Systematic Mapping of Uniform Recurrent Equation onto Linear

This paper presents a systematic procedure to mapping uniform recurrent equations on to linear systolic arrays. In this method, a bijective transformation must be computed first. Then, by applying this transformation over the N-Dim index space of a uniform recurrent equation, a new index space will be obtained that includes a 1-Dim location and a (N-1)-Dim time. By using the location and time dimensions, the location of performing each computation of uniform recurrent equations on the processor array and the time to performing each computation can be determined, respectively. Each computation time is specified by a (N-1)-Dim vector. Therefore, for reaching the scalar times correspond to logical times, we can use full trees first. Then we compute the times correspond to time vectors.