Monte Carlo Dose Calculation using GPU-Based parallel processing

Abstract

Recently, it became possible to operate physical phenomenon using Graphics Processing Unit (GPU), and Monte Carlo calculation methods came to be researched about shortening the computing time using GPU positively. This report shows how to significantly accelerate 3D dose calculation of photon beam using Graphics Processing Unit (GPU). We describe GPU parallel processing method for dose simulation based on NRCC DOSXYZnrc. The GPU version of dose calculation algorithm was implemented using NVIDIA CUDA programming environment and executed on an NVIDIA 8800GTX GPU. And the streaming version for Single Instruction Multiple Data (SIMD) in CPU was implemented using Intel C++ and executed on the CPU above mentioned. For dose calculation, 403 cm3 of cube water phantom were coded using DOSXYZnrc, the GPU and the streaming version. In addition, because the calculation of one particle was independent in the Monte Carlo method of EGSnrcMP4), the parallel computation inter PC is possible. Then, three dose calculation systems of same spec were constructed. In this report, dose calculation time of each algorithm was measured. The increase of the dose calculation speed using the streaming version was about three times against normal version. And it was consumed that the calculation speed of GPU version was about four times against normal one. Moreover, it was shown to be able to increase the calculation speed by computing in parallel with the PC equipped with proposed method.