Is there any way for VTune (or anything else) to measure the number of different cache lines a section of code (say a method) accesses, on an Ivy Bridge CPU (or a different model)? When I say "cache line", I mean how many different 64-byte blocks.
I suppose that such a functionality if exist should be exposed to the software by dedicated counters.
You may give Intel PIN a try to solve that. Using PIN you can monitor all load/store instructions executed.
As for what can be done with the performance counters (or PMU) I find this text of David Levinthal very comprehensive:
Thanks for mentioning PIN tool.It really grabbed my attention.
For more complete information about compiler optimizations, see our Optimization Notice.