Recently released SPEC MPI2007 2.0 benchmark adds support for testing much larger clusters, up to 2,048 ranks (cores). This represents a big leap over (the currently shipping with SPEC MPI version 1.1) Medium Data Set, which is designed to scale to 128 processing cores. The Large Data Set enables fair comparison of modern HPC systems as they have outgrown the Medium Data Set. SPEC awarded six engineers for their work on this benchmark.

Andrey Naraikin - Intel
Huiyu Feng ? SGI
Pavel Shelepugin - Intel

When creating new benchmarks, testing is vital, especially of a benchmark that can run into the thousands of cores. Andrey, Huiyu, and Pavel provided the testing needed to make sure our SPEC MPI2007 large dataset benchmark works at the largest sizes. They have provided the testing, problem identification, scaling analysis and bug fixes required to allow us to make progress on the benchmark.

Brian Whitney ? Oracle

Brian is HPG?s release manager who prepared many benchmark test kits integrating all the changes brought by the entire SPEC HPG group. He also implemented numerous infrastructure improvements.

Carl Ponder - IBM

Carl was a crucial contributor for the development and management of MPI2007 documentation, especially with respect to the run rules, FAQ, and configuration file

Cloyce Spradling ? Oracle

SPEC MPI2007 uses the tools that SPEC CPU2006 is based upon. Cloyce adapted the tools to the HPC needs and extended them to include new flexibility needed for MPI2007.

Photo: Klaus-Dieter Lange (SPEC Awards Committee Chair, right) presents award to Brian Whitney (Oracle)