The actual throughput, latency, and cost-effectiveness of feedback systems, concurrent recursive systems, and generalized concurrent systems are analyzed, and an overall design methodology is proposed. The differences between the throughput speedup, the iteration bound speedup, and the intrinsic throughput-latency tradeoffs in concurrent design are discussed. It is shown that the iteration bound can be relaxed at the expense of latency and that a concurrent system can achieve the relaxed iteration bound only if the recursion hardware is not pipelined and sufficient concurrency (sometimes higher than the improvement in the iteration bound) is available.
展开▼