
Real-Time GPU Fluid Dynamics 번역 (3)
·
GPU Programming/CUDA fluidsGL
4.4 Thread Configuration First, we choose a warp-size multiple for the block horizontal size: 32 · 2 = 64. Then we have to choose a block vertical size. Because the typical maximum threads per block is 1024, we cannot choose 64 for the vertical dimensions, as we would have 64 · 64 = 4096 threads! For this reason we choose a 64x4 block size (256 threads per block). We divide the domain size in 64..