GPUs for Science Day, Oct. 12, 2023

GPUs have been instrumental in ground-breaking innovations, from scientific simulations to generative AI. This year, NERSC is proud to host the annual GPUs for Science event in person. Our goal is to celebrate recent GPU-enabled scientific achievements and inspire future roadmaps. The day will start with an introduction to three DOE compute facilities (NERSC, ALCF, and OLCF), followed by a series of talks on GPU-accelerated scientific applications and emerging software programming models. The day will wrap up with a panel of leading industry experts discussing their vision for upcoming GPU ecosystems.

With increasing diversity in GPU hardware and software, users interested in performance and portability across DOE supercomputers are highly encouraged to join. The event is also open to everyone interested in learning about the exciting science in action. 

Logistics

• Meeting type: In-person (with optional Zoom attendance)
• Location: Lawrence Berkeley National Lab (Berkeley Lab), Bldg 59 (NERSC), Room 3101
• Date: October 12, 2023 (Thursday)
• Time: 8:30 a.m. - 5 p.m. (Pacific Daylight Time, UTC-7)
• Recording available on YouTube

Registration

• In-person attendee registration is now closed.
• Online attendee (live stream) registration is closed.

Agenda*

Title	Speakers	Affiliation	Time PDT (UTC-07)
Breakfast and Registration			8 a.m.
Opening Remarks	Jack Deslippe	LBNL	8:45 a.m.
Session 1: Compute Facilities			9 a.m.
Perlmutter system at NERSC	Debbie Bard Hannah Ross	LBNL	9:05 a.m.
The path to Aurora	Jae Hyuk Kwack	ANL	9:25 a.m.
Science from GPUs on the Frontier Exascale Supercomputer	Balint Joo	ORNL	9:45 a.m.
Break (10 mins)			10:05 a.m.
Session 2: Applications			10:15 a.m.
Harnessing GPUs for science: lessons from the Exascale Computing Project	Erik Draeger	LLNL	10:15 a.m.
Towards the Convergence of Data-Driven and Kinetic Models at Exascale	Axel Huebl	LBNL	10:35 a.m.
Global Cloud Resolving Modeling on Exascale Computers	Mark Taylor	Sandia	10:55 a.m.
ExaCA: exascale component-scale microstructure simulation	Sam Reeve	ORNL	11:15 a.m.
Exascale Flow Simulations for Energy Science	Paul Fischer	ANL	11:35 a.m.
Q&A Session over lunch (45 mins)			11:55 a.m.
Enhancing Science: Efficient Julia Workflows and GPU Use on Perlmutter with Jupyter	Marius Millea	UC Davis	12:40 p.m.
Accelerating Molecular Discovery and Design with AI: Insights from the Open Catalyst Project	Brandon Wood	Meta	1:00 p.m.
Then and Now: How GPUs Empower ECP Mathematical Libraries	Sherry Li	LBNL	1:20 p.m.
AMReX and selected AMReX-based applications	Ann Almgren Weiqun Zhang	LBNL	1:40 p.m.
Break (10 mins)			2 p.m.
Session 3: Programming Models			2:10 p.m.
A Task Graph-based Programming System for CPU-GPU Heterogeneous Computing	Dian-lun Lin	UW - Madison	2:10 p.m.
Stellar Mergers as a Benchmark: Performance of Octo-Tiger, HPX and Kokkos on Perlmutter	Gregor Daiß	University of Stuttgart / LSU	2:30 p.m.
Evaluating the Performance of One-sided Communication on CPUs and GPUs	Nan Ding	LBNL	2:50 p.m.
Break (20 mins)			3:10 p.m.
Session 4: Emerging Hardware and Software Ecosystems			3:30 p.m.
NVIDIA's Accelerated Computing Ecosystem	Scot Halverson	NVIDIA	3:30 p.m.
Software & Hardware Co-design for Exascale Scientific Computing	Michael Rowan	AMD	3:45 p.m.
Future-Proofing Your Science: Programming GPUs (and More!) with Open Standards	John Pennycook	Intel	4 p.m.
Panel Discussion (30 min)			4:15 p.m.
Closing Remarks	Weile Wei	LBNL	4:45 p.m.
Adjourn			4:55 p.m.

*Agenda subject to change

Visitor Information

Please refer to our visitor information page and an interactive map of the Berkeley Lab campus. An area map is shown below for quick reference.

Organizing Committee

Mukul Dave
Muhammad Haseeb
Dhruva Kulkarni
Weile Wei 

About Berkeley Lab

Lawrence Berkeley National Laboratory is a multidisciplinary national laboratory located in Berkeley, California, on a hillside directly above the campus of UC Berkeley. The site consists of 76 buildings located on 183 acres, which overlook both the campus and the San Francisco Bay.