Matthew Wilcox wrote:
> On Sun, Jul 07, 2002 at 11:06:32PM -0400, Alexander Viro wrote:
>
>>The thing being, if you are already contended you are playing "I'll release
>>CPU now" vs. "I'll spin in hope that contender will go away right now".
>>
>>IOW, it's a win only if you get contention often and for short intervals.
>>Which is a very good indication that something is rotten with your locking
>>scheme. Like, say it, having lost the control over the amount of locks
>>as the result of brainde^Woverenthusiastic belief that fine-grained ==
>>good. With everything that follows from that...
>
> So let's get some numbers. It really shouldn't be hard to make our
> current semaphores spin a little before they sleep. If we get some
> numbers showing it does help then either we need this change in mainline
> or we need to fix our locking.
Sounds good to me. Do you have any code, or a workload that you know
will trigger it? I have a feeling that Specweb will probably show
this behavior, but I want something simpler.
-- Dave Hansen haveblue@us.ibm.com- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jul 15 2002 - 22:00:13 EST