University of Waterloo will speak on ``Single Stepping in Event Visualisation Tools for Distributed Applications''. TIME: 2:30-3:30 p.m. ROOM: DC 1331 ABSTRACT Debugging distributed applications is a difficult but essential task. Event visualisation tools simplify this effort by providing a graphical view of distributed executions. Although such tools help understand distributed applications, they are frequently insufficient for full debugging purposes. The need for traditional debugging operations is often overlooked when building event visualisation tools. One of the most useful operations in traditional debuggers is single stepping. Such an operation is available in parallel debuggers but only on an individual-process basis. The difficulties faced when dealing with partially ordered executions have deterred the development of a single stepping facility that deals with multiple processes at once. In this presentation I will describe the theoretical foundations of such a facility as applied to event visualisation tools. These tools commonly use abstraction techniques to reduce the often overwhelming amount of detail presented to the user. As process clusters and abstract events form the basic building blocks for abstract visualisation, single stepping in their presence will also be detailed. DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF WATERLOO SEMINAR ACTIVITIES MASTER'S THESIS PRESENTATION -Thursday, August 15, 1996 Erik Demaine, graduate student, Dept. Comp. Sci., Univ. Waterloo, will speak on ``Simulation of Message-Passing in Distributed-Memory Architectures''. TIME: 2:30-3:30 p.m. ROOM: DC 1331 ABSTRACT This talk discusses my work on a project called PUPPET (Performance Under a Pseudo-Parallel EnvironmenT). It allows the evaluation of parallel-program performance using only a pseudo-parallel (or non-parallel) system. That is, researchers can use their workstations to develop and test their message-passing applications, and project the performance results to a real parallel computer. Similarly, universities can offer parallel programming courses without using valuable supercomputing resources. I will overview previous work in the area of simulating parallel computers. The rest of the talk will focus on new contributions to the area, including an approximate network model, support for collective and non-blocking communication, and simulation of multitasking.