1 cacheable to cacheable
2 cacheable to non - cacheable (for example, from the user memory to the DMA, the DMA useless cache)
3 non - cacheable cacheable (such as from the DMA to user memory)
4 non - cacheable to non - cacheable
Especially the 1 2 and 3
CodePudding user response:
Not necessarily, a small amount of data replication, of course, is the cache to cache the fastest, but large pieces of data replication (a few MB +), 4 in fact is the fastest, so just have MOVNTxxxx series instructions