You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In the current irq balance method, we can only limit the irq balancing range with specified numa node. When there exists various devices, the irq handlers' process time will be very different, and their real-time requirements will also be different. In our products, we need the various devices can bind to various cpu partitions. If re-add the affinity_hint feature, the devices can specified the irq balancing range by specified cpu partition which corresponding to the affinity_hint cpu mask.
I have read the rejected reason in #68. I think that the policyscript approach cannot satisfy my above demands.
The text was updated successfully, but these errors were encountered:
hejingxian123
changed the title
Re-add affinity_hint to support balance irqs within cpu partitions
Re-add affinity_hint to support balancing irqs within cpu partitions
Jan 9, 2021
Sorry, I'm not re adding affinity hint, it caused far too many issues. I also dont see how policyscript cant satisfy your needs. It allows for the overriding of numa node, and can easily be expanded to do just about anything, including specifying a complete balance mask
In the current irq balance method, we can only limit the irq balancing range with specified numa node. When there exists various devices, the irq handlers' process time will be very different, and their real-time requirements will also be different. In our products, we need the various devices can bind to various cpu partitions. If re-add the affinity_hint feature, the devices can specified the irq balancing range by specified cpu partition which corresponding to the affinity_hint cpu mask.
I have read the rejected reason in #68. I think that the policyscript approach cannot satisfy my above demands.
The text was updated successfully, but these errors were encountered: