Lock granularity...

[manfreds en colorfullife.com]

sgi has written a special patch that measures the lock
contention:
oss.sgi.com/projects/lockmeter

Another problem with the lock granularity is the use of
cli() instead of a private spinlock in many drivers,
although most of them are not performance critical.
[e.g. the complete ISDN subsystem].

Please note that fine grained locks are sometimes slower
that one big lock under realistic loads, e.g. I removed
the per-semaphore array spinlocks from the SysV code
because it was 5% slower than the simple lock.

--
	Manfred

-----------------------------------------------------
This mail sent through IMP: http://web.horde.org/imp/

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo en vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/