DGEMM benchmark is a simple, multi-threaded, dense-matrix multiply benchmark.
100K+
Description
This is a docker image for the DGEMM benchmark and the docker run command will execute DGEMM benchmark with a problem size of 1024.
DGEMM Description
The Crossroads/NERSC-9 Memory Bandwidth benchmark is a simple single-node multi-threaded dense-matrix multiply benchmark. The code is designed to demonstrate high floating-point compute rates on a machine under sustained computation. The Offeror is expected to run this benchmark to report compute performance and not to report peak hardware performance.
Required Problem Sets
The DGEMM benchmark shall be used to benchmark each type of node intended for computation in the proposed system. If a given node has configurable options, the benchmark will be used to report the performance of each configuration.
Authorship
This benchmark was developed at Sandia National Laboratories, led by Simon Hammond.
SOURCE: http://www.nersc.gov/research-and-development/apex/apex-benchmarks/dgemm/
Content type
Image
Digest
sha256:4917886a1…
Size
218.2 MB
Last updated
over 3 years ago
docker pull gouravr/dgemm:v2