Hi,
I did a fair amount of work getting Amber to compile with cray
compilers on NCSA Blue waters. However, I was under a time crunch at
the time (approaching a release) so I punted on a bunch of stuff and
focused on a subset of programs that were relatively easy to get
working. The stuff I couldn't get working: pbsa, gbnsr6, sander,
nmr_aux, nss, etc, mmpbsa, amberlite, quick
Stuff that works (or worked): pmemd (using an external FFTW),
antechamber, sqm, reduce, leap, addles, nmode, cpptraj, sff, rism,
nab. mdgx, xtalutil, saxs, FEW, paramfit, cphstats. This was all using
a system NetCDF as well as I couldn't get the bundled one to build
easily.
The reason I gave up on the other stuff (sander/pbsa in particular)
was that there was too much to do to get it to play nice with cray
compilers. I can try to help with this but unfortunately I don't have
as much time for development as I used to. If you're interested, maybe
we should take this off-list and start working in a branch in the main
Amber GIT tree so that any changes can eventually be incorporated into
the next release to help future generations.
-Dan
On Thu, Sep 21, 2017 at 5:15 PM, Timothy Schutt
<tschutt.mymail.mines.edu> wrote:
> Thanks for the responses!
>
> noemil lets it get past the configure step! It then crashes quickly in the
> "make install" with:
> ...[lots of lines]....
> mpicc -fPIC -O0 -h gnu -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> -DBINTRAJ -DHASGZ -D__PLUMED_HAS_DLOPEN -DMPI -c -o symtab.o symtab.c
> mpicc -fPIC -O0 -h gnu -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> -DBINTRAJ -DHASGZ -D__PLUMED_HAS_DLOPEN -DMPI -c -o verbose.o
> verbose.c
> mpicc -fPIC -O0 -h gnu -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> -DBINTRAJ -DHASGZ -D__PLUMED_HAS_DLOPEN -DMPI -c -o warshall.o
> warshall.c
> /opt/cray/pe/cce/8.5.7/cray-binutils/x86_64-pc-linux-gnu/bin/ld: cannot
> find -lmpi
> main.o: In function `create_file_names':
> /p/home/tschutt7/Software/amber16/AmberTools/src/byacc/main.c:268: warning:
> the use of `mktemp' is dangerous, better use `mkstemp'
> Makefile:34: recipe for target '/p/home/tschutt7/Software/amber16/bin/yacc'
> failed
> make[2]: *** [/p/home/tschutt7/Software/amber16/bin/yacc] Error 1
> make[2]: Leaving directory '/p/home/tschutt7/Software/
> amber16/AmberTools/src/byacc'
> Makefile:145: recipe for target 'cray_parallel' failed
> make[1]: *** [cray_parallel] Error 2
> make[1]: Leaving directory '/p/home/tschutt7/Software/
> amber16/AmberTools/src'
> Makefile:7: recipe for target 'install' failed
> make: *** [install] Error 2
>
> going to the sander directory - "make install" returns:
> make: *** No rule to make target 'cray_parallel', needed by 'install'. Stop.
>
> and attempting "make parallel" returns:
> ....[lots of lines]...
> ftn -DBINTRAJ -DMPI -DSANDER -c -O0 -fPIC -O0 -emf -f free
> -I/p/home/tschutt7/Software/amber16/include -I/opt/cray/pe/netcdf-
> hdf5parallel/4.4.1.1/CRAY/8.3/include -o np_force.SANDER.o np_force.F90
> ftn -DBINTRAJ -DMPI -DSANDER -c -fPIC -O0 -emf -f free
> -I/p/home/tschutt7/Software/amber16/include -I/opt/cray/pe/netcdf-
> hdf5parallel/4.4.1.1/CRAY/8.3/include -I../sander \
> -o density.SANDER.o density.F90
>
> _REAL_ function density_atom(dist)
> ^
> ftn-828 crayftn: ERROR DENSITY_ATOM, File = density.F90, Line = 763, Column
> = 30
> Procedure "DENSITY_ATOM" is defined in multiple places. Argument 1 has
> INTENT(IN) in one definition but not the other.
>
> Makefile:254: recipe for target 'density.SANDER.o' failed
> make[1]: *** [density.SANDER.o] Error 1
> make[1]: Leaving directory '/p/home/tschutt7/Software/
> amber16/AmberTools/src/pbsa'
> Makefile:492: recipe for target 'libpbsa' failed
> make: *** [libpbsa] Error 2
>
>
> I don't know if this is cray support issues in general or if our new
> machine is just That awkward or... Regardless, Thanks for helping! Let me
> know your thoughts on the most promising of the above three approach(es)?
>
> On Thu, Sep 21, 2017 at 3:46 PM, David A Case <david.case.rutgers.edu>
> wrote:
>
>> Sending this again, because my laptop crashed just as I was mailing this,
>> and I'm not sure if it was really mailed or not.....dac
>>
>>
>> ----- Forwarded message from David A Case <david.case.rutgers.edu> -----
>>
>> Date: Thu, 21 Sep 2017 22:43:37 +0200
>> From: David A Case <david.case.rutgers.edu>
>> To: AMBER Mailing List <amber.ambermd.org>
>> Subject: Re: [AMBER] Is it possible to compile without linking C++ and
>> fortran?
>>
>> On Thu, Sep 21, 2017, Timothy Schutt wrote:
>> >
>> > Sorry to be a hassle again. Still trying to get Amber to work on this
>> Cray
>> > XC40/50 system. Is it possible to compile amber without C++ linking to
>> > fortran, perhaps without cuda or EMIL features?
>>
>> Sorry crays are giving you so much trouble! You can use the "-noemil"
>> flag to configure to turn off emil compiling. Then, after configure try
>> going directly to AmberTools/src/sander and typing "make install" (or
>> perhaps "make parallel" -- I'm travelling right now. That should(!?!) get
>> you sander.MPI. Please don't be shy about letting us know whether this
>> does
>> or does not work....
>>
>> I'm not sure what to do about the cpptraj.MPI problem: maybe Dan will have
>> a good suggestion.
>>
>> >
>> > Error! Could not link C++ mpi code using mpi fortran linker:
>> > "ftn "
>> > [C++ / fortran / MPI] cross-linking is needed for cuda_parallel and EMIL
>> > builds.
>> > Try: "mpif90 -show"
>> > Configure failed due to the errors above!
>>
>> If you are lucky, adding the -noemil flag will eliminate this error.
>> If not, you might try going
>> to AMBERHOME/AmberTools/src, edit the configure2 script so that it doesn't
>> exit after the above message.
>>
>> > Is there a way to turn off cuda_parallel and EMIL builds? (I'm not asking
>> > for cuda in the configure flags but it's still partly trying apparently?)
>> > At the moment mostly I just need sander.mpi and cpptraj?
>>
>> As an aside: thanks for sending the information you have so far: it's
>> pretty
>> clear that we have let Cray support wither (wasn't that long ago that
>> things
>> worked there....)
>>
>> ....dac
>>
>>
>> ----- End forwarded message -----
>>
>> _______________________________________________
>> 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 Fri Sep 22 2017 - 07:30:01 PDT
