Re: [patch] O(1) scheduler, -G1, 2.5.2-pre10, 2.4.17 (fwd)

[Davide Libenzi]

On Fri, 11 Jan 2002, Ingo Molnar wrote:

> indeed. The question is, should we migrate processes around just to get
> 100% fairness in 'top' output? The (implicit) cost of a task migration
> (caused by the destruction & rebuilding of cache state) can be 10
> milliseconds easily on a system with big caches.

10 ms is exactly what i've observed while i was coding the BMQS balance
code. Leaving a cpu idle for more than 10ms will make real tests like
kernel builds to suffer performance degradation. By using 10ms i always
got the same time of the standard scheduler.

- Davide