Join BSC, Intel and INESC-ID to deep dive into memory system profiling and performance analysis.
The course is organized by the BSC Memory Systems Team, founded in 2014. Since its creation, the team has been actively exploring memory systems for high-performance computing (HPC) and AI. It has led multiple industrial projects with Huawei Technologies (China), Micron Technology (US and Italy), and Samsung Electronics (Korea).
The lecture on Roofline models will be delivered by experts from the Heterogeneous Computing and Performance Modeling Hub created by INESC-ID, Lisbon. The lecturers have developed multiple Roofline models, several of which are now used in production tools for Intel HPC servers.
The TopDown analysis will be presented by an expert from Intel with tens of years of experience in HPC application profiling.
The lecture on heterogeneous memory systems will be presented by the BSC team for Accelerators and Communications for HPC. The lecturers bring extensive experience with heterogeneous memory systems combining DDR4 and Intel Optane DIMMs, gained through direct collaboration with Intel.
The course will present the work on A Mess of Memory System Benchmarking, Simulation and Application Profiling developed in the context of the Barcelona Zettascale project. We will also show how this work complements two widely-used performance analysis methods: the Roofline model and the TopDown analysis.