Advanced HPC-CI Webinar Series: Portable Accelerator Programming with SYCL

Remote event

SYCL is an open standard for programming heterogeneous architectures in ISO C++. This webinar will give a high-level overview of the SYCL programming language and the software ecosystem to write and tune SYCL code for different accelerator architectures. We will focus on GPUs and discuss how we can provide SYCL performance portability across hardware from different vendors, including Intel, Nvidia, and AMD GPUs, by employing a single-source model based on a modern C++ standard. As an example of complex scientific software, we will demonstrate briefly how the Amber molecular dynamics software was ported from CUDA to SYCL using Intel oneAPI software development tools and Intel Xe architecture GPUs. We will discuss numerical results and benchmark data that demonstrate the accuracy and performance of the SYCL implementation on data center and consumer-grade GPU hardware.

Instructor

Andreas Goetz

Associate Research Scientist, SDSC

Andreas Goetz is an associate research scientist at SDSC, where he leads a research group in Quantum- and AI-enabled computational Chemistry, working at the intersection of chemistry, life sciences, and scientific computing. His research draws on quantum mechanics, statistical mechanics, molecular dynamics, and machine learning approaches to enable simulations of complex molecular systems on massively parallel computer architectures. He is a contributing author to the ADF and QUICK quantum chemistry software and the AMBER software for biomolecular simulations, which are widely used in academic and industrial research. He is also co-founder of ATTMOS Inc., which develops technologies for computational drug discovery. Andreas enjoys training the next generation of scientists in software engineering and numerical simulation methods via lectures, workshops, and the supervision of interns. He is the author of over 80 scientific publications and editor of the book 'Electronic Structure Calculations on Graphics Processing Units. Before joining SDSC in 2009, Andreas performed postdoctoral research at the VU University in Amsterdam and obtained his undergraduate and Ph.D. degrees in chemistry from the Friedrich-Alexander University in Erlangen, Germany.