R. Jesse Chaney, Lawrence A. Crowl, and Vikram A. Saletore, "What is the Price of Abstraction? PVM Versus Charm", Technical Report 94-80-15, Department of Computer Science, Oregon State University, Corvallis, Oregon, 97331-3202, November 1994
This report describes translations of an existing sequential program to the Charm parallel programming system and to the PVM parallel programming library. Charm provides a rich set of data and execution abstraction facilities, while PVM does not. One argument in favor of abstraction facilities is that it is better to express an algorithm in terms of the problem -- higher levels of abstraction promote algorithm expression closer to the form of the problem not in terms of the (machine) implementation. An argument against providing abstraction facilities is that they introduce significant execution overhead. Contrary to our expectations, experience translating our program supports the conclusion that, compared to PVM, Charm's abstraction facilities introduce no significant execution overhead, but that they introduce a significant translation (programming) overhead.