Room: AAPM ePoster Library
Purpose: develop a GPU based parallel computation algorithm for increasing Monte-Carlo calculation speed based on EGSnrc/DOSXYZnrc code package.
Methods: computational power by parallelizing the simulation with multiple GPU threads reduces the time required to reach the desired uncertainty in MC simulation. DOSXYZgpu, a GPU implementation of EGSnrc code written in CUDA Fortran as an algorithm. The work relies on a well-validated and popular code among medical physicists, EGSnrc/DOSXYZnrc. In order to transport particles between two consecutive interactions, we developed the included algorithm by handling several thousands of histories per warp (instead of 32 histories per warp) and adding an intermediate loop.
Results: several homogeneous as well as voxel phantoms, DOSXYZgpu implementation is compared with the original sequential EGSnrc/DOSXYZnrc. Maximum speed-up of 205 times achieved while the accuracy of the simulation is preserved.
Conclusion: t-test statistical analysis and gamma-index indicates that for more than 95% of the voxels, there is no significant difference between the results obtained from the GPU and the CPU.