[FFmpeg-devel] [PATCH] avutil/WIP: add AVAsync API

[wm4]

On Wed, 11 Nov 2015 08:45:52 +0100
Hendrik Leppkes <h.leppkes at gmail.com> wrote:

> On Wed, Nov 11, 2015 at 1:27 AM, Clément Bœsch <u at pkh.me> wrote:
> > From: Clément Bœsch <clement at stupeflix.com>

> > I hope this is going in a direction most developers like. If it's not
> > the case, I'm of course open to any reworking but please be specific.  
> 
> I'm quite confused by this approach, it doesn't really seem to
> correspond at all to what we talked about earlier.
> You can't really put a high-level async API on our low-level sync API,
> we should start re-thinking the low level API first.
> 
> The first goal should be to get to a new low level API that implements
> decoupled m:n output, then you could build a high-level API on top of
> that, if you wanted something like that.
> Such a low level API would easily integrate into other frameworks,
> while this async API seems to be rather rigid and may be rather
> troublesome to integrate into existing workflows.
> 

I have to agree. It's just not what I expected. While I think having a
higher level API in FFmpeg would be very good, the async m:n API is
what we really need.

Besides that, the presented API looks pretty rigid and inflexible. This
is basically the wrong way to start. If you start with a high level
API, all effort should be put into making that API _good_, and not
making its implementation a playground for low level API improvements.
If the goal is fixing VideoToolbox, ffplay.c could probably be used for
experiments. (It runs a separate decoder thread, so it might be ideal
for trying such async things.)