CISL SEMINAR SERIES: GPU-Accelerated Large-Eddy Simulation with FastEddy™

04/03/2019 - 10:00am
Mesa Lab, Main Seminar Room (MSR)

GPU-Accelerated Large-Eddy Simulation with FastEddyTM

Jeremy Sauer

https://staff.ucar.edu/users/jsauer

NCAR Research Applications Laboratory

 

Large-eddy simulation (LES) is a computationally intensive numerical modeling technique characterized by high-fidelity spatial and temporal resolution in the discretization of prognostic conservation equations for mass, momentum, and energy in order to explicitly resolve the dynamics of the most energetic scales of eddies in a turbulent flow field. This technique has proven particularly valuable in studies of the atmospheric boundary layer, where a turbulence resolving capacity is paramount to the study of myriad boundary layer processes and coupled phenomenology therein. CPU-based approaches to LES modeling require immense HPC resources for modern expansive numerical investigations, and are prohibitive as a predictive tool at all but the coarsest of LES resolutions. With the maturation of GPU hardware and programming techniques over the last decade, GPU-accelerated LES is poised to enable a disruptive advance. This talk provides an overview and early-milestones of FastEddyTM, a purpose-built, resident-GPU LES model, targeting performant, fine-scale, weather research and prediction.

 
Biography
Jeremy Sauer has been a scientific-modeling software engineer at NCAR’s Research Applications Laboratory since October 2016. His primary focus is GPU-accelerated model development and application. Prior to arriving at NCAR he spent 11 years at Los Alamos National Laboratory in the development and application of HPC-based models for multiphase atmosphere-coupled processes and electromagnetic wave propagation. Jeremy received his BA in Physics and Computer Science from the University of Montana, MS in Computational Science from the University of Montana, and PhD in Geophysical Fluid Dynamics from Florida State University.

 

Refreshments will be served at 9:45 a.m.!

Wednesday, April 3, 2019

10:00 a.m. - 11:00 a.m.

Mesa Lab, Main Seminar Room (MSR)