In this video, Pranay Reddy Kommera describes his prize-winning poster, “Design of a Distributed Memory, Multi-GPU Implementation of a 2D DG Core.”
Two University of Wyoming graduate students earned a trip to the SC16 conference in November by virtue of winning the poster contest at the recent Rocky Mountain Advanced Computing Consortium (RMACC) High Performance Computing Symposium.
As part of their prize, both students will attend the SC16 in Salt Lake City (Nov. 13-18), which will expose them to cutting-edge research being performed both in HPC and extreme computing,” says Suresh Muknahallipatna, a UW professor in the Department of Electrical and Computer Engineering, and the students’ mentor.
The 2016 RMACC Poster winners:
- Pranay Reddy Kommera, a first-year doctoral student from India majoring in electrical engineering, took first prize with his poster project, titled “Design and Optimization of a Distributed Memory, Multi-GPU Implementation of a 2-D Discontinuous Galerkin Atmospheric Dynamical Core.”
- Supreeth Madapur Suresh, a second-year master’s student from Bangalore, India, majoring in electrical and computer engineering, won third place with his poster project, titled “An OpenACC Implementation of a Dynamical Core Component of the MPAS Atmospheric Model.”
Kommera’s poster focused on the design and optimization of multiple graphics processing units (GPU) for a mathematical method — called Discontinuous Galerkin — that is used for high-resolution meteorological modeling and weather forecasting. This method is used for an atmospheric model (called a non-hydrostatic model) that uses the latest computer hardware and involves central processing units (Intel CPUs) and graphic cards (NVIDIA GPUs).
His work is designed to improve the parallel scaling efficiency using the Open Accelerators (OpenACC) Programming model.
In today’s world, accurate weather and climate forecasting prediction is of high importance. Atmospheric researchers develop many models (simulations) involving mathematical equations to predict weather accurately,” Kommera says. “These atmospheric models are highly computational and require lot of time to execute. With the advent of sophisticated computer hardware and accelerators, or graphic cards, it was made possible to implement the models in less time than before.”
Suresh’s poster focused on refactoring of the existing Model for Prediction Across Scales (MPAS) source code — for both CPU and GPU execution — using the OpenACC programming model. His work looks primarily at the portability between CPU and GPU, and improving performance by loop optimization.
The MPAS is a collaborative project for developing atmosphere, ocean and other earth-system simulation components for use in climate, regional climate and weather studies, Suresh says. OpenACC is a directive-based approach, which adds a few lines of code to work on both CPU and GPU. By adding OpenACC directives and performing a few basic optimization techniques, Suresh says he was able to achieve a speed-up of almost 4.5 times on CPU and eight times on GPU.
Both students presented their posters based on research connected with a summer 2016 internship at the National Center for Atmospheric Research (NCAR) in Boulder, Colo., and research grant work of Muknahallipatna that also was funded by NCAR.
Sponsored by RMACC, of which UW is a member, the annual event brings together designers and users of high-performance computing systems from universities, government laboratories and industry throughout the Rocky Mountain region. Research, education and best practices will be discussed. UW hosted the conference in 2013.
RMACC is a collaboration among academic and research institutions located throughout the intermountain states. Its mission is to facilitate widespread effective use of high-performance computing throughout the Rocky Mountain region.
This year’s symposium — which took place in Fort Collins — featured sessions designed specifically for students, who had the opportunity to learn about careers in the field of high-performance computing, interview and resume skills, and strong production skills. It also featured special beginner-level tutorials for students who wish to learn how to use a variety of advanced computing skills in their research.
The symposium regularly features a series of tutorials covering many aspects of high-performance computing system design and use, nationally recognized speakers and group breakout sessions.
Both students view the upcoming supercomputing conference as a good opportunity to learn more about their chosen field and make valuable contacts.
I hope to receive good exposure to the most recent advancements in the field of high-performance computing,” Kommera says. “I also will be looking into work of other participants in the field of parallel computing and computer vision, as these are my Ph.D. research interests. It also will be a platform to meet many other researchers and understand the real-world problems they are trying to solve.”
Suresh had similar thoughts, saying, in the field of high-performance computing, the supercomputing conference is one of the most important events.
All top companies in the field of high-performance computing will showcase their upcoming technologies,” he says. “This exposes me to all the cutting-edge technologies in the field. The supercomputing conference conducts workshops and training sessions by industry experts. I believe, by attending these sessions, I can improve my knowledge and keep track of all of the new innovations in the field.”
Additionally, Vivaswat Shastry, a UW undergraduate research assistant, also will attend the supercomputing conference, Muknahallipatna says. Shastry was selected to attend the conference based on his HPC research work using NVIDIA Jetson for embedded computing. His attendance is fully sponsored by the supercomputing conference.
Source: University of Wyoming