creating pipelines on the fly.

[Tim Müller]

On Fri, 2016-07-08 at 08:07 -0700, Julia Smith wrote:

Hi Julia,

> I am curious what kind of costs (and limitations) there might be of
> creating pipelines on the fly, the scenario being playing a video,
> then adding another pipeline to play a completely different piece of
> video content. The output of the pipelines could be mixed or read off
> to independent consumers (files or to textures).
> 
> Is there any general (aka heavy) overhead in defining a pipeline but
> leaving it idle?

There shouldn't be a heavy overhead in defining pipelines but leaving
them idle (ie. in NULL state). They'll use up some memory for the
objects, but most resources (buffers, threads, filedescriptors, etc.)
will only be allocated/opened once you start up the pipeline.

Creating pipelines "on the fly" should also not be a problem.

It depends a bit on your streams and the target hardware how many
concurrent pipelines you can run of course, but it's not an unusual
thing to do at all.

Cheers
 -Tim

-- 
Tim Müller, Centricular Ltd - http://www.centricular.com