Efficient implementation of bidirectional associative memories on the extended hypercube

Implementation of bidirectional associative memories (BAMs) on various multiprocessor topologies is discussed. Studies reveal that BAMs can be implemented efficiently on the extended hypercube (EH) topology since the performance of the extended hypercube is better than that of the binary hypercube topology. The architecture of the BAM network and implementation aspects of BAMs on multiprocessor topologies are discussed. An overview of the EH topology and efficient execution of multinode broadcast and total exchange problems on the EH are considered. Results of the simulation studies are presented.<<ETX>>