[kepler-dev] Actor is not ready to fire messages

Christopher Brooks cxh at eecs.berkeley.edu
Tue Jan 8 12:34:12 PST 2008


More specifically, the issue is that some actors (BooleanSwitch)
can work in SDF under certain circumstances.  So, we could have
BooleanSwitch implement an interface that signifies that it is
not-ok-with SDF, but that would deny model developers use of
BooleanSwitch under situations where it would be ok.

Another alternative would be to have an annoying dialog box.
This would be bad.

A better idea would be to borrow something from Lisp, where we have
speed and safety settings.  If the safety setting was high, we would
throw an error if an SDF model used this actor.

_Christopher
--------

    
    
    This is actually a rather interesting research problem.
    We have two major sets of (partial) results on this:
    
     - The behavioral interfaces that Yuhong Xiong developed
       in his PhD thesis and several papers on the Ptolemy website.
       These use interface automata to specify actor and director
       assumptions and behaviors, and formal composition to check
       compatibility. It is effectively a "behavioral type system."
       This has never really been "productized" in the Ptolemy tree.
       It's just a concept demonstration phase.
    
     - The Cal actor language, developed by Joern Janneck while he
       was a postdoc here. One of the main motivations of Cal was that
       actor definitions could be statically analyzed for token
       production/consumption behavior (unlike Java actors, where
       the question is actually undecidable).  This static analysis
       has never been implemented, but there would be considerable
       interest in doing so.  Cal has been adopted by MPEG as the
       specification language for all video and audio coding standards
       (!) and is also being extensively developed at Xilinx as a
       hardware design language for FPGAs.
    
    Anybody want a PhD thesis topic?
    
    Edward
    
    At 11:08 AM 1/8/2008, Timothy McPhillips wrote:
    >Hi Christopher,
    >
    >The idea of requiring that actors declare SDF-safety by implementing  
    >a Java interface is interesting.  There seem to be a number of  
    >related issues it'd be nice to address at some point.
    >
    >A general issue, it seems to me, is that it can be hard for users  
    >(developers included) to know what components (actors, directors,  
    >etc) can be used productively together. Knowing what actors are  
    >compatible with which directors is one example of this issue.   
    >Another example is that different actors may be incompatible with  
    >each other due to differences in how data is managed (ports vs  
    >parameters, sequences of tokens vs arrays of tokens when handling  
    >variably-sized collections of data, etc).  One may find an actor that  
    >sounds like it does what one desires, but due to differences in how  
    >data is consumed or produced it may turn out that the actor won't  
    >work with the actors (or director) one has already selected for use  
    >in a workflow.   So there may be multiple dimensions in which  
    >component incompatibility may arise.
    >
    >Examining the documentation or source code for actors can reveal why  
    >certain components don't work well together, but I wonder if there  
    >isn't a more general solution that makes it easy to see what actors  
    >and directors are meant to work well together in the first place.
    >
    >What do you think?
    >
    >Tim
    >
    >On Jan 8, 2008, at 7:51 AM, Christopher Brooks wrote:
    >
    >> Are you using the SDF director?
    >> My suggestion is to not use BooleanSwitch in SDF.
    >> As to why, see Dan Higgins' explanation in
    >> http://www.kepler-project.org/Wiki.jsp?page=DirectorsInKepler
    >>
    >> My understanding is that BooleanSwitch may be used in SDF, but only
    >> under carefully controlled situations.  This issue seems to come up
    >> fairly regularly with BooleanSwitch, the documentation should be more
    >> clear, I'll see what I can do.  Maybe we should disable BooleanSwitch
    >> under SDF, or at least have it print a message or something.  Perhaps
    >> actors that are SDF safe should implement a specific interface, or
    >> actors in actor/lib that are not SDF safe should implement a specific
    >> interface.
    >>
    >> _Christopher
    >>
    >> --------
    >>
    >>     Hi,
    >>
    >>     I create two custom actor, one of them reads a binary file and  
    >> other
    >>     writes the input into a netcdf file (i also update my actors  
    >> according
    >>     to earlier suggestions about stopping workflow) and the workflow
    >>     structure is similar to BinaryFileRead and BinaryFileWrite but  
    >> in my
    >>     case there is only one file which has multiple time step data.  
    >> I am
    >>     using Boolean Switch actor to stop my work flow but at the last
    >>     iteration it gives me "Actor is not ready to fire" messages.  
    >> So, I check
    >>     the "end of file" port of my custom binary file reader actor using
    >>     display actor and it seems ok.
    >>
    >>     any suggestion will be helpful,
    >>
    >>     best wishes
    >>
    >>     --ufuk
    >>     _______________________________________________
    >>     Kepler-dev mailing list
    >>     Kepler-dev at ecoinformatics.org
    >>     http://mercury.nceas.ucsb.edu/ecoinformatics/mailman/listinfo/ 
    >> kepler-dev
    >> --------
    >> _______________________________________________
    >> Kepler-dev mailing list
    >> Kepler-dev at ecoinformatics.org
    >> http://mercury.nceas.ucsb.edu/ecoinformatics/mailman/listinfo/ 
    >> kepler-dev
    >
    >_______________________________________________
    >Kepler-dev mailing list
    >Kepler-dev at ecoinformatics.org
    >http://mercury.nceas.ucsb.edu/ecoinformatics/mailman/listinfo/kepler-dev
    
    ------------ 
    Edward A. Lee
    Chair of EECS and Robert S. Pepper Distinguished Professor
    231 Cory Hall, UC Berkeley, Berkeley, CA 94720-1770
    phone: 510-642-0253, fax: 510-642-2845
    eal at eecs.Berkeley.EDU, http://www.eecs.berkeley.edu/Faculty/Homepages/lee.h
   tml  
--------



More information about the Kepler-dev mailing list