Re: AMBER: problem to compile PMEMD?

From: Robert Duke <rduke.email.unc.edu>
Date: Tue, 29 Jan 2008 21:10:49 -0500

Okay, two possibilities come to mind:
1) the value you specified for MPI_HOME is not actually correct (seems
plausible but I can't tell for sure without system access). Enter 'which
mpif77' to check this out (and it should give you a path which minus the
bin/mpif77 is MPI_HOME.
2) the libraries actually needed may have changed a bit compared to the
mpich I had available when the configure script and data were written. To
check this out, enter 'mpif77 -link_info', and correct the MPI_LIBS line if
necessary.

I believe these suggestion are made in the README. Note that pmemd install
is not as bombproof as sander install; this is the case because 1) pmemd has
more targets, many of which may not be the build machine, 2) performance is
of utmost importance for pmemd, so I really don't try to do any
autoconfigure stuff, which would really screw up if you are building on a
machine that is not the run machine, and 3) time is mostly devoted to making
the code fast... In the case of sander, it actually uses link info
information to find the correct libraries, so looking at the library and
header locations and library names in the sander configure will probably
also be useful (though the mpif77 step should work). I would note, however,
that nothing is bombproof in terms of confuration (build mpi incorrectly,
not specifying the libraries needed and all bets are off - by the way, that
is another possibility here - did mpi get built with ifort specified as the
f77 and f90 compiler?). I deal with several facilities where standard mpi
files get renamed, put in the wrong locations, etc., and you have to go sort
out what happened, either for pmemd or sander...

Regards - Bob


----- Original Message -----
From: <fatima.chami.durham.ac.uk>
To: <amber.scripps.edu>
Sent: Tuesday, January 29, 2008 7:10 PM
Subject: Re: AMBER: problem to compile PMEMD?


> Sander.MPI runs smoothly but for some reasons PMEMD DOESNT EVEN COMPILE
> CORRECTLY !
> I dont know why it did not compile at least the uniprocessor pmemd.
> I didnt edit the config.h file which looks like:
> ===================================================================
> ./configure linux64_opteron ifort mpich
> Intel ifort compiler found; version information:
> Version 9.0
> Use Intel MKL? (better performance) (answer yes or no):
>
>
> ./configure: line 356: [: too many arguments
> ./configure: line 367: [: =: unary operator expected
> ./configure: line 376: [: =: unary operator expected
> File config_data/fft.pubfft being used...
> File config_data/linux64_opteron.ifort being used...
> File config_data/interconnect.mpich being used...
> configure assumes mpich files are in /usr/local/Cluster-
> Apps/mpich/mx/intel/64/1.2.6..0.94/.
> PMEMD Configurate successfully completed.
> ====================================================================
> more config.h :
> ---------------------------------------------------------------------------
> MATH_DEFINES =
> MATH_LIBS =
> IFORT_RPATH = /usr/local/lib:/usr/X11R6/lib:/usr/local/Cluster-
> Apps/sge/lib/lx26-amd64:/usr/local/Cluster-Apps/intel/cce/9.1
> .046/lib:/usr/local/Cluster-Apps/intel/fce/9.0//lib:/usr/local/Cluster-
> Apps/intel/idbe/9.0//lib:/usr/local/Cluster-Apps/mx/m
> x-1.1.1/lib64:/usr/local/Cluster-Apps/mpich/mx/intel/64/1.2.6..0.94//lib/shared
> FFT_DEFINES = -DPUBFFT
> FFT_INCLUDE =
> FFT_LIBS =
> NETCDF_HOME =
> NETCDF_DEFINES =
> NETCDF_MOD =
> NETCDF_LIBS =
> MPI_HOME = /usr/local/Cluster-Apps/mpich/mx/intel/64/1.2.6..0.94/
> MPI_DEFINES = -DMPI -DSLOW_NONBLOCKING_MPI
> MPI_INCLUDE = -I$(MPI_HOME)/include
> MPI_LIBDIR = $(MPI_HOME)/lib
> MPI_LIBS = -L$(MPI_LIBDIR) -lmpich
> DIRFRC_DEFINES = -DDIRFRC_EFS -DDIRFRC_NOVEC
> CPP = /lib/cpp
> CPPFLAGS = -traditional -P
> F90_DEFINES = -DFFTLOADBAL_2PROC
>
> F90 = ifort
> MODULE_SUFFIX = mod
> F90FLAGS = -c -auto
> F90_OPT_DBG = -g -traceback
> F90_OPT_LO = -tpp7 -O0
> F90_OPT_MED = -tpp7 -O2
> F90_OPT_HI = -tpp7 -xW -ip -O3
> F90_OPT_DFLT = $(F90_OPT_HI)
>
> CC = gcc
> CFLAGS =
>
> LOAD = ifort
> LOADFLAGS =
> LOADLIBS = -limf -lsvml -Wl,-rpath=$(IFORT_RPATH)
> hamilton [x86_64.linux] 173% vim config.h
> MATH_DEFINES =
> MATH_LIBS =
> IFORT_RPATH = /usr/local/lib:/usr/X11R6/lib:/usr/local/Cluster-
> Apps/sge/lib/lx26-amd64:/usr/local/Cluster-
> Apps/intel/cce/9.1.046/lib:/usr/local/Cluster-
> Apps/intel/fce/9.0//lib:/usr/local/Cluster-
> Apps/intel/idbe/9.0//lib:/usr/local/Cluster-Apps/mx/mx-
> 1.1.1/lib64:/usr/local/Cluster-Apps/mpich/mx/intel/64/1.2.6..0.94//lib/shared
> FFT_DEFINES = -DPUBFFT
> FFT_INCLUDE =
> FFT_LIBS =
> NETCDF_HOME =
> NETCDF_DEFINES =
> NETCDF_MOD =
> NETCDF_LIBS =
> MPI_HOME = /usr/local/Cluster-Apps/mpich/mx/intel/64/1.2.6..0.94/
> MPI_DEFINES = -DMPI -DSLOW_NONBLOCKING_MPI
> MPI_INCLUDE = -I$(MPI_HOME)/include
> MPI_LIBDIR = $(MPI_HOME)/lib
> MPI_LIBS = -L$(MPI_LIBDIR) -lmpich
> DIRFRC_DEFINES = -DDIRFRC_EFS -DDIRFRC_NOVEC
> CPP = /lib/cpp
> CPPFLAGS = -traditional -P
> F90_DEFINES = -DFFTLOADBAL_2PROC
>
> F90 = ifort
> MODULE_SUFFIX = mod
> F90FLAGS = -c -auto
> F90_OPT_DBG = -g -traceback
> F90_OPT_LO = -tpp7 -O0
> F90_OPT_MED = -tpp7 -O2
> F90_OPT_HI = -tpp7 -xW -ip -O3
> ------------------------------------------------------------------------------
> Quoting Robert Duke <rduke.email.unc.edu>:
>
>>
>> Not sure what you have done here. You configure for uniprocessor, and
>> then
>> are getting errors over mpi stuff not being defined. Did you hand-edit
>> the
>> config.h for pmemd to include -DMPI (not a good idea...)? Something else
>> nonstandard? Did you read the README under amber9/src/pmemd? If you
>> send
>> your config.h with info about your machine, I'll look at it, but you are
>> really not providing much info here, and to do the parallel build, it
>> really
>>
>> helps to have somebody at your facility that knows how to build/test mpi.
>> Now, if you got parallel sander running okay, it should be relatively
>> easy
>> to get pmemd running too...
>> Regards - Bob Duke
>>
>> ----- Original Message -----
>> From: <fatima.chami.durham.ac.uk>
>> To: <amber.scripps.edu>
>> Sent: Tuesday, January 29, 2008 5:45 PM
>> Subject: AMBER: problem to compile PMEMD?
>>
>>
>> > Dear Amber's users,
>> >
>> > I configured PMEMD but failed to install it !!
>> > I tried both nopar and mpich and got the same error
>> >
>> > any help !
>> > thanks
>> > fatima
>> >
>> > -------------------------------------------------------------
>> >
>> > ./configure linux64_opteron ifort nopar
>> > Intel ifort compiler found; version information:
>> > Version 9.0
>> > Use Intel MKL? (better performance) (answer yes or no):
>> > no
>> > File config_data/fft.pubfft being used...
>> > File config_data/linux64_opteron.ifort being used...
>> > File config_data/interconnect.nopar being used...
>> > PMEMD Configurate successfully completed.
>> > hamilton [x86_64.linux] 78% make install
>> > .....
>> > pmemd.f90:(.text+0x58): undefined reference to `mpi_init_'
>> > pmemd.f90:(.text+0x6e): undefined reference to `mpi_comm_rank_'
>> > pmemd.f90:(.text+0x84): undefined reference to `mpi_comm_size_'
>> > pmemd.f90:(.text+0x9a): undefined reference to `mpi_comm_group_'
>> > pmemd.f90:(.text+0x6ca): undefined reference to `mpi_bcast_'
>> > nmr_calls.o: In function `nmr_calls_mod_mp_bcast_nmr_dat_':
>> > nmr_calls.f90:(.text+0x37f): undefined reference to `mpi_bcast_'
>> > nmr_calls.f90:(.text+0x3a6): undefined reference to `mpi_bcast_'
>> > nmr_calls.f90:(.text+0x5f8): undefined reference to `mpi_bcast_'
>> > nmr_calls.f90:(.text+0x621): undefined reference to `mpi_bcast_'
>> > ene_frc_splines.o:ene_frc_splines.f90:(.text+0x1e0): more undefined
>> > references
>> > to `mpi_bcast_' follow
>> > make[1]: *** [pmemd] Error 1
>> > make[1]: Leaving directory
>> > `/data/hamilton/dch1fc/amber/amber9/src/pmemd/src'
>> > make: *** [install] Error 2
>> > -----------------------------------------------------------------------
>> > The AMBER Mail Reflector
>> > To post, send mail to amber.scripps.edu
>> > To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>> >
>>
>>
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber.scripps.edu
>> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>>
>>
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>


-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Jan 30 2008 - 06:07:32 PST
Custom Search