Well, the sizes of the matrices involved may vary from about 3 to 1000, but it's not necessarily that my implementation is not fast, it's just that I thought/hoped there'd be a very Maple-esque way without for loops.

Well, the sizes of the matrices involved may vary from about 3 to 1000, but it's not necessarily that my implementation is not fast, it's just that I thought/hoped there'd be a very Maple-esque way without for loops.