Using cuda-memcheck with Matlab CUDA mex-files

I have been able to profile Matlab CUDA mex-files by the procedure in

Profiling CUDA Matlab Mex file codes

Now, I'm trying to use cuda-memcheck with Matlab CUDA mex-files by a procedure which is conceptually the same, but up to now I haven't been fully successful. Let me explain the problem better.

I'm concerned with the following simple kernel

__global__ void square_array(double *a, int N)
{
  int idx = blockIdx.x * blockDim.x + threadIdx.x;
  if (idx<N) a[idx] = a[idx] * a[idx];
}

I'm running cuda-memcheck by the following command line

cuda-memcheck "C:\Program Files\MATLAB\R2010b\bin\win64\MATLAB.exe" -nojvm -nosplash -r TestCudaMemCheck

Everything works fine. But if I artificially introduce the following error

if (idx<N) a[5000] = a[idx] * a[idx];

when the array size is less than 5000, then the Matlab window is stuck and becomes not responding.

I'm using CUDA 5.0 under Windows 7.

How could I solve the problem? Thanks in advance.