The Standard Performance Evaluation Corporation [SPEC] has released a new version of its MPI2007 benchmark. The new release adds a large data set for machines with 64 to 2048 cores. The SPEC MPI2007 workload was designed by the High Performance Group at SPEC in order to measure “the performance of parallel computing systems and clusters running actual end-user Message-Passing Interface (MPI) applications.” It provides performance metrics that can be used to compare different hardware architectures (SMP, NUMA, clusters) and interconnects, processors, memory hierarchy, compilers, and MPI implementations.
The cliché ‘it’s a small world’ certainly doesn’t apply to data, which just keeps expanding,” says Kalyan Kumaran, Ph.D., SPEC HPG chair and manager of applications performance engineering at Argonne National Laboratory. “With SPEC MPI2007 V2.0, we’re fulfilling the need of the research and science communities for a standardized benchmark suite based on real MPI applications using large datasets.”
The V.2.0 of MPI2007 was developed by SPEC members with members AMD, Argonne National Laboratory, Fujitsu, IBM, Indiana University, Intel, Platform Computing, QLogic, SGI, Sun, and Technische Universität Dresden. For more info on all the message passing goodness, read their full release here.