function inside the cuda kernel

Is there any ways i can have a function inside cuda kernel. I mean my cuda kernel gets pretty long and har开发者_JAVA百科d to debug at one point. Thanks.

yes, just mark function with __device__ and it will be callable only from GPU. Check CUDA Programming guide, section B.1 Here is the direct link