[gmx-users] Problem with Gromacs installation with GPU.

jesmin jahan shraban03 at gmail.com
Tue Aug 21 18:21:42 CEST 2012

Dear All,

I have installed gromacs 4.5.3  on a cluster. I downloaded the
gromacs-4.5-GPU-beta2-X86_64 binaries and followed the following

0. Prerequisites:
    - OpenMM (included in the binary release)
    - NVIDIA CUDA libraries (version >=3.0);
    - NVIDIA driver (for details on compatiblity consult
    - NVIDIA CUDA-enabled GPU (for compatiblity list see

1. Download and unpack the binary package for the respective
OS and architecture. Copy the content of the package to your
normal GROMACS installation directory (or to a custom location).
Note that as the distributed Gromacs-GPU packages do not contain
the entire set of tools and utilities included in a full GROAMCS
installation. Therefore, it is recommended to have a ≥v4.5
standard Gromacs installation along the GPU accelerated one.
e.g. on unix:

tar -xvf gromacs-4.5-GPU.tar.gz

2. Add the openmm/lib directory to your library path, e.g. in bash:


If there are other OpenMM versions installed, make sure that the
supplied libraries have preference when running mdrun-gpu.
Also, make sure that the CUDA libraries installed match the version
of CUDA with which GROMACS-GPU is compiled.

3. Set the OPENMM_PLUGIN_DIR environment variable to contain the path
to the openmm/lib/plugins directory, e.g. in bash:

export OPENMM_PLUGIN_DIR=PATH_TO_GROMACS/openmm/lib/plugins

4. At this point, running the command:

PATH_TO_GROMACS/bin/mdrun-gpu -h

should display the standard mdrun help which means that all the
necessary libraries are accessible.


But running the PATH_TO_GROMACS/bin/mdrun-gpu -h command gives me

/home1/01945/jesmin/Gromacs/4.5.3/bin/mdrun-gpu: error while loading
shared libraries: libcudart.so.3: cannot open shared object file: No
such file or directory

So, if I actually look into the cuda library directory I found the
following files:

libcublas.so        libcudart.so        libcufft.so        libcuinj.so
       libcurand.so        libcusparse.so        libnpp.so
libcublas.so.4      libcudart.so.4      libcufft.so.4
libcuinj.so.4      libcurand.so.4      libcusparse.so.4
libcublas.so.4.2.9  libcudart.so.4.2.9  libcufft.so.4.2.9
libcuinj.so.4.2.9  libcurand.so.4.2.9  libcusparse.so.4.2.9

Is there anyone who knows about how to fix this problem?

Please let me know.

Then, I installed the Gromacs-openmm precompiled version which used
cuda 4.1  following the instructions

To install Gromacs-OpenMM, follow these steps.

1. Install Gromacs 4.

2. Install OpenMM.

3. Copy mdrun-openmm to gromacs/bin, where "gromacs" is the root
directory of your Gromacs installation (e.g. /usr/local/gromacs).

4. Copy params.agb to gromacs/share/gromacs/top.

5. Add the OpenMM lib directory to your library path (PATH on Windows,

6. If you have an Nvidia GPU, install CUDA
(http://www.nvidia.com/object/cuda_get.html).  Make sure the CUDA lib
directory is in your library path.

7. If you installed OpenMM in the default location (/usr/local/openmm
on Mac OS X or Linux, Program Files\OpenMM on Windows), no further
steps are required.  If you installed it in a different location, set
the environment variable OPENMM_PLUGIN_DIR to point to the OpenMM
plugin directory (e.g. /usr/local/openmm/lib/plugins).

Note that I do not have any /usr/local directory. my home is

Now if I type

mdrun-openmm -h, I get the following error:

/home1/01945/jesmin/cilk/lib64/libstdc++.so.6: version
`GLIBCXX_3.4.11' not found (required by

Can anyone help me to about how to solve this problem?

Jesmin Jahan Tithi
PhD Student, CS
Stony Brook University, NY-11790.

More information about the gromacs.org_gmx-users mailing list