I talk about about schedule algorithm in OS level in multi-core architecture(a single thread multi cores).
I know some parameters which could reflect load situation of core,such as temperature of core,the weight of all the tasks each core,anyone others?
when we find OS is in load imbalance state in multi-core architecture,we could migrate threads according to temperature of core to other idle core or core of lower temperature.
the weight of runqueue in each core also reflect load situation when load imbalance. we could migrate threads from the core of highest weight .
maybe we also make some measurements to reduce "pingpang" to improve performance.how to reduce it ?
have you other ways to improve performance when OS happen load imbalance ?
thank you for reading and we talk about the topic together.
looking forward your reply.