Hi Paul and Case,
Thank you for the responses.
The changed flags below did the trick to compile Amber successfully with
openmpi and intel.
All the AmberTools test.parallel passed, while all but 1 of Amber11 passed.
This is the failed one:
==============================================================
export TESTsander='../../exe/pmemd.MPI'; cd tip5p && ./Run.tip5p
diffing mdout.tip5p.save with mdout.tip5p
possible FAILURE: check mdout.tip5p.dif
==============================================================
Let me know if I should be concerned with this single test failure.
Hadrian.
On Mon, Oct 4, 2010 at 12:40 PM, Paul S. Nerenberg <psn.berkeley.edu> wrote:
> Hi Hadrian,
>
> The last problem (compiling pmemd) is now a known issue with the intel
> compilers.* It revolves around the -fast optimization flag and the
> other flags it is automatically calling (-static) and the type of
> libraries you have on your system (looks like dynamic). Try editing
> your $AMBERHOME/src/config.h (and $AMBERHOME/AmberTools/src/config.h)
> file and replacing '-fast' with '-ipo -O3 -no-prec-div -xHost'. Then
> do a "make clean" and recompile the parallel version.
>
> Best,
>
> Paul
>
> *It's not really a bug, you just have to be "heads up" about whether
> you've got static or dynamic libraries.
>
>
> On Oct 3, 2010, at 10:30 PM, Hadrian Djohari wrote:
>
> > Hi everyone,
> >
> >
> >> I have some difficulty installing Amber 11 in our cluster, so I would
> >> appreciate everyone's help. I would like the answers to all the
> >> questions
> >> below, but if someone can help me with the answer for the specific
> >> openmpi/intel installation right at the bottom part, I would be
> >> really
> >> grateful.
> >>
> >> Amber installation questions:
> >>
> >> 1. The instruction in the Amber and AmberTools document seem to
> >> direct
> >> people to have both the source and the compiled binaries in the
> >> same
> >> directory by using only "make" instead of "make install". Is this
> >> the only
> >> way? We prefer to have the source and the compiled binaries in
> >> different
> >> directories (/usr/local/src/amber and /usr/local/amber). If we
> >> want separate
> >> directories, which one should $AMBERHOME point to?
> >>
> >> 2. The installer should be "root" or a user? I compiled "make
> >> serial"
> >> just fine as a user, but found some errors compiling as "root",
> >> probably due
> >> to some paths not being recognized.
> >>
> >> 3. Since I'm installing Amber, I should also install AmberTools.
> >> And
> >> I'd like to install the parallel version. So, is there a step-by-
> >> step
> >> instructions on how to install Amber (with AmberTools) as
> >> parallel code
> >> (please see no.1 too)? That would be EXTREMELY HELPFUL. Some of
> >> these
> >> questions needed answers (why do I have to install the serial
> >> versions
> >> first? how do I link to an existing libraries (fftw, mpi, mkl)?
> >> what paths
> >> are needed by the installer to execute?). These small things can
> >> lead to an
> >> unsuccessful compilation, so I would really like some more detailed
> >> information about this.
> >>
> >> 4. Once the code is compiled, is there a short test job that use
> >> PBS
> >> submission, not the "make test" that is available in the
> >> directories?
> >>
> >>
> >> Also, I have tried this approach to install MD on a RHEL5.5 server:
> > export AMBERHOME=/usr/local/amber/amber11
> > cd $AMBERHOME/AmberTools/src
> > ./configure -mpi intel
> > make parallel
> > This is successful.
> >
> > And then:
> > cd $AMBERHOME/src
> > make parallel
> >
> > But it produces this error;
> > ...
> > mpif90 -fast -c charmm_gold.f90
> > mpif90 -fast -o pmemd.MPI gbl_constants.o gbl_datatypes.o
> > state_info.o
> > file_io_dat.o mdin_ctrl_dat.o mdin_ewald_dat.o mdin_debugf_dat.o
> > prmtop_dat.o inpcrd_dat.o dynamics_dat.o img.o parallel_dat.o
> > parallel.o
> > gb_parallel.o pme_direct.o pme_recip_dat.o pme_slab_recip.o
> > pme_blk_recip.o
> > pme_slab_fft.o pme_blk_fft.o pme_fft_dat.o fft1d.o bspline.o
> > pme_force.o
> > pbc.o nb_pairlist.o nb_exclusions.o cit.o dynamics.o bonds.o angles.o
> > dihedrals.o extra_pnts_nb14.o runmd.o loadbal.o shake.o prfs.o
> > mol_list.o
> > runmin.o constraints.o axis_optimize.o gb_ene.o veclib.o gb_force.o
> > timers.o
> > pmemd_lib.o runfiles.o file_io.o bintraj.o pmemd_clib.o pmemd.o
> > random.o
> > degcnt.o erfcfun.o nmr_calls.o nmr_lib.o get_cmdline.o master_setup.o
> > pme_alltasks_setup.o pme_setup.o ene_frc_splines.o gb_alltasks_setup.o
> > nextprmtop_section.o angles_ub.o dihedrals_imp.o cmap.o charmm.o
> > charmm_gold.o ../../netcdf/lib/libnetcdf.a
> > ipo: remark #11000: performing multi-file optimizations
> > ipo: remark #11005: generating object file /tmp/ipo_ifortPoYvAz.o
> > /usr/bin/ld: cannot find -lmpi_f90
> > make[2]: *** [pmemd.MPI] Error 1
> > make[2]: Leaving directory `/usr/local/amber/amber11/src/pmemd/src'
> > make[1]: *** [parallel] Error 2
> > make[1]: Leaving directory `/usr/local/amber/amber11/src/pmemd'
> > make: *** [parallel] Error 2
> >
> > I don't know why it cannot find -lmpi_f90 since libmpi_f90.* are in
> > the
> > directory that is in the ld_library path
> > hxd58.login:/usr/local/amber/amber11/src$ echo $LD_LIBRARY_PATH
> > /usr/local/openmpi/openmpi-intel/lib/:/usr/local/lib:/usr/local/
> > intel/compilers/11.1.056/lib/intel64:/usr/local/intel/compilers/
> > 11.1.056/idb/lib/intel64
> > hxd58.login:/usr/local/amber/amber11/src$ ll
> > /usr/local/openmpi/openmpi-intel/lib
> > ...
> > -rwxr-xr-x 1 root root 1129 Mar 26 2010 libmpi_f90.la
> > lrwxrwxrwx 1 root root 19 Mar 26 2010 libmpi_f90.so ->
> > libmpi_f90.so.0.0.0
> > lrwxrwxrwx 1 root root 19 Mar 26 2010 libmpi_f90.so.0 ->
> > libmpi_f90.so.0.0.0
> > -rwxr-xr-x 1 root root 17140 Mar 26 2010 libmpi_f90.so.0.0.0
> > ...
> >
> > Thank you,
> >
> > --
> > Hadrian Djohari
> > HPCC Manager
> > Case Western Reserve University
> > (W): 216-368-0395
> > (M): 216-798-7490
> >
> >
> >
> > --
> > Hadrian Djohari
> > HPCC Manager
> > Case Western Reserve University
> > (W): 216-368-0395
> > (M): 216-798-7490
> > _______________________________________________
> > 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
>
--
Hadrian Djohari
HPCC Manager
Case Western Reserve University
(W): 216-368-0395
(M): 216-798-7490
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Oct 04 2010 - 22:00:04 PDT
