It seems the processing of the lines is independent of each other. If the output is as well, you might give each fork a different starting line (0 to 3) and let them read the same file but only process every fourth line. No need to split the source file then.