On Sun, 2007-09-23 at 13:58 +0200, Ingo Molnar wrote:

quoted text
> i'm also wondering - do we even need sync_vruntime() at all? Now that we 
> fix up vruntime while we migrate tasks across rqs, the global notion of 
> vruntime makes less and less of a sense.

Yeah, I was wondering that too, but it does seem to be needed.  If I
make it a noop, and run the pinned tasks with an unpinned make -j2 test,
Xorg at nice -5 still takes hits...

se.wait_max              :            11.513462
se.wait_max              :             6.250911
se.wait_max              :            15.181511
se.wait_max              :           188.963625
se.wait_max              :             4.951569
se.wait_max              :            11.391749

...which begs the question "gee, what happens if I pin a nice 0 hog to
CPU0, and a nice 19 hog to CPU1 with the queues never ever syncing?".

The answer to that one seems to be "Don't even think about it!".  This
needs more investigation... there's gotta be a 32 bit thingie hiding in
there.  I lost control for a few seconds, got it back, then lost it
again forever (sysrq-b worked though, hmm).

	-Mike

-
unsubscribe notice
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/