-
Amira Abdel-Rahman (admin) authoredAmira Abdel-Rahman (admin) authored
numbers.md 6.58 KiB
------------------------------------------------
V100
------------------------------------------------
7 TFlop double
14 TFlop single
32 GB/sec PCIe
300 GB/sec NVLINK
------------------------------------------------
1x1x4 chiral
------------------------------------------------
input 1058444 particles
mps: 254 srt: 625 frc: 3521 int: 106 cpy: 2424 out: 2.8e+05 (us))
------------------------------------------------
coupon
------------------------------------------------
input 986154 particles
mps: 277 srt: 289 frc: 3242 int: 91 cpy: 2251 out: 2.72e+05 (us))
------------------------------------------------
simulate
------------------------------------------------
14 TFlop / 277 Mpps = 50541 ops/particle
------------------------------------------------
integrate
------------------------------------------------
986154 particles / 91e-6 s = 10.8 G/s
14 TFlop / (986154 particles / 91e-6 s) = 1291 ops/integrate
------------------------------------------------
copy
------------------------------------------------
986154 particles * 7 arrays * 4 bytes / 2251 us
= 12266 = 12 GB/s
------------------------------------------------
GPU_check
------------------------------------------------
peer access:
from 0 to 1: yes
from 0 to 2: yes
from 0 to 3: yes
from 0 to 4: yes
from 0 to 5: no
from 0 to 6: no
from 0 to 7: no
from 1 to 0: yes
from 1 to 2: yes
from 1 to 3: yes
from 1 to 4: no
from 1 to 5: yes
from 1 to 6: no
from 1 to 7: no
from 2 to 0: yes
from 2 to 1: yes
from 2 to 3: yes
from 2 to 4: no
from 2 to 5: no
from 2 to 6: yes
from 2 to 7: no
from 3 to 0: yes
from 3 to 1: yes
from 3 to 2: yes
from 3 to 4: no
from 3 to 5: no
from 3 to 6: no
from 3 to 7: yes
from 4 to 0: yes
from 4 to 1: no
from 4 to 2: no
from 4 to 3: no
from 4 to 5: yes
from 4 to 6: yes
from 4 to 7: yes
from 5 to 0: no
from 5 to 1: yes
from 5 to 2: no
from 5 to 3: no
from 5 to 4: yes
from 5 to 6: yes
from 5 to 7: yes
from 6 to 0: no
from 6 to 1: no
from 6 to 2: yes
from 6 to 3: no
from 6 to 4: yes
from 6 to 5: yes
from 6 to 7: yes
from 7 to 0: no
from 7 to 1: no
from 7 to 2: no
from 7 to 3: yes
from 7 to 4: yes
from 7 to 5: yes
from 7 to 6: yes
GPUs:
number: 0
name: Tesla V100-SXM2-16GB
global memory: 16945512448
max grid size: 2147483647
max threads per block: 1024
max threads dimension: 1024
multiprocessor count: 80
max threads per multiprocessor: 2048
number: 1
name: Tesla V100-SXM2-16GB
global memory: 16945512448
max grid size: 2147483647
max threads per block: 1024
max threads dimension: 1024
multiprocessor count: 80
max threads per multiprocessor: 2048
number: 2
name: Tesla V100-SXM2-16GB
global memory: 16945512448
max grid size: 2147483647
max threads per block: 1024
max threads dimension: 1024
multiprocessor count: 80
max threads per multiprocessor: 2048
number: 3
name: Tesla V100-SXM2-16GB
global memory: 16945512448
max grid size: 2147483647
max threads per block: 1024
max threads dimension: 1024
multiprocessor count: 80
max threads per multiprocessor: 2048
number: 4
name: Tesla V100-SXM2-16GB
global memory: 16945512448
max grid size: 2147483647
max threads per block: 1024
max threads dimension: 1024
multiprocessor count: 80
max threads per multiprocessor: 2048
number: 5
name: Tesla V100-SXM2-16GB
global memory: 16945512448
max grid size: 2147483647
max threads per block: 1024
max threads dimension: 1024
multiprocessor count: 80
max threads per multiprocessor: 2048
number: 6
name: Tesla V100-SXM2-16GB
global memory: 16945512448
max grid size: 2147483647
max threads per block: 1024
max threads dimension: 1024
multiprocessor count: 80
max threads per multiprocessor: 2048
number: 7
name: Tesla V100-SXM2-16GB
global memory: 16945512448
max grid size: 2147483647
max threads per block: 1024
max threads dimension: 1024
multiprocessor count: 80
max threads per multiprocessor: 2048
copy 5000000 floats from CPU to GPU
6854.000000 us, 2.918e+09 B/s
1846.000000 us, 1.08342e+10 B/s pinned
copy 5000000 floats from GPU to GPU 0:
GPU 1: 849.000000 us, 2.35571e+10 B/s
GPU 2: 851.000000 us, 2.35018e+10 B/s
GPU 3: 443.000000 us, 4.51467e+10 B/s
GPU 4: 444.000000 us, 4.5045e+10 B/s
GPU 5: 2051.000000 us, 9.75134e+09 B/s
GPU 6: 2026.000000 us, 9.87167e+09 B/s
GPU 7: 2017.000000 us, 9.91572e+09 B/s
add 5000000x5000000 floats:
3.747896 s, 6670.409500 G/s
peer add 5000000x5000000 GPU 0 floats:
GPU 1: 3.755309 s, 6657.241500 G/s
GPU 2: 3.731777 s, 6699.221000 G/s
GPU 3: 3.735422 s, 6692.684500 G/s
GPU 4: 3.736765 s, 6690.278500 G/s
parallel peer add 5000000x5000000 GPU 0 floats:
5 GPUS: 0.783563 s, 19940.962000 G/s
------------------------------------------------
CPU_check
------------------------------------------------
processor : 95
vendor_id : GenuineIntel
cpu family : 6
model : 85
model name : Intel(R) Xeon(R) Platinum 8275CL CPU @ 3.00GHz
stepping : 7
microcode : 0x5002f01
cpu MHz : 1221.613
cache size : 36608 KB
physical id : 1
siblings : 48
core id : 23
cpu cores : 24
apicid : 111
initial apicid : 111
fpu : yes
fpu_exception : yes
cpuid level : 13
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single pti fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves ida arat pku ospke avx512_vnni
bugs : cpu_meltdown spectre_v1 spectre_v2 spec_store_bypass l1tf mds swapgs itlb_multihit
bogomips : 5999.99
clflush size : 64
cache_alignment : 64
address sizes : 46 bits physical, 48 bits virtual
power management:
add 500000x500000 floats with 1 thread:
65.058972 s, 3.842667 G/s
add 500000x500000 floats with 96 threads:
1.376176 s, 181.662813 G/s
------------------------------------------------
pipe_check
------------------------------------------------
send: 100000000 points
receive: 4.301932 s, 9.29815e+07 B/s