Low power architectures for digital signal processing algorithms requiring
inner product computation are presented. In the first step a power efficien
t memory organization exploiting data reuse is determined. In the second st
ep an order of evaluation of the partial products that reduces the switchin
g activity at the inputs of the computational units is derived, Information
related to both coefficients which are static and data which are dynamic,
is used to drive the reordering of computation. Experimental results for se
veral signal processing algorithms prove that the proposed techniques lead
to significant savings in net switching activity and thus in power consumpt
ion. (C) 2000 Elsevier Science B.V. All rights reserved.