> in my case the working set is small enough and the processes are top-priority so they probably stay in the L2 if not the L1.
Maybe! Maybe not. If it's top priority on core X but something else with a much better (or cache-unfriendly) dataset is on the hyperthread-sibling core then your high priority process can still have cache misses.
Maybe! Maybe not. If it's top priority on core X but something else with a much better (or cache-unfriendly) dataset is on the hyperthread-sibling core then your high priority process can still have cache misses.