Hi Ata --
I was trying to read between the lines to determine what performance questions you were trying to answer.
Performance counter events can definitely measure data cache and instruction cache accesses (Event selects 0x040 and 0x080). You can also measure the number of DRAM accesses (Event select 0x0E0). I think this is consistent with what you have mentioned in your posting.
However, here's a few fine points. It sounds like you would like to measure page faults and disk I/O. Unfortunately, it's hard to infer information about performance at this level of abstraction through the use of performance counter events. You will probably need to use OS-level monitoring tools (like top on Linux or the Task Manager, etc. on Windows) to look at page faults and disk I/O.
The "page hit" and "page miss" refer to memory controller/DRAM pages which are a different concept than virtual memory pages. It's confusing, but modern memory controller design has the concept of open and closed pages which is separate from the notion of virtual memory pages.
Hope this helps!
Thanks for your reply. So far I am not able to find suitable events to map disk I/O performance, thus had to fall back to linux utility programs to perform the desired.