Optimizing HPC Performance: Locality and Affinity in Practice

Who Should Attend

• HPC users running applications on multi-core or multi-node systems.
• Researchers aiming to improve scalability and memory efficiency.

What You’ll Learn

• Manage thread and process affinity for cache and memory efficiency.
• Address NUMA effects with locality-aware data placement.
• Use OpenMP optimizations for affinity and scheduling.
• Configure OpenMPI process mapping and binding across nodes.

Level

• Intermediate (basic OpenMP/MPI helpful; session includes a brief intro).