False sharing here: even if the two processors are not sharing the same data, they might share the same cache unit (due to the mechanism mapping address to cache unit), which means that the data could be switched in and out the cache even if processors are not actually sharing the same address.

@jt is this due to the granularity of pages being a lot larger than most cache lines? Also, what does cache unit mean here?

@ufxela I think the meaning of cache unit varies in different contexts: https://www.usenix.org/legacy/publications/library/proceedings/sedms4/full_papers/bolosky.txt However, the idea of false sharing applies regardless of whether the "unit" is cache line or pages.