Co-design of Parallel Numerical Methods for Plasma Physics and Astrophysics

Glinskiy, Boris; Kulikov, Igor; Снытников, А. В.; Romanenko, Alexey; Chernykh, Igor; Вшивков, В. А.

doi:10.14529/jsfi140305

Cited by 5 publications

(1 citation statement)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our approach also goes beyond known solutions to performance analysis and co-design in various application fields (see, e.g., [8,35], and references therein), which usually concentrate on very specific stencil structures but lack our rigorous focus on reproducibility and modeling.…”

Section: Related Workmentioning

confidence: 99%

Collecting and Presenting Reproducible Intranode Stencil Performance: INSPECT

Hornich

Hammer

Hager

et al. 2019

JSFI

View full text Add to dashboard Cite

Stencil algorithms have been receiving considerable interest in HPC research for decades. The techniques used to approach multi-core stencil performance modeling and engineering span basic runtime measurements, elaborate performance models, detailed hardware counter analysis, and thorough scaling behavior evaluation. Due to the plurality of approaches and stencil patterns, we set out to develop a generalizable methodology for reproducible measurements accompanied by state-of-the-art performance models. Our open-source toolchain and collected results are publicly available in the "Intranode Stencil Performance Evaluation Collection" (INSPECT). We present the underlying methods, models and tools involved in gathering and documenting the performance behavior of a collection of typical stencil patterns across multiple architectures and hardware configuration options. Our aim is to endow performance-aware application developers with reproducible baseline performance data and validated models to initiate a well-defined process of performance assessment and optimization. All data is available for inspection: source code, produced assembly, performance measurements, hardware performance counter data, single-core and multicore Roofline and ECM (execution-cache-memory) performance models, and machine properties. Deviations between measured performance and performance models become immediately evident and can be investigated. We also give hints as to how INSPECT can be used in practice for custom code analysis.

show abstract

Section: Related Workmentioning

confidence: 99%