December 2025: NUMA-Aware Scheduling Enhancement delivered for sched-ext/scx. Implemented NUMA support for older kernels by adding a CPU->node map in BPF and initializing it from user-space NUMA topology data collected via the topology crate. Enabled NUMA by default and optimized perf-counter based scheduling to select the least-busy CPU from the same NUMA node as the previous CPU, enhancing locality and throughput. This work broadens deployment options, improves performance for NUMA workloads, and reduces cross-node migrations. All changes tracked in commit 738e356afc3eaab7cfde78e1c034d5370615276d (Andrea Righi, Balbir Singh).

November 2025 (sched-ext/scx): Delivered Adaptive CPU-aware Task Scheduling with per-CPU performance counter tracking and policy-based dispatch. Implemented logic to keep tasks on the same CPU or distribute them based on perf events, and added a mechanism to select the least busy (idle) CPU for task dispatching to reduce latency and improve resource utilization. Enabled perf-counter collection in perfmon mode (-e) and used per-CPU usage data to drive dynamic scheduling decisions, including safeguards to never migrate migration-disabled tasks. This work enhances scheduling efficiency, reduces contention, and improves CPU utilization across varied workloads.