Real-time visualization of concurrent processes

Abstract

Parallel programs usually have complex execution behavior which can be studied by visualizing their execution. The run-time visualization captures the dynamic behavior of the program which is otherwise hidden by postmortem visualization. The visualization effort for hierarchical systems such as Cedar requires parallelism to be detected at cluster and processor level simultaneously. We examine the real-time visualization methodology for shared memory multiprocessors. Two applications, visualizing the concurrent processes and matrix-related computations, are used to highlight the importance of visualization in understanding parallel program execution on shared memory multiprocessors.