An ever-increasing number of small, medium, and large computers is becoming available as tools for the automotive engineer. These machines offer a bewildering variety of architectural features. Whereas in the early days of computing, floating-point computations were performed on scalar hardware, now vector and a variety of parallel architectures are available. It is therefore important that the automotive engineer be aware of the tradeoffs offered by these new architectures so that he can effectively match a computational problem to the appropriate class of machines. Guidelines are presented to aid in improving computer performance. Effects on program efficiency are discussed. (A)
Abstract