P06: GPU Performance of the Simple Cloud Resolving E3SM Atmosphere Model (SCREAM)

GPU Performance of the Simple Cloud Resolving E3SM Atmosphere Model (SCREAM)

• Mark Taylor

• mataylo@sandia.gov

ANL: Jayesh Krishna, Danqing Wu, Xingqiu Yuan,  LANL: Rob Aulwes, Philip Jones, Bob Robey,   LBNL: Noel Keen,   BNL: Wuyin Lin

LLNL: David Bader, Peter Bogenschutz, H. Beydoun, Peter Caldwell, Aaron Donahue, Chris Golaz, Walter Hannah, Jungmin Lee, Paul Ullrich

ORNL: Matthew Norman, Sarat Sreepathi, Isaac Lyngass, PNNL: Chris Terai, Ruby Leung, Mikhail Ovchinnikov, Kyle Pressel, Qiu Yang, Guangxing Lin

SNL:    Luca Bertagna, Andrew Bradley, Conrad Clevenger, Chris Eldred, Jim Foucar, Oksana Guba, Ben Hillman, Mark Taylor, Maciej Waruszewski

‘Performance - Algorithms - Tools’

E3SM

The Simple Cloud Resolving E3SM Atmosphere Model (SCREAM) is a full
featured atmospheric global circulation model with the HOMMEXX
nonhydrostatic dynamical core and state-of-the-art parameterizations
for microphysics (SHOC), moist turbulence (P3) and radiation
(RTMGP++). All components have been written from scratch in C++ with
the Kokkos library used to abstract the on-node execution model for
both CPUs and GPUs. We present performance results from SCREAM’s
DYAMOND (cloud resolving model intercomparison project) simulations
running at 3.25 km global resolution on CPUs, NVIDIA GPUs and AMD
GPUs. Our fastest results were obtained on the Exascale system
Frontier, using 32K AMD MI250 GPUs and obtaining atmosphere component
speeds greater than 1 SYPD at 3.25 km resolution.

yes

mataylo@sandia.gov