Abstract

We use a form of dynamic code generation, called dynamic instrumentation, to collect data about the execution of an application program. Dynamic instrumentation allows us to instrument running programs to collect performance and other types of information. The instrumentation code is generated incrementally and can be inserted and removed at any time. Our instrumentation currently runs on the SPARC, PA-RISC, Power2, Alpha, and x86 architectures. Specification of what data to collect are written in a specialized language called the Metric Description Language, that is part of the Paradyn Parallel Performance Tools. This language allows platform-independent descriptions of how to collect performance data. It also provides a concise way to specify how to constrain performance data to particular resources such as modules, procedures, nodes, files, or message channels (or combinations of these resources). We also describe the details of how we weave instrumentation into a running program.

Citations