<title>Problem Adaptation To Systolic Arrays</title>

Abstract

The systolic array architecture is known to make highly efficient use of hardware in evaluating certain matrix products provided that the matrices are banded strongly.However, this high efficiency can be degraded significantly if the matrices to be processed do not possess the narrow bandwidth feature, but assume a more general structure. This paper introduces and evaluates two techniques which in some instances can enhance systolic array efficiency.The approach effectively reduces to adapting problem structure so that it more naturally fits the systolic array architecture. Potential benefits from this approach are quantified and presented in graphical form.

Modified engagement method for matrix operation

“…In order to give clearly a comprehensive comparison of the performances of the systolic array and the engagement array, we now consider the matrix-vector multiplication case. Four systolic@engagement matrix-vector multipliers are considered here, i.e., the original systolic array, the Priester et al improved systolic array, 21 the Speiser-Whitehouse engagement method, 11 and the improved engagement version.…”

“…Priester et al 21 have given a detailed analysis of the original systolic method and their improved systolic method. For a thorough comparison between the Speiser-Whitehouse engagement method and the improved engagement version, we illustrate their arrangements in Figs.…”