Introduction
This page hosts the CTS-2 benchmarks. CTS-2 is the next Commodity Technology Systems procurment for Los Alamos, Sandia and Lawrence Livermore National Laboratories. The benchmarks are being used as one component of system evaulation.
Over the next few months we will update this site as needed. While we do not anticipate changing the FOM Benchmarks they could change. In addition, the microbenchmarks will likely evolve. Finally, as needed we will update both the benchmarks and their documents if bugs are found or clarification of run rules required.
For now some benchmark applications have only websites provided. Once we have a reference figure of merit (FOM) for an application we will include a static source of that application on this site and a document describing the benchmarking procedures for the application. Reference FOMs will be run on our CTS-1 systems. Technical specs for Quartz one of the largest of these machines can be found here.
The CTS-2 RFI page is https://hpc.llnl.gov/cts-2-rfi. The CTS-2 Benchmark appendix be submitted with a response.
Slides presented during vendor briefings are here. If you want to schedule a vendor briefing please e-mail the cts2 benchmark team cts2-benchmarks@llnl.gov.
Questions?
For questions about the benchmarks and other information on this site please contact cts2-benchmarks@llnl.gov.
FOM Benchmarks
Benchmark | Lines of Code | Parallelism | Language | Code Description |
---|---|---|---|---|
|
10,000 | MPI+OpenMP | C++ | HPCG is a software package that performs a fixed number of multigrid preconditioned (using a symmetric Gauss-Seidel smoother) conjugate gradient (PCG) iterations using double precision (64 bit) floating point values. |
2,000 + dependency on MFEM |
MPI+RAJA | C++ | Laghos solves the time-dependent Euler equation of compressible gas dynamics in a moving Lagrangian frame using unstructured high-order finite element spatial discretization and explicit high-order time-stepping. It is built on top of a general discretization library (MFEM) and supports two modes: *full assembly*, where performance is limited by the data motion in a conjugate gradient (CG) solve, and *partial assembly*, where performance depends mostly on small dense matrix operations and the CG solve communication. | |
10,000 | MPI+OpenMP | C++ | Monte Carlo transport benchmark with multi-group cross section lookups. Stresses memory latency, significant branching, and one large kernel that is 1000’s of lines big. | |
3,000 | MPI+OpenMP | Fortran | SNAP serves as a proxy application to model the performance of a modern discrete ordinates neutral particle transport application |
Micro Benchmarks
Links to example benchmarks for microbenchmarks. Offerors can use their others versions of these benchmarks, but should provide a link to an open reference version their bid is based from. Optimized code may be provided in the bid, but is not required.
Benchmark | Parallelism | Description |
---|---|---|
Stream | OpenMP | Memory subsystem bandwidth performance tests. |
DGEMM | OpenMP | CPU FLOPs performance test |
OSU | MPI | MPI Performance Benchmark |
DAXPY | Serial | Single Core performance test |