Analysis and Modeling of Advanced PIM Architecture Design Tradeoffs
Proceedings of the 2004 ACM/IEEE conference on Supercomputing
Bristlecone: A Language for Robust Software Systems
ECOOP '08 Proceedings of the 22nd European conference on Object-Oriented Programming
Bamboo: a data-centric, object-oriented approach to many-core software
PLDI '10 Proceedings of the 2010 ACM SIGPLAN conference on Programming language design and implementation
Recovery tasks: an automated approach to failure recovery
RV'10 Proceedings of the First international conference on Runtime verification
The structure of authority: why security is not a separable concern
MOZ'04 Proceedings of the Second international conference on Multiparadigm Programming in Mozart/Oz
Integrating task parallelism with actors
Proceedings of the ACM international conference on Object oriented programming systems languages and applications
| Hi-index | 0.00 |
This paper presents precise versions of some "laws" that must be satisfied by computations involving communicating parallel processes. The laws take the form of stating plausible restrictions on the histories of computations that are physically realizable. The laws are very general in that they are obeyed by parallel processes executing on a time varying number of distributed physical processors. For example, some of the processors might be in orbiting satellites. The laws are justified by appeal to physical intuition and are to be regarded as falsifiable assertions about the kinds of computations that occur in nature rather than as proved theorems in mathematics. The laws are intended to be used to analyze the mechanisms by which multiple processes can communicate to work effectively together to solve difficult problems. The laws presented in this paper are intended to be applied to the design and analysis of systems consisting of large numbers of physical processors. The development of such systems is becoming economical because of rapid progress in the development of large scale integrated circuits. We generalize the usual notion of the history of a computation as a sequence for events for the notion of a partial order of events. Partial orders of events seem better suited to expressing the causality involved in parallel computations than totally ordered sequences of events obtained by "considering all shuffles" of the elementary steps of the various parallel processes [2l, 22]. The utility of partial orders is demonstrated by using them to express our laws for distributed computation. These laws in turn can be used to prove the usual induction rules for proving properties of procedures. They can also be used to derive the continuity criterion for graphs of functions studied in the Scott-Strachey model of computation. The graph of a function is simply the set of all input output pairs for the function. We can prove that the graph of any physically realizable procedure p that behaves like a mathematical function is the limit of a continuous functional F such that: i graph(p) = F ({}) i N In other words the graph of p is the limit of the n-fold compositions of F with itself beginning with the empty graph.