[prev in list] [next in list] [prev in thread] [next in thread] 

List:       fuse-devel
Subject:    Re: [fuse-devel] stat performance on multicore systems: measurements
From:       Stefan Westerfeld <stefan () space ! twc ! de>
Date:       2012-08-01 16:41:51
Message-ID: 20120801164151.GA26837 () space ! twc ! de
[Download RAW message or body]

   Hi!

On Tue, Jul 31, 2012 at 05:38:51PM +0200, Miklos Szeredi wrote:
> Stefan Westerfeld <stefan@space.twc.de> writes:
> > On Mon, Jul 30, 2012 at 05:21:49PM +0200, Miklos Szeredi wrote:
> >> Stefan Westerfeld <stefan@space.twc.de> writes:
> >> 
> >> > The way I think it should work is that the FUSE fs will have different worker
> >> > threads which are bound to one cpu per thread. If two processes running on cpu
> >> > 1 and 3 access the filesystem, FUSE should use a worker thread bound to cpu 1
> >> > to process requests from the userspace process running on cpu 1 and another
> >> > worker thread bound to cpu 3 to process requests from the userspace process
> >> > running on cpu 3.
> >> 
> >> Does the following patch help?
> >> 
> >> The wake_up_sync() call should prefer scheduling the woken up task on
> >> the same CPU as the waker.
> >
> > No, the performance stays the same, and the filesystem thread(s) and the
> > userspace client still get executed on different cpus most of the
> > time.
> Hmm, looks like this is because the wake_up() selects the first task on
> the list and *then* checks if the CPU is the same.
> 
> What we need instead is something that tries to find a fuse task that is
> already on the same CPU and wake that up.
> 
> Could you please try this patch?  It's a gigantic hack, but should prove
> or disprove the theory.

Your patch doesn't compile here:

Linux ubuntu 3.2.21 #2 SMP Fri Jul 27 12:56:14 CEST 2012 x86_64 x86_64 x86_64 GNU/Linux

root@ubuntu:/usr/src/linux-source-3.2.0/linux-source-3.2.0# LANG=C make
[...]
  CC [M]  fs/fuse/dev.o
fs/fuse/dev.c: In function 'queue_request':
fs/fuse/dev.c:271:3: error: too many arguments to function '__wake_up_locked'
include/linux/wait.h:160:6: note: declared here

> The multithreaded loop should also be modified to start each thread with
> an affinity to a different CPU.

Yes, that sounds good to me.

   Cu... Stefan
-- 
Stefan Westerfeld, Hamburg/Germany, http://space.twc.de/~stefan

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
fuse-devel mailing list
fuse-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/fuse-devel
[prev in list] [next in list] [prev in thread] [next in thread]