Multithreaded systolic computation

In this paper we propose a synergy of processing on parallel processor arrays (systolic or SIMD) and multithreading, termed multithreaded systolic computation. The multithreaded systolic computation principle is demonstrated on a programmable systolic array executing a set of linear algebra algorithms. We demonstrate that multithreaded systolic computation can provide throughput improvements that asymptotically approach the number of simultaneously executable threads.