If block size increases, main memory tag size decreases (because word offset size will increase and we are assuming increasing block size does not increase overall size of ram)
in full associative cache CM tag size = MM tag size. So, CM tag size decreases => comparison time decreases => Hit/Miss TIME decreases.
For Direct/set associative it will NOT affect the hit/miss time, as tag size does not change as shown by you in the link to other question.
But S3 has ‘miss rate’ not miss time. So I think, tag sizes are of no use for us in S3.
When block size increases, No of cache lines will decrease, so I thought that with lesser cache lines, miss rate will increase. BUT, there is a catch….
When block size increased, it so did for MM also. there too number of blocks decreased. So, actually it is completely dependent on access sequence, whether miss rate will decrease or not.
We cannot say that miss rate decreases ALWAYS.
Eg:
case 1 → CM : 4 lines and MM : 16 blocks : access sequence 3, 2, 5, 4, 6, 8, 14, 13, 7 → miss rate 9/9
after doubling block size → CM : 2 lines and MM : 8 blocks : access sequence becomes 1, 1, 2, 2, 3, 4, 7, 6, 3 → miss rate 7/9
case 2 → CM : 4 lines and MM : 16 blocks : access sequence 0, 4, 8, 12 → miss rate 4/4
after doubling block size → CM : 2 lines and MM : 8 blocks : access sequence becomes 0, 2, 4, 6 → miss rate 4/4
case 1 where access sequence implies increased spatial locality of reference, we see a decrease in miss rate. but when accesses are apart, like in case 2, we see no change. hence nothing can be surely said about miss rate without looking at an access sequence.
hence, s3 is false.
s1 and s2 are true.
PS : cool question, lets discuss thoughts about this if any… I may wrong. What is the answer anyway?
