Home » conference-talk

conference-talk

Separating physics and performance: Using Python to implement fast and maintainable climate models.

Location: 
CG Auditorium
Speaker: 
Rhea George

Most global climate models have been optimized for performance on a specific hardware architecture, and typically cannot run efficiently on GPUs without extensive effort that increases the complexity of the code. This inhibits our ability to leverage emerging supercomputers to run at high (e.g. storm-resolving scale) resolutions that help reduce uncertainty in global climate projections.

Speaker Description: 

Rhea George has a doctorate in Atmospheric Sciences from the University of Washington, an undergraduate Computer Science bachelor’s degree from University of California Berkeley, and worked as a software engineer in renewable energy before joining Vulcan for this climate modeling project. 

Event Category:

Modern Tools for Lidar and Radar Data Processing and Visualization for the Research Community

Location: 
CG Auditorium
Speaker: 
Brenda Javornik

What are the challenges faced as we take 30 years of legacy code, update it, make it easy for the research community to access, use, install, and add their own contributions.  

Speaker Description: 

Brenda Javornik is a software engineer working on the Lidar Radar Open Software Environment (LROSE) in the Earth Observing Laboratory at UCAR.  LROSE is a collaborative project with Colorado State University and UCAR.  

Event Category:

What if?: Exploring masking of conditionals for performance portability

Location: 
CG Auditorium
Speaker: 
Jon Rood

In the GPU programming paradigm, loops are typically "hoisted" into what are hopefully perfectly nested collapsable loops. This technique usually increases the size of loop bodies along the way. This allows for maximum parallelism and work for each GPU thread which results in maximum device utilization. However, for good performance on the CPU, loops are typically "lowered" into several very simple vectorizable loop bodies. Some performance portability tools attempt to do this classical transformation automatically at compile time.

Speaker Description: 

Jon Rood is a Compuatational Scientist at the National Renewable Energy Laboratory. He currently focuses on performance engineering, performance portability, and software quality assurance for two DOE applications under the Exascale Computing Project involving modeling of wind energy and combustion energy.

Event Category:

From PeleC to PeleACC, to PeleC++: What we learned porting our AMReX application to two modern GPU programming models

Location: 
CG Auditorium
Speaker: 
Jon Rood

PeleC is an Exascale Computing Project application for simulating compressible combustion in complex geometries. It has been built on top of the popular AMReX library. In the beginning of the Exascale Computing Project, PeleC was focused on KNL. It uses a mixture of C++, C, and kernels written in Fortran to obtain performance by focusing on vectorization. Recently we have taken two approaches in deciding PeleC's future for obtaining performance on exascale GPU machines. In the first programming model, we decorated the Fortran kernels with OpenACC directives.

Speaker Description: 

Jon Rood is a Computational Scientist at the National Renewable Energy Laboratory. He currently focuses on performance engineering, performance portability, and software quality assurance for two DOE Exascale Computing Project applications involving modeling of wind energy and combustion energy.

Event Category:

A Deep Learning Approach for Intelligent Thinning of Satellite Data

Location: 
CG Auditorium
Speaker: 
Sarvesh Garimella

As new observation platforms are launched into orbit, the amount of satellite data generated globally is increasing rapidly. Transmission of large amounts of data to ground stations is limited by available bandwidth. Also, the amount of data that can be ingested into numerical models is limited, especially if the models are run within operational time constraints. In this study, a deep learning approach is used to compress satellite data intelligently into a specified dimensionality.

Speaker Description: 

Dr. Sarvesh Garimella is the Chief Scientist and Chief Operating Officer at ACME AtronOmatic, LLC. His work is at the interface of artificial intelligence and atmospheric science, and he leads the research and operational efforts at ACME to create innovative new features for the users of the MyRadar app. With a background in planetary science and environmental engineering as an undergraduate at Caltech (BS ’11), his graduate career focused on clarifying the role of anthropogenic emissions of particulate matter on clouds and climate. He completed his doctorate in Climate Physics and Chemistry at MIT (PhD ’16), where his research focused on representing the microphysical underpinnings of ice cloud formation in global climate models with a machine learning approach. As part of his affiliation with the MIT Center for Global Change Science, his work also examined the the policy implications of this research from both a climate and human health perspective. In addition to his scientific interests, Sarvesh is an avid jazz trumpet player in several groups in the Portland area. He is currently the President of the NoPo Big Band, a community organization dedicated to providing high quality jazz music to the Portland community. His other hobbies include exploring the outdoors and listening to live music.

Event Category:

Trans-Disciplinary Insights for Climate Science Modeling and Big Data

Location: 
CG Auditorium
Speaker: 
Seth McGinnis

Philosophers of science study how science is practiced and how scientists develop new understanding of the systems they investigate. NCAR has recently had a number of philosophers of science as visiting scholars, who have collaborated with climate scientists and others in the organization to study how we work with numerical models and big data.

Speaker Description: 

Seth McGinnis is an Associate Scientist IV at NCAR with joint appointments in CISL and RAL.  As the Data Manager for the NARCCAP and NA-CORDEX data collections, he makes the output from regional climate models usable by and available to people who need information about climate change in North America.  His research focuses on bias correction, interpolation, data access, and other issues affecting the practical use of model output by non-specialists.

Event Category:

Performance Analysis of NALU on aarch64

Location: 
CG Auditorium
Speaker: 
Srinath Vadlamani

Arm based SOCs have proven to be viable platforms for the complicated workloads in scientific HPC computing.   We will present performance analysis of the SIERRA Low Mach Module: Nalu (henceforth referred to as Nalu), developed at Sandia National Labs, applications on arm based ThunderX2, AWS Graviton2 and A64FX platforms (as long as access is given).  Single node memory, compute and thread performance will be presented for a production relevant test case.  We will present scaling performance on many-node systems.

Speaker Description: 

Practitioner of parallel HPC computing with a focus on scientific applications while considering algorithm enhancements acknowledging hardware performance efficacy. 

Event Category:

Taking Advantage of Machine Learning for Fire Radiative Power Modeling

Location: 
CG Auditorium
Speaker: 
Christina Kumler

Currently, the High Resolution Rapid Refresh (HRRR) coupled with Smoke (HRRR-Smoke) and many other smoke forecasting models represent diurnal cycle of fire emissions by using a one-curve fits-all approach for modeling future FRP with a climatological curve. As is, it is quite challenging to accurately forecast fire emissions and smoke concentrations in rapidly changing fire behavior and weather conditions. Further, sometimes FRP values may not be available or are missing for an active fire.

Speaker Description: 

Christina did her undergraduate work in applied mathematics at CU Boulder, then got her masters in atmospheric and oceanic sciences at University of Miami, Florida. She has been working on machine learning projects as applied to earth sciences for the last couple of years. When not working, she's into racing triathlons, photography, exploring the outdoors, and eating tasty food with her husband and dog.

Event Category:

CESM in the Cloud

Location: 
CG Auditorium
Speaker: 
Brian Dobbins

I'll add to this later -- but in short:

Speaker Description: 

Software Engineer @ NCAR working in Technology Development

Event Category:

Maintainable HPC with Python and C++

Location: 
CG Auditorium
Speaker: 
Srijith Rajamohan

High-Performance software usually requires the use of a low-level language such as C or C++ with various message passing or shared memory paradigms such as MPI or OpenMP. With the popularity of Python in the world of Data Science a plethora of tools have evolved to make data analysis easier and reduce the mean time to discovery. A lot of these frameworks such as Python Numba can be used for Scientific Computing as well. However, when optimizing compute-intensive portions of the code a language such as C or C++ is desirable.

Speaker Description: 

I am currently employed at Virginia Tech as a Computational Scientist since September 2014. My research interests lie in Numerical Methods, Computational Science and Machine Learning using High-Performance Computing with traditional and novel accelerator technologies. Since joining VT, I have also had a chance to hone my skills at Data Visualization. My background during my Doctoral work at the SimCenter: National Center for Computational Engineering, was in CFD and Electromagnetics using the Finite-Element Method.

I have a Doctorate in Computational Engineering from the University of Tennessee and a Masters in Electrical Engineering from the The Pennsylvania State University.

Please find more about me at srijithr.gitlab.io and a web resume at https://srijithr-newresume.netlify.com/

Event Category:

Pages

Subscribe to conference-talk