A Tools Information Interface for OpenSHMEM

The Partitioned Global Address Space (PGAS) programming model, OpenSHMEM, is getting more traction as a useful method for parallel programming on future-generation platforms. However, very few works have explored on the enabling of external tools to analyze and control performance behavior of OpenSHMEM runtimes. While the OpenSHMEM standard recently introduced the profiling interface allowing tools to collect and monitor performance, it still does not define a mechanism through which an implementation can expose its internal performance knobs and metrics to the end users. To write OpenSHMEM programs that perform efficiently in a uniform manner across different platforms, it is necessary to understand and control these internal performance metrics. Early work reveals that OpenSHMEM performance variables can provide insights that are crucial to performance debugging, analysis, and optimization. In this paper, we propose a generic tools information interface with flexible and portable variable representation and a set of APIs that provide users the capability to analyze and control the performance behavior. The goal of this paper is to establish the usefulness and feasibility of such an API that users can leverage to better understand the internal details of the runtime.