doPerfStatSKL

perf stat \
-e task-clock -e cycles -e cycle_activity.stalls_total -e cycle_activity.stalls_mem_any \
-e icache_16b.ifdata_stall \
-e iTLB-load-misses -e iTLB-loads \
-e rs_events.empty_cycles -e uops_executed.stall_cycles \
-e uops_executed.cycles_ge_1_uop_exec \
-e uops_executed.cycles_ge_2_uops_exec \
-e uops_executed.cycles_ge_3_uops_exec \
-e uops_executed.cycles_ge_4_uops_exec \
-e instructions -e branch-instructions -e branch-misses \
-e arith.divider_active \
-e fp_arith_inst_retired.512b_packed_single -e fp_arith_inst_retired.512b_packed_double \
-e fp_arith_inst_retired.256b_packed_single -e fp_arith_inst_retired.256b_packed_double \
-e fp_arith_inst_retired.128b_packed_single -e fp_arith_inst_retired.128b_packed_double \
-e fp_arith_inst_retired.scalar_single -e fp_arith_inst_retired.scalar_double \
-e mem_load_retired.l1_hit -e mem_load_retired.l2_hit -e mem_load_retired.l3_hit \
-e mem_load_retired.l3_miss \
-e cycle_activity.stalls_l3_miss \
-e core_power.lvl0_turbo_license -e core_power.lvl1_turbo_license -e core_power.lvl2_turbo_license \
-e dTLB-load-misses -e dTLB-loads \
-e kmem:mm_page_alloc_extfrag -e kmem:mm_page_alloc -e kmem:mm_page_free \
-e compaction:mm_compaction_try_to_compact_pages \
-e huge_memory:mm_collapse_huge_page \
$1