I am working on GPGPU development.
Most of my work's in Gentoo and Suse 10.2 and NVidia
I was searching for help on Gentoo LAM-MPI install after emerge.
didn't find much so I continued on my own, documenting how I got LAM-MPI operating on a small 2 node Gentoo X86-64 cluster. Thought its the least I could do, contribute back to the Gentoo community. Hope someone finds this information useful...............


LAM-MPI uses SSH to pass messages to/from cluster nodes.
SSH and LAM must be installed on all cluster nodes, Master & Slaves.
A cluster Master-node runs lamd (LAM daemon).
Cluster Slave-nodes run sshd (SSH daemon).
The Master-node is an ssh client to all Slave-nodes, ssh servers.
The lamd on the Master-node controls messaging to/from Slave-nodes.
A set of MPI utilities (i.e. mpirun) on the Master-node
will distribute execution messages to Slave-nodes.

Setup Gentoo systems to run LAM-MPI:
LAM network uses secondary ethernet at 1Gb

Slave-nodes:
> # /etc/init.d/sshd start
( and to run on boot )
> # rc-update add sshd default
> # useradd -m lamboot
> # passwd lamboot
( lamboot )

Master-node:
> # useradd -m lamboot
> # passwd lamboot
( lamboot )
> # login lamboot
> # ssh-keygen
( converts password to public/private key pair)
> # ssh-copy-id -i .ssh/id_rsa.pub 169.254.46.208
( copies keys to all Slave-nodes thru 1Gb net)
( duplicate for all other Slave-nodes )
> # ssh 169.254.46.208 'cat /proc/partitions'
( test ssh thru the cross-over 1Gb net )
( this must connect and execute without password entry )
Create a LAM host description file on the Master-node in /home/lamboot/lamhosts
169.254.68.28 cpu=4
169.254.46.208 cpu=2
( include all Slave-nodes, Master is first node )
> # lamboot -v lamhost
( boot up the LAM cluster environment )
( this needs automated at boot time )
( previous commands were one-time only )
> # lamnodes
( test that all nodes are running )
> # mpirun -sa -s n0 C lamhello
( execute a C, C++, or Fortran compiled binary, lamhello )
( this code was compiled with appropriate MPI lib, and utils )
( binary executable was placed on the Master-node in /home/lamboot/ )


After setup, only the following commands will be executed
on the Master-node when entire cluster has finished reboot:

> # login lamboot
> # lamboot -v lamhosts
(then verify LAM )
> # lamnodes
> # mpirun -s n0 C lamhello