Select query output in parallel threads

I need to write over 20 million records in a flat file. A direct selection request will use time. I feel the need to generate output in parallel based on pieces of data, that is, having 10 queries select more than 10% of the data each in parallel. Then sort and merge on UNIX.

I can use rownum for this, but it would be tedious, static and need to be updated every time rownum changes me.

Is there a better alternative?