Special Issue: European–American Working Group on Automatic Performance Analysis (APART)
✍ Scribed by Michael Gerndt; John Gurd
- Publisher
- John Wiley and Sons
- Year
- 2007
- Tongue
- English
- Weight
- 75 KB
- Volume
- 19
- Category
- Article
- ISSN
- 1532-0626
- DOI
- 10.1002/cpe.1122
No coin nor oath required. For personal study only.
✦ Synopsis
This special issue is devoted to research undertaken by the European-American Working Group on Automatic Performance Analysis (APART). APART (http://www.fz-juelich.de/apart) was established in 1999 as an EU-funded working group with more than 15 partners from Europe and the U.S.A., plus more than 10 associate partners from academia and industry. APART was funded by the European Commission until 2004 and continues to function as an informal framework for research on tools for automatic performance analysis.
The goals of APART are to support the development of automatic performance analysis for parallel computers and to act as a focal point for research and industry groups working on associated tools.
While the constituent research groups in APART primarily develop their own tools, such as Paradyn (http://www.cs.wisc.edu/paradyn), KOJAK (http://www.fz-juelich.de/zam/kojak), Periscope (http://www.lrr.in.tum.de/∼gerndt/home/Research/PERISCOPE/Periscope.htm) and others, APART forms a framework in which to exchange information among the groups and to support standardization and interoperability.
Most notably, APART has defined the APART Specification Language (ASL), which allows performance properties to be formally specified, an important requirement for any automation of associated tools.
APART also defined the Standard Intermediate Program Representation (SIR), an XML-based representation of C, C++, Java and Fortran programs. The representation is similar to abstract syntax trees, but includes only information useful for performance tools; it thus enables performance tools to take into account static information about programs.
In order to increase the portability of performance tools, the Monitoring Request Interface (MRI) was defined. This interface provides a means to formulate requests for dynamic information that is being observed using run-time monitors. It also allows access to the monitored data and control of the execution of the underlying application program. These features are designed to enable offline and online tools to communicate with different monitoring systems implementing the MRI interface.