Name: Li Tang
Pronouns: he/him/his
Biography:
Dr. Tang is a Computer Scientist in the Applied Computer Science group at Loa Alamos National Laboratory (LANL). Before joining LANL, he was a Research Associate in Computational Science Initiative (CSI) at Brookhaven National Laboratory (BNL). Dr. Tang received his Ph.D. from the University of Notre Dame under the supervision of Dr. Xiaobo Sharon Hu in 2017 and worked as a Research Intern at Sandia National Laboratories (SNL) in 2012. His PhD work was selected as a doctoral showcase in SC’16. Dr. Tang has a broad background of High-performance Computing (HPC) co-design. He has been involved in the Department of Energy (DOE) co-design efforts since 2011 and has over 14 years of experience working on accelerating applications with novel computing devices. His current research interests include HPC co-design, novel computing beyond Moore’s Law, programming models, and in-situ data systems. His recent papers of novel computing and in-situ data systems received two ISAV Best Paper awards and one ExHET Best Paper award.
Institution/Lab: Los Alamos National Laboratory
Website: ltang85.github.io
SRP Collaboration Topic/Title: High-Performance Computing with Python
Field or research area: High-Performance Computing
Please select all the topical areas that apply to your project:
Computational Science Applications (i.e., bioscience, cosmology, chemistry, environmental science, nanotechnology, climate, etc.); 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:
LANL’s Venado, the first NVIDIA Grace-Hopper supercomputer in the US, will be ready in early 2024. NVIDIA’s Grace-Hopper integrates its latest Grace CPU and H100 GPU on the same chip for faster CPU-GPU data movement, and Venado provides an exotic liquid cooling system for maximized system performance and reliability. To run large-scale simulations using GPUs on Venado, conventional HPC programming solutions include the mix of CUDA (GPU programming) and MPI (node communication), and the Kokkos ecosystem. However, these solutions usually require significant programming (e.g., C++ and MPI) and hardware expertise (e.g., GPU architecture). To strengthen LANL’s HPC productivity on the rapidly evolving heterogeneous supercomputers with GPUs, we will evaluate an innovative HPC programming paradigm, programming physics simulations using NumPy, by accelerating LANL’s physics simulations on hundreds of Venado GPU nodes.
Desired relevant skills, background, or interests:
Python, NumPy
Other comments:
Do any special requirements apply? In-Person Only; Permanent Resident OK; International OK
Other, specify:
Keywords:
HPC;Ocean Modeling;Python;GPU
Lightning Talk Title: Physics Modeling with Python on GPUs