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 hard to debug at one point. Thanks.