OpenVKL does “volume computational kernels” as part of Intel’s rendering toolkit. There are two workloads (and a third that does SYCL that didn’t run). Overall metrics are more middle of the road with some backend memory stalls but also ok retirement rate. Pattern below suggests multiple phases through the code.
AMD metrics show floating point intensive codes with smaller amount of branches.
elapsed 3971.149
on_cpu 0.896 # 14.34 / 16 cores
utime 56359.739
stime 584.664
nvcsw 33937687 # 98.60%
nivcsw 480386 # 1.40%
inblock 1002920 # 252.55/sec
onblock 8052352 # 2027.71/sec
cpu-clock 56931734359547 # 56931.734 seconds
task-clock 56941310564491 # 56941.311 seconds
page faults 47461092 # 833.509/sec
context switches 34437705 # 604.793/sec
cpu migrations 20220 # 0.355/sec
major page faults 5466 # 0.096/sec
minor page faults 47455626 # 833.413/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 19400954559475 # 97.035 branches per 1000 inst
branch misses 201690377818 # 1.04% branch miss
conditional 12895106123407 # 64.496 conditional branches per 1000 inst
indirect 1374941034553 # 6.877 indirect branches per 1000 inst
cpu-cycles 235999440720760 # 3.72 GHz
instructions 199532654165200 # 0.85 IPC
slots 471738528929274 #
retiring 103684313307317 # 22.0% (27.1%)
-- ucode 587326337589 # 0.1%
-- fastpath 103096986969728 # 21.9%
frontend 46597727613920 # 9.9% (12.2%)
-- latency 41563283533494 # 8.8%
-- bandwidth 5034444080426 # 1.1%
backend 226997644094787 # 48.1% (59.2%)
-- cpu 88712172916793 # 18.8%
-- memory 138285471177994 # 29.3%
speculation 5850505760480 # 1.2% ( 1.5%)
-- branch mispredict 4911986409490 # 1.0%
-- pipeline restart 938519350990 # 0.2%
smt-contention 88602923627941 # 18.8% ( 0.0%)
cpu-cycles 234813534923853 # 3.71 GHz
instructions 199261913372212 # 0.85 IPC
instructions 66407166959113 # 17.629 l2 access per 1000 inst
l2 hit from l1 905877907882 # 24.18% l2 miss
l2 miss from l1 164919857111 #
l2 hit from l2 pf 146645173962 #
l3 hit from l2 pf 48609945732 #
l3 miss from l2 pf 69549635740 #
instructions 66387630998801 # 375.847 float per 1000 inst
float 512 93 # 0.000 AVX-512 per 1000 inst
float 256 27363772230 # 0.412 AVX-256 per 1000 inst
float 128 24924227430959 # 375.435 AVX-128 per 1000 inst
float MMX 0 # 0.000 MMX per 1000 inst
float scalar 3 # 0.000 scalar per 1000 inst
Intel metrics
elapsed 4136.591
on_cpu 0.896 # 14.33 / 16 cores
utime 57965.710
stime 1331.660
nvcsw 76665884 # 99.28%
nivcsw 555305 # 0.72%
inblock 816520 # 197.39/sec
onblock 7819888 # 1890.42/sec
cpu-clock 59247135542528 # 59247.136 seconds
task-clock 59259522665649 # 59259.523 seconds
page faults 43320780 # 731.035/sec
context switches 77241651 # 1303.447/sec
cpu migrations 46304 # 0.781/sec
major page faults 4528 # 0.076/sec
minor page faults 43316252 # 730.959/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 19225950316371 # 91.942 branches per 1000 inst
branch misses 198690011214 # 1.03% branch miss
conditional 19225950333715 # 91.942 conditional branches per 1000 inst
indirect 4849736895446 # 23.192 indirect branches per 1000 inst
slots 282541754736980 #
retiring 146147493214299 # 51.7% (51.7%)
-- ucode 14751838264754 # 5.2%
-- fastpath 131395654949545 # 46.5%
frontend 55659811470922 # 19.7% (19.7%)
-- latency 42073001368270 # 14.9%
-- bandwidth 13586810102652 # 4.8%
backend 69719964764269 # 24.7% (24.7%)
-- cpu 36329134258130 # 12.9%
-- memory 33390830506139 # 11.8%
speculation 14637555457175 # 5.2% ( 5.2%)
-- branch mispredict 14202906582505 # 5.0%
-- pipeline restart 434648874670 # 0.2%
smt-contention 0 # 0.0% ( 0.0%)
cpu-cycles 182201387562091 # 2.77 GHz
instructions 253179753855479 # 1.39 IPC
l2 access 1615030149531 # 12.736 l2 access per 1000 inst
l2 miss 684417918973 # 42.38% l2 miss
Process information crashed part way through so was incomplete but pieces shown below.
328 processes
80 vklBenchmarkCPU 743574.56 8707.04
34 clinfo 9.58 3.34
19 vulkaninfo 0.19 0.96
3 glxinfo:gdrv0 0.11 0.01
6 clang 0.06 0.06
1 glxinfo 0.05 0.01
1 glxinfo:cs0 0.05 0.01
1 glxinfo:disk$0 0.05 0.01
1 glxinfo:sh0 0.05 0.01
1 glxinfo:shlo0 0.05 0.01
2 vulkani:disk$0 0.02 0.11
1 llvmpipe-1 0.01 0.06
1 llvmpipe-10 0.01 0.06
1 llvmpipe-11 0.01 0.06
1 llvmpipe-12 0.01 0.06
1 llvmpipe-13 0.01 0.06
1 llvmpipe-14 0.01 0.06
1 llvmpipe-15 0.01 0.06
1 llvmpipe-2 0.01 0.06
1 llvmpipe-3 0.01 0.06
1 llvmpipe-4 0.01 0.06
1 llvmpipe-5 0.01 0.06
1 llvmpipe-6 0.01 0.06
1 llvmpipe-7 0.01 0.06
1 llvmpipe-8 0.01 0.06
1 llvmpipe-9 0.01 0.06
1 llvmpipe-0 0.01 0.05
1 ps 0.00 0.01
59 sh 0.00 0.00
13 gsettings 0.00 0.00
12 gcc 0.00 0.00
8 systemd-detect- 0.00 0.00
7 stat 0.00 0.00
6 llvm-link 0.00 0.00
5 openvkl 0.00 0.00
4 phoronix-test-s 0.00 0.00
2 gmain 0.00 0.00
2 which 0.00 0.00
1 date 0.00 0.00
1 dirname 0.00 0.00
1 grep 0.00 0.00
1 ifconfig 0.00 0.00
1 ip 0.00 0.00
1 lscpu 0.00 0.00
1 mktemp 0.00 0.00
1 qdbus 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 stty 0.00 0.00
1 systemctl 0.00 0.00
1 template.sh 0.00 0.00
1 wc 0.00 0.00
1 xrandr 0.00 0.00
1 xset 0.00 0.00
26 processes running
47 maximum processes
Core computation blocks look as follows
6550) openvkl cpu=0 start=648.38 finish=1286.06
6551) vklBenchmarkCPU cpu=12 start=648.38 finish=1286.01
6552) vklBenchmarkCPU cpu=7 start=648.42 finish=1286.01
6554) vklBenchmarkCPU cpu=6 start=648.42 finish=1286.01
6557) vklBenchmarkCPU cpu=3 start=648.42 finish=1286.01
6565) vklBenchmarkCPU cpu=9 start=648.42 finish=1286.01
6566) vklBenchmarkCPU cpu=15 start=648.42 finish=1286.01
6558) vklBenchmarkCPU cpu=14 start=648.42 finish=1286.01
6556) vklBenchmarkCPU cpu=11 start=648.42 finish=1286.01
6561) vklBenchmarkCPU cpu=2 start=648.42 finish=1286.01
6563) vklBenchmarkCPU cpu=5 start=648.42 finish=1286.01
6553) vklBenchmarkCPU cpu=4 start=648.42 finish=1286.01
6555) vklBenchmarkCPU cpu=1 start=648.42 finish=1286.01
6560) vklBenchmarkCPU cpu=8 start=648.42 finish=1286.01
6562) vklBenchmarkCPU cpu=0 start=648.42 finish=1286.01
6564) vklBenchmarkCPU cpu=13 start=648.42 finish=1286.01
6559) vklBenchmarkCPU cpu=10 start=648.42 finish=1286.01