I don't think you should read it line by line. Just read the whole thing, and process it line by line. 1300 rows are peanuts, memory or speed should not be an issue.
Looking at your actual process though, I would process the whole matrix at once. Maybe I misunderstood your calculations, but I don't see a reason for a loop (except for the cumulative sum in the end).
Probably fastest:
1. Read whole file
2. Compute x.*y for whole matrix
3. cumsum over results
