Lines Matching refs:clusters

10 … scheduler** extends on the clutch scheduler design to support multiple clusters of different perf…
70 …rms, the `sched_clutch_bucket_group` represents threads enqueued on all clusters on the platform. …
175 * **Rapidly open up clusters** if the workload can use them (e.g., for parallelism, to mitigate pas…
179 …s should migrate to clusters where execution efficiency won’t be significantly worse. This might b…
183 …ncepts largely apply to a platform with several homogenous clusters and heterogenous clusters with…
187 …the ability to influence the width of the system (in terms of number of clusters), while retaining…
191 …p i.e. per sched\_clutch\_bucket\_group. The ability to prefer specific clusters rather than clust…
195 …endations for latency sensitive workloads from efficient to performance clusters and ensure that t…
217 … delta that should exist between the source and destination nodes (i.e. clusters) for the thread t…
245 …clusters before asymmetric clusters to ensure the threads migrate to idle clusters with similar pe…
249 … the steal operation is to find other threads running/runnable in other clusters which should be e…
253 …oot. This foreign priority queue allows other clusters to easily rebalance threads from asymmetric…
262 * (3) Check if foreign threads are running on the non-native clusters (sched_edge_foreign_running_t…
266 …ch thread is found, then the scheduler aims to reduce the load on other clusters by stealing threa…
274 …clusters in order to reduce contention on the shared resources. It achieves that by maintaining th…
281 …clusters irrespective of type. The idea is that this scheduling policy will put a shared resource …
283 …uler since this always tries to fill up the native clusters to capacity even when non-native clust…
287 …clusters of various types to ensure that all threads of the workload make equal progress aka "stir…