Trade-Offs In Mapping FFT Computations onto Fixed Size Mesh Processor Array