2019 MultiCore Workshop

September 25, 2019 (Wednesday)

8:00am - In-person registration begins / Coffee/Tea Service, Bagels, and Pastries to be provided.

8:30 - Workshop Introduction, John Dennis (NCAR)

8:45 - Keynote: Accelerating Earth System Models with machine learning, Kelly Kochanski (University of Colorado)

9:45 - Break (15 min)

Panel (Jed Brown moderator):

10:00 - Future of Machine Learning and Physically-based Methods

• Panelists: Claire Monteleoni (CU), Kelly Kochanski (CU), DJ Gagne (NCAR), David Hall (NVIDIA)

Session 1:  Machine and Deep Learning (Jed Brown)

10:45 - Incorporating Python Machine Learning Parameterizations into Fortran Models: A Tale of Two Frameworks, DJ Gagne (NCAR)

11:10 - NVIDIA: Deep Learning Projects in Weather, Climate, and Space, David Hall (NVIDIA)

11:35 - GPU Benefits for Earth System Science, Stan Posey (NVIDIA), Raghu Kumar (NVIDIA)

12:00pm - Lunch (1 hour)

Session 2:  Numerical methods and solvers (Bryan Flynt)

1:00 - Multigrid and Mixed-Precision: Fast Solvers for Weather and Climate Models, Christopher Maynard (Met Office / University of Reading)

1:25 - Recent Advances in Sparse Linear System Solver Stacks for Exascale, Steve Thomas (NREL)

1:50 - Preconditioning Matrix-Free Operators in libCEED, Jeremy L. Thompson (University of Colorado)

2:15 - Algorithmic Choices that Improve Hardware Utilization and Accuracy, Matt Norman (ORNL)

2:40pm - Break (20 minutes) 

Session 3: Porting and Optimization (John Dennis)

3:00 - Microarchitecture Analysis and Optimization Techniques, Gunther Huebler (UWM)

3:25 - Accelerating Solar Physics Application MURaM for GPU, Eric Wright (University of  Delaware)

3:50 - Accelerating cloud schemes within the Unified Model for CPU-GPU based high-performance computing systems, Wei Zhang (Oak Ridge National Laboratory)

4:15 - Porting the RTE+RRTMGP radiative transfer package for next-generation supercomputers, Benjamin Hillman (Sandia National Laboratories)

4:40pm - Adjourn 

September 26, 2019 (Thursday)

08:00am - Coffee/Tea Service, Bagels, and Pastries to be provided.=

Session 4: Vendors (Chris Maynard)

8:30 - Cray Shasta for Earth System Modeling, Ilene Carpenter (Cray Inc.)

8:55 - AMD Epyc for High Performance Computing workloads, Martin Hilgeman (AMD)

9:20 - In-Network Computing Technology and Performance Advantages, Gerardo Cisneros-Stoianowski (Mellanox)

9:45 - Insights into Intel innovations for HPC and AI, Andrey Ovsyannikov (Intel)

10:10 - Break (15 min) 

Session 5: Model development and correctness (Ilene Carpenter)

10:25 - Developing a Terrestrial Dynamical Core for E3SM, Jed Brown (University of Colorado)

10:50 Evaluating the Spread of Climate Model Ensembles Based on Computing Environment Selection, Tom Robinson (SAIC/GFDL)

11:15 - Experience with mixed-precision within physics parameterizations, John Dennis (NCAR)

11:40 - MPAS-A on GPUs using OpenACC, Supreeth Suresh (NCAR)

12:05 - Lunch (55 minutes)

Session 6: Language and Tools (Matthew Norman)

1:00 - Progress in Porting the LFRic Weather and Climate Model to FPGAs using C and OpenCL, Mike Ashworth (University of Manchester)

1:25 - Building a High-Performance Earth System Model in Julia, Maciej Waruszewski (Naval Postgraduate School)

1:50 - A GPU performance analysis library providing arbitrary granularity in time and thread count, Jim Rosinski (UCAR/CPAESS)

2:15pm - Break (15 minutes)

2:30 - Keynote: The ECMWF Scalability Programme - Progress and Plans, Peter Bauer (ECMWF)

Panel (Chris Maynard moderator)

3:30 - Will we ever achieve a 1 km global model?

4:15–4:50 - Workshop Wrap Up