kernel launch: invalid argument error

I’m just trying to manipulate the RGB values of each pixel in a photo. But when I try to launch a kernel I’m getting a “invalid argument” error.
this my kernel code:

__global__ void addKernel(uchar* ptr_BGR, uchar* ptr_GREY, int num_rows, int num_cols)
	int x = threadIdx.x;
	int y = blockIdx.y;
	int absolute_pos = num_cols*y*3 + x;
	uchar b = ptr_BGR[absolute_pos];
	uchar g = ptr_BGR[absolute_pos+1];
	uchar r = ptr_BGR[absolute_pos+2];
	ptr_GREY[absolute_pos] =  255-b;
	ptr_GREY[absolute_pos+1] =  255-g;
	ptr_GREY[absolute_pos+2] =  255-r;

and this is where I launch the kernel.

const dim3 blockSize(1, num_rows, 1);
	const dim3 threadsPerBlock(num_cols, 1,1);
	addKernel<<<blockSize, threadsPerBlock>>>(d_pt_BGR, d_pt_GREY, num_rows, num_cols);

    // Check for any errors launching the kernel
    cudaStatus = cudaGetLastError();
    if (cudaStatus != cudaSuccess) {
        fprintf(stderr, "addKernel launch failed: %s\n", cudaGetErrorString(cudaStatus));
        //goto Error;

the values of the num_rows and num_cols are both 900 for the images I use. Thanks in advance.