2015/11/04 by Andreas Müller, Michal A. Kopera, Müller, Andreas +9 · 3 citations
Earth and Planetary Sciences · Environmental Science · #Atmospheric and Oceanic Physics (physics.ao-ph) #Climate variability and models #D.2.8 #Distributed #FOS: Computer and information sciences #FOS: Physical sciences #Fluid Dynamics (physics.flu-dyn) #G.1.8 #G.4 #Geophysics (physics.geo-ph) #Meteorological Phenomena and Simulations #Parallel #Precipitation Measurement and Analysis #Software Engineering (cs.SE) #and Cluster Computing (cs.DC)
paper · pdf · doi:10.48550/arxiv.1511.01561
openalex publication_date 2015/11/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Numerical weather prediction (NWP) has proven to be computationally\nchallenging due to its inherent multiscale nature. Currently, the highest\nresolution NWP models use a horizontal resolution of about 10km. In order to\nincrease the resolution of NWP models highly scalable atmospheric models are\nneeded.\n The Non-hydrostatic Unified Model of the Atmosphere (NUMA), developed by the\nauthors at the Naval Postgraduate School, was designed to achieve this purpose.\nNUMA is used by the Naval Research Laboratory, Monterey as the engine inside\nits next generation weather prediction system NEPTUNE. NUMA solves the fully\ncompressible Navier-Stokes equations by means of high-order Galerkin methods\n(both spectral element as well as discontinuous Galerkin methods can be used).\nMesh generation is done using the p4est library. NUMA is capable of running\nmiddle and upper atmosphere simulations since it does not make use of the\nshallow-atmosphere approximation.\n This paper presents the performance analysis and optimization of the spectral\nelement version of NUMA. The performance at different optimization stages is\nanalyzed using a theoretical performance model as well as measurements via\nhardware counters. Machine independent optimization is compared to machine\nspecific optimization using BG/Q vector intrinsics. By using vector intrinsics\nthe main computations reach 1.2 PFlops on the entire machine Mira (12% of the\ntheoretical peak performance). The paper also presents scalability studies for\ntwo idealized test cases that are relevant for NWP applications. The\natmospheric model NUMA delivers an excellent strong scaling efficiency of 99%\non the entire supercomputer Mira using a mesh with 1.8 billion grid points.\nThis allows to run a global forecast of a baroclinic wave test case at 3km\nuniform horizontal resolution and double precision within the time frame\nrequired for operational weather prediction.\n