On Sun, 2011-02-13 at 22:05 +0200, Colin Guthrie wrote: > With this push based approach, you do loose some individual granularity, > but the net volume of the underlying h/w should be the same as your > approach. What granularity would I lose? I think your suggested logic would be quite equivalent to the one that I originally proposed. > The concern I have with the approach outlined, is that it adds > complexity to the core and I'm not 100% sure how far the chain can go > (e.g. can you have a filter-sink1->filter-sink2->filter-sink3->hw-sink > pipeline? - with a push model this is possible). It's possible with the pull model too - the filter sinks are always traversed recursively. About complexity - I haven't done a thorough analysis of your suggestion, but I would guess that it would be a little bit simpler. There would still be a significant amount of added complexity in the core, though. I'll finish the patch using the original logic first, and if you want, I can probably do another version to see how much the push model will differ. -- Tanu
