First off, be wary of any approach that claims "instant parallelism." Changing sequential code to truly parallel code is non-trivial and is currently, in my opinion, not a task that should be automated.

The first major question to ask yourself is: "Can my program be subdivided into independent modules?" If so, the next question is "Can I do it easily?"

You say that you''re writing a chess program? Well, searching a chess game tree is a task which can (and should) be done in parallel over as many CPU''s as you can get.

A good parallel message passing library that I''ve used many times is PVM (http://www.epm.ornl.gov/pvm/pvm_home.html). Another alternative is MPI; I''ve heard that it''s better but I''ve never used it myself.

PVM works by separating large jobs into smaller tasks and then magically dispatching them over a network of hetergeneous machines. You can do this for your chess program with a minimal amount of coding on your part (i.e. you can probably make a few changes to one of the sample programs). I did this once to "parallelize" an automated theorem proving program, it was not hard at all.

Good luck.