The multilevel fast physical optics (MLPO) algorithm attains a computational complexity comparable to that of the fast Fourier transform (FFT) based techniques by using hierarchical domain decomposition and phase compensated interpolation approach. In this communication we present an optimized distributed memory algorithm, obtained by partitioning not only the radiating aperture but also the grid of far field directions. Such a scheme leads to improved speed and reduced memory requirements. The performance of the proposed approach is evaluated in terms of load balance and communication cost, and tested in the context of very large antenna problems.