Hi Matteo,
Do you really need AmberTools in parallel? - I would suggest just skipping
AmberTools parallel and building AMBER in parallel instead.
cd $AMBERHOME/AmberTools/src
./configure -mpi gnu
make clean
cd ../../
./AT15_Amber.py
cd src
make clean
make parallel
This will build you sander.MPI and pmemd.MPI which are the programs that
most benefit from running in parallel and chances are they will be much
easier to build than AMBERTools.
It will also likely be much easier to debug compilation problems if there
are failures here.
All the best
Ross
> -----Original Message-----
> From: Matteo Guglielmi [mailto:matteo.guglielmi.epfl.ch]
> Sent: Friday, September 23, 2011 12:07 PM
> To: amber.ambermd.org
> Subject: Re: [AMBER] AmberTools Parallel Compilation Issue
>
> Ok, no intel compilers but with the same impi (3.2.2) set to use
> gfortran, gcc and g++:
>
> ambertools (serial - OK)
> amber (serial - OK)
> ambertools (parallel - ERROR)
>
> and here will follow the error message.
>
> Is it Debian so Bad?
>
> BTW compiling openmpi 1.4.3 and using the very same gnu compilers
> everything get compiled correctly... but there I have the problem
> that infiniband becomes slow while with impi infiniband is just
> right.
>
> ...
> a - dsarpack.o
> a - wallclock.o
> ranlib arpack.a
> mv arpack.a /software/amber11/lib
> make[2]: Leaving directory
> `/share/lcbc/software/amber11/AmberTools/src/arpack'
> mpif90 -ffree-form -o pbsa.MPI dsvdc.o pbsa.o pb_read.o pb_write.o
> getcoor.o runmd.o runmin.o force.o pb_init.o sa_driver.o np_force.o
> pb_force.o pb_fddrv.o pb_iimdrv.o pb_nlsolver.o pb_exmol.o pb_mpfrc.o
> pb_direct.o pb_list.o timer.o egb.o ene.o locmem.o myopen.o rdparm.o
> pb_lsolver.o decomp.o rgroup.o rfree.o debug.o coed20.o indexg.o
> irre32.o phidev.o project.o qld.o coed6.o irre31.o miniop.o
> prodis.o qint.o wint.o transf.o curv.o GrToPr.o jumps.o problem.o
> IIM.o regular.o ctime.o gmres.o daxpy.o ddot.o dheqr.o dpigmr.o
> dscal.o dslui2.o dxlcal.o dchkw.o dgmres.o dnrm2.o drlcal.o
> dsilus.o dslui.o isdgmr.o dcopy.o dhels.o dorth.o ds2y.o
> dslugm.o dsmv.o qs2i1d.o d1mach.o fdump.o i1mach.o j4save.o
> xercnt.o xerhlt.o xermsg.o xerprn.o xersve.o xgetua.o amg.o
> aux1r5.o
> amg1r5.o bicg.o dbcg.o dslui4.o dsluti.o dsmtv.o dslubc.o isdbcg.o
> pb_chkfrc.o svdcmp.o svbksb.o pythag.o \
> ../lib/nxtsec.o ../lib/random.o \
> /software/amber11/lib/arpack.a
> /software/amber11/lib/lapack.a
> /software/amber11/lib/blas.a
> /bin/mv pbsa.MPI /software/amber11/bin
> make[1]: Leaving directory
> `/share/lcbc/software/amber11/AmberTools/src/pbsa'
> (cd nab && make parallel )
> make[1]: Entering directory
> `/share/lcbc/software/amber11/AmberTools/src/nab'
> mpicc -DBINDIR='"/software/amber11/bin"' \
> -DINCDIR='"/software/amber11/include"' \
> -DLIBDIR='"/software/amber11/lib"' \
> -DCC='"mpicc"' -DCPP='"/software/amber11/bin/ucpp -l"'
> -DFLIBS='"-L/software/amber11/lib -lsff_mpi -lpbsa
> /software/amber11/lib/arpack.a /software/amber11/lib/lapack.a
> /software/amber11/lib/blas.a /software/amber11/lib/libnetcdf.a
> -lgfortran "' \
> -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -DBINTRAJ -DMPI
> \
> -o /software/amber11/bin/mpinab nab.c
> gcc: ": No such file or directory
> <command-line>: warning: missing terminating " character
> <command-line>: warning: missing terminating " character
> nab.c: In function 'n2c':
> nab.c:159: error: missing terminating " character
> nab.c:159: error: expected expression before ',' token
> nab.c: In function 'cc':
> nab.c:261: error: missing terminating " character
> nab.c:261: error: expected expression before ')' token
> make[1]: *** [mpinab] Error 1
> make[1]: Leaving directory
> `/share/lcbc/software/amber11/AmberTools/src/nab'
> make: *** [parallel] Error 2
>
>
> On 09/23/11 17:24, Carlos Sosa wrote:
> > Hi Matteo,
> >
> > Since potentially the issue is with the compiler I only suggested
> GNU, not
> > the OS.
> >
> > On Fri, Sep 23, 2011 at 10:21 AM, Matteo
> Guglielmi<matteo.guglielmi.epfl.ch
> >> wrote:
> >
> >> Carlos)
> >>
> >> Unfortunately I cannot move to RedHat, the cluster is already up and
> >> running with lots of codes installed.
> >>
> >> DAC)
> >>
> >> Just changing the compilers to 11.1 (fresh installation from .tgz
> files)
> >> makes ambertools (serial) basically not compilable at all (it
> crashes
> >> right after the compilation of python):
> >>
> >> ### ERRORS with intel 11.1, impi 3.2.2, ambertools serial ###
> >> ...
> >> changing mode of /software/amber11/lib/python2.6/lib-
> dynload/audioop.so
> >> to 755
> >> changing mode of /software/amber11/lib/python2.6/lib-dynload/ to 755
> >> running install_scripts
> >> copying build/scripts-2.6/idle -> /software/amber11/bin
> >> copying build/scripts-2.6/pydoc -> /software/amber11/bin
> >> copying build/scripts-2.6/2to3 -> /software/amber11/bin
> >> copying build/scripts-2.6/smtpd.py -> /software/amber11/bin
> >> changing mode of /software/amber11/bin/idle to 755
> >> changing mode of /software/amber11/bin/pydoc to 755
> >> changing mode of /software/amber11/bin/2to3 to 755
> >> changing mode of /software/amber11/bin/smtpd.py to 755
> >> running install_egg_info
> >> Writing
> >> /software/amber11/lib/python2.6/lib-dynload/Python-2.6.6-py2.6.egg-
> info
> >> if test -f /software/amber11/bin/python -o -h
> >> /software/amber11/bin/python; \
> >> then rm -f /software/amber11/bin/python; \
> >> else true; \
> >> fi
> >> (cd /software/amber11/bin; ln python2.6 python)
> >> rm -f /software/amber11/bin/python-config
> >> (cd /software/amber11/bin; ln -s python2.6-config python-config)
> >> /usr/bin/install -c -m 644 ./Misc/python.man \
> >> /software/amber11/share/man/man1/python.1
> >> make[1]: Leaving directory
> >> `/share/lcbc/software/amber11/AmberTools/src/etc/Python-2.6.6'
> >> Starting installation of MTK++ at Fri Sep 23 16:17:02 CEST 2011.
> >> (cd mtkpp&& make install )
> >> make[1]: Entering directory
> >> `/share/lcbc/software/amber11/AmberTools/src/mtkpp'
> >> Making install in src
> >> make[2]: Entering directory
> >> `/share/lcbc/software/amber11/AmberTools/src/mtkpp/src'
> >> Making install in Diagnostics
> >> make[3]: Entering directory
> >> `/share/lcbc/software/amber11/AmberTools/src/mtkpp/src/Diagnostics'
> >> /bin/bash ../../libtool --tag=CXX --mode=compile icpc -
> DHAVE_CONFIG_H
> >> -I. -I../.. -I/software/amber11/AmberTools/src/boost-1.38.0 -I./..
> >> -std=c++0x -DNDEBUG -DUSE_TINYXML -DTIXML_USE_STL -DC99_OK -O2 -
> fPIC
> >> -DHAVE_ZLIB -DHAVE_ZLIB -MT MTKException.lo -MD -MP -MF
> >> .deps/MTKException.Tpo -c -o MTKException.lo MTKException.cpp
> >> libtool: compile: icpc -DHAVE_CONFIG_H -I. -I../..
> >> -I/software/amber11/AmberTools/src/boost-1.38.0 -I./.. -std=c++0x
> >> -DNDEBUG -DUSE_TINYXML -DTIXML_USE_STL -DC99_OK -O2 -fPIC -
> DHAVE_ZLIB
> >> -DHAVE_ZLIB -MT MTKException.lo -MD -MP -MF .deps/MTKException.Tpo -
> c
> >> MTKException.cpp -o MTKException.o
> >> /usr/include/c++/4.4.5/bits/stringfwd.h(67): error: identifier
> >> "char16_t" is undefined
> >> template<> struct char_traits<char16_t>;
> >> ^
> >>
> >> /usr/include/c++/4.4.5/bits/stringfwd.h(68): error: identifier
> >> "char32_t" is undefined
> >> template<> struct char_traits<char32_t>;
> >> ...
> >> #### END ####
> >>
> >> Now fresh installation again from .tgz files, same setup but
> >> using intel compilers 11.0:
> >>
> >> FLIBS= -L$(LIBDIR) -lsff -lpbsa -lrism -ldrfftw -ldfftw
> >> $(LIBDIR)/arpack.a $(LIBDIR)/lapack.a $(LIBDIR)/blas.a
> >> $(LIBDIR)/libnetcdf.a -
> L/software/intel/Compiler/11.0/084/lib/intel64/
> >> -lifport -lifcore -lsvml
> >>
> >> ambertools (serial - OK)
> >> amber (serial - OK)
> >> ambertools (parallel with "make nabonly" - ERROR)
> >>
> >> I get exactly the same error message (what to do next then?):
> >>
> >> ....
> >> a - amber_rism_interface.o
> >> ranlib /software/amber11/lib/libsff.a
> >> mpicc -o /software/amber11/bin/hcp_getpdb hcp_getpdb.c
> >> make[1]: Leaving directory
> >> `/share/lcbc/software/amber11/AmberTools/src/sff'
> >> (cd nab&& make install )
> >> make[1]: Entering directory
> >> `/share/lcbc/software/amber11/AmberTools/src/nab'
> >> mpicc -shared-intel -DBINDIR='"."' -DINCDIR='"."' -DLIBDIR='"."'
> \
> >> -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -
> DBINTRAJ -DMPI
> >> \
> >> -DCC='"mpicc"' -DCPP='"/software/amber11/bin/ucpp -
> l"'
> >> -DFLIBS='"-L/software/amber11/lib -lsff_mpi -lpbsa
> >> /software/amber11/lib/arpack.a /software/amber11/lib/lapack.a
> >> /software/amber11/lib/blas.a /software/amber11/lib/libnetcdf.a
> >> -L/software/intel/Compiler/11.0/084/lib/intel64/ -lifport -lifcore
> >> -lsvml "' -o nab nab.c
> >> nab.c(159): error: missing closing quote
> >> CPP, cppstring, INCDIR,
> >> ^
> >>
> >> nab.c(261): error: missing closing quote
> >> sprintf( word, " %s ", FLIBS );
> >> ^
> >>
> >> compilation aborted for nab.c (code 2)
> >> make[1]: *** [nab] Error 2
> >> make[1]: Leaving directory
> >> `/share/lcbc/software/amber11/AmberTools/src/nab'
> >> make: *** [nabonly] Error 2
> >>
> >>
> >>
> >>
> >>
> >> On 09/23/11 15:51, Carlos Sosa wrote:
> >>> This is not a solution but a workaround. I used Intel MPI and the
> GNU
> >>> compiler on RedHat and did not encounter any problems.
> >>>
> >>> On Fri, Sep 23, 2011 at 8:15 AM, David A
> Case<case.biomaps.rutgers.edu
> >>> wrote:
> >>>
> >>>> On Fri, Sep 23, 2011, Matteo Guglielmi wrote:
> >>>>
> >>>>> ### ERROR MESSAGE ###
> >>>>>
> >>>>> nab.c(261): error: missing closing quote
> >>>>> sprintf( word, " %s ", FLIBS );
> >>>>> ^
> >>>>
> >>>> This sort of error has been reported several times before. As far
> as I
> >> can
> >>>> see, no one has ever found anything wrong with the Amber codes,
> and it
> >> is
> >>>> frustrating, since the code itself above looks fine. You might
> play
> >> with
> >>>> the
> >>>> FLIBS variable in config.h (is it perhaps empty?), to see if
> anything
> >> pops
> >>>> out.
> >>>>
> >>>>> PS: using the intel compilers 11.1 insead of 11.0 I cannot even
> >>>>> compile the serial version of ambertools... I get very
> similar
> >>>>> error messages which I'm not reporting here.
> >>>>
> >>>> Can you do a fresh reinstall, configure (serial) and "make
> nabonly",
> >> then
> >>>> report the exact error results (and complete version information
> on the
> >>>> compiler and OS)?
> >>>>
> >>>> ...thx...dac
> >>>>
> >>>> _______________________________________________
> >>>> 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
Received on Fri Sep 23 2011 - 12:30:05 PDT
