[molpro-user] issue with Molpro 2010.1 (patch 2): ABDATA missing

Manhui Wang wangm9 at cardiff.ac.uk
Wed Sep 22 18:14:58 BST 2010


Dear Pedro,

Yes, the mpd is quite annoying. The default process management system in
mpich2-1.2.1p1 is not hydra, but mpd. If you use the latest one
MPICH2-1.3rc1(I can't guarantee it is stable enough), you will not
suffer such problem. Fortunately, MPICH2, IntelMPI etc will move to
Hydra as default startup instead of mpd in the future.

Best wishes,
Manhui

Pedro Jorge Caridade wrote:
> Dear Manhui
> 
> Thanks for your reply. I'm really becoming exhausted with the molpro
> compilation. We had a big problem with disks and I had to rebuild
> everything from scratch.
> The mpich2 is mpich2-1.2.1p1.tar.gz, but since the mpd leaves me
> zombies processes when a sister or the mother superior dies, I have
> change to mpiexec as submission process. In this case (I think as the
> same problem as mpi from intel with hydra) the ABDATA is not created
> since root cannot submit the processes, and the molpro script is not
> appropriate to this. Any how, I've manage in 2009 to adapt the script
> to run but never had this strange problem with ABDATA. I even try to
> rebuild the library with mpd deamon on. It compiles but when
> submitting I always get this strange error.
> 
> Best wishes
> Pedro
> 
> 
> On Wed, Sep 22, 2010 at 5:42 PM, Manhui Wang <wangm9 at cardiff.ac.uk> wrote:
>> Dear Pedro,
>>
>> Which version of MPICH2 are you using? If the process management system
>> for starting parallel jobs is mpd, then you have to start mpd before
>> rebuilding the library data:
>>
>> make -C lib veryclean
>> make
>>
>> Best wishes,
>> Manhui
>> Pedro Jorge Caridade wrote:
>>> Hello
>>>
>>> I'm having the same problem since we are using mpiexec rather the
>>> global MPD ring daemon. I compile molpro with TCGMSG-MPI, MPICH2 and
>>> ACML library. I manage to compile (hopefully!) everything but now got
>>> some error when submitting the jobs:
>>>
>>>  ?PROGRAMMING ERROR IN DECOMPRESSION    131060     65536    133123    131059
>>>  Please contact Molpro support
>>>
>>> Any suggestion?
>>> Pedro
>>>
>>>
>>>
>>>
>>> On Wed, Sep 22, 2010 at 11:00 AM, Manhui Wang <wangm9 at cardiff.ac.uk> wrote:
>>>> Hi,
>>>>
>>>> Reuti wrote:
>>>>> Am 22.09.2010 um 10:21 schrieb Manhui Wang:
>>>>>
>>>>>> Dear Stefan,
>>>>>>
>>>>>> The possible reason is that your Intel MPI mpd daemon
>>>>> As MPICH2 will move to Hydra (i.e. w/o any permanent daemons) as default startup (incl. tight integration into certain queuing systems) and remove former startup methods in the future, Intel MPI will most likely do the same. It was recently on the MPICH2 list, that any compilation for mpd startup can also use the Hydra startup right now (the one from the latest MPICH2 release). As I don't have Intel MPI, I can't test though.
>>>>>
>>>> I have looked at the Intel MPI 4.0. mpiexec.hydra now has been included
>>>> in the library. First test shows it works fine. However, it is still
>>>> experimental, and they will improve it in next release. See the discussion
>>>>
>>>> http://software.intel.com/en-us/forums/showthread.php?t=76537
>>>>
>>>> When hydra of Intel MPI is mature in the future, we will switch to it
>>>> since the current mpd system is very inconvenient.
>>>>
>>>>
>>>> Manhui
>>>>
>>>>> -- Reuti
>>>>>
>>>>>
>>>>>> happened not to
>>>>>> run in the background when you built the library data.
>>>>>> You can fix it by rebuilding the library data (not compiling the whole
>>>>>> code):
>>>>>>
>>>>>> make -C lib veryclean
>>>>>> make
>>>>>>
>>>>>> But ensure your mpd is running in the background before doing this.
>>>>>>
>>>>>>
>>>>>> Best wishes,
>>>>>> Manhui
>>>>>>
>>>>>>
>>>>>> Stefan Becuwe wrote:
>>>>>>> Hello,
>>>>>>>
>>>>>>> I compiled/installed Molpro 2010.1 in the same way as 2009.1: GA 4.3.2,
>>>>>>> TCGMSG-MPI, Intel compilers and Intel MPI.  Running the tests as follows
>>>>>>>
>>>>>>>   make MOLPRO_OPTIONS=-n8 test
>>>>>>>
>>>>>>> results in
>>>>>>>
>>>>>>> openc: Error in opening file [...]/molpro2010.1/testjobs/../lib/ABDATA
>>>>>>>
>>>>>>>  GLOBAL ERROR fehler on processor   0
>>>>>>> make[1]: *** Deleting file `Cs_DKH10.out'
>>>>>>> make: *** [test] Interrupt
>>>>>>>
>>>>>>> It seems ABDATA is missing (is present in 2009.1).  Any idea how to fix
>>>>>>> this problem?  Thanks in advance!
>>>>>>>
>>>>>>> Regards,
>>>>>>>
>>>>>>>
>>>>>>> Stefan
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Molpro-user mailing list
>>>>>>> Molpro-user at molpro.net
>>>>>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>>>>> --
>>>>>> -----------
>>>>>> Manhui  Wang
>>>>>> School of Chemistry, Cardiff University,
>>>>>> Main Building, Park Place,
>>>>>> Cardiff CF10 3AT, UK
>>>>>> Telephone: +44 (0)29208 76637
>>>>>> _______________________________________________
>>>>>> Molpro-user mailing list
>>>>>> Molpro-user at molpro.net
>>>>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>>> --
>>>> -----------
>>>> Manhui  Wang
>>>> School of Chemistry, Cardiff University,
>>>> Main Building, Park Place,
>>>> Cardiff CF10 3AT, UK
>>>> Telephone: +44 (0)29208 76637
>>>> _______________________________________________
>>>> Molpro-user mailing list
>>>> Molpro-user at molpro.net
>>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>>>
>>> _______________________________________________
>>> Molpro-user mailing list
>>> Molpro-user at molpro.net
>>> http://www.molpro.net/mailman/listinfo/molpro-user
>> --
>> -----------
>> Manhui  Wang
>> School of Chemistry, Cardiff University,
>> Main Building, Park Place,
>> Cardiff CF10 3AT, UK
>> Telephone: +44 (0)29208 76637
>>

-- 
-----------
Manhui  Wang
School of Chemistry, Cardiff University,
Main Building, Park Place,
Cardiff CF10 3AT, UK
Telephone: +44 (0)29208 76637



More information about the Molpro-user mailing list