2-D FFTs of Large Images with the AP-120B

摘要

The issues involved in implementing a 2-D FFT on the FPS AP-120B array processor when the data memory available is less than the image size are investigated. After a brief review of the alternative techniques that have been proposed in the literature (with matrix transposition, without matrix transposition, vector radix), a recently developed two-level implementation is described. An analysis of the CPU and I/O requirements is given, showing that this algorithm is significantly superior to existing methods due to the reduced I/O requirements. 2 figures, 1 table. (ERA citation 05:017305)