On the Intel x86 system, the 'cache-misses' event is mapped to LONGEST_LAT_CACHE.MISS, which counts core-originated cacheable requests that miss the L3 cache (Longest Latency cache). Since L3 cache resides in the uncore and uncore PMU is shared by all cores, then how does perf count the miss for a specific process?
For example, does below only count all L3 miss from process 'ls' or system-wide?
$ perf stat -e cache-misses -- ls