Answer from cs61c-cb (minh uyen nguyen 16765774) for Question 2
This change is made based on the concept of spatial locality.  In the beginning, whenever we are in the innner loop, we have to find a way to access data in the memory 200 times continuously.  It means that we have to find a way to store 200 data in our cache => hit time and miss penalty are longer than when we have only 20 data to deal with (when the inner loop runs only 20 times).