In many scientific simulation codes, the bulk of the floating-point arithmetic required is done by a small number of compact computational kernels. In this paper, we explore the potential use of configurable computers to ...