A Unified Approach to a Class of Data Movements on an Array Processor
IEEE Transactions on Computers
Ordered fast Fourier transforms on a massively parallel hypercube multiprocessor
Journal of Parallel and Distributed Computing
Unified Architecture for Divide and Conquer Based Tridiagonal System Solvers
IEEE Transactions on Computers
Array Permutation by Index-Digit Permutation
Journal of the ACM (JACM)
A VLSI Constant Geometry Architecture for the Fast Hartley and Fourier Transforms
IEEE Transactions on Parallel and Distributed Systems
Parallel Architecture for Fast Transforms with Trigonometric Kernel
IEEE Transactions on Parallel and Distributed Systems
Notes on Shuffle/Exchange-Type Switching Networks
IEEE Transactions on Computers
Theory and Applications of Digital Speech Processing
Theory and Applications of Digital Speech Processing
Application-specific architecture for fast transforms based on thesuccessive doubling method
IEEE Transactions on Signal Processing
ACM Transactions on Design Automation of Electronic Systems (TODAES)
Hi-index | 0.00 |
The most effective use of mesh connected computers is achieved by paying careful attention to the organization of the storage and movement of data. For an important class of algorithms the formalization of the different operations they perform lead to an unified treatment for them and may result in important simplifications. In this work we apply this point of view to the Fast Fourier Transform (FFT). In particular, we present a unified view of a set of FFT algorithms on mesh connected computers with non-shared memory. To this end we use a combination of two techniques, 'mapping vector' and index-digit permutations, which allow us to define the organization of the storage and movement of data for any FFT algorithm whose radix is a power of 2. The methodology we have employed is general and can be applied to other algorithms obtained through the divide and conquer strategy.