AMBER Archive (2006)

Subject: Re: AMBER: failure in testing amber8 parallel parts on IBM-SP4 machines

From: Rachel (comeonsos_at_googlemail.com)
Date: Fri Nov 24 2006 - 13:53:29 CST

Hi, Nicolas,

Thanks a lot for your reply, I am sorry I didn't read carefullly before I
sent out last email. Actually I have succeeded in running some parallel jobs
using the 'version' I compiled myself (using poe), therefore I think the
compilation should be OK. However, I still want to finish the parallel
tests. After I did the setting as you suggested (e.g. setenv DO_PARALLEL
poe) etc. and ran the test, I got:

#############################################
ERROR: 0031-808 Hostfile or pool must be used to request nodes
#############################################

As I am running on the national HPCx, we need to supply an account number
(in the submit script) in order to run the jobs, and I think this is why I
got this message? if so, how can I do the tests? I am sorry but I am really
newbie.

Best regards,
Rachel

On 11/24/06, Nicolas Lux Fawzi <fawzin_at_berkeley.edu> wrote:
>
> Hi Rachel
> Previously, I succeeded in getting amber (9) to run and pass the test
> on AIX (v 5.something, on power3 and power5?) and followed the
> procedure that I think was already described on this list in response
> to one of your previous emails.
> I think Ross Walker already answered this question earlier, but there
> were additional problems that you solved first.
> i've pasted in the response below. In summary, I haven't seen AIX
> configured to use "mpirun" as the program to run mpi parallel programs,
> but rather I've seen it configured with poe. I've copied Ross Walker's
> answer below and highlighted the part that I think you still need to
> fix:
>
> > mpirun: Command not found.
> > I part of your problem. On IBM Power 4 machines running AIX (I assume
> > you are running AIX) one typically uses IBM's parallel operating
> > environment "POE".
> > You would typically run a parallel job setting:
> >
> > setenv MP_NODES 1
> > setenv MP_TASKS_PER_NODE 8
> > poe $AMBERHOME/exe/sander.MPI -O blah blah blah
> >
> > poe reads the environment variables to determine the number of cpus to
> > use rather than using a -np x option like mpirun would.
> > Thus to run the parallel test cases in amber9 you would set the above
> > environment variables to the combination of nodes and tasks you want
> > and then run:
> >
> > setenv DO_PARALLEL poe
> > make test.sander.parallel
>
> Those commands work for c-shell (csh, tcsh, etc.) If you are using sh,
> bash, etc, use the "export" command instead. Let us know if you need
> help with that. If you type "env" at your shell, you should see if
> DO_PARALLEL is set to poe or to mpirun. If "which poe" doesn't turn up
> the path to poe, then your machine may not be configured for it, or
> it's simply not in your path and you should look for poe.
>
> Hope this is helpful!
>
> -Nick
>
>
>
>
> Begin forwarded message:
>
> > From: "Ross Walker" <ross_at_rosswalker.co.uk>
> > Date: November 22, 2006 9:55:45 AM PST
> > To: <amber_at_scripps.edu>
> > Subject: RE: AMBER: compile amber8 on IBM-sp4
> > Reply-To: amber_at_scripps.edu
> >
> > Dear Rachel,
> >
> > The fact that you are seeing the error:
> >
> > mpirun: Command not found.
> > I part of your problem. On IBM Power 4 machines running AIX (I assume
> > you are running AIX) one typically uses IBM's parallel operating
> > environment "POE".
> >
> > You would typically run a parallel job setting:
> >
> > setenv MP_NODES 1
> > setenv MP_TASKS_PER_NODE 8
> > poe $AMBERHOME/exe/sander.MPI -O blah blah blah
> >
> > poe reads the environment variables to determine the number of cpus to
> > use rather than using a -np x option like mpirun would.
> >
> > Thus to run the parallel test cases in amber9 you would set the above
> > environment variables to the combination of nodes and tasks you want
> > and then run:
> >
> > setenv DO_PARALLEL poe
> > make test.sander.parallel
> >
> > However, I believe the main issue you are having is with compiling and
> > from your earlier emails this seemed to be an issue with the mpi
> > installation:
> >
> > #include file "mpif.h" not found.
> > This could come from two situations. Firstly IBM's mpi might never
> > have been installed on your machine or secondly your environment
> > variables might not be configured correctly.
> >
> > Try 'which mpxlf90' and see if it returns /usr/bin/mpxlf90
> >
> > if it says 'mpxlf90: Command not found.' then I suspect mpi is not
> > installed or possibly installed in a non standard place. Check with
> > whoever setup the machine to find out which mpi they installed and
> > where.
> >
> > For the moment you should be able to build and test the serial
> > version. I would advise you to do this first to make sure things are
> > working. Do:
> >
> > cd $AMBERHOME/src
> > ./configure -nopar xlf90_aix
> > make clean
> > make
> > cd ../test
> > make clean
> > make
> >
> > All the best
> > Ross
> >
> > /\
> > \/
> > |\oss Walker
> >
> > | HPC Consultant and Staff Scientist |
> > | San Diego Supercomputer Center |
> > | Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
> > | http://www.rosswalker.co.uk | PGP Key available on request |
> >
> > Note: Electronic Mail is not secure, has no guarantee of delivery, may
> > not be read every day, and should not be used for urgent or sensitive
> > issues.
> >
> >
> >> From: owner-amber_at_scripps.edu [mailto:owner-amber_at_scripps.edu] On
> >> Behalf Of Rachel
> >> Sent: Wednesday, November 22, 2006 02:09
> >> To: amber_at_scripps.edu
> >> Subject: Re: AMBER: compile amber8 on IBM-sp4
> >>
> >> Hi, Scott,
> >>
> >> Thanks very much for your reply, however, i am still not clear how to
> >> solve the problem, you suggested to use 'make -n'? and the results
> >> are as following:
> >> ----------------------------------------------------------------------
> >> cd dmp; ./Run.dmp
> >> cd adenine; ./Run.adenine
> >> cd cytosine; ./Run.cytosine
> >> cd nonper; ./Run.nonper
> >> cd nonper; ./Run.nonper.belly
> >> cd nonper; ./Run.nonper.belly.mask
> >> cd nonper; ./Run.nonper.min
> >> cd nonper; ./Run.cap
> >> cd nonper; ./Run.nonper.nocut
> >> cd tip4p; ./Run.tip4p
> >> cd tip5p; ./Run.tip5p
> >> cd 4096wat; ./Run.pure_wat
> >> cd dhfr; ./Run.dhfr
> >> cd dhfr; ./Run.dhfr.noshake
> >> cd dhfr; ./Run.dhfr.min
> >> cd gb_rna; ./Run.gbrna
> >> cd gb_rna; ./Run.gbrna.min
> >> cd gb_rna; ./Run.gbrna.ln
> >> cd gbsa_xfin; ./Run.gbsa
> >> cd polarizable_water; ./Run.pol_wat
> >> cd ubiquitin; ./Run.ubiquitin
> >> cd dna_pol; ./Run.dna_pol
> >> cd aspash; ./Run.aspash
> >> cd circ_dna; ./Run.circdna
> >> cd gb2_trx; ./Run.trxox
> >> cd trx; ./Run.trx
> >> cd trx; ./Run.trx.cpln
> >> cd cnstph; ./Run.cnstph
> >> cd rdc; ./Run.dip
> >> cd tgtmd/change_target; ./Run.tgtmd
> >> cd tgtmd/change_target.rms; ./Run.tgtmd
> >> cd tgtmd/change_target.ntr; ./Run.tgtmd
> >> cd tgtmd/conserve_ene; ./Run.tgtmd
> >> cd tgtmd/minimize; ./Run.tgtmin
> >> cd tgtmd/PME; ./Run.tgtPME
> >> cd trajene; ./Run.trajene
> >> cd pheTI; ./Run.0; ./Run.1; ./Run.p0; ./Run.p1
> >> cd ion_wat; ./Run.ion_wat
> >> cd alp; ./Run.alp
> >> cd pb_pgb; ./Run.pbpgb
> >> cd umbrella; ./Run.umbrella
> >> cd LES_noPME; ./Run.LESmd
> >> cd LES_noPME; ./Run.LESmd.rdiel
> >> cd LES; ./Run.PME_LES
> >> cd LES_CUT; ./Run.LES
> >> cd LES_TEMP; ./Run.2temp
> >> cd LES_GB; ./Run.LES
> >> make: 1254-002 Cannot find a rule to create target \ from
> >> dependencies.
> >> Stop.
> >> ----------------------------------------------------------------------
> >> ----------------------------------
> >>
> >> As you said the compilation line is broken, how can I make sure i
> >> provided enough details? and if i use 'make test.sander', etc.
> >> instead of make test.parallel, then i got:
> >> ----------------------------------------------------------------------
> >> ---------------------------------
> >> cd dmp; ./Run.dmp
> >> This test not set up for parallel
> >> cannot run in parallel with #residues < #pes
> >> cd adenine; ./Run.adenine
> >> This test not set up for parallel
> >> cannot run in parallel with #residues < #pes
> >> ==============================================================
> >> cd cytosine; ./Run.cytosine
> >> mpirun: Command not found.
> >> ./Run.cytosine: Program error
> >> make: 1254-004 The error code from the last command is 1.
> >> ----------------------------------------------------------------------
> >> ---------------------------------
> >>
> >> This is the first time I compile amber myself, and as it is IBM-SP4
> >> machine, so what i did was just follow Carlos Sosa's Running Amber on
> >> IBM systems (
> >> http://www.msi.umn.edu/~cpsosa/ChemApps/MolMech/amber/patches/amber8/
> >> INSTALL_ibm) on AMBER website, I really appreciate your help if
> >> anyone can let me knowsome more details of how to do it.
> >>
> >> Best regards,
> >> Rachel
> >>
> >>
> >>
> >> On 11/21/06, Scott Brozell <sbrozell_at_scripps.edu> wrote: Hi,
> >>>
> >>> On Tue, 21 Nov 2006, Rachelwrote:
> >>>
> >>> > Dear all,
> >>> >
> >>> > I am trying to compile amber8 on the IBM-SP4 machines, I used
> >>> './configure
> >>> > -mpi xlf90_aix', after i used 'make paralle', i got the following
> >>> error
> >>> > message:
> >>>
> >>> > "egb.f", line 131.12: 1506-296 (S) #include file "mpif.h" not
> >>> found.
> >>> > make: 1254-004 The error code from the last command is 1.
> >>>
> >>> It looks like the compilation line is broken.
> >>> You didn't provide enough detail, so the full command
> >>> invoked by make is not clear.
> >>>
> >>> make should be invoking something like, eg
> >>>
> >>> /lib/cpp -traditional -I/usr/lpp/ppe.poe/include -P
> >>> -I/usr/local/srb/9/src/include -DMPI-DNMLEQ -DCLINK_PLAIN -Drs6000
> >>> -DPOE sander.f > _sander.f
> >>> mpxlf90_r -bmaxdata:0x80000000 -c -qfixed -c -qfree-o sander.o
> >>> _sander.f
> >>>
> >>> Compare this with the output of make -n
> >>> What happens if you enter mpxlf90_r -show -help ?
> >>>
> >>> > and then if i go to the $AMBERHOME/test directory and use 'make
> >>> > test.parallel', it says:
> >>>
> >>> See page 7 errata
> >>> http://amber.scripps.edu/doc8/errata.html
> >>>
> >>> Scott
> >>>
> >>> ---------------------------------------------------------------------
> >>> --
> >>> The AMBER Mail Reflector
> >>> To post, send mail to amber_at_scripps.edu
> >>> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
> On Nov 24, 2006, at 9:58 AM, Rachel wrote:
>
> > Dear all,
> >
> > When I use 'make test.sander' in $AMBERHOME/test for the amber8
> > parallel parts on IBM_SP4 machines, I got the following error
> > messages, I searched in the mailing list, and found out that the
> > should be related to "mpi installation and compilation". Thesystem I
> > am using is a standard IBM's AIX operating system. Can anyone suggest
> > what is the reason for the errors and how can I correctly test the
> > parallel part? Thanks a lot.
> >
> > ######################################################################
> > cd dmp; ./Run.dmp
> > This test not set up for parallel
> > cannot run in parallel with #residues < #pes
> > cd adenine; ./Run.adenine
> > This test not set up for parallel
> > cannot run in parallel with #residues < #pes
> > ==============================================================
> > cd cytosine; ./Run.cytosine
> > mpirun: Command not found.
> > ./Run.cytosine: Program error
> > make: 1254-004 The error code from the last command is 1.
> >
> >
> > Stop.
> > bash-2.05a$ make test.sander
> > cd dmp; ./Run.dmp
> > This test not set up for parallel
> > cannot run in parallel with #residues < #pes
> > cd adenine; ./Run.adenine
> > This test not set up for parallel
> > cannot run in parallel with #residues < #pes
> > ==============================================================
> > cd cytosine; ./Run.cytosine
> > mpirun: Command not found.
> > ./Run.cytosine: Program error
> > make: 1254-004 The error code from the last command is 1.
> >
> > Stop.
> > ######################################################################
> >
> > With my best regards,
> > Rachel
> >
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu