On Sun, Nov 17, 2002 at 12:11:35AM +0000, Alan Cox wrote:
> Bill - so what happens if you trim down the aio, event and ksoftirqd
> threads to a sane size (you might also want to do something about the
> fact 2.5 still runs ksoftirq too easily). Intuitively I'd go for a
> square root of the number of processors + 1 sort of function but what do
> the benchmarks say ?
Both reorganizing the per-cpu thread pools as state machines and
inserting new locking look like work-intensive projects...
It's not become explosively bad yet (1MB of overhead is eyebrow-raising
but not particularly damaging) so there's no rush to trim this down,
but I'm at least thinking about doing this later. One of the major
obstacles for the state machine approach is that the migration threads
run at RT priority while the rest do not, and of course the greater
than per-cpu granularity approach suffers from additional locking.
Bill
-
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 : Sat Nov 23 2002 - 22:00:18 EST