The architecture of MANIP: a parallel computer system for solving NP-complete problems

In this paper, we study the network architecture of MANIP, a parallel Machine for processing Non-deterministic Polynomial complete problems. The most general technique that can be used to solve a wide variety of NP-complete problems on a uni-processor system, optimally or sub-optimally, is the branch and bound algorithm. We have adapted and extended the branch and bound algorithm for parallel processing. The parallel branch and bound algorithm requires a combination of sorting and merging. A common memory to sort for a large number of processors can become a bottleneck in the system. We have proposed a system with distributed intelligence so that sorting can be carried out in a distributed fashion. A uni-directional ring network is proved to be-the optimal and most cost-effective inter-processor communication network when sorting is done by a hardware priority queue in each processor.