Scheduling pipelined communication in distributed memory multiprocessors for real-time applications

This paper investigates communication in distributed memory multiprocessors to support tasklevel parallelism for real-time applications. It is shown that wormhole routing, used in second generation multicomputers, does not support task-level pipelining because its oblivious contention resolution leads to output inconsistency in which a constant throughput is not guaranteed. We propose scheduled routing which guarantees constant throughputs by integrating task specifications with flow-control. In this routing technique, communication processors provide explicit flowcontrol by independently executing switching schedules computed at compile-time. It is deadlock-free, contention-free, does not load the intermediate node memory, and makes use of the multiple equivalent paths between non-adjacent nodes. The resource allocation and scheduling problems resulting from such routing are formulated and related implementation issues are anal yzed. A comparison with wormhole routing for various generalized hyp ercubes and tori shows that scheduled routing is effective in providing a constant throughput when wormhole routing does not and enables pipelining at higher input arrival rates.