Prior to the start of this award, we had developed out-of-core algorithms for one-dimensional FFTs on a uniprocessor with multiple disks and for one-dimensional FFTs on a distributed-memory cluster. During the award period, we extended our results to multidimensional FFTs on a distributed-memory cluster. We examined both the dimensional method [BC99], which works on one dimension at a time, and the vector-radix method [Rin01], which works on a small piece of all dimensions at once. We designed and implemented efficient out-of-core algorithms for each of these cases. Subsequently, we answered theoretical questions about how to order and group the individual dimensions in the dimensional method [Fin01].