Effect of different arbitration priority strategies on throughput of a memory-coupled multiprocessor with HBO

The effect of various arbitration priority strategies on the overall throughput of a multiprocessor with hierarchical bus organization (HBO) is considered. Each bus in the hierarchy is memory coupled so that the lower level buses or processors connected to it can access the shared memory according to some priority scheme. The performance of the multiprocessor system is investigated for four priority allocation schemes, namely, round-robin, FIFO (first in, first out), rotating, and fixed priority. Fixed priority shows the best performance as revealed by simulation results.<>