2021 ALCF Computational Performance Workshop

Workshop
2021 ALCF Computational Performance Workshop

Attend our virtual workshop to boost application performance and achieve computational readiness for INCITE and other ALCF allocation programs.

The 2021 ALCF Computational Performance Workshop will be hosted as an online-only event.

Join us May 4-6, 2021, for an experience designed to help you boost application performance and achieve computational readiness on ALCF systems. Participants will have an opportunity to:

  • Work with ALCF and industry professionals through collaborative online sessions
  • Participate in guided hands-on exercises
  • Benchmark and debug your application with exclusive reservations on ALCF computing systems
  • Learn how to improve utilization on Theta/ThetaGPU/Cooley and prepare for future ALCF computing resources
  • Explore advanced techniques and tools to enhance application performance and expand your data science skills
  • Prepare for a major allocation award (e.g., INCITE, ALCC, ALCF Data Science Program)

For questions or concerns, please email us at comp-perf-support@alcf.anl.gov

Workshop GitHub Repository

Agenda

Day 1, Tuesday, May 4

Topic (most sessions include a hands-on component)

Speakers

10:00-10:05am(CST) Welcome Mike Papka (CELS ALD and ALCF Director)
10:05-10:15am(CST) Onboarding, Slides Yasaman Ghadar (Argonne)
10:15-11:15am (CST) Workshop Overview (Slides,Video) Getting Started on Theta (Slides, Video) and ThetaGPU (Slides, Video) Chris Knight, Graham Fletcher, Corey Adams, Misha Salim (Argonne)
10:15-11:15am(CST) Overview of TAU at Breakout Room, (Slides) Sameer Shende(TAU)
10:15-11:15am(CST) Overview of HPC ToolKit at Breakout Room(Slides) Mark Krentel (HPCToolkit)
11:15-11:30am (CST) Break  
11:30am-12:30pm (CST) Enabling Workflows with Balsam (Video) and Parsl (Video, Slides Misha Salim, Antonio Villareal, Tom Uram (Argonne)
12:30-1:00pm (CST) Lunch Break  
1:00-1:30pm (CST) Future ALCF Systems (Slides, Video) Scott Parker (Argonne)

1:30-2:30pm (CST)

Portability with OpenMP (Slides, Video) Colleen Bertoni, Ye Luo (Argonne)
2:30-2:45pm (CST) Break  
2:45-3:45pm (CST) Visualizing Your Data (Slides,Video) Joseph Insley, Silvio Rizzi, Janet Knowles, Victor Mateevitsi (Argonne)
3:45-4:15pm (CST) Allocation Programs at ALCF (Slides, Video) Katherine Riley (Argonne)
4:15-4:45pm (CST) Containers (Slides, Video) Romit Maulik, Taylor Childers (Argonne)
4:45-5:00pm (CST)  Day 1 Feedback  

Day 2, Wednesday, May 5

 
10:00-11:00am (CST) NVIDIA Ecosystem (Slides, Video) Max Katz (NVIDIA)
11:00-11:15am (CST) Break  
11:15am-12:15pm (CST) Scaling Deep Learning Applications (Video) Sam Foreman, Huihuo Zheng (Argonne)
12:15-12:45pm (CST) Lunch Break  
12:45-1:45pm (CST) Portability with Kokkos (Video, Hands-on Video) and Raja (Video) Nevin Liber, Patrick Steinbrecher, Brian Homerding (Argonne)
1:45-2:00pm (CST) Break  
2:00-2:45pm (CST) Data Transfer and Sharing with Globus, (Slides, Video

 

Gregory Nawrocki (Globus)

2:45-3:00pm (CST) Data Transfer and Sharing with Eagle, (Slides, Video) Avanthi Mantrala (Argonne)
3:00-4:15pm (CST) Overview of VTune/Advisor (Slides, Video) Matthew Cordery (Intel)
4:15-4:30pm (CST) Day 2 Feedback  

Day 3, Thursday, May 6

   
10:00-11:00am (CST) Profiling Frameworks with TensorFlow(Video) and Pytorch (Slides, Video Denis Boyda, Corey Adams, Filippo Simni (Argonne)
11:00-11:15am (CST) Break  
11:15am-12:15pm (CST) Reduced Precision in TF, Torch (Video) Kyle Felker, Murali Krishna Emani, Bethany Lusch (Argonne)
12:15-12:45pm (CST) Lunch Break  
12:45-1:45pm (CST) File Systems and I/O Performance (Slides, Video) Gordon McPheeters (Argonne)
1:45-2:00pm (CST) Break  
2:00-2:45pm (CST) MPI/Network (Slides, Video Sudheer Chunduri (Argonne)
2:45-3:30pm (CST) Portability with SYCL/oneAPI(Slides, Video, Hands-on Video) Thomas Applencourt (Argonne)
3:30-4:00pm (CT) MPS/MIG Mode on A100 (Video) Murali Krishna Emani, Venkatram Vishwanath (Argonne)
4:00-4:15pm (CST) Next Steps (DD) (Slides) Yasaman Ghadar (Argonne)
4:15-4:30pm (CST) Day 3 Feedback and Adjourn