__global__  void ArrayMul( float *A, float *B, float *C )
{
	__shared__ float prods[BLOCKSIZE];

	unsigned int numItems = blockDim.x;
	unsigned int tnum = threadIdx.x;
	unsigned int wgNum = blockIdx.x;
	unsigned int gid = blockIdx.x*blockDim.x + threadIdx.x;

	prods[tnum] = A[gid] * B[gid];

	for (int offset = 1; offset < numItems; offset *= 2)
	{
		int mask = 2 * offset - 1;
		__syncthreads();
		if ((tnum & mask) == 0)
		{
			prods[tnum] += prods[tnum + offset];
		}
	}

	__syncthreads();
	if (tnum == 0)
		C[wgNum] = prods[0];
}