I have an algorithm whose computations are realised using MATLAB. Now, I am porting this algorithm to GPU.
Since inputs are read using MATLAB, and data transfer from MATLAB to GPU card is performed by “mexcuda”.
After this using CUDA 9.0 version kernels are developed.
Now, I would like to invoke profiler (nvprof) to profile tasks run by MATLAB and CUDA 9.0. Is this possible?
Could anyone suggest how to profile MATLAB computations using nvprof?