Re: [PATCH] load_balance: "busiest CPU" -> "busier CPUs"

[Nick Piggin <nickpiggin@yahoo.com.au>]

hawkes@SGI.com wrote:
> Second try, incorporating Suresh Siddha's suggestion to avoid adding
> a field to the runqueue struct.

Sorry for not getting back to you earlier.

Looks like it should work, although bviously exclusive cpusets are
still the preferred way to do exclusive partitioning. Not sure what
the others think about adding complexity and overhead for a
relatively uncommon case, I'm not a huge fan.

I have a suggestion for an improvement - cycle through all CPUs
in the busiest group before moving on to find_busiest_group again?

Do you really need the schedstat field? You'll have to bump the
schedstat version.

The idea I had when thinking about how to handle this, is to have
something like your `consider_cpus` field in the sched domain
itself (which can help you not do too much work at one time, but
will have other complexities like when to reset the mask).

--
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.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/