Name: Nan Ding
Pronouns: she/her/hers
Biography:
Nan Ding is a Research Scientist in the Performance and Algorithms group of the Computer Science Department at Lawrence Berkeley National Laboratory. Her research interests include high-performance computing, performance modeling, and performance optimization. Nan received her Ph.D. in computer science from Tsinghua University, Beijing, China, in 2018.
Institution/Lab: Lawrence Berkeley National Laboratory
Website: https://crd.lbl.gov/divisions/amcr/computer-science-amcr/par/members/staff/nan-ding/
SRP Collaboration Topic/Title: Explore the performance of GPU-initiated communications and CPU-initiated communication on heterogeneous architectures
Field or research area: HPC Communications
Please select all the topical areas that apply to your project:
Computer Science (i.e., architectures, compilers/languages, networks, workflow/edge, experiment automation, containers, neuromorphic computing, programming models, operating systems, sustainable software); High-Performance Computing
Brief Abstract:
Effective programming models offer programmers the ability to harness the capabilities of the underlying platform. For decades, the CPU-initiated Message Passing Interface (MPI) has become a de facto standard for communication among processes running on distributed memory systems. As high-performance GPU computing becomes the trend, GPU-initiated communication becomes a viable solution for multi-GPU scaling. However, the lack of deep understanding of GPU-initiated communication performance and its impact on an application’s performance becomes a hurdle. As such, the proposed topic is to explore the performance of GPU-initiated communications and CPU-initiated communication by using or creating representative benchmarks on Perlmutter, Frontier, and possibly Aurora. The work may includes writing benchmark using CPU- and GPU-initiated communication.
Desired relevant skills, background, or interests:
knows MPI programming Passion about research work
Other comments:
Do any special requirements apply? Minimum GPA (specify what GPA in comments below); In-Person Only; U.S. Citizen Only; Permanent Resident OK; International OK
Other, specify:
Keywords:
one-sided MPI, GPU clusters, point-to-point communication, collective communication
Lightning Talk Title: A message roofline model to characterize communication performance