I am looking for a model (or "law", as in Moore's law) to describe the acceleration of sequential computation over time. By sequential computation I mean, for example, repeated applications of a hash function (e.g. SHA256). (Another example would be repeated multiplication modulo an unfactorisable RSA modulus.)
In his keynote David Patterson has a chart about "single program speed" (see here) which more or less captures what I am looking for. Does this metric have other names, and have there been studies about it?
Below is an image from "Computer Architecture: A Quantitative Approach":