CUDA中的基本块

Question

Is there a way to see the basic-block information for a CUDA source code? I am trying to compile a CUDA code so that the output it gives me is a file with all the basic-blocks listed. But so far,I haven't been able to find any option in nvcc which allows me to do that.

If we can extract the basic-block information from the IR (PTX), that also helps.

Answer 1

I was looking for a way to extract the basic-blocks or the control flow graph of a CUDA source. It seems there is a way, but we get the control flow in the form of an intermediate representation (PTX).

The "nvdisasm" has a way to extract the control flow graph of a kernel. More information can be found here: http://docs.nvidia.com/cuda/cuda-binary-utilities/index.html#axzz3lONRPxm5