The Lawrence Livermore National Laboratory (LLNL)–led Center for Efficient Exascale Discretizations (CEED) co-design effort will develop next-generation discretization software and algorithms—which deliver a significant performance gain over conventional low-order methods—to enable a wide range of DOE and National Nuclear Security Administration (NNSA) applications to run efficiently on future exascale hardware. CEED is a research partnership involving 30+ computational scientists from two DOE labs and five universities, including UTK.
For UTK’s part, ICL is instrumental in identifying, developing, and optimizing tensor contractions that are essential building blocks for these kinds of DOE/NNSA applications. The ICL team will also play an integral role in co-designing application programming interfaces (APIs) with the LLNL scientists, external partners, and vendors. It will deliver a high-performance tensor contractions package through the Matrix Algebra on GPU and Multicore Architectures (MAGMA) library.
Find out more at http://ceed.exascaleproject.org/
In Collaboration With
- Lawrence Livermore National Laboratory