Open aayasin opened 1 year ago
Here is output for a 5 seconds multicore run where toplev reports 80 seconds.
ayasin@icl-adl-01:/iusers/ayasin/perf-tools$ ./do.py profile --perf /iusers/ayasin/perf --toplev-args ' --global' --tune :sample:3 ":perf-pebs:'-b -e cpu_core/event=0xc5,umask=0x11,name=BR_MISP_RETIRED.COND/ppp -c 20003'" :perf-pebs-top:-1 -a './adobe-HQ.sh pgo1 t73' -pm 12 -v1 INFO: App: ./adobe-HQ.sh pgo1 t73 . per-app counting 3 runs .. /iusers/ayasin/perf stat -r3 --log-fd=1 --td-level=2 -e "cpu-clock,context-switches,cpu-migrations,page-faults,instructions,cycles,ref-cycles,branches,branch-misses,cycles:k,{slots,cpu_core/topdown-retiring/,cpu_core/topdown-bad-spec/,cpu_core/topdown-fe-bound/,cpu_core/topdown-be-bound/,cpu_core/topdown-heavy-ops/,cpu_core/topdown-br-mispredict/,cpu_core/topdown-fetch-lat/,cpu_core/topdown-mem-bound/},cpu_core/event=0xc4,umask=0x40,name=System-entries/u,r2424,r0160,r0262" -- ./adobe-HQ.sh pgo1 t73 | tee adobe-HQ-pgo1-t73.perf_stat-r3.log | egrep 'seconds [st]|CPUs|GHz|insn|topdown|Work|System|all branches' | uniq 77,336.34 msec cpu-clock # 15.745 CPUs utilized ( +- 2.65% ) 419,131,098,203 cpu_core/topdown-retiring/ # 5.556 G/sec ( +- 1.58% ) 195,044,038,033 cpu_core/topdown-bad-spec/ # 2.586 G/sec ( +- 0.98% ) 92,150,614,839 cpu_core/topdown-fe-bound/ # 1.222 G/sec ( +- 3.74% ) 110,879,392,005 cpu_core/topdown-be-bound/ # 1.470 G/sec ( +- 12.46% ) 16,877,205,681 cpu_core/topdown-heavy-ops/ # 223.734 M/sec ( +- 2.74% ) 193,658,937,149 cpu_core/topdown-br-mispredict/ # 2.567 G/sec ( +- 1.00% ) 59,771,314,506 cpu_core/topdown-fetch-lat/ # 792.364 M/sec ( +- 3.72% ) 54,177,283,436 cpu_core/topdown-mem-bound/ # 718.206 M/sec ( +- 20.09% ) 4,747,438 System-entries # 62.935 K/sec ( +- 8.44% ) 4.912 +- 0.158 seconds time elapsed ( +- 3.23% ) topdown full tree + All Bottlenecks .. PERF=/iusers/ayasin/perf /usr/bin/python /iusers/ayasin/perf-tools/pmu-tools/toplev.py --cputype=core --no-desc -vl6 --nodes '+IPC,+Instructions,+UopPI,+Time,+SLOTS,+CLKS,+Mispredictions,+Big_Code,+Instruction_Fetch_BW,+Branching_Overhead,+DSB_Misses,+Memory_Bandwidth,+Memory_Latency,+Memory_Data_TLBs,+Core_Bound_Likely' -V adobe-HQ-pgo1-t73.toplev-vl6-perf.csv --global --tune 'DEDUP_NODE = "Other_Light_Ops,Lock_Latency,Contested_Accesses,Data_Sharing,FP_Arith"' -- ./adobe-HQ.sh pgo1 t73 2>&1 | tee adobe-HQ-pgo1-t73.toplev-vl6.log | egrep '<==|MUX|Info(\.Bot|.*Time)|warning.*zero' | sort BAD Bad_Speculation.Branch_Mispredicts % Slots 24.4 [ 4.0%]<== Info.Botlnk.L0 Core_Bound_Likely Metric 0.00 Info.Botlnk.L2 DSB_Misses Scaled_Slots 2.51 [ 2.1%] Info.Bottleneck Big_Code Scaled_Slots 1.01 [ 3.8%] Info.Bottleneck Branching_Overhead Scaled_Slots 7.93 [ 3.8%] Info.Bottleneck Instruction_Fetch_BW Scaled_Slots 5.33 [ 3.8%] Info.Bottleneck Memory_Bandwidth Scaled_Slots 1.75 [ 3.8%] Info.Bottleneck Memory_Data_TLBs Scaled_Slots 0.90 [ 3.8%] Info.Bottleneck Memory_Latency Scaled_Slots 0.14 [ 3.8%] Info.Bottleneck Mispredictions Scaled_Slots 27.60 [ 3.8%] Info.System Time Seconds 79.94 MUX % 2.00 warning: 5 nodes had zero counts: Contested_Accesses Core_Bound_Likely DSB_Switches Data_Sharing UopPI ERROR: Too many metrics with zero counts ( Contested_Accesses Core_Bound_Likely DSB_Switches Data_Sharing UopPI). Run longer or use: --toplev-args ' --no-multiplex' !
Here is output for a 5 seconds multicore run where toplev reports 80 seconds.