https://hgpu.org/?p=16838
gpuSPHASE - A shared memory caching implementation for 2D SPH using CUDA