readme.tx

1) Forking multiple threads for Readers not beneficial.
2) Using parallel sections for readers and mappers not beneficial.
3) can't really have reducer in parallel with anything else. Besides, reduction is a very tiny percentage of the total time.

4) cache competition on L3 might be causing the 16 threads to be missing more and values to messup.