Parallel programming is difficult due to the complexity of dealing with conventional lock-based synchronization. To simplify parallel programming, there have been a number of proposals to support transactions directly in hardware and eliminate locks completely. Although hardware support for transactions has the potential to completely change the way parallel programs are written, initially tran...

A number of programming language models, including actors, provide inherent concurrency. We are developing high-level language constructs using actors and studying their implementation on multiprocessor architectures. This report describes our experience with programming in actors by means of a speci c example of scienti c computation. We also discuss work in progress on language annotations an...

Debugging MIMD programs is often a delicate job. As a matter of fact, they can have diierent behaviors in successive executions. So, cyclic debugging is not applicable. To make it available for parallel programmers , we propose execution replay (full and partial) for our multi-threaded execution model, the Communicating Active Components (CAC). CAC/s have been deened to implement Parallel Objec...

Memoization is an optimization that provides asymptotic speedups, automatically achieving many of the benefits of dynamic programming. Memoization, however, trades off reduced execution time for additional required storage. This additional storage requirement can be reduced somewhat by using several techniques from incrementalization. The first technique, called pruning, statically identifies t...

Reproducing the execution of a concurrent program is important in debugging and testing. It requires that, regardless of the actual order in which processes may execute, the reproduced execution is identical, with respect to the order in which certain activities occur, to a previously recorded execution. This paper presents a solution to the reproducibility problem for programs written in the S...