kernel void bubbleSort2(global int* A, global int* B, local int* scratch) {
	int id = get_global_id(0);
	int lid = get_local_id(0);
	int N = get_global_size(0);
	double temp;

	scratch[lid] = A[id];
	barrier(CLK_LOCAL_MEM_FENCE);

	for (int i = N-1; i > 0; i--)
	{
		for (int j = 0; j <= i-1; j++)
		{
			if (A[j] > A[j+1])
			{
				temp = A[j];
				A[j] = A[j+1];
				A[j+1] = temp;
			}
			barrier(CLK_LOCAL_MEM_FENCE);
		}
	}
	B[id] = A[id];
}