A low-delay common subexpression elimination algorithm for constant matrix multiplications over GF(2m)

Abstract

In this work, a low-delay common subexpression elimination (LDCSE) algorithm is proposed. Unlike the previous common subexpression elimination (CSE) algorithms, which mostly focus on area, the aim of the LDCSE algorithm is to optimize both area and delays in hardware implementations of constant matrix multiplication over the GF(2m). A lower computational complexity gate-level delay computing method is proposed to compute delays according to the transformed constant matrices. The new CSE algorithm employs greedy algorithm and proposed delay computing method to search the set of common subexpressions with minimal area-delay-product (ADP). The Experimental results have shown that the LDCSE algorithm achieves smaller ADP compared with previous works.