Audio to text translation using the OpenAI Whisper model and Whisper.cpp. Runs mostly on physical cores. There are three workloads: tiny, small and medium.
Topdown profile is dominated by backend stalls with not much frontend stalls.
AMD metrics confirm a backend bound application with twice as many memory stalls as CPU stalls. Frontend stalls are also very low.
elapsed 3281.110
on_cpu 0.469 # 7.50 / 16 cores
utime 24529.851
stime 71.720
nvcsw 789353 # 67.13%
nivcsw 386482 # 32.87%
inblock 272 # 0.08/sec
onblock 2000952 # 609.84/sec
cpu-clock 24748102587968 # 24748.103 seconds
task-clock 24748961686997 # 24748.962 seconds
page faults 11397536 # 460.526/sec
context switches 1191423 # 48.140/sec
cpu migrations 32105 # 1.297/sec
major page faults 2 # 0.000/sec
minor page faults 11397534 # 460.526/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 4572575710604 # 46.718 branches per 1000 inst
branch misses 14777271136 # 0.32% branch miss
conditional 4115973292256 # 42.053 conditional branches per 1000 inst
indirect 17517967739 # 0.179 indirect branches per 1000 inst
cpu-cycles 102097628650614 # 1.94 GHz
instructions 97507810361983 # 0.96 IPC
slots 205099946303142 #
retiring 37214286254367 # 18.1% (18.2%)
-- ucode 48968915545 # 0.0%
-- fastpath 37165317338822 # 18.1%
frontend 4284690737113 # 2.1% ( 2.1%) low
-- latency 3315556843362 # 1.6%
-- bandwidth 969133893751 # 0.5%
backend 162775146273176 # 79.4% (79.4%) high
-- cpu 47232427007135 # 23.0%
-- memory 115542719266041 # 56.3%
speculation 616667516467 # 0.3% ( 0.3%) low
-- branch mispredict 409532861682 # 0.2%
-- pipeline restart 207134654785 # 0.1%
smt-contention 209032937338 # 0.1% ( 0.0%)
cpu-cycles 101963426481978 # 1.94 GHz
instructions 97517015919776 # 0.96 IPC
instructions 32634325078974 # 131.264 l2 access per 1000 inst
l2 hit from l1 2265760655143 # 24.68% l2 miss
l2 miss from l1 283693287681 #
l2 hit from l2 pf 1244526315275 #
l3 hit from l2 pf 667413553522 #
l3 miss from l2 pf 106025961425 #
instructions 32607938489954 # 134.540 float per 1000 inst
float 512 76 # 0.000 AVX-512 per 1000 inst
float 256 637404 # 0.000 AVX-256 per 1000 inst
float 128 4387069981977 # 134.540 AVX-128 per 1000 inst
float MMX 0 # 0.000 MMX per 1000 inst
float scalar 0 # 0.000 scalar per 1000 inst
instructions 97825072073572 #
opcache 11470979872220 # 117.260 opcache per 1000 inst
opcache miss 293294675378 # 2.6% opcache miss rate
l1 dTLB miss 448855689540 # 4.588 L1 dTLB per 1000 inst
l2 dTLB miss 48011900496 # 0.491 L2 dTLB per 1000 inst
instructions 97843884531235 #
icache 458280019597 # 4.684 icache per 1000 inst
icache miss 76535669308 # 16.7% icache miss rate
l1 iTLB miss 35830986 # 0.000 L1 iTLB per 1000 inst
l2 iTLB miss 0 # 0.000 L2 iTLB per 1000 inst
tlb flush 676808 # 0.000 TLB flush per 1000 inst
Intel metrics show most memory stalls at the L2 level.
elapsed 6401.384
on_cpu 0.241 # 3.85 / 16 cores
utime 24581.149
stime 58.903
nvcsw 554296 # 80.38%
nivcsw 135330 # 19.62%
inblock 1965872 # 307.10/sec
onblock 1992264 # 311.22/sec
cpu-clock 24794386618554 # 24794.387 seconds
task-clock 24795763519490 # 24795.764 seconds
page faults 7532655 # 303.788/sec
context switches 720042 # 29.039/sec
cpu migrations 94201 # 3.799/sec
major page faults 91 # 0.004/sec
minor page faults 7532564 # 303.784/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 10341718642383 # 54.805 branches per 1000 inst
branch misses 9565794658 # 0.09% branch miss
conditional 10341734517295 # 54.805 conditional branches per 1000 inst
indirect 156368123592 # 0.829 indirect branches per 1000 inst
slots 532475805237926 #
retiring 251720082877249 # 47.3% (47.3%)
-- ucode 45341165034866 # 8.5%
-- fastpath 206378917842383 # 38.8%
frontend 39863638341533 # 7.5% ( 7.5%)
-- latency 2430803551860 # 0.5%
-- bandwidth 37432834789673 # 7.0%
backend 239695063602645 # 45.0% (45.0%)
-- cpu 184700781665297 # 34.7%
-- memory 54994281937348 # 10.3%
speculation 2347427794172 # 0.4% ( 0.4%) low
-- branch mispredict 826280176476 # 0.2%
-- pipeline restart 1521147617696 # 0.3%
smt-contention 0 # 0.0% ( 0.0%)
cpu-cycles 89422295522279 # 0.87 GHz
instructions 188568355555583 # 2.11 IPC
l2 access 9948057087090 # 53.102 l2 access per 1000 inst
l2 miss 4720943665861 # 47.46% l2 miss
cpu-cycles 88972264948894 # 18.1% memory latency
load stalls 15534009333939 # 0.0% l1 bound
l1 miss 35174852346207 # 24.3% l2 bound
l2 miss 13537304518777 # 5.0% l3 bound
l3 miss 9109399710512 # 10.2% dram bound
store_stalls 533063233128 # 0.6% store bound
Process summary shows almost all time in a .ape-1.10 program with many processes launched
651721 processes
648384 .ape-1.10 16390354.57 2833993.32
36 clinfo 4.10 2.25
38 vulkaninfo 1.33 0.96
6 php 0.15 0.89
4 vulkani:disk$0 0.14 0.11
2 llvmpipe-0 0.07 0.05
2 llvmpipe-1 0.07 0.05
2 llvmpipe-10 0.07 0.05
2 llvmpipe-11 0.07 0.05
2 llvmpipe-12 0.07 0.05
2 llvmpipe-13 0.07 0.05
2 llvmpipe-14 0.07 0.05
2 llvmpipe-15 0.07 0.05
2 llvmpipe-2 0.07 0.05
2 llvmpipe-3 0.07 0.05
2 llvmpipe-4 0.07 0.05
2 llvmpipe-5 0.07 0.05
2 llvmpipe-6 0.07 0.05
2 llvmpipe-7 0.07 0.05
2 llvmpipe-8 0.07 0.05
2 llvmpipe-9 0.07 0.05
6 clang 0.06 0.06
3 rocminfo 0.03 0.00
1 lspci 0.00 0.02
1 ps 0.00 0.01
97 sh 0.00 0.00
12 gcc 0.00 0.00
9 whisperfile 0.00 0.00
8 gsettings 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 gmain 0.00 0.00
5 phoronix-test-s 0.00 0.00
3 dconf worker 0.00 0.00
2 dmesg 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 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 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
3015 processes running
3062 maximum processes