On Wed, 21 Jun 2000, Mike Galbraith wrote:
>was consistently beating ac22.riel by a full minute (8.5 vs 9.5). Adding
>this patch to the mix took another solid 30 seconds off build time.
Ok. I'm actually fixing an elevator bug (spotted by Ingo Molnar), so I
only had time to do a fast test and also here applying the patch
apparently ported the classzone kernel to its more normal behaviour.
I also dropped the same stuff from 2.2.17pre4 and I'm waiting feedback
about the machines that was stalling, I'll have more info soon (stay
tuned).
The fact is that there's no way, to fix the race in the allocator without
hurting performance we have to define a per-task freelist (as I originally
described on l-k as alternate and more correct fix but maybe too complex
and not necessary, but it's going to be really necessary instead). I'll
attempt to do that when the elevator issue is solved (maybe in 2.2.x we
can live with the bug... after all unless there's heavy swapping going on
it's unlikely to hit the race considering the most of the stuff is
protected by the kernel lock and the atomic allocation can eat all the
free-poll anyway ;).
Andrea
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Jun 23 2000 - 21:00:22 EST