On Thu, 1 Jun 2000 kumon@flab.fujitsu.co.jp wrote:
> test1-D stext_lock breakdown(unit is us/transaction)
> 35.80 TOTAL Lock Var
> 7.52 schedule+1684 kernel_flag
> 4.64 __fput+49 kernel_flag
> 3.52 tcp_v4_rcv+610 0x2c(%ebx)
> 2.86 dentry_open+264 kernel_flag
> 2.63 dput+19 dcache_lock
> 2.57 sys_fcntl+142 kernel_flag
> 2.41 d_lookup+91 dcache_lock
> 1.40 old_mmap+309 kernel_flag
> 1.12 sock_close+137 kernel_flag
> 1.11 wait_for_connect+704 0x2c(%ebx)
> 1.07 tcp_accept+38 0x2c(%esi)
> 0.67 schedule+195 runqueue_lock
> 0.58 __wake_up+302 runqueue_lock
Thanks a lot for testing ;-) I've got something that _may_ make life
better (SMP-safe mmap() + {get,put}_write_access() + do_pipe() +
->follow_link() and ->readlink()). I'll put it in the same place when it
will go through the tests I'm running.
... OK, it got through the parallel build of itself, XFree86, glibc and
gcc + some network activity + X, so it looks like the thing is not
completely unstable. Admittedly, that's one hell of a skewed test, but
contention data is interesting - #1 is runqueue_lock in do_fork(), #2 -
bdev_lock in bdget() (my fault, is there anybody out there with a decent
ready-to-use hash function for device numbers?), #3 - spinlock in
do_IRQ(), #4 - big lock in sys_msync(), #5 and 6 - uses of console_lock in
do_syslog(). Hmm...
Anyway, the next variant is on ftp.math.psu.edu/pub/viro/dcache-ac7-G.gz
and folks are welcome to try it - it may be interesting.
Cheers,
Al
-
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 : Wed Jun 07 2000 - 21:00:12 EST