Re: kernbench on 512p

[Martin J. Bligh]

--On Thursday, August 19, 2004 17:24:04 -0400 Jesse Barnes <jbarnes@xxxxxxxxxxxx> wrote:

> On Thursday, August 19, 2004 5:11 pm, Jesse Barnes wrote:
>> The output is attached (my mailer insists on wrapping it if I inline it). 
>> I used 'lockstat -w'.
> 
> The highlights:
> 
>  nw   spin   rjct  lock & function
> 19.0% 81.0%    0%  dcache_lock
>  3.3% 96.7%    0%    d_alloc+0x270
>  2.7% 97.3%    0%    d_delete+0x40
> 18.3% 81.7%    0%    d_instantiate+0x90
>  4.7% 95.3%    0%    d_move+0x60
> 34.6% 65.4%    0%    d_rehash+0xe0
> 19.1% 80.9%    0%    dput+0x40
> 10.5% 89.5%    0%    link_path_walk+0xef0
>    0%  100%    0%    sys_getcwd+0x210
> 
> 41.4% 58.6%    0%  rcu_state
> 61.3% 38.7%    0%    __rcu_process_callbacks+0x260
> 41.4% 58.6%    0%    rcu_check_quiescent_state+0xf0
> 
> So it looks like the dcache lock is the biggest problem on this system with 
> this load.  And although the rcu stuff has improved tremendously for this 
> system, it's still highly contended.

Hmmm. dcache_lock is known-fucked, though I'm suprised at d_rehash
(file deletion)?

RCU stuff is a bit sad.

M.

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