Parallelization of the streamline simulation based on CUDA

To accelerate the streamline simulation and satisfy the real-time demands, in this paper, we proposed a method based on GPUs to parallelize the streamline simulation. CUDA architecture was used to implement the parallel algorithm on a single GPU and a multi-GPU computer. In our method, a grid is organized into a 2D array of blocks, and all threads in a block are organized into a 1D array, such that each thread in a block computes one streamline. To implement the method on multiple GPUs, the physical cell model is divided into sub-models to make the number of sub-models equal to the number of GPUs. The algorithm is applied to a Tóthian basin as an example. The experimental analysis shows that the parallel algorithm based on different numbers of GPUs has different accelerations. For a single GPU, the speedup reaches 170 times; and for five GPUs, it is 808 times, for a physical model with 40×106 cells. The conclusion is that GPUs can greatly accelerate the streamline simulation.