Measurements of parallelism in ordinary FORTRAN programs

In the folklore of computer architecture there has been much speculation about the effectiveness of various machines in performing various computations. While it is easy to design a machine (or part of a machine) and study its effectiveness on this algorithm or that, it is rather difficult to make general effectiveness statements about classes of algorithms and machines. We are attempting to move in this direction, and this paper contains experimental measurements of a rather wide class of algorithms. Such measurements should be helpful in establishing some parameters of machine organization.