Re: AIM7 40% regression with 2.6.26-rc1

[Ingo Molnar]

* Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:

> In contrast, these "try to emulate bad behavior with the old known-ok 
> semaphores" don't show anything AT ALL. We already know it's related 
> to semaphores. And your patches aren't even guaranteed to show the 
> same issue.

yeah, i was just trying to come up with patches to probe which one of 
the following two possibilities is actually the case:

 - if the regression is due to the difference in scheduling behavior of 
   new semaphores (different wakeup patterns, etc.), that's fixable in 
   the new semaphore code => then the BKL code need not change.

 - if the regression is due due to difference in the fastpath cost, then 
   the new semaphores can probably not be improved (much of their appeal 
   comes from them not being complex and not being in assembly) => then 
   the BKL code needs to change to become cheaper [i.e. then we want 
   your patch].

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