A CPU and GPU numeric processing library, using both built-in CPU and OpenCL benchmarks. All run on my AMD system and the OpenCL fp16 fails on my Intel system. The OpenCL fp32 passes on Intel The AMD is considerably faster, so curious if I am getting some GPU? Looks like the first two workloads are multi-threaded and the rest are single-threaded.
Topdown profile shows frontend stalls as high and some variation between workloads and over time.
AMD metrics show little floating point and moderate numbers of branches. Some L2 access though not particularly high backend stalls.
elapsed 255.589
on_cpu 0.316 # 5.05 / 16 cores
utime 769.161
stime 522.034
nvcsw 84146 # 54.56%
nivcsw 70084 # 45.44%
inblock 0 # 0.00/sec
onblock 156304 # 611.54/sec
cpu-clock 1297086975088 # 1297.087 seconds
task-clock 1297163581400 # 1297.164 seconds
page faults 1828517 # 1409.627/sec
context switches 155290 # 119.715/sec
cpu migrations 1592 # 1.227/sec
major page faults 155 # 0.119/sec
minor page faults 1828362 # 1409.508/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 566661991334 # 91.881 branches per 1000 inst
branch misses 61259641104 # 10.81% branch miss
conditional 252994830141 # 41.022 conditional branches per 1000 inst
indirect 24440357358 # 3.963 indirect branches per 1000 inst
cpu-cycles 5083213642342 # 1.21 GHz
instructions 6172430147473 # 1.21 IPC
slots 10170139884144 #
retiring 2242627936735 # 22.1% (29.2%)
-- ucode 8206700190 # 0.1%
-- fastpath 2234421236545 # 22.0%
frontend 3184418817599 # 31.3% (41.5%)
-- latency 2649897791970 # 26.1%
-- bandwidth 534521025629 # 5.3%
backend 2238686350080 # 22.0% (29.2%)
-- cpu 1024364604864 # 10.1%
-- memory 1214321745216 # 11.9%
speculation 10863172263 # 0.1% ( 0.1%) low
-- branch mispredict 10849523258 # 0.1%
-- pipeline restart 13649005 # 0.0%
smt-contention 2493533444045 # 24.5% ( 0.0%)
cpu-cycles 5022114467177 # 1.22 GHz
instructions 6157130266342 # 1.23 IPC
instructions 2057106720487 # 49.333 l2 access per 1000 inst
l2 hit from l1 86581943288 # 9.06% l2 miss
l2 miss from l1 2752199852 #
l2 hit from l2 pf 8456824069 #
l3 hit from l2 pf 5607231135 #
l3 miss from l2 pf 836878029 #
instructions 2053827320707 # 21.171 float per 1000 inst
float 512 83 # 0.000 AVX-512 per 1000 inst
float 256 508 # 0.000 AVX-256 per 1000 inst
float 128 43480827313 # 21.171 AVX-128 per 1000 inst
float MMX 0 # 0.000 MMX per 1000 inst
float scalar 0 # 0.000 scalar per 1000 inst
instructions 2665431 #
opcache 988658 # 370.919 opcache per 1000 inst
opcache miss 530873 # 53.7% opcache miss rate
l1 dTLB miss 5558 # 2.085 L1 dTLB per 1000 inst
l2 dTLB miss 1178 # 0.442 L2 dTLB per 1000 inst
instructions 2715463 #
icache 1322587 # 487.058 icache per 1000 inst
icache miss 112382 # 8.5% icache miss rate
l1 iTLB miss 14 # 0.005 L1 iTLB per 1000 inst
l2 iTLB miss 0 # 0.000 L2 iTLB per 1000 inst
tlb flush 19 # 0.007 TLB flush per 1000 inst
Intel metrics show lower on-cpu and both L2 and dram stalls.
elapsed 749.501
on_cpu 0.116 # 1.85 / 16 cores
utime 1332.666
stime 55.026
nvcsw 515494 # 6.33%
nivcsw 7622108 # 93.67%
inblock 15760 # 21.03/sec
onblock 10456 # 13.95/sec
cpu-clock 1384483475706 # 1384.483 seconds
task-clock 1384896944201 # 1384.897 seconds
page faults 5647449 # 4077.884/sec
context switches 8141137 # 5878.515/sec
cpu migrations 120081 # 86.708/sec
major page faults 163 # 0.118/sec
minor page faults 5647286 # 4077.766/sec
alignment faults 0 # 0.000/sec
emulation faults 0 # 0.000/sec
branches 327470422030 # 29.701 branches per 1000 inst
branch misses 2265053860 # 0.69% branch miss
conditional 327470459310 # 29.701 conditional branches per 1000 inst
indirect 45635483920 # 4.139 indirect branches per 1000 inst
slots 21933232930778 #
retiring 12264288512056 # 55.9% (55.9%) high
-- ucode 701472801784 # 3.2%
-- fastpath 11562815710272 # 52.7%
frontend 2803267333224 # 12.8% (12.8%)
-- latency 1783243291316 # 8.1%
-- bandwidth 1020024041908 # 4.7%
backend 6871443097914 # 31.3% (31.3%)
-- cpu 3842821497310 # 17.5%
-- memory 3028621600604 # 13.8%
speculation 1485815766476 # 6.8% ( 6.8%)
-- branch mispredict 1451257286366 # 6.6%
-- pipeline restart 34558480110 # 0.2%
smt-contention 0 # 0.0% ( 0.0%)
cpu-cycles 5173265968393 # 0.39 GHz
instructions 14473440438126 # 2.80 IPC
l2 access 178180311971 # 17.066 l2 access per 1000 inst
l2 miss 83307874133 # 46.75% l2 miss
cpu-cycles 4579755625449 # 28.7% memory latency
load stalls 1298392731852 # 0.5% l1 bound
l1 miss 1273798954929 # 10.5% l2 bound
l2 miss 794601654431 # 5.6% l3 bound
l3 miss 539003203401 # 11.8% dram bound
store_stalls 17972241280 # 0.4% store bound
AMD metrics show most of the time in the blas_cpu process.
1198 processes
102 blas_cpu 12168.09 8389.18
51 cg_cpu 695.13 183.60
306 blas_opencl 373.02 658.65
153 cg_opencl 140.62 150.05
272 clinfo 74.48 24.64
38 vulkaninfo 1.14 1.14
6 glxinfo:gdrv0 0.13 0.03
6 glxinfo:gl0 0.13 0.03
4 vulkani:disk$0 0.12 0.12
6 php 0.10 0.15
2 glxinfo 0.08 0.02
2 glxinfo:cs0 0.08 0.02
2 glxinfo:disk$0 0.07 0.02
2 glxinfo:sh0 0.07 0.01
2 glxinfo:shlo0 0.07 0.01
2 llvmpipe-0 0.06 0.06
2 llvmpipe-1 0.06 0.06
2 llvmpipe-10 0.06 0.06
2 llvmpipe-11 0.06 0.06
2 llvmpipe-12 0.06 0.06
2 llvmpipe-13 0.06 0.06
2 llvmpipe-14 0.06 0.06
2 llvmpipe-15 0.06 0.06
2 llvmpipe-2 0.06 0.06
2 llvmpipe-3 0.06 0.06
2 llvmpipe-4 0.06 0.06
2 llvmpipe-5 0.06 0.06
2 llvmpipe-6 0.06 0.06
2 llvmpipe-7 0.06 0.06
2 llvmpipe-8 0.06 0.06
2 llvmpipe-9 0.06 0.06
6 clang 0.05 0.05
3 rocminfo 0.03 0.03
1 lspci 0.01 0.02
1 ps 0.00 0.01
98 sh 0.00 0.00
18 arrayfire 0.00 0.00
13 gcc 0.00 0.00
11 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 phoronix-test-s 0.00 0.00
3 gmain 0.00 0.00
2 cc 0.00 0.00
2 dconf worker 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 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
0 processes running
59 maximum processes
Computation blocks look as follows
363808) arrayfire cpu=5 start=6.95 finish=23.75
363809) blas_cpu cpu=7 start=6.95 finish=23.73
363810) blas_cpu cpu=14 start=6.96 finish=23.73
363811) blas_cpu cpu=10 start=6.96 finish=23.73
363812) blas_cpu cpu=4 start=6.96 finish=23.73
363813) blas_cpu cpu=9 start=6.96 finish=23.73
363814) blas_cpu cpu=8 start=6.96 finish=23.73
363815) blas_cpu cpu=3 start=6.96 finish=23.73
363816) blas_cpu cpu=5 start=6.96 finish=23.73
363817) blas_cpu cpu=15 start=6.96 finish=23.73
363818) blas_cpu cpu=0 start=6.96 finish=23.73
363819) blas_cpu cpu=6 start=6.96 finish=23.73
363820) blas_cpu cpu=12 start=6.96 finish=23.73
363821) blas_cpu cpu=1 start=6.96 finish=23.73
363822) blas_cpu cpu=11 start=6.96 finish=23.73
363823) blas_cpu cpu=2 start=6.96 finish=23.72
363824) blas_cpu cpu=13 start=6.96 finish=23.72
363825) blas_cpu cpu=7 start=6.96 finish=23.73