Question: Why should I use parallel lil-gp on a single CPU machine?
Answer: Well if you need to adapt your program to work with the new kernel it is a good idea to give it a tired on 1 CPU first. Do this by putting only one machine in your PVM.Invoke the parallel kernel by specifying more host using the -h option.Question: My program works fine in sequential mode but not when I use the parallel mode of the kernel. What changed?
Answer: Are you reading information directly from a file in your program? The slave nodes will do the same! The thing to check is if your file is accessible to the slaves.Question: The file(s) my program needs are available to the slaves they still do not work!
Answer: Do all the slaves access the same file? OR do they use a copy? Check the permissions. If the permissions are OK how do you open these files? The slaves want access the very same file you better open it in read-only mode or you will get race conditions (hard to find)!Question: I got the demo applications to compiled but now when I am trying to run them it doesn't seem to do anything?
Answer: PVM requires that the programs it should start should be put in a very specific directory. On un*x machine this would relative to the users HOME. More precisely in the pvm3/bin. Inside this bin directory there should be a directory named after the operating system you are using. Check the PVM manual for more accurate information.Question: My program still doesn't work when I running it on our cluster! But it works with 1 CPU in parallel mode.
Answer: Well the fact that it runs on 1 CPU in parallel mode is good start.
- Are you reading information directly from a file? If so you should realize that this file should be accessible to all the slaves (ask your sys. admin. for more details).So maybe it is not available on the cluster.
- Is your program available to the nodes of your cluster? Again the slave program should be put in a directory where PVM can find it.