Cache related problems Will rate according to correctness of the answer. Thanks
ID: 2249650 • Letter: C
Question
Cache related problems
Will rate according to correctness of the answer. Thanks in advance
Write no-allocate cache (1) Never writes into the cache, independent of hits or misses (2) Does not bring the block into the cache upon a write miss (3) Always brings the block from memory upon write miss and writes into the block (4) Does not bring the block into the cache upon a read miss (5) None of the above Compared to a writeback cache, a write through cache typically (1) Uses less memory bandwidth (2) Uses more memory bandwidth (3) Uses equal memory bandwidth (4) Is faster (5) None of the above The average memory access time is given by (1) hit-time hit-rate + miss-penalty"miss-rate (2) Hit-time + miss-rate miss-penalty (3) Hit rate miss-rate * miss-penalty (4) Hit-time miss penalty (5) None of the above As the block size increases while holding the cache size constant, the miss rate (1) Keeps increasing (2) Keeps decreasing (3) first increases then decreases (4) first decreases and then increases (5) None of the above Cache coherence (1) Makes writes atomic (2) Ensures that writes to block are seen in the same order by all cores (3) Ensures a read retrieves the value of the last write (4) All of the above (5) None of the aboveExplanation / Answer
---> Write no-allocate cache does not bring the block into the cache upon a write miss. The block is modified in the main memory only and not loaded into the cache. Hence option(2) is correct.
---> Compared to a writeback cache, a write through cache typically uses more memory bandwidth. Here, every write needs a main memory access so bandwidth is more. Hence option(2) is correct.
---> Average memory access time= hit time * hit rate + miss rate * miss penalty. Hence option(1) is correct.
---> The miss rate first decreases and then increases with the increase in block size.Hence option(4) is correct.
---> Cache coherence makes the write atomic, ensures that writes to block are seen in the same order by all the cores and ensures that a read retrieves the value of the last write. hence option(4) is correct.