Consider a cache with a line size of 16 words with each word containing 4 bytes. Assume that 20% of all memory accesses requested by the CPU are write operations. We would like to compare the performance of the write-through with no write-allocate policy (abbreviated as WT) with the write- back with write-allocate policy (abbreviated as WB). Data from memory is read into cache one line at a time. For WT, a single word is written to main memory for each write operation. For WB, the entire line must be written back if it is dirty at the time of replacement; you may assume that on average 25% of the lines in the cache are dirty. A) For both of these policies, find the average amount of main memory traffic (measured as bytes per CPU's memory access). Your answer in each case should be an expression in terms of the hit ratio H.
B) Now compute the average main memory traffic for each policy for the following hit ratios: 0.92, 0.95, 0.98. Also write the main memory traffic per memory access for the case where there is no cache. Show all your results in a tabular form.
C) What conclusions can you draw from the above results?