[Clam-devel] NaiveFlowControl: Missing case
David García Garzón
dgarcia at iua.upf.edu
Wed Feb 28 04:36:22 PST 2007
* Blocked source
Because not all buffers are emptied it may happen that at the time of
doing the single execution of a source it is not able to execute because
output buffer is full. This happens on 'Branch saturation' case but i might
be possible in other cases where the tokens are not consumed. But the
solution of this case might be also a solution of the Branch saturation case.
On Wednesday 28 February 2007 10:49:44 David García Garzón wrote:
> Some more use case:
>
> * Branch saturation
> Example: A branch ending in a monitor which has a lower Do
> granularity (it needs more Do's to process the same) than the branch ending
> on the Sink. FlowControl Do ends as soon as the Sink is executed and this
> happens on less loops than the monitor branch. So data in monitor branch
> don't get consumed and at last writting on the buffer to get into the
> branch is not possible.
>
> Any other one?
>
> On Dimarts 27 Febrer 2007, David García Garzón wrote:
> > Just to remember it for tomorrow. The current NaiveFlowControl we
> > implemented today does not support the following cases:
> >
> > * Filter with bigger size + a ever runable processing
> > Ever runnable processings can be a set of processing not connected
> > to callback interfaces (FileIn+Oscillator) or a port less processing such
> > control senders. The filter with bigger size makes some iterations end
> > without filling the sink but as always there are some processing to run,
> > Do() will run with no end.
> >
> > * Network without a sink
> > Like analysis networks or fileOut recording. As the finalization
> > condition is right now not having any sink. The flow control will just
> > exit without running any other processing than sources.
> >
> > David.
> >
> >
> >
> >
> >
> > _______________________________________________
> > Clam-devel mailing list
> > Clam-devel at llistes.projectes.lafarga.org
> > https://llistes.projectes.lafarga.org/cgi-bin/mailman/listinfo/clam-devel
More information about the clam-devel
mailing list