Implementing SAR image processing using backprojection on the Cell Broadband Engine

Many compute intense algorithms once deemed impractical are now being found realizable on modern multicore processors due to their massive compute power. An example of such an algorithm is the backprojection method for synthetic aperture radar (SAR) image processing. Backprojection provides many benefits compared to the other method of SAR processing, the polar format algorithm (PFA). This paper studies the implementation of convolution backprojection (CBP) on the Cell Broadband Enginetrade (Cell/B.E.) architecture. The paper will study how the implementation of CBP must be changed to deal with the programming issues related to using the Cell/B.E. processor. The programming issues that are studied include partitioning the work across a heterogeneous set of processors (SPEs and PPEs), managing the heterogeneous memories (small local storage and large off-chip XDR memory), overlapping processing with communication and read/writes from the XDR, and optimizing the SPE and PPE code to highly utilize all ALU paths. The Gedae programming language and multithreading compiler is used to address these programming issues. Performance of the CBP implementation on the Cell/B.E. processor using Gedae is presented.