$ cat t571.cu
#include <stdio.h>
#define DSIZE 16
#define BLACK 0
#define RED 1

__global__ void red_black_ordering(float *d_grid, int grid_row, int value, float *d_diff1)

{
        int index = blockIdx.x * blockDim.x + threadIdx.x;
        float temp = 0.0;
        int row = index/grid_row; // what row we're currently on
        int col = index - (row * grid_row); // what col we're currently on

        if (row != 0 && col != 0 && row != grid_row-1 && col != grid_row-1 && (row+col)%2 == value) {
                temp = d_grid[index];
                d_grid[index] = 99;
                //d_grid[index] = .2 * (d_grid[index] + d_grid[index-1] + d_grid[index+1] + d_grid[index-grid_row] + d_grid[index+grid_row]);
                d_diff1[index] = abs(d_grid[index] - temp);
        }
}

int main(){

  float *d_grid, *h_grid, *d_diff, *h_diff;
  int size = DSIZE*DSIZE;
  int row = DSIZE;
  h_grid = (float *)malloc(size*sizeof(float));
  h_diff = (float *)malloc(size*sizeof(float));
  cudaMalloc(&d_grid, size*sizeof(float));
  cudaMalloc(&d_diff, size*sizeof(float));
  cudaMemset(d_grid, 0, size*sizeof(float));
  cudaMemset(d_diff, 0, size*sizeof(float));
  int threads_per_block = 256;
  int blocks_per_grid = (int)( ( size + threads_per_block - 1) / threads_per_block);
  red_black_ordering <<< blocks_per_grid, threads_per_block >>> (d_grid, row, BLACK, d_diff);
  cudaMemcpy(h_grid, d_grid, size*sizeof(float), cudaMemcpyDeviceToHost);
  cudaMemcpy(h_diff, d_diff, size*sizeof(float), cudaMemcpyDeviceToHost);
  for (int i = 0; i < row; i++){
    for (int j = 0; j < row; j++)
      printf("%.0f ", h_diff[(i*row)+j]);
    printf("\n");}
  return 0;
}
$ nvcc -arch=sm_35 -o t571 t571.cu
$ ./t571
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 99 0 99 0 99 0 99 0 99 0 99 0 99 0 0
0 0 99 0 99 0 99 0 99 0 99 0 99 0 99 0
0 99 0 99 0 99 0 99 0 99 0 99 0 99 0 0
0 0 99 0 99 0 99 0 99 0 99 0 99 0 99 0
0 99 0 99 0 99 0 99 0 99 0 99 0 99 0 0
0 0 99 0 99 0 99 0 99 0 99 0 99 0 99 0
0 99 0 99 0 99 0 99 0 99 0 99 0 99 0 0
0 0 99 0 99 0 99 0 99 0 99 0 99 0 99 0
0 99 0 99 0 99 0 99 0 99 0 99 0 99 0 0
0 0 99 0 99 0 99 0 99 0 99 0 99 0 99 0
0 99 0 99 0 99 0 99 0 99 0 99 0 99 0 0
0 0 99 0 99 0 99 0 99 0 99 0 99 0 99 0
0 99 0 99 0 99 0 99 0 99 0 99 0 99 0 0
0 0 99 0 99 0 99 0 99 0 99 0 99 0 99 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
$