Re: [Jack-Devel] issues with latency management with "complex" clients

On Wed, Feb 16, 2011 at 12:18:01PM -0600, Chris Caudle wrote:
> On Wed, February 16, 2011 9:30 am, Paul wrote:
> > JACK tends to assume that as long as port connections stay the
> > same, the graph is essentially static and unchanging. ...
> > As the track is switched between these two modes, the answers
> > to the two questions regarding latency change.
> ...
> > Client's like Ardour effectively change where the "edge" of the graph is
> 
> 
> Seems like changing modes is actually a special case of graph change,
> where the graph changes but the connected ports do not change.

its a latency recalculation. 
a graph change just includes a latency recalculation.

> 
> Would it be reasonable for applications to ignore it by default, but tell
> jack that the graph changed in cases where the app cares?  In the given
> example, Ardour would decide whether it mattered or not, and if it did
> call to jack to say "the graph changed, recalculate."  Other apps would
> not care, so they would never make the call.

we are finding an answer to whether ardour should tell jack whether the
latencies changed. thats what we are talking about.

> 
> When you say expensive, what kind of time are you talking about?  Going
> from playback to record takes a couple of clicks, so if you are talking
> seconds it would annoy a user, if you are talking tenths of a second it
> might not be noticeable.

we are talking about microseconds. highly unlikely that it takes a
milisecond.

> 
> Is part of the concern that it would cause clicks on jack 1, or that for
> large configurations the recalculation time scales with the graph size, so
> it might not bother me with a handful of tracks, but it would definitely
> bother someone with lots of tracks and dozens of soft synths connected?

it might cause a click. it involves locking the graph.

> 
> -- 
> Chris Caudle
> 
> 
> 
> 
> 
> Jack-Devel mailing list
> [hidden]
> http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org

-- 
torben Hohn

1297888508.869_0.ltw:2,a <20110216203448.GB8926 at siel dot b>