Phoronix tensorflow with resnet50 batch size 64
AMD metrics shows a workload with a lot of memory operations including L2 misses combined with cpu core operations with a fair amount of floating point.
elapsed 863.009
on_cpu 0.905 # 14.49 / 16 cores
utime 12411.355
stime 91.535
nvcsw 3282710 # 91.85%
nivcsw 291211 # 8.15%
inblock 5720
onblock 1200
cpu-clock 12516157413304 # 12516.157 seconds
task-clock 12518177989728 # 12518.178 seconds
page faults 11064521 # 883.876/sec
context switches 3578045 # 285.828/sec
cpu migrations 541658 # 43.270/sec
major page faults 42 # 0.003/sec
minor page faults 11064479 # 883.873/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 584022412717 # 20.641 branches per 1000 inst
branch misses 5377049822 # 0.92% branch miss
conditional 536426064913 # 18.959 conditional branches per 1000 inst
indirect 3308982806 # 0.117 indirect branches per 1000 inst
cpu-cycles 53588811756063 # 3.87 GHz
instructions 28290781703108 # 0.53 IPC
slots 107133025023402 #
retiring 9524494731791 # 8.9% (10.4%)
-- ucode 997841396 # 0.0%
-- fastpath 9523496890395 # 8.9%
frontend 1938764873769 # 1.8% ( 2.1%)
-- latency 1594472883906 # 1.5%
-- bandwidth 344291989863 # 0.3%
backend 79921103165442 # 74.6% (87.4%)
-- cpu 34642755177137 # 32.3%
-- memory 45278347988305 # 42.3%
speculation 45476223603 # 0.0% ( 0.0%)
-- branch mispredict 38988923174 # 0.0%
-- pipeline restart 6487300429 # 0.0%
smt-contention 15702947676761 # 14.7% ( 0.0%)
cpu-cycles 53578106185765 # 3.87 GHz
instructions 28287157572549 # 0.53 IPC
instructions 9432751214605 # 134.834 l2 access per 1000 inst
l2 hit from l1 928033007528 # 11.50% l2 miss
l2 miss from l1 52635401579 #
l2 hit from l2 pf 250202870552 #
l3 hit from l2 pf 47147988792 #
l3 miss from l2 pf 46468462841 #
instructions 9428428106186 # 37.224 float per 1000 inst
float 512 68 # 0.000 AVX-512 per 1000 inst
float 256 787391345 # 0.084 AVX-256 per 1000 inst
float 128 350180974896 # 37.141 AVX-128 per 1000 inst
float MMX 0 # 0.000 MMX per 1000 inst
float scalar 31 # 0.000 scalar per 1000 inst
Intel metrics shows less time on the cores and also a longer overall runtime.
elapsed 1665.194
on_cpu 0.621 # 9.94 / 16 cores
utime 16471.812
stime 74.065
nvcsw 2445852 # 86.38%
nivcsw 385496 # 13.62%
inblock 6928
onblock 1168
cpu-clock 16522570247592 # 16522.570 seconds
task-clock 16525566278147 # 16525.566 seconds
page faults 9772881 # 591.379/sec
context switches 2839550 # 171.828/sec
cpu migrations 742592 # 44.936/sec
major page faults 51 # 0.003/sec
minor page faults 9772830 # 591.376/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 1315588911649 # 22.142 branches per 1000 inst
branch misses 4086514516 # 0.31% branch miss
conditional 1315588926369 # 22.142 conditional branches per 1000 inst
indirect 470231770021 # 7.914 indirect branches per 1000 inst
slots 81158674362062 #
retiring 30632591230285 # 37.7% (37.7%)
-- ucode 1473030581277 # 1.8%
-- fastpath 29159560649008 # 35.9%
frontend 9540960332730 # 11.8% (11.8%)
-- latency 8398912584891 # 10.3%
-- bandwidth 1142047747839 # 1.4%
backend 40569942688186 # 50.0% (50.0%)
-- cpu 14005646405876 # 17.3%
-- memory 26564296282310 # 32.7%
speculation 1651850357581 # 2.0% ( 2.0%)
-- branch mispredict 1271607577543 # 1.6%
-- pipeline restart 380242780038 # 0.5%
smt-contention 0 # 0.0% ( 0.0%)
cpu-cycles 54996415778281 # 2.07 GHz
instructions 77038363659819 # 1.40 IPC
l2 access 1914216855761 # 62.810 l2 access per 1000 inst
l2 miss 344524930940 # 18.00% l2 missThe process tree time mostly concentrated in python code with a large number of parallel threads started, more than the # of cores. Here is a summary of processes started.
470 processes
105 tf_cnn_benchmar 422684.85 2544.84
91 python3 372951.14 2240.24
38 vulkaninfo 0.76 1.14
6 php 0.19 0.14
4 vulkani:disk$0 0.08 0.12
2 llvmpipe-0 0.04 0.06
2 llvmpipe-1 0.04 0.06
2 llvmpipe-10 0.04 0.06
2 llvmpipe-11 0.04 0.06
2 llvmpipe-12 0.04 0.06
2 llvmpipe-13 0.04 0.06
2 llvmpipe-14 0.04 0.06
2 llvmpipe-15 0.04 0.06
2 llvmpipe-2 0.04 0.06
2 llvmpipe-3 0.04 0.06
2 llvmpipe-4 0.04 0.06
2 llvmpipe-5 0.04 0.06
2 llvmpipe-6 0.04 0.06
2 llvmpipe-7 0.04 0.06
2 llvmpipe-8 0.04 0.06
2 llvmpipe-9 0.04 0.06
6 clang 0.03 0.04
1 lspci 0.00 0.04
88 sh 0.00 0.00
12 gcc 0.00 0.00
11 gsettings 0.00 0.00
9 stty 0.00 0.00
8 stat 0.00 0.00
8 systemd-detect- 0.00 0.00
6 llvm-link 0.00 0.00
5 glxinfo 0.00 0.00
5 phoronix-test-s 0.00 0.00
3 gmain 0.00 0.00
3 tensorflow 0.00 0.00
2 dconf worker 0.00 0.00
2 grep 0.00 0.00
2 lscpu 0.00 0.00
2 setterm 0.00 0.00
2 uname 0.00 0.00
2 which 0.00 0.00
1 cc 0.00 0.00
1 date 0.00 0.00
1 dirname 0.00 0.00
1 dmesg 0.00 0.00
1 dmidecode 0.00 0.00
1 ifconfig 0.00 0.00
1 ip 0.00 0.00
1 lsmod 0.00 0.00
1 mktemp 0.00 0.00
1 ps 0.00 0.00
1 readlink 0.00 0.00
1 realpath 0.00 0.00
1 sed 0.00 0.00
1 sort 0.00 0.00
1 systemctl 0.00 0.00
1 template.sh 0.00 0.00
1 wc 0.00 0.00
0 processes running
75 maximum processesThe process tree shows the bulk of time spent in a parallel block with many more threads started than cores
1915428) tensorflow start=291.21 finish=574.01
1915429) tf_cnn_benchmar start=291.22 finish=573.63
1915430) python3 start=291.34 finish=573.62
1915431) python3 start=291.34 finish=573.62
1915432) python3 start=291.34 finish=573.62
1915433) python3 start=291.34 finish=573.62
1915434) python3 start=291.34 finish=573.62
1915435) python3 start=291.34 finish=573.62
1915436) python3 start=291.34 finish=573.62
1915437) python3 start=291.34 finish=573.62
1915438) python3 start=291.34 finish=573.62
1915439) python3 start=291.34 finish=573.62
1915440) python3 start=291.34 finish=573.62
1915441) python3 start=291.34 finish=573.62
1915442) python3 start=291.34 finish=573.62
1915443) python3 start=291.35 finish=573.62
1915444) python3 start=291.35 finish=573.62
1915445) python3 start=291.78 finish=573.63
1915446) python3 start=291.78 finish=573.63
1915447) python3 start=291.78 finish=573.63
1915448) python3 start=291.78 finish=573.63
1915449) python3 start=291.78 finish=573.63
1915450) python3 start=291.78 finish=573.63
1915451) python3 start=291.78 finish=573.63
1915452) python3 start=291.78 finish=573.63
1915453) python3 start=291.78 finish=573.63
1915454) python3 start=291.78 finish=573.63
1915455) python3 start=291.78 finish=573.63
1915456) python3 start=291.78 finish=573.63
1915457) python3 start=291.78 finish=573.63
1915458) python3 start=291.79 finish=573.63
1915459) python3 start=291.79 finish=573.63
1915460) tf_cnn_benchmar start=292.10 finish=573.63
1915461) tf_cnn_benchmar start=292.10 finish=573.63
1915462) tf_cnn_benchmar start=292.10 finish=573.63
1915463) tf_cnn_benchmar start=292.10 finish=573.63
1915464) tf_cnn_benchmar start=292.10 finish=573.63
1915465) tf_cnn_benchmar start=292.10 finish=573.63
1915466) tf_cnn_benchmar start=292.10 finish=573.63
1915467) tf_cnn_benchmar start=292.10 finish=573.63
1915468) tf_cnn_benchmar start=292.10 finish=573.63
1915469) tf_cnn_benchmar start=292.10 finish=573.63
1915470) tf_cnn_benchmar start=292.10 finish=573.63
1915471) tf_cnn_benchmar start=292.10 finish=573.63
1915472) tf_cnn_benchmar start=292.10 finish=573.63
1915473) tf_cnn_benchmar start=292.10 finish=573.63
1915474) tf_cnn_benchmar start=292.10 finish=573.63
1915475) tf_cnn_benchmar start=292.10 finish=573.63
1915476) tf_cnn_benchmar start=292.10 finish=573.63
1915477) tf_cnn_benchmar start=292.10 finish=573.63
1915478) tf_cnn_benchmar start=292.10 finish=573.63
1915479) tf_cnn_benchmar start=292.10 finish=573.63
1915480) tf_cnn_benchmar start=292.10 finish=573.63
1915481) tf_cnn_benchmar start=292.11 finish=573.63
1915482) tf_cnn_benchmar start=292.11 finish=573.63
1915483) tf_cnn_benchmar start=292.11 finish=573.63
1915484) tf_cnn_benchmar start=292.11 finish=573.63
1915485) tf_cnn_benchmar start=292.11 finish=573.63
1915486) tf_cnn_benchmar start=292.11 finish=573.63
1915487) tf_cnn_benchmar start=292.11 finish=573.63
1915488) tf_cnn_benchmar start=292.11 finish=573.63
1915489) tf_cnn_benchmar start=292.11 finish=573.63
1915490) tf_cnn_benchmar start=292.11 finish=573.63
1915491) tf_cnn_benchmar start=292.11 finish=573.63
1915492) tf_cnn_benchmar start=292.11 finish=573.63
1915493) tf_cnn_benchmar start=293.26 finish=293.26