Benchmark test:

Compare performance of regular calculation in C, versus parallel FBP
calculation where each operator is implemented as a separate process.
I will use UNIX procs and pipes for this first test.

I have a 2x core netbook, so the parallel system should have some advantage
there.  The regular C system should also have an advantage, due to less IO.

The calculations to be tested:


1. simple adding, two inputs each with 1,000,000 integers in binary, add to
   give one output of 1,000,000 integers.
   - try in C with arrays not files
   - try in C with files
   - try also with decoding from text / output text, will be slower I guess