Takeaways from the 2022 PaCER Mentored Sprints
Learn from the experiences of others, as they port applications to AMD GPUs!
This is an online session. Please complete the form below to receive dial-in details.
This talk summarize the PaCER mentored sprints and the experience of porting PaCER applications to AMD GPU hardware.
We will cover highlights of porting existing OpenACC and CUDA applications to AMD GPUS. Particularly, we will look at initial runtime comparisons of a number of applications on Topaz and Mulan in addition to a number of strategies for profiling applications on AMD GPU hardware. We’ll cover common porting activities and best practices for sprints based on developer experiences, and discuss currently available approaches for performance optimization on AMD GPUs.
Speaker Joseph Schoonover, Founder & Operations Lead and Lead Software Engineer, Fluid Numerics
 |
Joe Schoonover is a scientist and research software engineer at Fluid Numerics. From 2016-2017, Dr. Schoonover held a post-doctoral research position at Los Alamos National Laboratory’s Center for Nonlinear Studies, where he studied topographic wave arrest as a mechanism for western boundary current separation, transport operator diagnosis through impulse functions and graph theory, climate model equilibration acceleration strategies, and high performance computing.
Joe’s current research focus is in spectral element methods, physical oceanography, and environmental fluid dynamics and he has a continued interest in portable GPU programming and scientific application modernization. |