[gmx-users] configuration, make, installation problem

[Mark Abraham]

Nevena Todorova wrote:
> Hi there,
> i am trying to install (double precision, mpi )Gromacs 3.3 on an SGI Altix system and during this process i encounter few problems.

I observe that you are trying to install 3.3 on the APAC ac machine. If 
you want a double precision MPI version, I'm sure the admins will do 
that for you in no time flat - they've always done so for me, as 
evidenced by the plethora of pre-release 3.3 versions they still have 
lying around! (see "module avail") Unless you need 3.3, you may as well 
ask for 3.3.1 too!

> I configure the system using the commmand:
> ./configure --enable-mpi --enable-shared --disable-nice --disable-float --prefix=$HOME/g-3.3 --program-suffix=_d_mpi --with-fft=fftw2
> 
> and  it seems to be running without problems untill the end where it stops without an output line for a successful configuration. It is not giving me an errors for missing programs, or environment set ups.
> The end look like this:
> "config.status: creating share/top/Makefile
> config.status: creating share/html/Makefile
> config.status: creating share/html/images/Makefile
> config.status: creating share/html/online/Makefile
> config.status: creating man/Makefile
> config.status: creating man/man1/Makefile
> config.status: creating src/config.h
> config.status: executing depfiles commands

That is a normal finish.

> If i try to run "make" after this, it is compiling for some time until it stops with the following output:
> ": multiple definition of `fftwnd_f77_destroy_plan_'
> nonbonded/.libs/libnonbonded.a(fftwf77.o)(.text+0x680): first defined here
> /opt/fftw-2.1.5/intel-9.0/lib/libdfftw.a(fftwf77.o)(.text+0x700): In function `fftwnd_f77_':
> : multiple definition of `fftwnd_f77_'

My guess is that all of this gibberish is some kind of library/compiler 
unavailablity or incompatibility. You want to be sure you have loaded a 
compatible set of modules for the compiler and fftw library.

I use modules intel-cc/9.0.031 and fftw/2.1.5-intel-9.0 and scsl/1.6.0.0 
(the latter means you are using linear algebra stuff optimized for this 
machine instead of generic BLAS/LAPACK... not hugely important though) 
and for non-MPI have used

./configure --prefix=/home/224/mxa224/progs --with-fft=fftw2 
--without-motif --with-external-blas --with-external-lapack 
CPPFLAGS=-I/opt/fftw-2.1.5/intel-9.0/include 
LDFLAGS='-L/opt/fftw-2.1.5/intel-9.0/lib/ -lscs'

and for MPI have used

./configure --prefix=/home/224/mxa224/progs --with-fft=fftw2 
--without-motif --with-external-blas --with-external-lapack 
CPPFLAGS=-I/opt/fftw-2.1.5/intel-9.0/include 
LDFLAGS='-L/opt/fftw-2.1.5/intel-9.0/lib/ -lscs -lmpi' 
--program-suffix=_mpi --enable-mpi

all successfully.

Mark