AMBER Archive (2009)

Subject: RE: Re: Re: [AMBER] Not able to compile pmemd with openmpi in Amber9

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Wed Nov 25 2009 - 14:24:23 CST

Hi Kito,

MKL Helps with performance somewhat, although more with implicit solvent
than explicit solvent. Intel continually, for reasons known only to them,
change the way one has to link to the MKL libraries. Thus when AMBER 9 was
released it used the correct linking. New versions of MKL since then have
changed things which has broken this. AMBER 10 have been updated to work
with newer MKL versions via bugfixes but AMBER 9 was not.

Your options are therefore to use AMBER 10 with the latest bug fixes or
modify the link line to contain the correct linking for MKL (or just do away
with MKL as you did).

Which MKL version are you using?

for 9.0 this should be:

libmkl_lapack.a libmkl_em64t.a

For 10.x or later (although they may have changed it AGAIN for 11.x) it
should be:

libmkl_intel_lp64.a libmkl_sequential.a libmkl_core.a

Also make sure you run the test cases. Each new version of the Intel
compilers introduce new compiler bugs which can lead to incorrect answers
being generated.

11.1.046 is pretty bleeding edge. I still use 10.1.018 since I know it works
and I trust it to get the answers correct.

Good luck,
Ross

> -----Original Message-----
> From: amber-bounces_at_ambermd.org [mailto:amber-bounces_at_ambermd.org] On
> Behalf Of Jason Swails
> Sent: Wednesday, November 25, 2009 8:51 AM
> To: AMBER Mailing List
> Subject: Re: Re: Re: [AMBER] Not able to compile pmemd with openmpi in
> Amber9
>
> Kito,
>
> You want to be careful here. Run the tests to make sure it worked.
> However, that is not what I suggested. try using this config.h file
> instead:
>
> MATH_DEFINES =
>
> MATH_LIBS =
>
> IFORT_RPATH = /opt/gridengine/lib/lx26-
> amd64:/usr/lib/perl5/5.8.8
>
> MATH_DEFINES = -DMKL
>
> MATH_LIBS = -L$(MKL_HOME)/lib/em64t -lguide -lpthread -lmkl_core
> -lmkl_sequential -lmkl_intel_lp64
>
> FFT_DEFINES = -DPUBFFT
>
> FFT_INCLUDE =
>
> FFT_LIBS =
>
> NETCDF_HOME =
>
> NETCDF_DEFINES =
>
> NETCDF_MOD =
>
> NETCDF_LIBS =
>
> MPI_HOME = /opt/mpi/openmpi/1.3.3/intel/
>
> MPI_DEFINES = -DMPI -DSLOW_NONBLOCKING_MPI
>
> MPI_INCLUDE = -I$(MPI_HOME)/include
>
> MPI_LIBDIR = $(MPI_HOME)/lib
>
> MPI_LIBS = -L$(MPI_LIBDIR)
>
> DIRFRC_DEFINES = -DDIRFRC_EFS -DDIRFRC_COMTRANS -DDIRFRC_NOVEC
>
> CPP = /lib/cpp
>
> CPPFLAGS = -traditional -P
>
> F90_DEFINES = -DFFTLOADBAL_2PROC
>
> F90 = mpif90
>
> MODULE_SUFFIX = mod
>
> F90FLAGS = -c
>
> F90_OPT_DBG = -g -traceback
>
> F90_OPT_LO = -O0
>
> F90_OPT_MED = -O2
>
> F90_OPT_HI = -axWPT -ip -O3
>
> F90_OPT_DFLT = $(F90_OPT_HI)
>
> CC = mpicc
>
> CFLAGS =
>
> LOAD = mpif90
>
> LOADFLAGS =
>
> LOADLIBS = -limf -lsvml -Wl,-rpath=$(IFORT_RPATH)
>
>
> What you did was to prevent the pmemd installation from using the intel
> MKL
> at all. However, without changing the MATH_LIBS line, you were still
> trying
> to link to lguide (which is an MKL library). When you removed both,
> you
> told pmemd not to use intel MKL. However, the intel MKL performs well
> and
> will probably make your installation of pmemd faster. Also, when you
> start
> changing preprocessor items (like removing -DMKL), you'll need to make
> clean
> before you try to compile with the new options. Since you're not even
> compiling with MKL, you should not need libguide, so linking it in the
> MATH_LIBS line is unnecessary.
>
> Try making clean and using the above config.h file and see if that
> works.
> Also, don't forget to run the test cases to make sure you have a
> working
> executable.
>
> Good luck!
> Jason
>
>
> On Wed, Nov 25, 2009 at 3:12 AM, Kito <kitobhai_at_rediffmail.com> wrote:
>
> > I was able to solve the problem
> >
> > "/usr/lib64/libguide.so: no version information available (required
> by
> > ./pmemd)"
> >
> > It looked like the "no version information available" means that the
> > library version number is lower on the shared object so I created a
> soft
> > link of /opt/intel/Compiler/11.1/046/lib/intel64/libguide.so in
> /usr/lib64
> > and it worked.
> >
> >
> >
> > Thanks all
> >
> >
> >
> >
> >
> >
> >
> > On Wed, 25 Nov 2009 13:21:41 +0530 wrote
> >
> > >Thanks Jason,
> >
> >
> >
> > When I changed MATH_DEFINES = -DMKL
> >
> >
> >
> > to MATH_DEFINES =
> >
> >
> >
> > it worked and I was able to link the object files.
> >
> >
> >
> > But on checking library dependencies, I have problem with libguide.so
> > (first line on the output below)
> >
> >
> >
> >
> >
> >
> >
> > $> ldd pmemd
> >
> >
> >
> > ./pmemd: /usr/lib64/libguide.so: no version information available
> (required
> > by ./pmemd)
> >
> >
> >
> > libguide.so => /usr/lib64/libguide.so (0x00002b1a6e4ca000)
> >
> >
> >
> > libpthread.so.0 => /lib64/libpthread.so.0 (0x00000032fdc00000)
> >
> >
> >
> > libimf.so => /usr/lib64/libimf.so (0x00002b1a6e607000)
> >
> >
> >
> > libsvml.so => /usr/lib64/libsvml.so (0x00002b1a6e89f000)
> >
> >
> >
> > libmpi_f90.so.0 =>
> /opt/mpi/openmpi/1.3.3/intel/lib/libmpi_f90.so.0
> > (0x00002b1a6e9e1000)
> >
> >
> >
> > libmpi_f77.so.0 =>
> /opt/mpi/openmpi/1.3.3/intel/lib/libmpi_f77.so.0
> > (0x00002b1a6ebe4000)
> >
> >
> >
> > libmpi.so.0 => /opt/mpi/openmpi/1.3.3/intel/lib/libmpi.so.0
> > (0x00002b1a6ee1c000)
> >
> >
> >
> > libopen-rte.so.0 => /opt/mpi/openmpi/1.3.3/intel/lib/libopen-
> rte.so.0
> > (0x00002b1a6f0ec000)
> >
> >
> >
> > libopen-pal.so.0 => /opt/mpi/openmpi/1.3.3/intel/lib/libopen-
> pal.so.0
> > (0x00002b1a6f34c000)
> >
> >
> >
> > libdl.so.2 => /lib64/libdl.so.2 (0x00000032fd800000)
> >
> >
> >
> > libnsl.so.1 => /lib64/libnsl.so.1 (0x0000003300400000)
> >
> >
> >
> > libutil.so.1 => /lib64/libutil.so.1 (0x000000330a600000)
> >
> >
> >
> > libm.so.6 => /lib64/libm.so.6 (0x00000032fd400000)
> >
> >
> >
> > libc.so.6 => /lib64/libc.so.6 (0x00000032fd000000)
> >
> >
> >
> > libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x000000330ee00000)
> >
> >
> >
> > /lib64/ld-linux-x86-64.so.2 (0x00000032fcc00000)
> >
> >
> >
> > libirc.so => /usr/lib64/libirc.so (0x00002b1a6f5e8000)
> >
> >
> >
> > libifport.so.5 =>
> > /opt/intel/Compiler/11.1/046/lib/intel64/libifport.so.5
> (0x00002b1a6f71e000)
> >
> >
> >
> > libifcoremt.so.5 =>
> > /opt/intel/Compiler/11.1/046/lib/intel64/libifcoremt.so.5
> > (0x00002b1a6f857000)
> >
> >
> >
> > libintlc.so.5 =>
> /opt/intel/Compiler/11.1/046/lib/intel64/libintlc.so.5
> > (0x00002b1a6fafb000)
> >
> >
> >
> >
> >
> >
> >
> > Thanks again
> >
> >
> >
> > Kito
> >
> >
> >
> >
> >
> >
> >
> > On Wed, 25 Nov 2009 12:34:22 +0530 wrote
> >
> >
> >
> > >This appears to be an MKL issue. What version of MKL are you using?
> It
> >
> >
> >
> > would appear like the MKL libraries you're linking to no longer
> contain
> >
> >
> >
> > vd___MATHFUNCTION__ functions/subroutines anymore, so you have to
> find
> > which
> >
> >
> >
> > library is missing and link to that one as well. You can try adding
> >
> >
> >
> > -lmkl_intel_lp64 -lmkl_core and/or -lmkl_sequential to the MATH_LIBS
> line.
> >
> >
> >
> > I believe those are the only three libraries that pmemd will need,
> but this
> >
> >
> >
> > is for Intel MKL 10 or 11. For older MKL versions, the config file
> you have
> >
> >
> >
> > above should work...
> >
> >
> >
> >
> >
> >
> >
> > I just now tried compiling pmemd9 with the ifort11 compiler and MKL
> 11. It
> >
> >
> >
> > failed with the default config.h file (slightly different errors than
> > yours,
> >
> >
> >
> > but still MKL-related). Adding the three above -lmkl... statements
> above to
> >
> >
> >
> > MATH_LIBS fixed the problems.
> >
> >
> >
> >
> >
> >
> >
> > Again, though, this is dependent on the version of MKL that you're
> using.
> >
> >
> >
> >
> >
> >
> >
> > Good luck, and I'd be interested to hear any success/failure (along
> with
> >
> >
> >
> > which MKL you have installed :) )
> >
> >
> >
> >
> >
> >
> >
> > All the best,
> >
> >
> >
> > Jason
> >
> >
> >
> >
> >
> >
> >
> > On Wed, Nov 25, 2009 at 1:39 AM, Kito wrote:
> >
> >
> >
> >
> >
> >
> >
> > > Hi Amber Experts/Users,
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > > I am having difficulties with pmemd installation
> >
> >
> >
> > >
> >
> >
> >
> > > My h/w and s/w configuration is
> >
> >
> >
> > >
> >
> >
> >
> > > Linux Enterprise, OpenMPI, AMD x86_64, Intel Compiliers
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > > My config.h file is
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > > MATH_DEFINES =
> >
> >
> >
> > >
> >
> >
> >
> > > MATH_LIBS =
> >
> >
> >
> > >
> >
> >
> >
> > > IFORT_RPATH = /opt/gridengine/lib/lx26-amd64:/usr/lib/perl5/5.8.8
> >
> >
> >
> > >
> >
> >
> >
> > > MATH_DEFINES = -DMKL
> >
> >
> >
> > >
> >
> >
> >
> > > MATH_LIBS = -L$(MKL_HOME)/lib/em64t -lguide -lpthread
> >
> >
> >
> > >
> >
> >
> >
> > > FFT_DEFINES = -DPUBFFT
> >
> >
> >
> > >
> >
> >
> >
> > > FFT_INCLUDE =
> >
> >
> >
> > >
> >
> >
> >
> > > FFT_LIBS =
> >
> >
> >
> > >
> >
> >
> >
> > > NETCDF_HOME =
> >
> >
> >
> > >
> >
> >
> >
> > > NETCDF_DEFINES =
> >
> >
> >
> > >
> >
> >
> >
> > > NETCDF_MOD =
> >
> >
> >
> > >
> >
> >
> >
> > > NETCDF_LIBS =
> >
> >
> >
> > >
> >
> >
> >
> > > MPI_HOME = /opt/mpi/openmpi/1.3.3/intel/
> >
> >
> >
> > >
> >
> >
> >
> > > MPI_DEFINES = -DMPI -DSLOW_NONBLOCKING_MPI
> >
> >
> >
> > >
> >
> >
> >
> > > MPI_INCLUDE = -I$(MPI_HOME)/include
> >
> >
> >
> > >
> >
> >
> >
> > > MPI_LIBDIR = $(MPI_HOME)/lib
> >
> >
> >
> > >
> >
> >
> >
> > > MPI_LIBS = -L$(MPI_LIBDIR)
> >
> >
> >
> > >
> >
> >
> >
> > > DIRFRC_DEFINES = -DDIRFRC_EFS -DDIRFRC_COMTRANS -DDIRFRC_NOVEC
> >
> >
> >
> > >
> >
> >
> >
> > > CPP = /lib/cpp
> >
> >
> >
> > >
> >
> >
> >
> > > CPPFLAGS = -traditional -P
> >
> >
> >
> > >
> >
> >
> >
> > > F90_DEFINES = -DFFTLOADBAL_2PROC
> >
> >
> >
> > >
> >
> >
> >
> > > F90 = mpif90
> >
> >
> >
> > >
> >
> >
> >
> > > MODULE_SUFFIX = mod
> >
> >
> >
> > >
> >
> >
> >
> > > F90FLAGS = -c
> >
> >
> >
> > >
> >
> >
> >
> > > F90_OPT_DBG = -g -traceback
> >
> >
> >
> > >
> >
> >
> >
> > > F90_OPT_LO = -O0
> >
> >
> >
> > >
> >
> >
> >
> > > F90_OPT_MED = -O2
> >
> >
> >
> > >
> >
> >
> >
> > > F90_OPT_HI = -axWPT -ip -O3
> >
> >
> >
> > >
> >
> >
> >
> > > F90_OPT_DFLT = $(F90_OPT_HI)
> >
> >
> >
> > >
> >
> >
> >
> > > CC = mpicc
> >
> >
> >
> > >
> >
> >
> >
> > > CFLAGS =
> >
> >
> >
> > >
> >
> >
> >
> > > LOAD = mpif90
> >
> >
> >
> > >
> >
> >
> >
> > > LOADFLAGS =
> >
> >
> >
> > >
> >
> >
> >
> > > LOADLIBS = -limf -lsvml -Wl,-rpath=$(IFORT_RPATH)
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > > The error I get is
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > > mpif90 -o pmemd gbl_constants.o gbl_datatypes.o state_info.o
> > file_io_dat.o
> >
> >
> >
> > > parallel_dat.o mdin_ctrl_dat.o mdin_ewald_dat.o prmtop_dat.o
> inpcrd_dat.o
> >
> >
> >
> > > dynamics_dat.o img.o parallel.o pme_direct.o pme_recip.o pme_fft.o
> > fft1d.o
> >
> >
> >
> > > bspline.o pme_force.o pbc.o nb_pairlist.o cit.o dynamics.o bonds.o
> > angles.o
> >
> >
> >
> > > dihedrals.o runmd.o loadbal.o shake.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 alltasks_setup.o pme_setup.o
> >
> >
> >
> > > ene_frc_splines.o nextprmtop_section.o
> >
> >
> >
> > > -L/opt/intel/Compiler/11.1/046/mkl//lib/em64t -lguide -lpthread
> >
> >
> >
> > > -L/opt/mpi/openmpi/1.3.3/intel//lib -limf -lsvml
> >
> >
> >
> > > -Wl,-rpath=/opt/gridengine/lib/lx26-amd64:/usr/lib/perl5/5.8.8
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.o: In function `gb_ene_mod_mp_gb_ene_':
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x19de): undefined reference to `vdinvsqrt_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x1c93): undefined reference to `vdinv_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x1cac): undefined reference to `vdinv_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x1fde): undefined reference to `vdln_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x1ff7): undefined reference to `vdln_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x3b98): undefined reference to `vdinv_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x3dd9): undefined reference to `vdexp_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x4254): undefined reference to `vdinvsqrt_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x4289): undefined reference to `vdinv_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x43df): undefined reference to `vdexp_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x43fc): undefined reference to `vdinvsqrt_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x5403): undefined reference to `vdinvsqrt_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x55b1): undefined reference to `vdinv_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x55ca): undefined reference to `vdinv_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x57b2): undefined reference to `vdln_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.o: In function `gb_ene_mod_mp_calc_born_radii_':
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x678a): undefined reference to `vdinvsqrt_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x6a3f): undefined reference to `vdinv_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x6a58): undefined reference to `vdinv_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x6d8a): undefined reference to `vdln_'
> >
> >
> >
> > >
> >
> >
> >
> > > gb_ene.f90:(.text+0x6da3): undefined reference to `vdln_'
> >
> >
> >
> > >
> >
> >
> >
> > > make[1]: *** [pmemd] Error 1
> >
> >
> >
> > >
> >
> >
> >
> > > make[1]: Leaving directory `/opt/apps/amber/9/intel-
> mkl/src/pmemd/src'
> >
> >
> >
> > >
> >
> >
> >
> > > make: *** [all] Error 2
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > > Please help... I have gone through the archives and got similar
> problems
> >
> >
> >
> > > but their said solution's doesnt seems to help.
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > >
> >
> >
> >
> > > Thanks in advance
> >
> >
> >
> > >
> >
> >
> >
> > > Kito
> >
> >
> >
> > > _______________________________________________
> >
> >
> >
> > > AMBER mailing list
> >
> >
> >
> > > AMBER_at_ambermd.org
> >
> >
> >
> > > http://lists.ambermd.org/mailman/listinfo/amber
> >
> >
> >
> > >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > --
> >
> >
> >
> > ---------------------------------------
> >
> >
> >
> > Jason M. Swails
> >
> >
> >
> > Quantum Theory Project,
> >
> >
> >
> > University of Florida
> >
> >
> >
> > Ph.D. Graduate Student
> >
> >
> >
> > 352-392-4032
> >
> >
> >
> > _______________________________________________
> >
> >
> >
> > AMBER mailing list
> >
> >
> >
> > AMBER_at_ambermd.org
> >
> >
> >
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> >
> >
> > _______________________________________________
> >
> > AMBER mailing list
> >
> > AMBER_at_ambermd.org
> >
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER_at_ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
>
> --
> ---------------------------------------
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Graduate Student
> 352-392-4032
> _______________________________________________
> AMBER mailing list
> AMBER_at_ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber