Performance penalty when invoking a cuda kernel

Question

I\'m wondering what the overhead of performing a cuda kernel call is in C/C++ such as the following:

somekernel1<<>>(args);
som         


        

Answer 1

If you are using Visual Studio Pro on Windows I sugest you run a test application using NVidia's Parallel NSight, I think it can tell you the time stamps from the method call to the real execution, in any case a penalty is inherent, but it will be negligible if your kernels lasts long enought.