pvm_xstar
After the dependencies described below are met, installing pvm_xstar should
be easy because it does not involve any software compilation. The package
consists of 4 scripts:
pvm_xstar Bourne shell script
pvm_xstar_wrap Bourne shell script
pvm_xstar_master S-Lang script
pvm_xstar_slave S-Lang script
although end-users generally need to concern themselves only with the first.
Simply use the standard GNU configure and make commands within
a UNIX-like (e.g. Linux, Mac OS/X terminal, or Cygwin) environment:
./configure [options]
make install
You may specify the ---prefix option to configure to steer the
installation
to a non-default location, or use ---help to get more information.
It is strongly encouraged that you test your PVM setup and pvm_xstar
both before and after installation. These can be done from the top of
your pvm_xstar distribution with:
make test_local: a smoketest-style of testing which operates
directly from your unpacked distribution, and can be done prior to installationmake test: operational test, to be done after installation, to
ensure that your PATH is set properly, your PVM hosts file enables PVM to
find pvm_xstar and its components at runtime, and so forth.tests subdirectory
for more details.
pvm_xstar requires that the following software is installed on your system:
Yum, apt-get
or MacPorts (which for Mac OS/X we recommend over Fink). The
S-Lang PVM module is very small, and should also
take little time to download and install. Although XSTAR may be the
largest download of any of these dependencies, much of that is due to
the data required for its operation; as a standard Fortran 77 code that
has been tested for many years, XSTAR builds cleanly on most Unix-like systems.
While PVM may be installed easily on modern systems via binary package management, additional user-specific configuration is usually required before it can be employed. Common PVM configuration problems include
Unable to run pvm_xstar_slave on XXX.YYY.ZZZ: not found
If you are using PVM on a single machine, such as a multicore workstation or massively-parallel-processor (MPP), then you will not need to perform remote logins and can skip this section.
A common way of automating remote SSH logins involves generating keys on the master,
master% ssh-keygen -t dsa
...
copying the public keyfile that's generated to each slave host,
master% scp ~/.ssh/id_dsa.pub slave-host-1:master.key
adding it to the authorized keys file for that host,
slave-host-1% cd ~/.ssh
slave-host-1% cat ~/master.key >> authorized_keys
and ensuring that authorized_keys has the required permissions:
slave-host-1% chmod 600 authorized_keys
Another option is to use ssh-agent.
Enabling the PVM daemon to access your pvm_xstar scripts can be done in
several ways, such as installing them into a system-wide location
accessible to all users at login, but in our experience one of the
most reliable methods has been to add a line like
* ep=$PATH
or
* ep=<path/to/your/pvm_xstar/scripts>
to your PVM hosts file, as described in the pvm_intro(1) and pvmd3 man
page documentation. When the * ep= option appears at the top of
your hosts file it applies to all hosts in the virtual machine, but PVM
allows different values to be used simultaneously by defining ep
on the hostname line, such as
mylinux.mysite ep=/nfs/local/bin
mymac.mysite ep=/usr/local/bin