4 million iteration is nothing much in the scheme of things. I did a little finacial brute force matching problem that did 3.5 million in under a minute with about 50 lines of munging in the loops.

Gererally this sort of problem comes down to a few simple things. Disk are slow. Databases are slow. Memory, hashes, arrays (perhaps) are fast. Pull everything into memory, hash or array structures depending. Use hash tables to index the interactions. Process. Get results. Write to disk at the end. If you can't pull everyting into memory get more memory. Memory, processor, disk in that order on the hardware side. Algorithm, algoritm, algorithm on the code side.

If you can post some exapmle data/and or code that illustrates the problem I am sure it will be possible to make if much faster.