[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20091012110044.GC9228@kernel.dk>
Date: Mon, 12 Oct 2009 13:00:45 +0200
From: Jens Axboe <jens.axboe@...cle.com>
To: Nick Piggin <npiggin@...e.de>
Cc: Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
linux-fsdevel@...r.kernel.org,
Ravikiran G Thirumalai <kiran@...lex86.org>,
Peter Zijlstra <peterz@...radead.org>,
Linus Torvalds <torvalds@...ux-foundation.org>,
samba-technical@...ts.samba.org
Subject: Re: [rfc][patch] store-free path walking
On Mon, Oct 12 2009, Jens Axboe wrote:
> > Anyway, if you do get a chance to run dbench with this
> > modification, I would appreciate seeing a profile with clal
> > traces (my bigger system is ia64 which doesn't do perf yet).
>
> For what number of clients?
Attached are call graphs for 8 and 32 clients, with your latest patch
applied. At 8 clients, we're already at 80% spinlock time. While running
the 32 client record, I wanted to open a file in the tmpfs mount where
dbench was running. bash was spinning until dbench completed.
At 32 clients, throughput is 75% of the 8 client case. And spinlock time
there is at 90%.
I have attached both outputs.
--
Jens Axboe
Download attachment "8client.txt.bz2" of type "application/octet-stream" (53343 bytes)
Download attachment "32client.txt.bz2" of type "application/octet-stream" (65409 bytes)
Powered by blists - more mailing lists