Installation Guide

Contents:

System Requirements for executing HOOMD on a GPU

  • OS: Linux, Mac OS X, Windows Vista, or Windows XP
    • Note: Mac OS X versions older than 10.5.4 are not supported
  • CPU: Any x86 or x86_64 processor
  • RAM: At least 1GB is recommended
  • GPU: Any CUDA capable card: http://www.nvidia.com/object/cuda_home.html
    • Recommended (listed here in the rough order of performance expected, see http://www.ameslab.gov/hoomd/benchmarks.html for actual performance numbers):
      • GTX 280 (fastest)
      • Tesla 10 series
      • GTX 260
      • Tesla 800 series
      • 8800 GTX
      • 8800 GTS 512MB
      • 8800 GT
    • NOTE! Some board manufacturers sell Factory Over-clocked boards. These may work fine in games where a few bit errors just change the color of an onscreen pixel slightly, but GPGPU applications are extremely sensitive to such errors. HOOMD and other CUDA apps are known to crash on over-clocked GPUs.
  • Motherboard: 16x PCIe slot for the graphics card.
    • If you are installing the card in a second slot, double check you motherboard manual first. Many motherboards will decrease the slots to 8x with 2 cards installed. While HOOMD will still run in such a configuration, it is not recommended for performance reasons.
  • Power supply: Check with the GPU manufacturer for the power supply requirements for your GPU. If you have a store-bought system, its power supply is likely inadequate and will require upgrading.

Installing HOOMD in Windows

  1. Download the HOOMD installer from http://www.ameslab.gov/hoomd
  2. Install Python 2.5 using the installer at http://www.python.org/download/
  3. Install the corresponding drivers for your GPU. Go to http://www.nvidia.com and select download drivers.
  4. Uninstall any previous version of HOOMD before continuing
  5. Double click on the HOOMD installer and follow the on screen steps

HOOMD should now show up on your Start menu and the .hoomd file type is registered to execute the script when you double click on one. The command hoomd can also be run on the command line to start the HOOMD python interpreter.

The start menu includes links to benchmark scripts for checking the performance of your setup as well as some live demos that are fun to watch. In order to visualize the demos, you must have VMD 1.8.6 installed: http://www.ks.uiuc.edu/Research/vmd/

Check out the Quick Start Tutorial to learn how to use HOOMD.

Installing HOOMD in Linux

  1. Before you download HOOMD, check which python version you have by running 
     python -V
    in a terminal.
  2. Download the matching HOOMD package from http://www.ameslab.gov/hoomd
  3. Install the NVIDIA CUDA Toolkit using the installer from http://www.nvidia.com/object/cuda_get.html
    • Note: Make sure you install the toolkit version listed on the HOOMD download page.
    • Add /usr/local/cuda/lib to LD_LIBRARY_PATH (can be modified similar to PATH below)
  4. Install the corresponding drivers for your GPU. Go to http://www.nvidia.com and select download drivers.
    • Note: The drivers must match with the CUDA toolkit you installed. The HOOMD download page lists compatible driver version(s).
  5. Extract the HOOMD package using your favorite tool or by running 
     tar -xvjpf hoomd-0.7.1-Linux-x86_64-Python2.5.tar.bz2
    in a terminal (modify the filename to match the one you downloaded of course).
    • Note: If you extract this package in /opt as root, it will be installed in a system directory for any user on the system to run. If you prefer/need to have your own local copy, just extract to your home directory.
  6. Below, replace EXTRACTED_LOCATION with the location you extracted HOOMD to, i.e. /home/joaander/software/hoomd-0.7.1-Linux-x86_64-Python2.5
  7. Add EXTRACTED_LOCATION/bin to your $PATH. See http://www.newlinuxuser.com/howto-add-a-directory-to-my-path-statementvariable/ for instructions. You will probably need to log out and back in for the path setting to take effect.

After installation is complete, run 

hoomd
in any terminal. You should see something like this: 
Python 2.4.4 (#1, Nov  4 2007, 17:29:36)
[GCC 4.1.2 (Gentoo 4.1.2)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>>
Press CTRL-D to exit.

To test your installation further, you can execute benchmark scripts th check the performance of your setup or run some live demos that are fun to watch. You will find them in EXTRACTED_LOCATION/share/hoomd For example, to run the polymer benchmark on the GPU: 

cd EXTRACTED_LOCATION/share/hoomd/benchmarks/
./polymer_bmark.hoomd --mode=gpu

Note: to visualize the live demos, you need to have VMD 1.8.6 installed: http://www.ks.uiuc.edu/Research/vmd/

Check out the Quick Start Tutorial to learn how to use HOOMD.

Installing HOOMD in Mac OS X

  1. Download the HOOMD package from http://www.ameslab.gov/hoomd
  2. Install the NVIDIA CUDA Toolkit using the installer from http://www.nvidia.com/object/cuda_get.html
    • Note: Make sure you install the toolkit version listed on the HOOMD download page.
    • Note: Make sure to select the Customize button and ensure that CUDAKext is selected.
    • Add /usr/local/cuda/lib to DYLD_LIBRARY_PATH (see instructions for PATH below)
  3. Make sure that the Apple Software Update tool has installed all updates related to the video drivers.
  4. Extract the HOOMD package using your favorite tool or by running 
     tar -xvjpf hoomd-0.8.0-Darwin-i386-Python2.5.tar.bz2
    in a terminal (modify the filename to match the one you downloaded of course).
    • Note: If you extract this package in /opt as root, it will be installed in a system directory for any user on the system to run. If you prefer/need to have your own local copy, just extract to your home directory.
  5. Below, replace EXTRACTED_LOCATION with the location you extracted HOOMD to, i.e. /home/joaander/software/hoomd-0.8.0-Darwin-i386-Python2.5
  6. Add EXTRACTED_LOCATION/bin to your $PATH. See http://www.newlinuxuser.com/howto-add-a-directory-to-my-path-statementvariable/ for instructions. You will probably need to log out and back in for the path setting to take effect.

After installation is complete, run 

hoomd
in any terminal. You should see something like this: 
Python 2.5.1 (r251:54863, Apr 15 2008, 22:57:26) 
[GCC 4.0.1 (Apple Inc. build 5465)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>>
Press CTRL-D to exit.

To test your installation further, you can execute benchmark scripts th check the performance of your setup or run some live demos that are fun to watch. You will find them in EXTRACTED_LOCATION/share/hoomd For example, to run the polymer benchmark on the GPU: 

cd EXTRACTED_LOCATION/share/hoomd/benchmarks/
./polymer_bmark.hoomd --mode=gpu

Note: to visualize the live demos, you need to have VMD 1.8.6 installed: http://www.ks.uiuc.edu/Research/vmd/

Check out the Quick Start Tutorial to learn how to use HOOMD.

Generated on Mon Dec 22 10:16:19 2008 for HOOMD by doxygen 1.5.7.1

Home