Score-P: A Unified Performance Measurement System for Petascale Applications

Abstract

The rapidly growing number of cores on modern supercomputers imposes scalability demands not only on applications but also on the software tools needed for their development. At the same time, increasing application and system complexity makes the optimization of parallel codes more difficult, creating a need for scalable performance-analysis technology with advanced functionality. However, delivering such an expensive technology can hardly be accomplished by single tool developers and requires higher degrees of collaboration within the HPC community. The unified performance-measurement system Score-P is a joint effort of several academic performance-tool builders, funded under the BMBF program HPC-Software für skalierbare Parallelrechner in the SILC project (Skalierbare Infrastruktur zur automatischen Leistungsanalyse paralleler Codes). It is being developed with the objective of creating a common basis for several complementary optimization tools in the service of enhanced scalability, improved interoperability, and reduced maintenance cost.

Profiling of OpenMP Tasks with Score-P

“…For tied tasks, we introduced an extension to OPARI to track individual instances of tasks [10]. Its successor OPARI2 [13] includes this feature and provides automatic instrumentation of task constructs including support for task instance identification.…”

On Processing Extreme Data

“…New metrics, analysis techniques and models. Over the last years, a large number of performance measurement and analysis tools (like the ones described in [82,67,65]) have been created for the new heterogeneous, many-core, parallel computing architectures and systems. These tools support low-level instrumentation, performance measurement and analysis; they are based on hardware counters, time measurements, call graphs, and tools for tracing the operating system or measuring the performance of specific programming models; and they cover, for instance, multi-threaded, OpenMP, and MPI programs.…”

Performance Analysis Techniques for Task-Based OpenMP Applications

“…This mechanism was demonstrated via a prototypical extension of OPARI. In the meantime, the successor OPARI2 [2] was released, which-among other improvements-contains a production version of this instrumentation mechanism. Our work builds upon the OPARI2 instrumentation and uses it as a prerequisite for performance measurements.…”

“…For this purpose, the work presented in this paper leverages the Score-P [2] instrumentation and measurement system. To instrument OpenMP directives, Score-P utilizes the source-to-source instrumenter OPARI2, using the technique presented in [12] for task-related constructs.…”