In the world of computational fluid dynamics (CFD), end users are driven bythe need for fast, accurate and robust methods to deliver engineeringinsight to support design decisions. Traditional methods (typically secondorder finite volume or finite difference techniques) usually deliver two of these three requirements – i.e. you can have speed and accuracy in a highly tuned method for a specific application; accuracy and robustness can be delivered with a significant computational (and therefore financial) overhead, and of course speed and robustness can be delivered providing you don’t look too closely at the answers!