PDC Seminar: Optimized composition of performance-aware parallel components

Prof. Christoph Kessler

Linköping University

Abstract:

Programming parallel systems efficiently is notoriously difficult, and this is getting even worse with the trend towards more heterogeneous and hybrid parallel systems.

Components are a well-proven concept for managing software design and implementation complexity. However, to be reusable, components are often more general than necessary. Moreover, they hardcode and hide too many design decisions such as scheduling, resource management, algorithm selection or data representation selection, which could be bound better at a later point of time, e.g. at run-time when more information about currently available resources or problem parameters is known.

We show how such decisions can be guided at run-time with low overhead by look-up tables that are computed off-line during component deployment on the target system.

This context-aware composition is a powerful optimization technique that can be seen as a generalization of the currently very popular autotuning methods for special domains and library functions. While formulated for a parallel scenario, our method also applies to optimized composition of sequential programs as a special case.

We give also a short overview of the EU FP7 research project PEPPHER (PErformance Portability and Programmability of HEterogeneous many-core aRchitectures), which started in 2010. In one of the PEPPHER workpackages we elaborate and apply our approach towards a language-agnostic component framework that supports automatized performance portability across a major range of heterogeneous many-core systems.

Short bio

Christoph W. Kessler received a PhD degree in Computer Science in 1994 from the University of Saarbrücken, Germany, and a Habilitation degree in 2001 from the University of Trier, Germany, where he had worked for 6 years as assistant professor in computer science. In 2001 he joined Linköping university, Sweden, as associate professor at the programming environments laboratory (PELAB) of the computer science department (IDA).

In 2007 he was appointed full professor at Linköping university.

His research interests include parallel programming, compiler technology, code generation, optimization algorithms, and software composition.

He has published two books and more than 60 scientific papers in international journals and conferences. His contributions include e.g. the OPTIMIST retargetable optimizing integrated code generator for VLIW and DSP processors, the PARAMAT approach to pattern-based automatic parallelization, and the parallel programming languages Fork and NestStep.