The only thing I'll add to the "use a hash" advice, is don't read the data into two arrays first. You'll likely to run out of memory because of the duplication.

Far better to build a hash directly from the smaller set as you read it in. And the read the second set one-by-one and test it against the hash. For datasets of this size, that likely to save you a large amount of memory, and quite a bit of time.