Over the last decades a large variety of dataflow solutions emerged along with the proposed models of computation (MoC), namely the synchronous dataflows (SDF). These MoCs are widely used in streaming based systems su...
详细信息
ISBN:
(纸本)9783642547348;9783642547331
Over the last decades a large variety of dataflow solutions emerged along with the proposed models of computation (MoC), namely the synchronous dataflows (SDF). These MoCs are widely used in streaming based systems such as data and video dominated systems. The scope of our work will be on consistent dataflow properties that can be easily demystified and efficiently determined with the outlined mapping approach between dataflows and Petri nets. Along with this strategy, it is also highlighted that it's of a major relevance knowing in advance the proper initial conditions to start up any SDF avoiding buffer space over dimensioning. The methodology discussed in this paper improves the outcomes produced so far (in Petri net domain) at design stage aiming at knowing the amount of storage resource required, as well as has a substantial impact in the foreseen allocated memory resources by any signal processing system at the starting point and also points out new directions to minimize the buffer requirements at design stage.
Dataflow process networks lead to different theoretical model approaches and have demonstrated their adequacy in data-dominated intensive systems, namely synchronous dataflows. Since their appearance, dataflow models ...
详细信息
ISBN:
(纸本)9781457704345
Dataflow process networks lead to different theoretical model approaches and have demonstrated their adequacy in data-dominated intensive systems, namely synchronous dataflows. Since their appearance, dataflow models became too focused and specialized in their target applications. The paper presents a set of translating mechanisms allowing the mapping from dataflow models into Petri nets. This mapping allows taking advantage of Petri nets well-known properties verification capabilities and enriching dataflow models concerning scheduler information and resource allocation. This allows one to find out some hidden embedded features (model semantics and syntax) not normally addressed in dataflow analysis tools, which is briefly characterized. Dataflow model translation into Petri net domain give support to attain the required resource allocation under dataflow static scheduling list. This scheme allows one to make conclusion in Petri net domain to be applied in dataflow models to foresee the necessary amount of storage resources for each arc. An application example is used to illustrate the concept and effectiveness of the outlined approach.
暂无评论