right now only nvdec has its own picture pool, so it's not a 
> deal breaker.

This solution seems to work and is much less intrusive.