Performance Analysis
The Jülich Supercomputing Centre has a long tradition in the development of performance tools for parallel programs, e.g., the well-known MPI trace visualizer VAMPIR was invented here. The current focus is on the automation of the performance analysis process. In the
KOJAK project, we aimed at the
development of a generic automatic performance analysis environment for
parallel programs. Performance problems are specified in terms of execution
patterns that represent situations of inefficient behavior. These patterns
are input for an analysis process that recognizes and quantifies the
inefficient behavior in event traces. Mechanisms that hide the complex
relationships within event pattern specifications allow a simple
description of complex inefficient behavior on a high level of abstraction.
In the Scalasca, a successor to
KOJAK, the main focus is on scalability in order to support analysis
of parallel applications running on today's supercomputer consisting of
many thousand processor cores.
last change 16.03.2009 |
Anke Visser | Print
The Jülich Supercomputing Centre has a long tradition in the development of performance tools for parallel programs, e.g., the well-known MPI trace visualizer VAMPIR was invented here. The current focus is on the automation of the performance analysis process. In the
KOJAK project, we aimed at the
development of a generic automatic performance analysis environment for
parallel programs. Performance problems are specified in terms of execution
patterns that represent situations of inefficient behavior. These patterns
are input for an analysis process that recognizes and quantifies the
inefficient behavior in event traces. Mechanisms that hide the complex
relationships within event pattern specifications allow a simple
description of complex inefficient behavior on a high level of abstraction.
In the Scalasca, a successor to
KOJAK, the main focus is on scalability in order to support analysis
of parallel applications running on today's supercomputer consisting of
many thousand processor cores.
Collaborations and other projects:
- JSC is founding member and workpackage leader of the European
IST working group on automatic performance analysis
APART.
- In the KOJAK and Scalasca projects, we are collaborating with the
Innovative Computing Laboratory
of the University of Tennessee.
- In various German ( VI-HPS, SILC, eeClust) and EU ( ParMA) projects we
are working on improving performance analysis tools and methods for
parallel programs.
- We are also working very closely with the Dept. of Computer and Information
Science of the University of Oregon on the Program
Database Toolkit and TAU
projects.
last change 16.03.2009 |
Anke Visser | Print