Re: [uml-devel] [PATCH] host context switch reduction

[Laurent Vivier <LaurentVivier@wanadoo.fr>]

[-- Attachment #1: Type: text/plain, Size: 1411 bytes --]

Hi,

find hereunder the measurements I sent to Jeff.

If someone think this patch is really useful for 2.6 I can make the
port.

Regards,
Laurent
Please CC: me, as I'm not in the mailing list.

Le lun 17/05/2004 à 22:08, roland a écrit :
> hi !
> has this / will this be ported to 2.6 ?
> have there been made further performance comparisons ?
> regards
> roland
> 
> 
> >Well, we just looked at the context switches. We started 3 UMLs which
> >were doing a make world for xfree. Without patches the number of
> >context switches per second was around 60k, with the patch it was around
> >40k - so it's noticeable, at least from the cs.
> >Sven
> 
> ----- Original Message ----- 
> From: "Jeff Dike" <jdike@addtoit.com>
> To: <user-mode-linux-devel@lists.sourceforge.net>
> Sent: Sunday, February 29, 2004 6:35 AM
> Subject: [uml-devel] [PATCH] host context switch reduction
> 
> 
> > The patches below are from Laurent Vivier who didn't make them public.  They
> > add a new feature to ptrace on the host which cuts down on the number of
> > context switches needed for a UML system call, plus makes UML use it.
> > 
> > There's some interest in this, so I'm putting it out as-is.  I haven't played
> > with it.
> > 
> > Laurent did some performance testing and found ~40% speedup on a getpid
> > loop, and a 3.5% speedup on a kernel build.
> > 
> > Jeff
> > 
> > 

[-- Attachment #2: measure-getpid.txt --]
[-- Type: text/plain, Size: 1050 bytes --]

#include <stdio.h>

int main(int argc, char** argv)
{
        int i;
	int loop = atol(argv[1]);

        for (i = 0; i < loop; i++)
        {
                getpid();
        }
}

NATIF:

time ./getpid 10000000

real    0m7.920s
user    0m3.990s
sys     0m3.940s

UML with sysemu

real    3m55.052s
user    0m5.962s
sys     0m46.385s

real    3m56.964s
user    0m5.846s
sys     0m42.635s

real    3m54.179s
user    0m5.981s
sys     0m47.154s

UML w/o sysemu

real    6m16.956s
user    0m5.250s
sys     0m58.462s


real    6m17.126s
user    0m4.846s
sys     1m1.731s


real    6m16.461s
user    0m5.077s
sys     0m59.365s

---------------------------------------

original UML:

time make dep > /dev/null

real    4m51.669s
user    0m37.462s
sys     0m51.846s

time make bzImage modules > /dev/null

real    13m35.457s
user    4m18.885s
sys     1m40.481s

UML with sysemu:

time make dep > /dev/null

real    3m52.486s
user    0m37.981s
sys     0m50.808s

time make bzImage modules > /dev/null

real    13m5.980s
user    4m18.096s
sys     1m40.115s