Here, the actual computational work for each particle (and hence each thread) is constant as one can use the grid size and cell size to easily compute which cell a particle lies in. The major time cost comes from accessing and writing into the memory. Since the memory is shared and each cell is updated by multiple threads, these updates have to be done sequentially.
Here, the actual computational work for each particle (and hence each thread) is constant as one can use the grid size and cell size to easily compute which cell a particle lies in. The major time cost comes from accessing and writing into the memory. Since the memory is shared and each cell is updated by multiple threads, these updates have to be done sequentially.
This comment was marked helpful 0 times.