We propose a new algorithm and architecture for performing divisions i
n residue number systems (RNS). Our algorithm is suitable for RNS with
large moduli, with the aim of manipulating very large integers on a p
arallel computer or a special-purpose architecture. The two basic feat
ures of our algorithm are the use of a high-radix division method, and
the use of a floating-point arithmetic that should run in parallel wi
th the modular arithmetic.