2.2 Installation

The first step is to obtain SVM. The latest release can be downloaded from its website:

http://msdl.cs.mcgill.ca/people/tfeng/?research=svm

The .tar.gz packages are for Linux. Users may use the following command to extract those packages:

tar zxvf svm-xxxxx.tar.gz

(Here, xxxxx is the version number followed by the architecture, or src if the package is a source package.) A new directory will be created with SVM in it.

The .zip packages are for Windows. They can be unpacked with WinZip (http://www.winzip.com/) or WinRAR (http://www.rarlab.com/).

The users who want to experience the most cutting-edge features of SVM may download it from the CVS (Concurrent Versions System) server with the following commands (executed in Linux systems or in Cygwin in Windows):

export CVS_RSH="ssh"
cvs -z3 -d:ext:anoncvs@savannah.nongnu.org:/cvsroot/svm co svm

The CVS server is hosted by Savannah (https://savannah.nongnu.org/). SVM's project home on Savannah is at:

https://savannah.nongnu.org/projects/svm

Windows users are strongly advised to use binary distributions. Those distributions build in all the necessary libraries and do not require the Python environment. The users simply unpack the packages and execute svm.exe or scc.exe in it.

SVM developers or users of other operating systems (Linux, FreeBSD, SunOS, etc.) may download the source packages. svm.py and scc.py in the packages contain the main program of SVM and SCC, respectively. They are executable in the Python environment. For example, the following command starts SVM with no parameter. The command-line help will be printed to the console.

python svm.py

(Support svm.py is in the current directory.)

Start-up scripts svm and scc in the same directory as svm.py and scc.py are written for Linux users. By adding the SVM directory to the PATH environment variable, the users can invoke SVM and SCC simply by running svm and scc from any directory.

Similarly, scripts svm.bat and scc.bat are written for Windows users to invoke SVM and SCC from any directory.

If source packages or CVS versions are used, the users must manually build the PYPVM library before distributed simulation can be performed:

• Linux users.

  Execute ``make pypvm'' on the command-line in the SVM directory. This will build the PYPVM library with SCC and copy it to the SVM directory. Before executing the command, the users MUST confirm the following environment variables are set correctly:

  1. PVM_ROOT should be set to the path where PVM is installed. For RedHat Linux 9.0, this path is /usr/share/pvm3. It may be different for other Linux distributions.

  2. PVM_ARCH should be set to the ID of the architecture. This ID is equal to the name of the sub-directory of $PVM_ROOT/lib. Use the following command to retrieve this ID:

     ls -F $PVM_ROOT/lib | grep '/' | gawk 'sub(/\//, "", $0)'
     

• Windows users.

  The PYPVM library is prebuilt for Windows users and saved in the lib\win32\Release sub-directory of the SVM directory. Command ``make -f Makefile.win32 pypvm'' copies the library to the SVM directory.