Speaker: Mohamed Wahib, RIKEN Center for Computational Science, Japan
Date and time: Tuesday, September 29, 2026, at 2:00 p.m. CEST
Zoom link: https://go-nhr.de/perflab-seminar
Slides: https://hpc.fau.de/files/2026/10/Programming_GPU_Spet_2026.pdf
Abstract:
GPUs have become central to high-performance computing, but the way we program them is increasingly challenged by the evolution of their architecture. Traditional GPU programming emphasizes mapping parallel work onto threads, managing memory access, and optimizing individual kernels. Modern GPUs, increasingly shaped by AI workloads, combine specialized compute units, asynchronous data movement, and complex synchronization mechanisms, requiring us to rethink how scientific computation is organized. This talk traces the evolution of GPU programming through our research on accelerating HPC applications, from kernel optimization and automated kernel fusion to higher-level frameworks for adaptive scientific computation. Using examples from HPC applications, I will discuss how programming abstractions can improve productivity while preserving performance and where conventional approaches encounter limitations. I will then examine emerging directions beyond a homogeneous view of single-instruction, multiple-thread (SIMT) execution.
Short Bio:
Mohamed Wahib is a team principal (PI) of the “High Performance Artificial Intelligence Systems Research Team” at RIKEN Center for Computational Science (R-CCS), Kobe, Japan. Prior to that he worked as a senior scientist at AIST/TokyoTech Open Innovation Laboratory, Tokyo, Japan. His research interests revolve around the central topic of high-performance programming systems, in the context of HPC and AI. He is actively working on several projects including AI-based science, as well as high-level frameworks for programming traditional scientific applications.
For a list of past and upcoming NHR PerfLab seminar events, please see: https://hpc.fau.de/research/nhr-perflab-seminar-series/