pSather is a parallel extension of the existing object-oriented language Sather. It offers a shared-memory programming model which integrates both control- and dataparallel extensions. This integration increases the flexibility of the language to express different algorithms and data structures, especially on distributed-memory machines (e.g. CM-5). This report describes our design objectives and the programming language pSather in detail. ICSI and Eidgenossische Technische Hochschule (ETH), Zurich, Switzerland. E-mail: murer@icsi.berkeley.edu. y ICSI and Computer Science Division, U.C. Berkeley. E-mail: jfeldman@icsi.berkeley.edu. z ICSI and Computer Science Division, U.C. Berkeley. E-mail: clim@icsi.berkeley.edu. x ICSI E-mail: mseidel@icsi.berkeley.edu. ii Contents 1 Introduction 4 1.1 Roadmap of this Report : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 5 1.2 Grammar Notation : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 6 2...

Sather offers safety, elegance and interoperability to computational scientists. High performance scalable language extensions were a design goal of the serial language, not an afterthought. This is a DRAFT of an article to appear in the September/October issue of Computers in Physics. Introduction Sather is an object-oriented language first introduced in 1991. Since then, considerable practical experience has been obtained with the language by the hundreds of users making up the Sather community. Sather offers many safety and convenience features to help programmers avoid common errors and reuse code. Some of these were discussed in a previous Computers in Physics article 10

Thread-based concurrent languages currently do not provide much support to (a) avoid deadlocks, (b) treat competing threads in a fair way, and (c) allow branching depending on lock availability. This makes parallel programming difficult and error prone and thus reduces the programmer's productivity. In this paper we present a lock statement for fair atomic locking of several locks that supports (a), (b), and (c). We discuss the expressive power of the new lock statement and show the basic principles of an efficient implementation. 1 Introduction When extending a language for thread-based parallelism commonly several new features are introduced. The minimal set of features comprises mechanisms for thread creation and locking. However, by adding these features common thread problems are introduced and regrettably left for the programmer to solve. The most significant of these problems is caused by locking constructs based on the "one lock at a time" idea. With these constructs the prog...

The paper discusses the main aspects of a parallel debugger for the parallel object-oriented language pSather. PSather provides for a single shared-address space and for multiple threads per processor. Threads can arbitrarily migrate between processors. The debugger supports cyclic debugging which is a standard and quite effective technique for sequential programs. To address nondeterminism, deterministic replay is provided. For that reason, one program run is traced. The debugger uses these traces to repeatedly and identically reexecute the recorded program run. An efficient trace and replay scheme that is in particular suited for pSather and that reduces program perturbation and trace file length is presented. Furthermore, breakpoints and single-stepping for such a debugger using deterministic replay are discussed. 1 Introduction Debugging of sequential programs seems to be well understood. The standard technique is testing until an error manifests as a program failure and then to l...