2. Installation

The GriDock package includes executables for Linux 32 bit, Linux 64 bit, Windows 32 bit and Windows 64 bit. Each executable exists in two versions: the standard parallel version for unified memory systems (e.g. single workstation/PC with one ore more CPUs/cores) and the MPI (MPICH2) version for grid arrays.

 

2.1 Linux installation

Before to proceed with the installation, you must have the following packages:

GriDock_X.X.X.tar.gz Generic multiplatform GriDock package.
Vega_X.X.X_Linux_86-32.tar.gz VEGA command-line for Linux 32 bit.
Vega_X.X.X_Linux_86-64.tar.gz VEGA command-line for Linux 64 bit. This package is the alternative of the previous one if your system has a 64 bit operating system.

where X.X.X is the package version. You can download all needed packages from www.vegazz.net or www.ddl.unimi.it. The pre-compiled GriDock executables were built by CentOS 4.3 and they require the libc version 6.

 

2.1.1 Standard Linux installation

You must choose this installation procedure if you want to run GriDock locally on a single node (e.g. HPC system with shared memory architecture). All installed CPUs can be used by GriDock.

  1. Install VEGA command-line as explained in its manual.
  2. Copy the gridock executable from GriDock/Linux_86-32 or GriDock/Linux_86-64 folder of GriDock_X.X.X.tar.gz archive to VEGA installation directory.
  3. Copy the autodock4 and autogrid4 files from the AutoDock/Linux_86-32 or AutoDock/Linux_86-64 of GriDock_X.X.X.tar.gz archive to VEGA installation directory.
  4. Copy AutoDock directory from Vega/Data/AutoDock of GriDock_X.X.X.tar.gz archive to VEGA installation directory/Data.
  5. Copy gridock.xml file from Vega/Data of GriDock_X.X.X.tar.gz archive to VEGA installation directory/Data.
  6. Check/change the file permissions, typing in the command shell:
chmod 755 $VEGADIR/autodock4
chmod 755 $VEGADIR/autogrid4
chmod 755 $VEGADIR/gridock

To test the installation, type in the command shell:

vega
gridock

If the installation is right, no error messages will be shown.

 

2.1.2 MPI Linux installation with shared directories

This is the installation for grid array systems x86-based. MPICH2 is required to run this MPI version. If you have already installed MPICH2, skip this section and go to the next one.

 

2.1.2.1 MPICH2 installation

This is the procedure to install the MPICH2 package on a single node. You must repeat the installation for each node that you intend to use in the grid array. For more information about the installation, you can read the Installer's Guide available at the MPICH2 main site in the documentation section.

  1. Download the latest MPICH2 package from its home site (see downloads section), choosing the current stable version and clicking on UNIX and Windows source.
  2. Unpack the archive, typing in the command shell:
tar -zxvf mpich2-X.X.X.tar.gz

where X.X.X is the MPICH2 version.

  1. Change the current directory to mpich2-X.X.X:
cd mpich2-X.X.X
  1. Build the package, typing:
./configure
make
  1. Install the package:
make install
  1. Create the file .mpd.conf in home directory or /etc/mpd.conf. if your user is root.
cd $HOME
touch .mpd.conf
chmod 600 .mpd.conf

or

touch /etc/mpd.conf
chmod 600 mpd.conf
  1. Add the following line to the .mpd.conf or /etc/mpd.conf file with your preferred text editor (e.g. nedit, vi, etc):
secretword=<MY_PASSWORD>

the new MPICH2 releases require a different line:

MPI_SECRETWORD=<MY_PASSWORD>

where <MY_PASSWORD> is the access password and it must the same for all nodes that you want use in the grid.

  1. Run the MPI daemon:
mpd&
  1. Check the daemon sanity, typing a MPI command:
mpdtrace

the output must be localhost or any other host name.

  1. Try to run locally a non-MPI application in the MPI environment:
mpiexec -n 1 /bin/hostname

the output must be localhost.localdomain or any other host and domain names.

To stop the mpd daemon, you must use the mpdallexit command. To configure the remote hosts, you must consult the Installer's Guide.

 

2.1.2.2 VEGA and GriDock installation

  1. On your master node, create two shared directories: VEGA and GRIDOCK_DATA (other directory names can be used). They must be accessible to all nodes of your grid. The GRIDOCK_DATA directory is needed to put the data (sdf database and receptor files) to do the calculation. In that directory, you can find the results also.
  2. Unpack the content of Vega directory of Vega_X.X.X_Linux_86-32.tar.gz or Vega_X.X.X_Linux_86-64.tar.gz archive to VEGA shared folder.
  3. Copy the gridockmpi executable from GriDock/Linux_86-32 or GriDock/Linux_86-64 folder of GriDock_X.X.X.tar.gz archive to VEGA shared directory.
  4. Copy autodock4 and autogrid4 files from AutoDock/Linux_86-32 or AutoDock/Linux_86-64 of GriDock_X.X.X.tar.gz archive to VEGA shared directory.
  5. Copy Autodock directory from Vega/Data/Autodock of GriDock_X.X.X.tar.gz archive to VEGA shared directory/Data.
  6. Copy gridock.xml file from Vega/Data of GriDock_X.X.X.tar.gz archive to VEGA shared directory/Data.
  7. Check/change the file permissions, typing in the command shell:
chmod 755 VEGA_SHARED_PATH/vega
chmod 755 VEGA_SHARED_PATH/autodock4
chmod 755 VEGA_SHARED_PATH/autogrid4
chmod 755 VEGA_SHARED_PATH/gridockmpi
  1. To test locally GriDockMPI, you must type in the shell:
mpiexec -n <NUMBER_OF_PROCESSES> gridockmpi

where <NUMBER_OF_PROCESSES> is the total number of processes that you want create that is usually the number of CPUs/cores installed in your system.

WARNING:
The mpirun command has a buggy implementation of the MPI arguments that can be interpreted by GriDock as optional parameters. For this reason, when you use mpirun, you must specify always the receptor file name, the database and the AutoDock template.

 

2.2 Windows installation

GriDock and a special build of AutoDock 4 are included in VEGA ZZ package that can be downloaded from www.vegazz.net or www.ddl.unimi.it.