The Efficient Use of Graphic Accelerators to Implement High Accuracy FDTD Algorithms

The use of Graphics Processing Units (GPU) provided significant performance improvements to Electromagnetic simulations through the FDTD method. However, the current implementations only take into account the FDTD algorithms with low accuracy, as there are no reports in the literature of solutions that use more complex differential operators, capable of offering lower error rates. This work includes a study of how to implement FDTD on the GPU. Additionally, it presents a new programming technique focused on optimizing the use of complex differential operators. The practical tests show a performance improvement in the realization of high accuracy electromagnetic simulations.