Programming Techniques for Concurrent Supercomputers, Lecture by Justin Rattner
Part of the Intel Guest Lecture Series. Lecture Summary on front of tape: "Programming concurrent supercomputers is challenging but not nearly as difficult as orignally thought. This lecture will introduce, through a series of examples, the basic techniques of programming these distributed memory, message-passing multicomputers to achieve high performance. Concurrent decompostions for regular and irregular probelm domains will be introduced along with related methods for minimizing message paasing overhead and maintaining good load balance. To put concurrent programming in a real-worl context, several iPSC/2 application projects are described."