I ran a particular workload collecting the following performance counter results. The workload was designed to produce bad results.
But what I don't understand is how is it possible to have more stalled-cycles-frontend, than cycles? I've seen the same pattern on numerous occasions.
38,836,869,351 cycles
73,845,119,292 stalled-cycles-frontend # 190.14% frontend cycles idle
9,104,536,903 instructions # 0.23 insn per cycle
# 8.11 stalled cycles per insn
595,996,044 dtlb_load_misses.miss_causes_a_walk
100,735 itlb_misses.miss_causes_a_walk
