Lammps is molecular dynamics code. A steady profile with a very high amount of branch mis-prediction.
AMD metrics show a lot of floating point and a moderate number of branches with a high misprediction ratio and a relatively higher IPC
mev@augusta:~/source/wspy/workload/phoronix/lammps-perf$ more amdtopdown.txt
elapsed 3607.385
on_cpu 0.497 # 7.95 / 16 cores
utime 28422.382
stime 266.904
nvcsw 34931 # 31.82%
nivcsw 74834 # 68.18%
inblock 808 # 0.22/sec
onblock 149424 # 41.42/sec
cpu-clock 28689723667470 # 28689.724 seconds
task-clock 28689830702354 # 28689.831 seconds
page faults 320132 # 11.158/sec
context switches 127549 # 4.446/sec
cpu migrations 42780 # 1.491/sec
major page faults 318 # 0.011/sec
minor page faults 319814 # 11.147/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 23793575093943 # 98.234 branches per 1000 inst
branch misses 851710978452 # 3.58% branch miss
conditional 21073569115948 # 87.004 conditional branches per 1000 inst
indirect 230267785218 # 0.951 indirect branches per 1000 inst
cpu-cycles 121063342816018 # 2.09 GHz
instructions 243280310806043 # 2.01 IPC
slots 242118479094498 #
retiring 81755137984811 # 33.8% (33.8%)
-- ucode 12236037330 # 0.0%
-- fastpath 81742901947481 # 33.8%
frontend 23418316746824 # 9.7% ( 9.7%)
-- latency 12051836963268 # 5.0%
-- bandwidth 11366479783556 # 4.7%
backend 85433362902042 # 35.3% (35.3%)
-- cpu 51382962662189 # 21.2%
-- memory 34050400239853 # 14.1%
speculation 51385443415715 # 21.2% (21.2%)
-- branch mispredict 51082009983348 # 21.1%
-- pipeline restart 303433432367 # 0.1%
smt-contention 126136389092 # 0.1% ( 0.0%)
cpu-cycles 120737853365517 # 2.09 GHz
instructions 241988029059643 # 2.00 IPC
instructions 80671577903670 # 63.872 l2 access per 1000 inst
l2 hit from l1 3637543520009 # 3.39% l2 miss
l2 miss from l1 58163463293 #
l2 hit from l2 pf 1398780866072 #
l3 hit from l2 pf 61743676247 #
l3 miss from l2 pf 54570665759 #
instructions 80620315692299 # 332.278 float per 1000 inst
float 512 34 # 0.000 AVX-512 per 1000 inst
float 256 1002 # 0.000 AVX-256 per 1000 inst
float 128 26788381930153 # 332.278 AVX-128 per 1000 inst
float MMX 0 # 0.000 MMX per 1000 inst
float scalar 0 # 0.000 scalar per 1000 inst
Intel metrics show a particularly high IPC and more time overall in cores.
elapsed 5234.464
on_cpu 0.747 # 11.96 / 16 cores
utime 61938.747
stime 643.461
nvcsw 55070 # 27.73%
nivcsw 143501 # 72.27%
inblock 296264 # 56.60/sec
onblock 149624 # 28.58/sec
cpu-clock 62583082310865 # 62583.082 seconds
task-clock 62583209388366 # 62583.209 seconds
page faults 340254 # 5.437/sec
context switches 224519 # 3.588/sec
cpu migrations 45949 # 0.734/sec
major page faults 2768 # 0.044/sec
minor page faults 337486 # 5.393/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 48031836251892 # 124.988 branches per 1000 inst
branch misses 875770490567 # 1.82% branch miss
conditional 48031836267444 # 124.988 conditional branches per 1000 inst
indirect 10260173632800 # 26.699 indirect branches per 1000 inst
slots 359968213625054 #
retiring 207191541248343 # 57.6% (57.6%)
-- ucode 10580047628562 # 2.9%
-- fastpath 196611493619781 # 54.6%
frontend 24396382454844 # 6.8% ( 6.8%)
-- latency 9601924139766 # 2.7%
-- bandwidth 14794458315078 # 4.1%
backend 52986590450881 # 14.7% (14.7%)
-- cpu 45057192033341 # 12.5%
-- memory 7929398417540 # 2.2%
speculation 76099342557612 # 21.1% (21.1%)
-- branch mispredict 74687668037222 # 20.7%
-- pipeline restart 1411674520390 # 0.4%
smt-contention 0 # 0.0% ( 0.0%)
cpu-cycles 166960049481381 # 1.99 GHz
instructions 597991842690222 # 3.58 IPC
l2 access 4182134029230 # 19.354 l2 access per 1000 inst
l2 miss 218708314722 # 5.23% l2 missMost time spent in the lmp process; we do abort prematurely at end of the run
237 processes
24 lmp 28521.62 274.58
19 vulkaninfo 0.46 0.38
2 vulkani:disk$0 0.05 0.04
6 clang 0.05 0.01
1 llvmpipe-0 0.03 0.02
1 llvmpipe-1 0.03 0.02
1 llvmpipe-10 0.03 0.02
1 llvmpipe-11 0.03 0.02
1 llvmpipe-12 0.03 0.02
1 llvmpipe-13 0.03 0.02
1 llvmpipe-14 0.03 0.02
1 llvmpipe-15 0.03 0.02
1 llvmpipe-2 0.03 0.02
1 llvmpipe-3 0.03 0.02
1 llvmpipe-4 0.03 0.02
1 llvmpipe-5 0.03 0.02
1 llvmpipe-6 0.03 0.02
1 llvmpipe-7 0.03 0.02
1 llvmpipe-8 0.03 0.02
1 llvmpipe-9 0.03 0.02
68 sh 0.00 0.00
13 gcc 0.00 0.00
9 stty 0.00 0.00
8 systemd-detect- 0.00 0.00
7 gsettings 0.00 0.00
7 stat 0.00 0.00
6 llvm-link 0.00 0.00
5 gmain 0.00 0.00
4 glxinfo 0.00 0.00
4 phoronix-test-s 0.00 0.00
3 dconf worker 0.00 0.00
2 grep 0.00 0.00
2 which 0.00 0.00
1 cc 0.00 0.00
1 clinfo 0.00 0.00
1 date 0.00 0.00
1 dirname 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 ps 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 setterm 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
14 processes running
47 maximum processes
Following is the profile until the end.
7702) ?? start=4.82 finish=0.00
7703) ?? start=4.82 finish=0.00
7707) ?? start=5.53 finish=0.00
7708) ?? start=6.18 finish=0.00
7709) ?? start=6.18 finish=0.00
7710) lmp start=6.21 finish=1206.38
7712) lmp start=6.22 finish=1206.38
7716) lmp start=6.22 finish=1206.38
7711) lmp start=6.21 finish=1206.38
7714) lmp start=6.22 finish=1206.38
7719) lmp start=6.23 finish=1206.38
7713) lmp start=6.22 finish=1206.38
7717) lmp start=6.22 finish=1206.38
7721) lmp start=6.23 finish=1206.38
7715) lmp start=6.22 finish=1206.38
7722) lmp start=6.23 finish=1206.38
7725) lmp start=6.23 finish=1206.38
7718) lmp start=6.22 finish=1206.38
7724) lmp start=6.23 finish=1206.38
7728) lmp start=6.23 finish=1206.38
7720) lmp start=6.23 finish=1206.38
7727) lmp start=6.23 finish=1206.38
7730) lmp start=6.24 finish=1206.38
7723) lmp start=6.23 finish=1206.38
7729) lmp start=6.24 finish=1206.38
7732) lmp start=6.24 finish=1206.38
7726) lmp start=6.23 finish=1206.38
7731) lmp start=6.24 finish=1206.38
7733) lmp start=6.24 finish=1206.38