Stream processing has been a very active field in parallel programming for its suitability to expressthe concurrent architecture in embedded systems. Caused by its concurrent reasoning features,stream programming frameworks are built on some abstract models of computation (MoCs) tohandle the complexity and unpredictability. To allow us focus on the essential issues of time,communication and synchronisation of the parallel tasks, the support from a sound heterogeneousMoCs framework to stream application system is still in need. ForSyDe is our high levelexecutable design framework to express multi-computational-models, based on stream processingconcept. It is a heterogeneous diagram to describe intricate application behaviors, and offers crossdomain analysis features to support multi-domains integration and optimization. A case study inForSyDe framework shows that the communication structure of a stream application in SDFdomain could be migrated to the synchronous domain without any extra work on its computationfunctions. To integrate it with our work on a communication based NoC simulator, we believesome more interesting design exploration work could be done on the analysis of communicationand computation efforts, besides power issues.
[1]
Michael I. Gordon,et al.
Exploiting coarse-grained task, data, and pipeline parallelism in stream programs
,
2006,
ASPLOS XII.
[2]
Axel Jantsch,et al.
Modeling embedded systems and SoCs - concurrency and time in models of computation
,
2003,
The Morgan Kaufmann series in systems on silicon.
[3]
Sanjay Ghemawat,et al.
MapReduce: Simplified Data Processing on Large Clusters
,
2004,
OSDI.
[4]
Manuel M. T. Chakravarty,et al.
The haskell 98 foreign function interface 1
,
2003
.
[5]
Ingo Sander,et al.
System Modeling and Design Refinement in ForSyDe
,
2003
.
[6]
William Thies,et al.
StreamIt: A Language for Streaming Applications
,
2002,
CC.