netdev
[Top] [All Lists]

Re: [RFC] Use RCU for tcp_ehash lookup

To: Srivatsa Vaddagiri <vatsa@xxxxxxxxxx>
Subject: Re: [RFC] Use RCU for tcp_ehash lookup
From: Andi Kleen <ak@xxxxxxx>
Date: Thu, 2 Sep 2004 23:19:50 +0200
Cc: Andi Kleen <ak@xxxxxxx>, davem@xxxxxxxxxx, netdev@xxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, Dipankar <dipankar@xxxxxxxxxx>, paulmck@xxxxxxxxxx
In-reply-to: <20040901113641.GA3918@xxxxxxxxxx>
References: <20040831125941.GA5534@xxxxxxxxxx> <20040831135419.GA17642@xxxxxxxxxxxxx> <20040901113641.GA3918@xxxxxxxxxx>
Sender: netdev-bounce@xxxxxxxxxxx
On Wed, Sep 01, 2004 at 05:06:41PM +0530, Srivatsa Vaddagiri wrote:
>                            |  2.6.8.1               | 2.6.8.1 + my patch
> -------------------------------------------------------------------------------
> Average cycles                     |                        |
> spent in                   |                        |
> __tcp_v4_lookup_established  |        2970.65               | 668.227
>                            | (~3.3 micro-seconds)   | (~0.74 microseconds)
> -------------------------------------------------------------------------------
> 
> This repesents improvement by a factor of 77.5%!

Nice.

> 
> 
> > 
> > And it should also fix the performance problems with
> > cat /proc/net/tcp on ppc64/ia64 for large hash tables because the rw locks 
> > are gone.
> 
> But spinlocks are in! Would that still improve the performance compared to rw 
> locks?  (See me earlier note where I have explained that lookup done for 
> /proc/net/tcp is _not_ lock-free yet).

Yes, spinlocks are much faster than rwlocks.

-Andi


<Prev in Thread] Current Thread [Next in Thread>