...

Notes...

• Q:  how would miss-rate vary with block size?
• Q:  how might miss penalty for large blocks be reduced?
• Q:  what happens with a large block size, when a write miss occurs?
• Q:  Should the tag be the virtual address or the physical address?
• Q:  Should the second-level cache always contain everything held in the first-level cache (the ``multi-level inclusion'' property)
• Q:  How could one restructure a program to get better performance with a small cache?
• Q:  What kind of program would have a very high rate of conflict misses with a two-way set-associative cache?