Technical Papers Archive

The Design and Organization of the El Capitan Center of Excellence for Application Readiness


Authors: Constantinos Makrides (HPE); J. Austin Ellis (Advanced Micro Devices (AMD)); Stephen Abbott and Luke Roskop (HPE); Michael E. Rowan (Advanced Micro Devices (AMD)); Ramesh Pankajakshan (Lawrence Livermore National Laboratory (LLNL)); James Elliott (Sandia National Laboratories); Anna Pietarila Graham (Los Alamos National Laboratory (LANL)); and David F. Richards, Judith Hill, and Bronis R. de Supinski (Lawrence Livermore National Laboratory (LLNL))

Abstract: El Capitan is the first U.S. exascale system for national security. Its Center of Excellence (COE) enabled co-design and application readiness. The COE's central goal is to prepare applications to use the AMD Instinct MI300A APU and Cray Slingshot interconnect effectively on the first day of delivery, and continuously improve after delivery. Application engagements are driven by a team of eight engineers from HPE and AMD who support approximately 20 applications from three national laboratories. Quarterly hackathons had attendances reaching upwards of 50 participants.

We describe the technical scope, engagement model, and organizational structure that supported COE activities, emphasizing the key components of a high performance computing COE. We highlight five select case studies that demonstrate technical challenges in addressing functionality and performance across many software and hardware updates, often utilizing experimental software to overcome issues. We summarize lessons learned and AI impacts that may benefit future COEs.



Presentation: file

Back to Technical Papers Archive Listing