CUDA 学习笔记

二维grid下，二维block的线程全局id获取

线程分配如下所示：

线程分配格式为：

gridDim.x = 2
gridDim.y = 2

blockDim.x = 2
blockDim.y = 2

那么线程的global_id可以表示为：

global_index = row_offset + block_offset + tid

row_offset = number of threads in a row * blockIdx.y
= blockDim.x * blockDim.y * gridDim.x * blockIdx.y;
block_offset = number of threads in a block * blockIdx.x
= blockDim.x * blockDim.y * blockIdx.x