Re: [PATCH] semaphore fairness patch against test11-pre6

[Christoph Rohland <cr@sap.com>]

Hi David,

David Mansfield <lkml@dm.ultramaster.com> writes:
> If you can find the time to check this out more completely, I recommend
> it, because it seems like a great improvement to be able to accurately
> see vmstat numbers in times of system load.  I hope the other side
> effects are beneficial as well :-)

I wanted to point out that there may be some performance impacts by
this: We had exactly the new behaviour on SYSV semaphores. It led to
very bad behaviour in high load situations since for high frequency,
short critical paths this led to very high context switch rates
instead of using the available time slice for the program.

We changed the behaviour of SYSV semaphores to the current kernel sem
behaviour and never had problems with that change.

I still think that your change is right since this is kernel space and
you do not have the notion of a time slice.

        Christoph


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/