NERSCPowering Scientific Discovery Since 1974

GPUs for Science Day

Location: Lawrence Berkeley National Laboratory

July 2 & 3, 2019

In anticipation of our next system Perlmutter which will arrive in 2020, NERSC has a crucial role to play in helping the wider scientific community leverage and embrace GPUs for their simulation, data processing, and machine learning workloads. Part of this role includes hosting events like GPUs for Science Day, the goal of which is to help facilitate the transition to GPU systems by giving scientists the motivation, the tools, and the expertise they need to make this change possible.

The first day of the workshop featured overview talks, technical talks, and some case-studies. The second day of the workshop featured lightning talks and a mini GPU competition with prizes awarded to the team demonstrating the most improved performance.

Videos of the Day 1 talks (broken into AM and PM sessions) can be found here.

Agenda

Tuesday, July 2nd

Location: Building 50 Auditorium, 8:30am - 5:00pm

Day 1 Agenda (subject to change)

TitleSpeakerTime
Registration and Breakfast   8:00 AM (30 min)
Introduction and Welcome to GPUs for Science Day Jack Deslippe (NERSC) 8:30 AM (5 min)
Session: HPC + GPU Chairperson: Jack Deslippe  
NERSC to NVIDIA: A Perspective Zahra Ronaghi (NVIDIA) 8:35 AM (20 min)
Experiences on Summit  Jack Wells (ORNL) 8:55 AM (20 min)
Performance, Portability and Productivity Hai Ah Nam (LANL) 9:15 AM (25 min)
How to Run Fast on a GPU Max Katz (NVIDIA) 9:40 AM (25 min)
NERSC Application Readiness Process and Strategy Charlene Yang (NERSC) 10:05 AM (15 min)
Break   10:20 AM (10 min)
Session: Directive-Based Programming Chairperson: Helen He  
OpenMP: A Technical Overview Chris Daley (NERSC) 10:30 AM (25 min)
OpenMP 4.5 Case Study Rahul Gayatri (NERSC) 10:55 AM (15 min)
OpenACC for Science Jeff Larkin (NVIDIA) 11:10 AM (40 min)
Working Lunch   11:50 AM (45 min)
Session: Performance Portability APIs and Superfacility Chairperson: Brandon Cook  
The Kokkos C++ Performance Portability EcoSystem David Hollman (Sandia) 12:35 PM (25 min)
Kokkos: A Case Study Daniel Holladay (LANL) 1:00 PM (15 min)
RAJA: A Technical Perspective David Beckingsale (LLNL) 1:15 PM (25 min)
RAJA: An Applications Perspective Rich Hornung (LLNL) 1:40 PM (15 min)
The NERSC Superfacility Project: A Technical Overview Cory Snavely (NERSC) 1:55 PM (25 min)
Superfacility: The CAMERA Project as Case Study Alex Hexemer (LBL) 2:20 PM (15 min)
TomoPy: A CUDA Case Study Jonathan Madsen (NERSC) 2:35 PM (15 min)
Break   2:50 PM  (10 min)
Session: Bioinformatics and Machine Learning Chairperson: Steve Farrell  
Bioinformatics on GPUs Fernanda Foertter (NVIDIA) 3:00 PM (25 min)
Bioinformatics: A Case Study Bryce Foster (JGI) 3:25 PM (15 min)
High Performance Python Components Randy Gelhausen (NVIDIA) 3:40 PM (25 min)
Overview: Inelastic Neutron Scattering Claire Saunders (Caltech) 4:05 PM (5 min)
Accelerating Scientific Data Analysis with RAPIDS and CuPy Nick Becker (NVIDIA) 4:10 PM (10 min)
Machine Learning: A Technical Overview Mustafa Mustafa (NERSC) 4:20 PM (25 min)
Exascale Deep Learning for Climate Analytics Thorsten Kurth (NERSC) 4:45 PM (15 min) 

No-host dinner: 

  • Venue: Jupiter (map)

Wednesday, July 3rd

Location: Building 59, Room: 3101, 9:00am - 5pm

Day 2: Tutorials, Lightning talks, and GPU challenge

TitleSpeakerTime
Registration and Breakfast   8:30 AM (30 min)
Session: Application Readiness Chairperson: Thorsten Kurth   
Build Systems Jonathan Madsen (NERSC) 9:00 AM (30 min)
Docker/Shifter Containers Shane Cannon (NERSC) 9:30 AM (30 min)
Cori GPU nodes Brian Friesen (NERSC) 10:00 AM (30 min)
Profiling Tools Max Katz (NVIDIA) 10:30 AM (30 min)
Break   11:00 AM (15 min)
Hackathon overview and preparation Jack Deslippe (NERSC) 11:15 AM (45 min)
Working Lunch   12:00 PM (60 min)
Session: Lightning Talks Chairperson: Rollin Thomas  
Python + GPUs Laurie Stephey (NERSC) 12:10 PM (10 min) 
Ptychography Bjoern Enders (NERSC) 12:20 PM (10 min) 
AMReX Kevin Gott (NERSC) 12:30 PM (10 min)
WarpX: implementation and performance on GPU Rémi Lehe (LBL) 12:40 PM (10 min) 
BerkeleyGW Mauro Del Ben (LBL) 12:50 PM (10 min) 
Break   1:00 PM (15 min)
Hacking Competition*   1:15 PM - 4:00 PM

* Competition will be for most speed-up in a migration of provided mini-app to GPU and will be available in C++

Venue 

Lawrence Berkeley National Laboratory

Building 50 Auditorium and Building 59 Room 3010

Screen Shot 2019 06 24 at 1.07.57 PM

050 Aud 1

 

Site access:

If you are not an LBL employee and will attend in person, you will need to be pre-approved for site access. All registered participants should have been contacted by our administrator to help you in this process. 

Information about the free LBL shuttle: https://www2.lbl.gov/Workplace/Facilities/Support/Busses/all_routes.html

If you plan to ride the LBL shuttle, you must show your PRINTED site access pass and ID in order to board. The drivers will not accept the pass on your phone. Also, the LBL shuttle will not stop for you unless you wave since the stops are shared with UCB and city buses. Ride the LBL Blue Uphill shuttle route to building 65, which is the closest stop to both Building 50 (Day 1) and Building 59 (Day 2). 

Parking at LBL is free but it can be difficult to find a space unless you arrive very early. We recommend riding the shuttle if possible. 

1 Cyclotron Road, Bldg. 59
Berkeley, CA 94720
Tel: 510.486.7612
http://www.lbl.gov/

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