With the advent of any proposed language system for expressing parallel operation (whether as a true parallel language (e.g., ZPL [6]), parallel extensions to sequential language (e.g., UPC [4]), or parallel compiler directives (e.g., HPF [9])) questions soon arise regarding how performance instrumentation and measurement will be conducted, and how performance data will be analyzed and mapped to the language-level (high-level) parallel abstractions. Several issues make this an interesting problem. First, the language system implements a model for parallelism whose explicit parallel operation is generally hidden from the programmer. As such, parallel performance events may not be accessible directly, requiring instead support from underlying runtime software to observe them in full. When such support is unavailable, performance must be inferred from model properties. Second, the language system typically transforms the program into its parallel executable form, making it necessary to track code transformations closely so that performance data can be correctly mapped to the user-level source. The more complex the transformations, the more difficult the performance mapping will be. Last, high-level language expression of parallelism often goes hand-in-hand with an interest for cross-platform portability of the language system. Users will naturally desire the programming and performance tools to be portable as well.
For the performance tool developer, these issues complicate decisions regarding choice of tool technology and implementation approach. In this paper, we consider the problem of designing a performance tool interface for OpenMP. Three goals for a performance tool interface for OpenMP are considered:
Figure 1: OpenMP Parallel Region Operation Showing States and Events
While our study focuses mainly on the instrumentation interface, as that is where events are monitored and the operational state is queried, clearly the type of performance measurement will determine the scope of analyses possible. Ideally, the flexibility of the interface will support multiple measurement capabilities.
