Re: [AMBER] xgrid in opempi 1.4.1 still broken

From: Alan <alanwilter.gmail.com>
Date: Sun, 2 May 2010 18:28:04 +0100

Hi there,

I don't know how are yours Macs set up.

What I am trying to do is compiling Amber11 with Intel Compiler and MPI.

Xgrid is part of Mac OSX, but maybe only me here who have the controllers on
(sometimes I try to play something with Xgrid).

Anyway, for me, compiling OMPI for icc/ifort and trying ompi/examples
required that I have to get rid of $AMBERHOME/lib/openmpi/mca_plm_xgrid.*
either going to $AMBERHOME/AmberTools/src/openmpi-1.4.1/orte/mca/plm/xgrid
and doing 'make uninstall' or setting '--with-xgrid=no' or '--without-xgrid'
in configure_openmpi script.

If I don't do this, in ompi/examples:

$AMBERHOME/AmberTools/src/openmpi-1.4.1/examples% mpirun -c 2 hello_c
[amadeus.local:46005] [[11764,0],0] ORTE_ERROR_LOG: Unknown error: 1 in file
src/plm_xgrid_module.m at line 119
[amadeus.local:46005] [[11764,0],0] ORTE_ERROR_LOG: Unknown error: 1 in file
src/plm_xgrid_module.m at line 153
--------------------------------------------------------------------------
mpirun was unable to start the specified application as it encountered an
error.
More information may be available above.
--------------------------------------------------------------------------
2010-05-02 09:42:38.711 mpirun[46005:903] *** Terminating app due to
uncaught exception 'NSInvalidArgumentException', reason: '***
-[NSKVONotifying_XGConnection<0x100226a20> finalize]: called when collecting
not enabled'
*** Call stack at first throw:
(
0 CoreFoundation 0x00007fff87aeed24
__exceptionPreprocess + 180
1 libobjc.A.dylib 0x00007fff807790f3
objc_exception_throw + 45
 2 CoreFoundation 0x00007fff87b09a51
-[NSObject(NSObject) finalize] + 129
3 mca_plm_xgrid.so 0x00000001001657d0 -[PlmXGridClient
dealloc] + 64
 4 mca_plm_xgrid.so 0x0000000100164190
orte_plm_xgrid_finalize + 64
5 libopen-rte.0.dylib 0x0000000100047d5f
orte_plm_base_close + 31
 6 libopen-pal.0.dylib 0x0000000100095e58
event_process_active + 328
)
terminate called after throwing an instance of 'NSException'
[amadeus:46005] *** Process received signal ***
[amadeus:46005] Signal: Abort trap (6)
[amadeus:46005] Signal code: (0)
[amadeus:46005] *** End of error message ***
[1] 46005 abort mpirun -c 2 hello_c

Without xgrid lib:
 mpirun -c 2 hello_c
Hello, world, I am 0 of 2
Hello, world, I am 1 of 2

All fine.

Now compiling AmberTools and Amber11 with Intel and Amber's OMPI and
sander.MPI and pmemd.MPI are not working at all but this is for another
email I hope to send soon (I should be on holidays btw :-).

Cheers,

Alan

On Sun, May 2, 2010 at 03:28, David Watson <dewatson.olemiss.edu> wrote:

> On May 1, 2010, at 8:49 PM, Mengjuei Hsieh wrote:
>
> > On Fri, Apr 30, 2010 at 8:04 AM, Alan <alanwilter.gmail.com> wrote:
> >> Hi there,
> >> It's just to let those using Mac and using the script configure_openmpi
> that
> >> you need to amend:
> >>
> > [suggestion snipped]
> >>
> >> Indeed, this modification affects only Macs (and in a good way, all
> other OS
> >> ignores xgrid), so it would be safe to patch that to configure_openmpi
> until
> >> at least the day that openmpi fixes this issue with xgrid.
> >> Alan
> >
> > I know xgrid support has been broken (
> > https://svn.open-mpi.org/trac/ompi/ticket/1777 ) for some time. But
> > since I didn't have a chance to play xgrid so I haven't figured out
> > how can I get an error or something if I used the default building
> > argument. Do I have to install some kind Xgrid controller or activate
> > the Xgrid services in order to have it failed? :-)
> >
>
> No, you just have to have openmpi compiled with the "wrong" arguments and
> it will fail.
>
> I think macports take care of all of this, but if your openmpi is bailing
> due to xgrid, I remember there being some environmental variable you could
> set to make it NOT try to use xgrid.
> I am fairly certain no one wants Amber on xgrid unless they have a custom
> fiber channel setup.
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
Alan Wilter S. da Silva, D.Sc. - CCPN Research Associate
Department of Biochemistry, University of Cambridge.
80 Tennis Court Road, Cambridge CB2 1GA, UK.
>>http://www.bio.cam.ac.uk/~awd28<<
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun May 02 2010 - 10:30:03 PDT
Custom Search