cudaOccupancyMaxPotentialBlockSize ( int* minGridSize, int* blockSize, 
          T func, size_t dynamicSMemSize = 0, int  blockSizeLimit = 0 ) 

    Returns grid and block size that achieves 
    maximum potential occupancy for a device function func.

Parameters

    minGridSize
        - Returned minimum grid size needed to achieve 
          the best potential occupancy 
    blockSize
        - Returned block size 
    func
        - Device function symbol 
 
Function description:

    Returns in *minGridSize and *blocksize a suggested grid size and
    block size that achieves the best potential occupancy  

  int minB;  // Min grid size
           
  int T;     // block size = # threads in a thread block

  cudaOccupancyMaxPotentialBlockSize( &minB, &T, matrixMult, 0, 0);

  printf("Computed: minB = %d, T = %d\n", minB, T);

  int B = ceil((float) N*N / T ); // Round up to integral grid size

  // Run kernel on the GPU using NBlks block, K thread/per block
  matrixMult<<<B, T>>>(a, b, c, N);

  // Wait for GPU to finish before accessing on host
  cudaDeviceSynchronize();