Re: [AMBER] AmberTools16 Installation with openMPI fails?

From: Daniel Roe <daniel.r.roe.gmail.com>
Date: Wed, 12 Oct 2016 09:30:32 -0400

Hi,

This has been fixed in the GitHub version of cpptraj. A patch for
AmberTools may be soon depending on how much free time I have. An easy
workaround is to remove the #ifdef around the '#include <cstdio>' in
FileIO_MPI.cpp. As mentioned previously, an easier workaround is to
use a different version/flavor of MPI (this is what I would
recommend).

Since cpptraj has worked fine with older versions of OpenMPI, I blame
this issue on OpenMPI :-)

-Dan

On Mon, Oct 10, 2016 at 2:09 PM, Hai Nguyen <nhai.qn.gmail.com> wrote:
> .Ucisik: I can reproduce your error. So while waiting for a fix
> (hopefully), you can either follow suggestions from Marcelo or Alican.
>
> You can also turn off cpptraj build with mpi in AmberTools/src/Makefile
> (search for *cpptraj::* and comment related lines)
>
> cheers
> Hai
>
> On Mon, Oct 10, 2016 at 2:02 PM, Gulsevin,Alican <agulsevin.chem.ufl.edu>
> wrote:
>
>> >From what I understand, the problem is that C and OpenMPI libraries can
>> have overlapping stdio.h variables. Therefore, variables can mean different
>> things within their "scope" for some OpenMPI (and mpich) versions. The
>> older version suggested by Marcelo can work. If not, you can try the
>> version suggested by Amber itself.
>>
>> Best,
>> Alican
>>
>> ________________________________________
>> From: Marcelo Andrade Chagas <andrade.mchagas.gmail.com>
>> Sent: Monday, October 10, 2016 1:53 PM
>> To: AMBER Mailing List
>> Subject: Re: [AMBER] AmberTools16 Installation with openMPI fails?
>>
>> Dear,
>>
>> I found a problem once with the version of MPI.
>>
>> I managed to solve using this:
>>
>> openmpi-1.6.5.tar.gz
>>
>> It seems to have some bivliotecas files that were modified in the latest
>> versions and amber during the build does not find the necessary libraries.
>>
>> Best regards
>>
>> Marcelo Andrade Chagas, MSc
>> (PhD student)
>> Laboratório de Química Computacional e Modelagem Molecular - LQC-MM
>> * http://lqcmm.qui.ufmg.br/
>> Departamento de Química da Universidade Federal de Minas Gerais - UFMG
>> Tel:(31)3409-5776
>>
>> 2016-10-10 14:39 GMT-03:00 Ucisik, Melek Nihan <ucisik.illinois.edu>:
>>
>> > Hello,
>> >
>> > I wanted to follow up on this. I started over clean and did the serial
>> > installation first. No problems there. Using openMPI version 2.0.1, I
>> > attempted the parallel installation. mpif90, mpicc, mpiexec, and mpirun
>> > are all pointing to the correct directory.
>> >
>> > echo $PATH
>> > /opt/amber16/16.14/amber16/bin:/opt/amber16/16.14/
>> > amber16/bin:/opt/amber16/16.14/amber16/bin:/opt/amber16/
>> > 16.14/amber16/bin:/opt/openmpi/2.0.1/bin:/opt/pkg-
>> > config/0.28/bin:/opt/python/2.7/bin:/usr/local/brl/bin:/usr/
>> > lib64/qt-3.3/bin:/usr/local/bin:/bin:/usr/bin:/usr/local/
>> > sbin:/usr/sbin:/sbin:/home/cddadmin/bin:/opt/openmpi/2.0.1/include
>> >
>> > echo $LD_LIBRARY_PATH
>> > /opt/amber16/16.14/amber16/lib:/opt/amber16/16.14/
>> > amber16/lib:/opt/amber16/16.14/amber16/lib:/opt/amber16/
>> > 16.14/amber16/lib:/opt/openmpi/2.0.1:/opt/openmpi/2.
>> > 0.1/lib:/opt/python/2.7/lib:/usr/local/brl/local/lib:/usr/
>> > local/brl/local/lib64:
>> >
>> > make clean
>> > ./configure --with-python /opt/python/2.7/bin/python -mpi gnu
>> > make install
>> >
>> > Same errors as before showed up:
>> >
>> > FileIO_Mpi.cpp: In member function ‘virtual int FileIO_Mpi::Seek(off_t)’:
>> > FileIO_Mpi.cpp:33: error: ‘SEEK_SET’ was not declared in this scope
>> > FileIO_Mpi.cpp: In member function ‘virtual int FileIO_Mpi::Rewind()’:
>> > FileIO_Mpi.cpp:39: error: ‘SEEK_SET’ was not declared in this scope
>> > make[4]: *** [FileIO_Mpi.o] Error 1
>> > make[4]: Leaving directory `/usr/local/brl/opt/amber16/
>> > 16.14/amber16/AmberTools/src/cpptraj/src'
>> > make[3]: *** [parallel] Error 2
>> > make[3]: Leaving directory `/usr/local/brl/opt/amber16/
>> > 16.14/amber16/AmberTools/src/cpptraj'
>> > make[2]: *** [cpptraj] Error 2
>> > make[2]: Leaving directory `/usr/local/brl/opt/amber16/
>> > 16.14/amber16/AmberTools/src'
>> > make[1]: *** [parallel] Error 2
>> > make[1]: Leaving directory `/usr/local/brl/opt/amber16/
>> > 16.14/amber16/AmberTools/src'
>> > make: *** [install] Error 2
>> >
>> >
>> > Could this be a bug? The FileIO_Mpi.cpp at $AMBERHOME/AmberTools/src/
>> cpptraj/src
>> > seems to be the problem. Any ideas? I can’t get past this point.
>> >
>> > Thanks,
>> >
>> > M. Nihan Ucisik
>> >
>> >
>> > On Oct 8, 2016, at 11:56 AM, Ucisik, Melek Nihan <ucisik.illinois.edu
>> > <mailto:ucisik.illinois.edu>> wrote:
>> >
>> > Hi,
>> >
>> > I did build the serial first with no problems and I did use the normal
>> > configure script with the -mpi gnu flag after having checked where the
>> > mpif90 and mpirun were pointing. They point to the same openmpi
>> > installation which is in my path.
>> >
>> > “Make install” gave errors after using the normal configure script which
>> > is why I tried the configure_openmpi script. I thought the configure step
>> > failed with the normal script. Here are the errors I had gotten:
>> >
>> > FileIO_Mpi.cpp: In member function ‘virtual int FileIO_Mpi::Seek(off_t)’:
>> > FileIO_Mpi.cpp:33: error: ‘SEEK_SET’ was not declared in this scope
>> > FileIO_Mpi.cpp: In member function ‘virtual int FileIO_Mpi::Rewind()’:
>> > FileIO_Mpi.cpp:39: error: ‘SEEK_SET’ was not declared in this scope
>> > make[4]: *** [FileIO_Mpi.o] Error 1
>> > make[4]: Leaving directory `/usr/local/brl/opt/amber16/
>> > 16.14/AmberTools/src/cpptraj/src'
>> > make[3]: *** [parallel] Error 2
>> > make[3]: Leaving directory `/usr/local/brl/opt/amber16/
>> > 16.14/AmberTools/src/cpptraj'
>> > make[2]: *** [cpptraj] Error 2
>> > make[2]: Leaving directory `/usr/local/brl/opt/amber16/
>> > 16.14/AmberTools/src'
>> > make[1]: *** [parallel] Error 2
>> > make[1]: Leaving directory `/usr/local/brl/opt/amber16/
>> > 16.14/AmberTools/src'
>> > make: *** [install] Error 2
>> >
>> > I know this is some kind of openmpi/path problem but I wasn’t able to get
>> > it resolved as I am checking everything and nothing seems out of place in
>> > terms of paths etc. Another issue is that “make clean” gives an error
>> also:
>> >
>> > make[1]: Entering directory `/usr/local/brl/opt/amber16/16.14/src'
>> > make[1]: *** No rule to make target `clean'. Stop.
>> > make[1]: Leaving directory `/usr/local/brl/opt/amber16/16.14/src'
>> > make: [clean] Error 2 (ignored)
>> >
>> > I guess I’ll try to start over. I had installed Amber multiple times
>> > before (thanks to Jason’s wiki) on different platforms and didn’t have
>> > these problems, so I am confused as to what is going wrong this time.
>> >
>> > Anyway, thanks for the suggestions,
>> >
>> > Nihan
>> >
>> >
>> > On Oct 7, 2016, at 5:30 PM, Charles Lin <clin92.ucsd.edu<mailto:clin92
>> > .ucsd.edu><mailto:clin92.ucsd.edu>> wrote:
>> >
>> > Check if your mpif90 and mpirun point to the right executables.
>> >
>> > (which mpif90 and which mpirun).
>> >
>> > You should theoretically be able to use the normal configure script for
>> > this.
>> >
>> > Charlie
>> > ________________________________________
>> > From: Ucisik, Melek Nihan [ucisik.illinois.edu<mailto:uc
>> isik.illinois.edu
>> > ><mailto:ucisik.illinois.edu>]
>> > Sent: Friday, October 07, 2016 3:23 PM
>> > To: AMBER Mailing List
>> > Subject: [AMBER] AmberTools16 Installation with openMPI fails?
>> >
>> > Hi,
>> >
>> > I am trying to install AmberTools using openMPI. I did have the
>> > openmpi-2.0.1 module loaded before I started with the installation, so I
>> > had mpicc and mpif90 in my path. I tried using the “configure_openmpi”
>> > script as suggested in the manual and it failed multiple times, so I
>> tried
>> > tweaking it. The script advised me to have the openmpi directory also at
>> > $AMBERHOME/AmberTools/src, so I copied openmpi-2.0.1.tar.bz2 to there and
>> > extracted. I modified the configure_openmpi script in the line where
>> > “mpidirs” is defined. I ended up with this:
>> >
>> > “./configure_openmpi: line 119: YACC=/opt/amber16/16.14/bin/yacc: No
>> such
>> > file or directory
>> > openmpi configure failed, returning 127“
>> >
>> > This made no sense to me because /opt/amber16/16.14/bin/yacc was there.
>> >
>> > Then I removed the “$static” variable from line 119 of configure_openmpi
>> > and this is what I got:
>> >
>> > "configure: error: /bin/sh config/config.sub failed
>> > openmpi configure succeeded.
>> > (You may need to add /opt/amber16/16.14/lib to your LD_LIBRARY_PATH)
>> > make: *** No rule to make target `clean'. Stop.
>> > make: *** No rule to make target `install'. Stop.”
>> >
>> > So I am confused about this outcome because it says “error" in one line
>> > and “succeeded" in the next.
>> >
>> > Could someone tell me if I am good to continue from this point? Did the
>> > configuration fail or succeed?
>> >
>> > Thanks,
>> >
>> >
>> > M. Nihan Ucisik
>> > _______________________________________________
>> > AMBER mailing list
>> > AMBER.ambermd.org<mailto:AMBER.ambermd.org><mailto:AMBER.ambermd.org>
>> > http://lists.ambermd.org/mailman/listinfo/amber
>> >
>> > _______________________________________________
>> > AMBER mailing list
>> > AMBER.ambermd.org<mailto:AMBER.ambermd.org>
>> > http://lists.ambermd.org/mailman/listinfo/amber
>> >
>> > _______________________________________________
>> > AMBER mailing list
>> > AMBER.ambermd.org<mailto:AMBER.ambermd.org>
>> > http://lists.ambermd.org/mailman/listinfo/amber
>> >
>> > _______________________________________________
>> > AMBER mailing list
>> > AMBER.ambermd.org
>> > http://lists.ambermd.org/mailman/listinfo/amber
>> >
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber



-- 
-------------------------
Daniel R. Roe
Laboratory of Computational Biology
National Institutes of Health, NHLBI
5635 Fishers Ln, Rm T900
Rockville MD, 20852
https://www.lobos.nih.gov/lcb
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Oct 12 2016 - 07:00:02 PDT
Custom Search