P’Con – Experience with porting and scaling codes on AMD GPUs
9:30am - 12:00pm
As part of the first PaCER Conference – P’con – Setting the pace for exascale – Sunita Chandrasekaran, Associate Professor with the Department of Computer & Information Sciences CIS at the University of Delaware and Computational Scientist at Brookhaven National Lab will join Maicon Faria from HPC Now! to discuss and share their experiences porting and scaling codes on AMD GPUs.
The Pawsey Supercomputing Research Centre is undergoing a Capital Refresh, currently commissioning its Setonix supercomputer, which will be 30x more powerful than its current systems, Magnus and Galaxy. Setonix will feature 200k+ AMD EPYC™ Milan CPUs and 750+ AMD Instinct™ MI200 GPUs, configured such that it will provide a speed performance of 50 petaflops to Australian researchers, making it one of the largest supercomputers in the region.
To support researchers’ transition to the new technology and to create a direct pathway to achieve superior scale on next-generation supercomputers, the Centre launched the PaCER initiative. Through a competitive process, PaCER selected 10 research groups.
PaCER projects are gaining early access to supercomputing tools and infrastructure, training and exclusive hackathons focused on HPC performance at scale.
While P’con runs for a week, we will open the conference to the entire HPC and research community to participate in Thursday’s events.
Register to take the opportunity to learn from internationally recognised experts about working with GPU technologies.
Speaker: Prof. Sunita Chandrasekaran
Title: Preparing the software stack for the next-generation computing systems by Sunita Chandrasekaran
Abstract: This talk will tell stories from ECP SOLLVE and CAAR PIConGPU projects. The speaker will share programming challenges with respect to the software stack readiness for large complex scientific applications targeting systems like Frontier. The speaker will also share how the exascale supercomputer is offering scientists an opportunity to be able to prepare their applications in a manner that they have never been able to even think about until now. The stories will highlight novel and incremental achievements experienced by both the ECP SOLLVE and CAAR PIConGPU teams.
Sunita Chandrasekaran is an Associate Professor with the Department of Computer and Information Sciences at the University of Delaware, USA. She is also a computational scientist with Brookhaven National Laboratory. She is currently the PI for ECP SOLLVE and ORNL CAAR projects among others. She received her PhD in 2012 on Tools and Algorithms for High-Level Algorithm Mapping to FPGAs from the School of Computer Science and Engineering, Nanyang Technological University, Singapore. Her research spans High Performance Computing, exascale computing, parallel programming, benchmarking and data science. Applications of interest include scientific domains such as plasma physics, biophysics, solar physics and bioinformatics. She is a recipient of the 2016 IEEE-CS TCHPC Award for Excellence for Early Career Researchers in High Performance Computing. She has been involved with SC, ISC, IPDPS, IEEE Cluster, CCGrid, WACCPD, AsHES and P3MA in different capacities.
Speaker: Dr Maicon Faria
Title: Benchmarking, profiling and porting to AMD GPU architectures, use cases.
Abstract: The fast evolution pace of accelerators and APIs used on GPGPU aiming development create a complex scenario for planning long term support of software and keeping it in a state-of-art status. On this occasion, we review the AMD GPGPU environment covering the recent launch of CDNA2 GPU architecture, ROCm software stack, and HIP API in a comparative view against Nvidia CUDA. We will present performance benchmarks about the ROCm stack showing what we can expect from it on algebra routines, canonical science applications, and modern frameworks like Pytorch and TensorFlow.
Speaker’s Bio: PhD. in Physics at the University of São Paulo with an emphasis in simulations on Condensed Matter and Statistical Physics. A specialist on HPC for scientific research applications with experience in deploying HPC solutions, software development, and optimization. Expertise in intensive computational applications with GPGPU processing in Engineering and Science.