Problem with pgprof

Hello,

I have an application with some CUDA kernels that runs without problems. However, I am not able to profile it with pgprof that returns the following message:

0: ALLOCATE: 0 bytes requested; status = 0(no error)

What could be the reason for this message? I checked all the variables that are allocated in the device, and they are all correctly allocated.

Thanks

Hi Henrique,

What CUDA version did you build with and what CUDA driver version do you have?

It’s possible that if there’s a CUDA mismatch, this could cause this type of error.

Note that you may consider moving to using NVIDIA’s Nsight Systems profiler (NVIDIA Nsight Systems | NVIDIA Developer). Nvprof (pgprof is really a rebranded version of nvprof), is being discontinued and replaced by Nsight.

-Mat