3 ms·
This is kind of silly. The author has reduced the number of mutations per entry by a factor of up to 4. That's why contention is reduced (less M(O)ESI evictio
by loeg 7y ago
This is kind of silly. The author has reduced the number of mutations per entry by a factor of up to 4. That's why contention is reduced (less M(O)ESI eviction of shared lines). It happens to be that this problem is amenable to that kind of computation but it isn't always the case.