An Efficient Mapping of Directed Graph Based Computations onto SIMD Hypercube Arrays and Applications