corebinding

Corebinding, or core binding, is the practice of binding a process or thread to a specific set of CPU cores to control where its execution occurs, i.e., processor affinity. By restricting execution to designated cores, software can improve cache locality, reduce context switching, and sometimes meet real-time deadlines or energy goals. Corebinding can be applied at different granularity, from binding a whole process to a subset of cores to binding individual threads to particular cores.

Operating systems provide facilities for corebinding. In Linux, processor affinity is managed via sched_setaffinity or the taskset utility, and similar controls exist for individual threads. Windows offers SetThreadAffinityMask to bind threads to cores. macOS has more limited native support, with third-party or framework-specific approaches sometimes used. In high-performance computing, corebinding is often combined with NUMA awareness to keep memory accesses local, using libraries or runtimes that expose processor binding options.

Typical use cases include performance tuning of compute-bound workloads, real-time or deterministic systems, and scenarios seeking to reduce cross-core data movement or to reserve cores for OS or I/O tasks. It is also used in cloud and shared environments to improve isolation or to align with licensing or resource partitions.

Trade-offs and considerations include potential improvements in latency and cache efficiency versus risks of reduced overall throughput if the workload becomes poorly balanced across cores. Overly rigid bindings can hinder the operating system’s scheduler, cause underutilization, and complicate portable or scalable deployments. Power management and simultaneous multithreading (hyper-threading) can also influence the effectiveness of corebinding.

Related concepts include CPU affinity, thread pinning, processor binding, and NUMA-aware scheduling. See also: Linux sched_setaffinity, taskset; Windows SetThreadAffinityMask; cpuset; hyper-threading.