HPC Benchmarks
milan
Cluster Corp. is experienced in compiling and optimizing benchmarks on
a wide variety of platforms for performance runs and internal testing.
The ASCI Purple Benchmarks
The ASCI Purple Benchmarks were designed as are part of the ASCI Purple
Request for Proposal (RFP). There are three tiers. Tier 1 includes
sPPM, UMT2000, PRESTA, MEMBENCH, PARBENCCH, and SUPER.
http://www.llnl.gov/asci/platforms/purple/rfp/benchmarks/
HPC Challenge Benchmark
The HPC Challenge benchmark consists at this time of 7 benchmarks: HPL,
STREAM, RandomAccess, PTRANS, FFTE, DGEMM and b_eff Latency/Bandwidth.
HPL is the Linpack TPP benchmark. The test stresses the floating point
performance of a system. STREAM is a benchmark that measures
sustainable memory bandwidth (in GB/s), RandomAccess measures the rate
of random updates of memory. PTRANS measures the rate of transfer for
larges arrays of data from multiprocessor’s memory. Latency/Bandwidth
measures (as the name suggests) latency and bandwidth of communication
patterns of increasing complexity between as many nodes as is time-wise
feasible.
http://icl.cs.utk.edu/hpcc/
Intel MPI Benchmark (Pallas)
Intel® MPI Benchmarks 2.3 is a comprehensive set of MPI benchmarks,
formerly known as Pallas MPI Benchmarks (PMB).
http://www.intel.com/cd/software/products/asmo-na/eng/cluster//clustertoolkit/219848.htm
Bounce MPI Benchmark
The Stanford University Bounce (SUbounce) MPI Benchmark tests latency
and bandwidth of any MPICH derived implementation. It's a trivial
program, easy to 'compile and run' providing human readable results.
It's commonly used to verify the MPI F90 interface.
http://www.sourceforge.net/projects/subounce