[PATCH] scale down 074 on lower end machines

[PATCH] scale down 074 on lower end machines

[Christoph Hellwig]

074 takes ages to complete on my kvm test VM, but scaling it back to the
level used on IRIX makes it complete in slightly under 10 minutes.

I'm not sure if checking for UP vs SMP is the right way to go into slow
mode, but I couldn't think of anything better.


Signed-off-by: Christoph Hellwig <hch@lst.de>

Index: xfstests/074
===================================================================
RCS file: /cvs/xfs-cmds/xfstests/074,v
retrieving revision 1.7
diff -u -p -r1.7 074
--- xfstests/074	9 Nov 2005 02:49:08 -0000	1.7
+++ xfstests/074	15 May 2008 17:37:03 -0000
@@ -120,10 +120,17 @@ if [ $HOSTOS == "IRIX" ]; then 
         param_type="IRIX nondebug"
     fi
 elif [ $HOSTOS == "Linux" ]; then
-    numloops=10
-    numfiles=5
-    numchildren=3
-    param_type="Linux"
+    if uname -a | grep -q SMP; then
+        numloops=10
+        numfiles=5
+        numchildren=3
+        param_type="Linux SMP"
+    else
+        numloops=2
+        numfiles=3
+        numchildren=3
+        param_type="Linux UP"
+    fi
 else
     numloops=1
     numfiles=1

Re: [PATCH] scale down 074 on lower end machines

[David Chinner]

On Thu, May 15, 2008 at 07:39:13PM +0200, Christoph Hellwig wrote:
> 074 takes ages to complete on my kvm test VM, but scaling it back to the
> level used on IRIX makes it complete in slightly under 10 minutes.
> 
> I'm not sure if checking for UP vs SMP is the right way to go into slow
> mode, but I couldn't think of anything better.

Seems fair - I was wondering exactly the same thing yesterday
as this is another test that takes >15min to run on UML. I didn't
think of the UP vs SMP trigger, but it seems like a good way to
go given all of the real QA boxes we have are SMP.

I'll give it a run and see how it goes.

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: [PATCH] scale down 074 on lower end machines

[David Chinner]

On Fri, May 16, 2008 at 08:15:10AM +1000, David Chinner wrote:
> On Thu, May 15, 2008 at 07:39:13PM +0200, Christoph Hellwig wrote:
> > 074 takes ages to complete on my kvm test VM, but scaling it back to the
> > level used on IRIX makes it complete in slightly under 10 minutes.
> > 
> > I'm not sure if checking for UP vs SMP is the right way to go into slow
> > mode, but I couldn't think of anything better.
> 
> Seems fair - I was wondering exactly the same thing yesterday
> as this is another test that takes >15min to run on UML. I didn't
> think of the UP vs SMP trigger, but it seems like a good way to
> go given all of the real QA boxes we have are SMP.
> 
> I'll give it a run and see how it goes.

FWIW - runtime on UML has dropped from 918s to 88s. Very reasonable
now, and makes the maximum single test runtime in UML about 330s....

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: [PATCH] scale down 074 on lower end machines

[Christoph Hellwig]

ping?

On Thu, May 15, 2008 at 07:39:13PM +0200, Christoph Hellwig wrote:
> 074 takes ages to complete on my kvm test VM, but scaling it back to the
> level used on IRIX makes it complete in slightly under 10 minutes.
> 
> I'm not sure if checking for UP vs SMP is the right way to go into slow
> mode, but I couldn't think of anything better.
> 
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> Index: xfstests/074
> ===================================================================
> RCS file: /cvs/xfs-cmds/xfstests/074,v
> retrieving revision 1.7
> diff -u -p -r1.7 074
> --- xfstests/074	9 Nov 2005 02:49:08 -0000	1.7
> +++ xfstests/074	15 May 2008 17:37:03 -0000
> @@ -120,10 +120,17 @@ if [ $HOSTOS == "IRIX" ]; then 
>          param_type="IRIX nondebug"
>      fi
>  elif [ $HOSTOS == "Linux" ]; then
> -    numloops=10
> -    numfiles=5
> -    numchildren=3
> -    param_type="Linux"
> +    if uname -a | grep -q SMP; then
> +        numloops=10
> +        numfiles=5
> +        numchildren=3
> +        param_type="Linux SMP"
> +    else
> +        numloops=2
> +        numfiles=3
> +        numchildren=3
> +        param_type="Linux UP"
> +    fi
>  else
>      numloops=1
>      numfiles=1
---end quoted text---

Re: [PATCH] scale down 074 on lower end machines

[Christoph Hellwig]

ping^2

On Wed, May 21, 2008 at 10:15:52AM +0200, Christoph Hellwig wrote:
> ping?
> 
> On Thu, May 15, 2008 at 07:39:13PM +0200, Christoph Hellwig wrote:
> > 074 takes ages to complete on my kvm test VM, but scaling it back to the
> > level used on IRIX makes it complete in slightly under 10 minutes.
> > 
> > I'm not sure if checking for UP vs SMP is the right way to go into slow
> > mode, but I couldn't think of anything better.
> > 
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > 
> > Index: xfstests/074
> > ===================================================================
> > RCS file: /cvs/xfs-cmds/xfstests/074,v
> > retrieving revision 1.7
> > diff -u -p -r1.7 074
> > --- xfstests/074	9 Nov 2005 02:49:08 -0000	1.7
> > +++ xfstests/074	15 May 2008 17:37:03 -0000
> > @@ -120,10 +120,17 @@ if [ $HOSTOS == "IRIX" ]; then 
> >          param_type="IRIX nondebug"
> >      fi
> >  elif [ $HOSTOS == "Linux" ]; then
> > -    numloops=10
> > -    numfiles=5
> > -    numchildren=3
> > -    param_type="Linux"
> > +    if uname -a | grep -q SMP; then
> > +        numloops=10
> > +        numfiles=5
> > +        numchildren=3
> > +        param_type="Linux SMP"
> > +    else
> > +        numloops=2
> > +        numfiles=3
> > +        numchildren=3
> > +        param_type="Linux UP"
> > +    fi
> >  else
> >      numloops=1
> >      numfiles=1
> ---end quoted text---
---end quoted text---

Re: [PATCH] scale down 074 on lower end machines

[Niv Sardi]

Christoph Hellwig <hch@lst.de> writes:

> 074 takes ages to complete on my kvm test VM, but scaling it back to the
> level used on IRIX makes it complete in slightly under 10 minutes.
>
> I'm not sure if checking for UP vs SMP is the right way to go into slow
> mode, but I couldn't think of anything better.

Looks good enough for now.
-- 
Niv Sardi