Re: [AMBER] AMBER parallel version

From: ying yu <helen.yingyu.gmail.com>
Date: Sat, 16 Oct 2010 13:23:14 -0700

Thanks a lot,

I have no problem with serial version AMBER, also no problem with gcc
version mpi AMBER.

But only problem comes out when I try the intel compiler for the mpi AMBER.

I think my mpich2 is compiled by gcc not icc. I will try to recompile with
intel compiler.

Thanks,

Helen
On Sat, Oct 16, 2010 at 6:41 AM, Jason Swails <jason.swails.gmail.com>wrote:

> Hello,
>
>
> On Oct 16, 2010, at 12:40 AM, ying yu <helen.yingyu.gmail.com> wrote:
>
> > When I compile the mpi CPU version,
> >
> > In the $AMBERHOME/AmberTools/src
> >
> > ./configure –mpi gnu
> >
> > Make parallel
> >
> > The mpi version pmemd could be built successfully.
> >
> >
> >
> > But If I use:
> >
> > ./configure –mpi intel
> >
> > There is no error in this step,
> >
> > When I do : make parallel
> >
> > There will comes out:
>
> There are a lot of things that can go wrong here. First, does Amber
> compile ok in serial with the intel compilers? If not, it's likely that
> your environment is not set up correctly. If it does, then you can move on
> to the parallel compilation. However, the parallel build assumes several
> things. First, it assumes that you have a working MPI implementation that
> conforms to the MPI standard, such as openMPI, MPICH2, etc. Finally, and
> this is where I think you are getting hit, it assumes that mpif90 and mpicc
> invoke the compilers you specified on the command line. You can check which
> compilers are used by the mpi wrappers by typing "mpif90 -show" and "mpicc
> -show" into the terminal. If you see that it is invoking gfortran and/or
> gcc, then you have a problem. Typically, you need to maintain two separate
> MPI implementations to use two different compilers for parallel amber. You
> also have to take care to keep them separate so they don't clobber one
> another.
>
> You do, of course, have to do a make clean like Bill said before starting a
> new compilation, or it will attempt to use old object files created by the
> old compilers.
>
> Hope this helps,
> Jason
>
> >
> >
> >
> > gcc: language Host not recognized
> >
> > gcc: language Host not recognized
> >
> > gcc: rand2.c: linker input file unused because linking not done
> >
> > mpicc -c -Dflex -ip -O3 -xHost -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> > -DBINTRAJ -DMPI -o sasad.o sasad.c
> >
> > gcc: language Host not recognized
> >
> > gcc: language Host not recognized
> >
> > gcc: sasad.c: linker input file unused because linking not done
> >
> > flex -t mm_options.l | sed -e 's/{stdout}/{DONTchangeSTDOUThere}/' -e
> > 's/yy/mmo/g' -e 's/stdout/nabout/' -e 's/DONTchangeSTDOUThere/stdout/' >
> > lex.mm_options.c
> >
> > mpicc -c -Dflex -ip -O3 -xHost -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> > -DBINTRAJ -DMPI -o sff.o sff.c
> >
> > gcc: language Host not recognized
> >
> > gcc: language Host not recognized
> >
> > gcc: sff.c: linker input file unused because linking not done
> >
> > mpicc -c -Dflex -ip -O3 -xHost -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> > -DBINTRAJ -DMPI -o time.o time.c
> >
> > gcc: language Host not recognized
> >
> > gcc: language Host not recognized
> >
> > gcc: time.c: linker input file unused because linking not done
> >
> > mpicc -c -Dflex -ip -O3 -xHost -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> > -DBINTRAJ -DMPI -o traceback.o traceback.c
> >
> > gcc: language Host not recognized
> >
> > gcc: language Host not recognized
> >
> > gcc: traceback.c: linker input file unused because linking not done
> >
> > mpicc -c -Dflex -ip -O3 -xHost -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> > -DBINTRAJ -DMPI -o xminC.o xminC.c
> >
> > gcc: language Host not recognized
> >
> > gcc: language Host not recognized
> >
> > gcc: xminC.c: linker input file unused because linking not done
> >
> > rm -f libsff.a
> >
> > ar rv libsff.a binpos.o conjgrad.o lmodC.o memutil.o nblist.o newton.o
> > nmode.o prm.o rand2.o sasad.o sff.o time.o traceback.o xminC.o
> >
> > ar: creating libsff.a
> >
> > ar: binpos.o: No such file or directory
> >
> > make[2]: *** [libsff.a] Error 1
> >
> > make[2]: Leaving directory
> > `/home/yiyu/Amber-withlicense/amber/AmberTools/src/sff'
> >
> > make[1]: *** [libsff] Error 2
> >
> > make[1]: Leaving directory
> > `/home/yiyu/Amber-withlicense/amber/AmberTools/src/nab'
> >
> > make: *** [parallel] Error 2
> >
> >
> >
> > seems it doesn’t use icc as the compiler, did I miss something?
> >
> >
> >
> > Thanks,
> >
> >
> >
> > Ying
> >
> > --
> > Happy Helen
> >
> >
> >
> > --
> > Happy Helen
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
>
> --
> Jason Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Graduate Student
> 352-392-4032
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
Happy Helen
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Oct 16 2010 - 13:30:03 PDT
Custom Search