Output details

autopin – Automated Optimization of Thread-to-Core Pinning on Multicore Systems

<02> We present a framework that is capable of automatically detecting and applying optimal binding between threads of a running parallel application and processor cores (thread pinning) in a shared memory system. This is accomplished through hardware performance counters. We demonstrate that many applications from the SPEC OMP benchmark show quite sensitive runtime behaviour depending on the thread/core pinning used. We were able to find the best pinning in nearly all cases. The proposed framework is intended to supplement job scheduling systems for better automatic exploitation of multi-core systems as well as making programmers aware of this issue by providing measurement logs.