Compiling Slave Programs On Different Machines

We want to run slaves on goophy, a DEC alpha, mango and coconut, two IBM RS6000 computers. Creating all the necessary executables requires the following steps.
Step 1: Compile slave process on goophy
Log into goophy, change into the directory where your slave source code is and issue command,

cc -o answer c_slave.c -lpvm3

this creates the program answer in the current directory. Copy this program into the directory for goophy's PVM executables,

cp answer $HOME/pvm3/bin/ALPHA

Step 2: Compile slave process on mango
Log into mango, change into directory where your slave source code is and issue command,

cc -o answer c_slave.c -lpvm3

this creates the program answer in the current directory. We run a shared filesystem, so the current directory is the same on goophy and mango and the old program answer is overwritten by the new one. However the copy in $HOME/pvm3/bin/ALPHA is still there and this is the idea behind all this. Copy the new answer program into the directory for mango's PVM executables,

cp answer $HOME/pvm3/bin/RS6K

and we have two architecture specific programs with the same name in different directories where PVM will find them.

Step 3: Compile slave process on coconut
Coconut has the same architecture as mango and so you don't have to do anything on coconut.
Possible Difference
No shared file system
If your home directory is different on different computers, then your home directory is not shared between these computers. In this case you have to compile the slave program for every architecture and copy (ftp) the executable to every machine with this architecture. You should still put them into the directories $HOME/pvm3/bin/PVM_ARCH which you have to create on all machines so PVM can find these programs.
Back: The first program - communication is everything