Re: Current ULE state and plans (was Re: HMP scheduling on FreeBSD)

> Speaking as the guy who periodically bitched about what ULE was doing
> to misc/dnetc until someone pointed out idprio to me (a setting that
> 4BSD couldn't care less about but that ULE goes nuts over if you set it
> wrong), count me as a continuing user of 4BSD.

Rest assured you were counted in the 10/20. :-)

> Because it just works.

For now.

> Anecdotally, I don't personally observe the purported interactivity
> improvement ascribed to ULE, either, but I can't give you numbers on it.

ULE's benefits are much wider than just interactivity improvements.  Interactive behavior now probably needs some reassessment.

> And why shouldn't people try to create other new schedulers?
> And should we not at least keep testing that the scheduler interface is simple and
> promotes scheduler modularity?

People are free to try to create whatever scheduler they want.  What could be accepted in-tree is of course much more restricted and depends on a number of factors.  Nobody said that was an impossible nor a useless task though, but it's definitely not for the faint of heart.

The scheduler interface is not that big really.  It could be streamlined more, for the sake of better clarity of the current implementations, and that's also in the plans.  That's however not a real prerequisite to working on another scheduler.

-- 
Olivier Certner