"I may compare my implementation to what the compiler generates with a multi-dim."

I think that it is very likely that a decent compiler like Keil will do it better than you hand-coding pointer arithmetic in 'C'.

With a 2-dimensional array, the compiler knows exactly what it's doing and can use its best implementation; with hand-coded pointer arithmetic, the compiler does not know exactly what you're trying to do, so it may have to use some stuff which is sub-optimal for your particular case in order to be entirely general for all possibilities of what you might be trying to do...?

Hope that made sense?

Don't forget the maintainability issue: I'm sure the 2-dimensional array would be more maintainable than a load of messing about with pointers...