[vlc-devel] [PATCH 12/14] filter_chain: add a default implementation for the video_allocator

[Rémi Denis-Courmont]

Le tiistaina 30. heinäkuuta 2019, 10.14.17 EEST Steve Lhomme a écrit :
> On 2019-07-29 21:18, Rémi Denis-Courmont wrote:
> > Le maanantaina 29. heinäkuuta 2019, 22.08.58 EEST Alexandre Janniaux a 
écrit :
> >> While I agree that we probably don't need flow control and that image
> >> number is probably dominated, the starvation issue seems a real issue to
> >> tackle. What are the solution against starvation on Windows ? Should we
> >> just fail if we got starved ?
> > 
> > In general, it is the halting problemm which is to say that the general
> > case cannot be solved and assumptions have to be made.
> > 
> > If we need to make assumptions, we might just as well assume that filters
> > are well-behaved, as indeed they have so far been. Ideally, a filter that
> > needs to allocate surfaces (rather than modify input surface in-place)
> > should allocate a suitably large pool for itself. And that turns out
> > incompatible with filter_NewPicture().
> 
> I don't see how. This particular patch adds the possibility for filters
> to provide their own (output) allocator. A filter could very well create
> a pool and have its allocator pick pictures in that pool. That still
> goes through filter_NewPicture, because the filter doesn't know if it
> has to use pictures from the outside or it can use its own.

That makes no sense. The "outside" cannot know if the filter allocates or not. 
Already now, both cases have to work - it has more or less always been that 
way with filters.

So far only converters could be assumed to allocate pictures - and in the push 
model, that would be from the video context, not from the downstream/
filter_NewPicture().

> 
> > Since the filter chain is (partially) dynamic, we cannot even rely on
> > filters telling how many "extra" surfaces they need - the total value
> > could change in the middle of the stream. Instead, the filter ought to
> > allocate its surfaces during initialization, and potentially refuse to
> > start if it cannot succeed - but not break the whole pipeline.
> 
> I'm not sure we can estimate this amount easily. For example what would
> be the amount an adjust filter has to allocate ?

The adjust filter needs 0 surfaces since it operates in place.

Point being anyway that only the filter can know that, if anything. Since the 
filter chain is dynamic, it cannot be taken into account when the decoder and 
display are set up. It may be that even the filter does not know how many 
pictures it needs, but then there is nothing to fix; it will just maybe or 
maybe not work by allocating on the fly.


-- 
レミ・デニ-クールモン
http://www.remlab.net/