Analyse kernel executed from another kernel

Is it possible to analyse the performance of a kernel executed from another kernel on a Titan V in windows 7?

Yes, you should be able to profile kernels launched via dynamic parallelism. However, you won’t be able to select or profile only the device-launched kernel. For example, in the following scenario, you will only see profile results for a single kernel, “batch_launch”, but it will include data for the complete tree, including “entry”.

__global__ void
entry( int* foo )
{
    foo[threadIdx.x] = threadIdx.x;
}

__global__ void
batch_launch( int *foo )
{
    entry<<<1, N>>>( foo );
}