x265 benchmark with standard input and two workloads
AMD metrics show a 128-bit floating point program with some L2 misses and backend memory but otherwise medium range retiring. We spend ~50% of time on the CPU so threaded but not fully threaded operation?
elapsed 713.905
on_cpu 0.502 # 8.03 / 16 cores
utime 5668.028
stime 66.483
nvcsw 2807701 # 92.74%
nivcsw 219710 # 7.26%
inblock 6170832
onblock 27632
cpu-clock 5730360170277 # 5730.360 seconds
task-clock 5731795930958 # 5731.796 seconds
page faults 6120568 # 1067.827/sec
context switches 3030769 # 528.764/sec
cpu migrations 176614 # 30.813/sec
major page faults 67 # 0.012/sec
minor page faults 6120501 # 1067.816/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 2298223476101 # 64.254 branches per 1000 inst
branch misses 38765597182 # 1.69% branch miss
conditional 1447123909275 # 40.459 conditional branches per 1000 inst
indirect 184907459078 # 5.170 indirect branches per 1000 inst
cpu-cycles 6226498530298 # 1.98 GHz
instructions 9678813453851 # 1.55 IPC
slots 12467027650776 #
retiring 3419335741620 # 27.4% (35.4%)
-- ucode 47510862855 # 0.4%
-- fastpath 3371824878765 # 27.0%
frontend 1352978437698 # 10.9% (14.0%)
-- latency 861177646632 # 6.9%
-- bandwidth 491800791066 # 3.9%
backend 4618690961014 # 37.0% (47.8%)
-- cpu 1412014283462 # 11.3%
-- memory 3206676677552 # 25.7%
speculation 263046398653 # 2.1% ( 2.7%)
-- branch mispredict 242439249427 # 1.9%
-- pipeline restart 20607149226 # 0.2%
smt-contention 2812859705014 # 22.6% ( 0.0%)
cpu-cycles 7936497399423 # 2.00 GHz
instructions 12283162200513 # 1.55 IPC
instructions 4099519478946 # 39.677 l2 access per 1000 inst
l2 hit from l1 134192432419 # 12.06% l2 miss
l2 miss from l1 12511817133 #
l2 hit from l2 pf 21366966534 #
l3 hit from l2 pf 3874073579 #
l3 miss from l2 pf 3222493799 #
instructions 4091045784055 # 190.915 float per 1000 inst
float 512 71 # 0.000 AVX-512 per 1000 inst
float 256 1060 # 0.000 AVX-256 per 1000 inst
float 128 781040286672 # 190.915 AVX-128 per 1000 inst
float MMX 0 # 0.000 MMX per 1000 inst
float scalar 0 # 0.000 scalar per 1000 instCorresponding Intel metrics shows >50% retiring rate.
elapsed 788.657
on_cpu 0.610 # 9.76 / 16 cores
utime 7603.378
stime 90.091
nvcsw 3688077 # 90.89%
nivcsw 369791 # 9.11%
inblock 31549176
onblock 26776
cpu-clock 7679288766057 # 7679.289 seconds
task-clock 7681427192071 # 7681.427 seconds
page faults 6542506 # 851.731/sec
context switches 4061655 # 528.763/sec
cpu migrations 594059 # 77.337/sec
major page faults 793 # 0.103/sec
minor page faults 6541713 # 851.627/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 2477806622104 # 64.408 branches per 1000 inst
branch misses 43684720181 # 1.76% branch miss
conditional 2477806644216 # 64.408 conditional branches per 1000 inst
indirect 880974334257 # 22.900 indirect branches per 1000 inst
slots 10702457016518 #
retiring 5872152202319 # 54.9% (54.9%)
-- ucode 373372570662 # 3.5%
-- fastpath 5498779631657 # 51.4%
frontend 1885584589158 # 17.6% (17.6%)
-- latency 953145946637 # 8.9%
-- bandwidth 932438642521 # 8.7%
backend 2105534559715 # 19.7% (19.7%)
-- cpu 981934933436 # 9.2%
-- memory 1123599626279 # 10.5%
speculation 939803939689 # 8.8% ( 8.8%)
-- branch mispredict 884451371813 # 8.3%
-- pipeline restart 55352567876 # 0.5%
smt-contention 0 # 0.0% ( 0.0%)
cpu-cycles 6467518831703 # 1.91 GHz
instructions 12325634319096 # 1.91 IPC
l2 access 173564026661 # 31.787 l2 access per 1000 inst
l2 miss 36974658876 # 21.30% l2 miss
No surprises from the process summary, all time spent in x265 process
434 processes
150 x265 36013.72 343.12
38 vulkaninfo 0.76 1.14
6 glxinfo:gdrv0 0.16 0.13
2 glxinfo 0.12 0.05
2 glxinfo:cs0 0.12 0.05
2 glxinfo:disk$0 0.12 0.05
2 glxinfo:sh0 0.12 0.05
2 glxinfo:shlo0 0.12 0.05
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 php 0.03 0.14
6 clang 0.02 0.06
1 lspci 0.00 0.03
88 sh 0.00 0.00
13 gcc 0.00 0.00
9 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 phoronix-test-s 0.00 0.00
4 gmain 0.00 0.00
3 dconf worker 0.00 0.00
2 cc 0.00 0.00
2 lscpu 0.00 0.00
2 uname 0.00 0.00
2 which 0.00 0.00
2 xset 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 grep 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
1 xrandr 0.00 0.00
The core blocks start slightly more threads than cores
1666842) x265 start=4.92 finish=52.17
1666843) x265 start=4.92 finish=52.16
1666844) x265 start=4.92 finish=50.25
1666845) x265 start=4.92 finish=52.06
1666846) x265 start=4.92 finish=52.06
1666847) x265 start=4.92 finish=52.06
1666848) x265 start=4.92 finish=52.06
1666849) x265 start=4.92 finish=52.06
1666850) x265 start=4.92 finish=52.06
1666851) x265 start=4.92 finish=52.06
1666852) x265 start=4.92 finish=52.06
1666853) x265 start=4.92 finish=52.06
1666854) x265 start=4.92 finish=52.06
1666855) x265 start=4.92 finish=52.06
1666856) x265 start=4.92 finish=52.06
1666857) x265 start=4.92 finish=52.06
1666858) x265 start=4.92 finish=52.06
1666859) x265 start=4.92 finish=52.06
1666860) x265 start=4.92 finish=52.06
1666861) x265 start=4.92 finish=52.06
1666862) x265 start=4.94 finish=52.06
1666863) x265 start=4.94 finish=52.06
1666864) x265 start=4.94 finish=52.06
1666865) x265 start=4.94 finish=52.16
1666866) x265 start=4.94 finish=50.58What accounts for the relatively lower amount of on-core time? The AMD metrics provide one clue with the number of inblock/outblock operations (need to update the metrics to provide these as a rate).