The gravity parallel computation based on GPU

The high order spherical harmonic function is used to calculate the gravity, but the amount of computation is large and it takes a long time. This paper analyses the characteristics of high order spherical harmonic function calculation, discusses the CUDA (Compute Unified Device Architecture) parallel computing method for gravity, designs the kernel function and parallel computing process, and completes the gravity calculation based on the NVIDIA Corporation's GPU. Based on the speedup ratio index, the paper discusses the factors that affect the calculation speed. Besides GPU hardware, the number of threads per block and data exchange capacity between host and GPU device are also major factors. The computational bottleneck is calculation of Legendre coefficients. Speedup ratio shows that the work is valuable.