I'm having trouble understanding why this issue can't just be fixed by locks. If a segment of memory can be loaded into multiple caches, shouldn't a lock always be necessary to protect it?
taoy1
@carnegieigenrac, for 'we can't fix this problem', here the 'problem' means the cache coherence problem. This problem differs from the mutual exclusion problem taught in 15-213. Cache coherence problem is solved by hardware now, so we (programmers) don't need to worry about it. And it is because cache coherence problem has been solved by hardware, we can use lock to fix mutual exclusion problem.
I'm having trouble understanding why this issue can't just be fixed by locks. If a segment of memory can be loaded into multiple caches, shouldn't a lock always be necessary to protect it?
@carnegieigenrac, for 'we can't fix this problem', here the 'problem' means the cache coherence problem. This problem differs from the mutual exclusion problem taught in 15-213. Cache coherence problem is solved by hardware now, so we (programmers) don't need to worry about it. And it is because cache coherence problem has been solved by hardware, we can use lock to fix mutual exclusion problem.