meaning to create a process-to-cpucore mapping based on the measured behavior of a task. This would improve cache performance by grouping processes together which will step on each other's cache content (or maybe other criteria) as little as possible. Keeping a set of rules for frequently running tasks around may benefit short running processes or sorting these correctly in right from the start. Examples would be that running "cc1" does not like another cc1 on the same core but tolerates other processes which use huge amounts of cpu performance but keep a very small memory traffic footprint.
I thought about doing something like this, but I really lack the time
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
