OpenCL GPU 100× acceleration of physical optics

The application of the graphics processor to computing antenna radiation patterns provides a substantial acceleration in speed. The same brute force physical optics (PO) code; e.g., a symmetric single reflector of two hundred forty wavelengths diameter, that runs single-threaded C/C++ on an Intel Core 2 Q8200 CPU and compiled to be optimized for speed, runs one hundred times faster when ported to OpenCL and run on an AMD HD5850 GPU hosted by the CPU system. Porting the code to OpenCL involves opening the processing loops to be run in parallel and some additional code to host the GPU in the CPU system. The ability to accelerate antenna pattern computations by a factor of one hundred provides real substantial benefits when many patterns need to be computed repeatedly, such as when optimizing a reflector system.