CUDA Programming
host: CPU
device: GPU
nvcc: CUDA 编译器
gridDim: 1个grid中有多少个block
blockDim: 一个block中有多少个线程
blockIdx: 当前线程所属的block序号，同一个block中的所有线程共享这一个blockIdx
threadIdx: 当前线程在当前block中的序号
dim3
| Executed on the: | Only callable from the: | |
|---|---|---|
__device__ float DeviceFunc() |
Device | Device |
__global__ void KernelFunc() |
Device | Host |
__host__ float HostFunc() |
Host | Host |