Why is this interesting?
In Jacobi like algoritms, each processor stores its own xpoints by ypoints array of variables as well as guard rings containing the rows and columns from neighbours. One loads these guard rings at start of computation iteration and only updates points internal to array