A Schedulability-Preserving Transformation Scheme from Boolean- Controlled Dataflow Networks to Petri Nets

This paper presents a transformation scheme from Boolean-controlled dataflow (BDF) networks to Petri nets (PN) that has the following properties. A BDF has a bounded length schedule, if and only if the transformed Petri net has a bounded length schedule. A BDF has a bounded memory schedule if the transformed PN has a bounded memory schedule. The existence of bounded memory schedule of BDF is proved to be undecidable. This transformation scheme provides a way to construct a bounded memory schedule of a BDF using exsiting PN scheduling techniques. In BDF models tokens are associated with data, and actors communicate through FIFO channels. A general PN semantics does not distinguish tokens and does not preserve the order of tokens in a place. The proposed method uses different places to hold the Boolean tokens with different values. The ordering of Boolean tokens at each control input port is preserved by enforcing that new Boolean tokens are not produced until the previous Boolean token is consumed. Initial Boolean control tokens are handeled by use of a preamble. Furthermore, based on the transformation, the consistence notion originally defined in BDF is extended to PN.