NERSCPowering Scientific Discovery Since 1974

GPUs for Science day 2022, October 25th

Introduction

NERSC is proud to reboot the GPU for Science day event. Perlmutter has played a a crucial role in helping the wider scientific community leverage and embrace GPUs for their simulations. Our goal is to help facilitate the transition to GPU systems, and to offer users the motivation, tools, and expertise they need to make this change possible. The day will start with an introduction of Perlmutter, followed by the current state of multiple compilers and supported programming models for running on GPUs. The majority of the agenda will showcase about 10 science application teams to share their first-hand experiences and best practices on porting to GPUs. We hope this knowledge and lessons learned can be adapted to users’ own applications. The day will end with a panel on popularly used performance portable accelerator programming models.

With the announcement of Cori’s approaching retirement in January 2023, users migrating their applications from Cori to Perlmutter GPUs are highly encouraged to join. This event is also open to OLCF and ALCF users.

Date and Time: 8:30 am - 5:00 am (Pacific time), October 25 (Tuesday)

This will be a virtual event with presentations and video links made available shortly. Please join us for this event!

Agenda

Tuesday, October 25th

TitleSpeakerTime
Introduction and Welcome Jack Deslippe (NERSC)  8:30 AM (10 min)
Introduction to Perlmutter system
Jay Srinivasan (NERSC)  8:40 AM (10 min)
Current state of CUDA compilers and SDK Jeff Larkin (NVIDIA)  9:00 AM (30 min)
Current state of LLVM compiler Johannes Doerfert (LLVM)  9:30 AM (30 min)
Current state of CRAY GPU compiler John Levesque (HPE) 10:00 AM (30 min)
Coffee Break   10:30 AM (10 min)
Data Parallel C++ (DPC++) programming model Abhishek Bagusetty (Argonne) 10:40 AM (20 min)
Accelerating X-Ray tracing for Exascale systems using Kokkos Felix Wittwer (NERSC)
11:00 AM (20 min)
GPU based simulations with QMCPack
Ye Luo (Argonne) 11:20 AM (20 min)
Lunch Break   11:40 AM (50 min)
Massively parallel MD simulations using NAMD David Hardy (NAMD) 12:30 PM (20 min)
Lattice QCD (LQCD) project Steven Gottlieb (Indiana Univ.) 12:50 PM (20 min)
Massively parallel PIC using WARPX Andrew Myers (CRD, LBL) 01:10 PM (20 min)
Porting WEST code to GPUs
Victor Yu (Argonne) 01:30 PM (20 min)
Break   01:50 PM (20 min)
Metagenome analysis (MetaHipMer) for ExaBiome Muaaz Awan (NERSC) 02:10 PM (20 min)
XGC and WDMApp on Perlmutter
Aaron Scheinberg 02:30 PM (20 min)
Machine Learning applications to Collider Physics
Vinicius Mikuni (NERSC) 02:50 PM (20 min)
PASTIS @ ExaBiome Oguz Selvitopi 03:10 PM (20 min)
Break   03:30 PM (15 min)
Session: Performance Portability APIs Rahulkumar Gayatri, Brandon Cook, & Dossay Oryspayev 03:45 PM (45 min)
Closing remarks   04:30 PM  (10 min)

 

 

Registration

Registration is FREE and open for everyone interested, and users need to sign up at this link.

This event will be presented online using zoom.

 

Organizers

The members of the organizing committee for GPUs for Science 2022 are:

Dhruva Kulkarni

Neil Mehta

Helen He

Jack Deslippe