SEA Seminar - A comparison of climate applications on accelerated and conventional architectures

While the emergence of many-core technology from Intel and Nvidia has illustrated great potential, capitalizing on this potential presents considerable challenges for large scientific applications. In particular we focus on the domain of climate modeling. Climate models typically have very large code bases, > 1 million lines of code which makes support of multiple version of the code infeasible. Climate models are computationally very expensive which places a premium on model performance. Furthermore because of the large user base of these applications, there is a need to support a wide range of computational platforms. These three characteristics of climate modeling make it a particularly challenging application domain for which to apply accelerator technology.

We describe work to migrate important climate applications used at the National Center for Atmospheric Research (NCAR) to Intel Xeon Phi and Nvidia GPU. We also provide a comparison with more conventional Intel SandyBridge with Infiniband. In particular we concentrate on both kernels from the High Order Methods Model Environment (HOMME) as well as the complete application. For the kernels we explored several different programming paradigms, including OpenACC, Cuda C/Fortran, and OpenMP. We will illustrate our use of Extrae a tracing tool developed at the Barcelona Supercomputing Center and the profiling tool TAU to isolate sections of code for further improvement. We will also present initial evaluation of the Community Earth System Model in different executions modes on the Intel Xeon Phi system at TACC.

About the presenter
Presenter(s): 
Srinath Vadlamani & Youngsung Kim

Affiliation or organization:

Where

Building:

Room: 
Main Seminar Room

Lab/division or program:

Type of event:

When
Date and time: 
Thursday, August 29, 2013 - 3:00pm