It sounds like this might be a good job for PDL. The first few chapters of the book “PDL – Scientific Programming in Perl” book online in HTML format explain the method of doing it. The main point of PDL is to convert data to compact efficient sequential arrays for fast processing.

You would need to convert the hash to array of arrays, but it should be manageable. I've yet to really grasp the working of piddle slices, but you should be able to fly thru the data, look for the flag and count, then pull out it's piddle slice. Of course there are obstacles to overcome, like all data being numeric, so your ascii data will need to be converted to byte data types.