Efficient parallel architecture for highly coupled real-time linear system applications
C. C. Carroll, Sajib Barua, Abdollah Homaifar · 2003
The development of a systematic procedure for evolving a computational model that is particularly amenable for parallel processing in a multiprocessor environment is described. Any real-time system may be represented in general by a set of differential equations which govern the dynamic behavior of the system. As a specific example, a prototype real-time control problem is modeled as a set of differential equations. These are mapped onto a task graph which is then allocated to a set of processors in accordance with an allocation algorithm. This is followed by a verification and comparison stage wherein the results of such a mapping are compared with those of traditional uniprocessor methods in terms of speed-up ratio, efficiency, and average processor utilization. Results of simulation and scheduling are compared on the basis of standard performance indices.>