Propagation of Data Dependency through Distributed Cooperating Processes
RENSSELAER POLYTECHNIC INST TROY NY DEPT OF COMPUTER SCIENCE
Pagination or Media Count:
EPL is an equational programming language based on MODEL. In EPL, a computation to be performed by distributed cooperating processes is described as a set of specifications. The interconnections between processes are represented as a network graph. Processes data dependencies are derived from these interconnections. If cycles exist in the network graph which is the usual case these dependencies would impose additional internal scheduling constraints, created from a transitive closure of all internal dependencies in the processes in the cycle. Therefore there is a need to derive these additional internal scheduling constraints from the internal data dependencies of all the processes and the network interconnections between them. This task requires careful analysis, which in previous versions of EPL had to be performed by the user. This thesis describes an algorithm, and its implementation, to automatically derive the additional intra-process scheduling constraints to generate the correct implementation for the target machine. This algorithm can also be used in automatic repartitioning of a process specification.
- Computer Programming and Software