About ELPA¶
The publicly available ELPA library provides highly efficient and highly scalable direct eigensolvers for symmetric (hermitian) matrices.
The ELPA project has been started in 2008 with the aim to support the then uppcoming Petaflop HPC systems (the acronym originates from the German expression “Eigenwert Löser für Petaflop Anwendungen”; in English “Eigenvalue Solvers for Petaflop Applications”). During the last decade the ELPA-library has proven to provide highly efficient eigenvalue solvers for Petaflop systems. Nowadays, with (pre)-Exascale systems on the horizon, although keeping its well established name, the ELPA-library targets the Exascale range. The development of the ELPA-library has been supported by BMBF grants 01IH08007 (Dec. 2008 - Nov. 2011) and 01IH15001 (Feb. 2016 - Jan. 2019). During Oct. 2020 until April 2024 the development of the ELPA library has been supported by the European Center of Excellence project Nomad CoE.
Though especially designed for the use case of solving large problem sizes on massively parallel supercomputers, ELPA eigensolvers have proven to be also very efficient for smaller matrices.
The ELPA-library is mainly developed by the
Max Planck Society (Max Planck Computing and Data Facility (MPCDF) and Fritz-Haber-Institute)
Bergische Universität Wuppertal, Lehrstuhl für angewandte Informatik
Technische Universität München, Lehrstuhl für Informatik mit Schwerpunkt Wissenschaftliches Rechnen
ELPA supports all major HPC platforms, currently this includes:
TOP-500 Systems (known usage on systems with ranks 1-100):
LUMI, AMD EPYC CPUs and AMD Mi250x GPUs, CSC, Finnland
Supercomputer Fugaku, Fujitsu A64FX, RIKEN Center for Computational Science, Japan (Rank 1 of Top500 list (Nov. 2020))
Summit, IBM Power System AC922, IBM POWER9 22C 3.07GHz, NVIDIA Volta GV100, DOE/SC/Oak Ridge National Laboratory, USA (Rank 2 of Top500 list (Nov. 2020))
JUWELS Booster Module, Bull Sequana XH2000 , AMD EPYC 7402, NVIDIA A100, Forschungszentrum Jülich (FZJ), Germany (Rank 7 of Top500 list(Nov. 2020))
Piz Daint, Cray XC50, Xeon E5-2690v3 12C, NVIDIA Tesla P100, Swiss National Supercomputing Centre (CSCS), Switzerland (Rank 12 of Top500 list (Nov. 2020))
SuperMUC-NG, Lenovo ThinkSystem SD650, Xeon Platinum 8174 24C, Leibniz Rechenzentrum, Germany (Rank 15 of Top500 list (Nov. 2020))
COBRA, Intel Compute Module HNS2600BP, Xeon Gold 6148 20C, Max-Planck-Gesellschaft MPI/IPP, Germany (Rank 51 of Top500 list (Nov. 2020))
Architectures:
all X86_64 processors
Power8 and Power9 architecture
all Nvidia GPUs (starting at compute-capability “sm_35”)
AMD GPUs, especially AMD Mi250x
Intel GPUs (work in progress)
ARM arch64
ARM A64FX
Bluegene P and Q (retired)
K-computer (retired)
Intel KNL systems (retired)
NEX SX-Aurora accelerator cards
For development purposes, the ELPA library can also be used on consumer systems like
Linux PCs
Rasperry Pi (Gen 2 - 4)