Code Monkey home page Code Monkey logo

nvmlc's Introduction

nvmlc

Features

  1. latency matrix (excluding NUMA nodes) with default message size
  2. bandwidth matrix (excluding NUMA nodes) with default message size
    • bandwidth between host and gpuA
    • bandwidth between gpuA and gpuA
    • bandwidth between gpuA and gpuB

Compare

./p2pBandwidthLatencyTest

[P2P (Peer-to-Peer) GPU Bandwidth Latency Test]
Device: 0, NVIDIA A100-SXM4-80GB, pciBusID: 65, pciDeviceID: 1, pciDomainID:0
Device: 1, NVIDIA A100-SXM4-80GB, pciBusID: 65, pciDeviceID: 2, pciDomainID:0
Device: 2, NVIDIA A100-SXM4-80GB, pciBusID: 67, pciDeviceID: 1, pciDomainID:0
Device: 3, NVIDIA A100-SXM4-80GB, pciBusID: 67, pciDeviceID: 2, pciDomainID:0
Device: 4, NVIDIA A100-SXM4-80GB, pciBusID: 69, pciDeviceID: 1, pciDomainID:0
Device: 5, NVIDIA A100-SXM4-80GB, pciBusID: 69, pciDeviceID: 2, pciDomainID:0
Device: 6, NVIDIA A100-SXM4-80GB, pciBusID: 6b, pciDeviceID: 1, pciDomainID:0
Device: 7, NVIDIA A100-SXM4-80GB, pciBusID: 6b, pciDeviceID: 2, pciDomainID:0
Device=0 CAN Access Peer Device=1
Device=0 CAN Access Peer Device=2
Device=0 CAN Access Peer Device=3
Device=0 CAN Access Peer Device=4
Device=0 CAN Access Peer Device=5
Device=0 CAN Access Peer Device=6
Device=0 CAN Access Peer Device=7
Device=1 CAN Access Peer Device=0
Device=1 CAN Access Peer Device=2
Device=1 CAN Access Peer Device=3
Device=1 CAN Access Peer Device=4
Device=1 CAN Access Peer Device=5
Device=1 CAN Access Peer Device=6
Device=1 CAN Access Peer Device=7
Device=2 CAN Access Peer Device=0
Device=2 CAN Access Peer Device=1
Device=2 CAN Access Peer Device=3
Device=2 CAN Access Peer Device=4
Device=2 CAN Access Peer Device=5
Device=2 CAN Access Peer Device=6
Device=2 CAN Access Peer Device=7
Device=3 CAN Access Peer Device=0
Device=3 CAN Access Peer Device=1
Device=3 CAN Access Peer Device=2
Device=3 CAN Access Peer Device=4
Device=3 CAN Access Peer Device=5
Device=3 CAN Access Peer Device=6
Device=3 CAN Access Peer Device=7
Device=4 CAN Access Peer Device=0
Device=4 CAN Access Peer Device=1
Device=4 CAN Access Peer Device=2
Device=4 CAN Access Peer Device=3
Device=4 CAN Access Peer Device=5
Device=4 CAN Access Peer Device=6
Device=4 CAN Access Peer Device=7
Device=5 CAN Access Peer Device=0
Device=5 CAN Access Peer Device=1
Device=5 CAN Access Peer Device=2
Device=5 CAN Access Peer Device=3
Device=5 CAN Access Peer Device=4
Device=5 CAN Access Peer Device=6
Device=5 CAN Access Peer Device=7
Device=6 CAN Access Peer Device=0
Device=6 CAN Access Peer Device=1
Device=6 CAN Access Peer Device=2
Device=6 CAN Access Peer Device=3
Device=6 CAN Access Peer Device=4
Device=6 CAN Access Peer Device=5
Device=6 CAN Access Peer Device=7
Device=7 CAN Access Peer Device=0
Device=7 CAN Access Peer Device=1
Device=7 CAN Access Peer Device=2
Device=7 CAN Access Peer Device=3
Device=7 CAN Access Peer Device=4
Device=7 CAN Access Peer Device=5
Device=7 CAN Access Peer Device=6

***NOTE: In case a device doesn't have P2P access to other one, it falls back to normal memcopy procedure.
So you can see lesser Bandwidth (GB/s) and unstable Latency (us) in those cases.

P2P Connectivity Matrix
     D\D     0     1     2     3     4     5     6     7
     0	     1     1     1     1     1     1     1     1
     1	     1     1     1     1     1     1     1     1
     2	     1     1     1     1     1     1     1     1
     3	     1     1     1     1     1     1     1     1
     4	     1     1     1     1     1     1     1     1
     5	     1     1     1     1     1     1     1     1
     6	     1     1     1     1     1     1     1     1
     7	     1     1     1     1     1     1     1     1
Unidirectional P2P=Disabled Bandwidth Matrix (GB/s)
   D\D     0      1      2      3      4      5      6      7
     0 1524.39  16.19  15.63  15.14  13.76  13.87  14.31  13.36
     1  16.37 1536.38  15.58  15.16  13.72  13.80  14.33  13.37
     2  19.47  19.63 1545.50   9.76  13.69  13.85  14.29  13.34
     3  20.03  19.99  10.10 1568.78  13.71  13.84  14.41  13.38
     4  19.64  19.59  14.22  13.80 1573.51   8.16  13.33  12.47
     5  20.52  20.49  14.22  13.85   8.18 1562.50  13.36  12.48
     6  19.62  19.64  14.20  13.85  12.83  12.89 1560.94   7.82
     7  20.43  20.53  14.05  13.79  12.78  12.87   8.49 1567.20
Unidirectional P2P=Enabled Bandwidth (P2P Writes) Matrix (GB/s)
   D\D     0      1      2      3      4      5      6      7
     0 1521.42 274.23 272.17 273.72 273.60 271.96 275.08 275.02
     1 272.17 1591.14 273.70 271.87 275.84 275.32 275.10 273.31
     2 269.59 276.62 1548.56 270.21 275.31 274.83 274.62 271.72
     3 270.55 274.76 271.05 1550.10 273.97 275.01 275.05 273.35
     4 271.90 275.90 272.28 274.25 1586.29 275.49 273.56 273.70
     5 270.91 275.01 273.29 272.71 271.79 1583.08 275.52 274.73
     6 273.04 275.33 273.44 273.48 274.11 274.99 1579.88 274.45
     7 273.52 275.08 273.53 274.10 274.61 275.67 274.32 1576.69
Bidirectional P2P=Disabled Bandwidth Matrix (GB/s)
   D\D     0      1      2      3      4      5      6      7
     0 1550.87  15.77  21.91  21.20  19.98  20.28  20.71  20.04
     1  15.48 1599.28  21.80  21.14  19.86  20.17  20.64  20.06
     2  20.00  20.10 1563.28  10.85  15.23  15.37  15.57  14.43
     3  19.83  19.86  10.84 1599.28  14.95  15.09  15.32  14.19
     4  19.37  19.27  15.94  15.64 1598.47   8.96  14.80  13.69
     5  19.74  19.65  16.12  15.75   8.94 1595.20  14.84  13.80
     6  20.14  20.16  16.28  15.99  14.73  14.81 1596.83   8.50
     7  19.13  19.20  14.83  14.55  13.60  13.71   8.44 1592.76
Bidirectional P2P=Enabled Bandwidth Matrix (GB/s)
   D\D     0      1      2      3      4      5      6      7
     0 1551.64 424.90 425.24 424.32 424.53 424.99 425.31 424.54
     1 426.59 1608.34 514.94 518.54 518.17 519.89 519.59 517.15
     2 426.83 519.05 1591.14 512.07 513.74 513.39 512.56 513.41
     3 425.89 519.39 515.84 1600.10 519.72 522.48 519.72 520.25
     4 426.34 519.19 514.97 519.38 1595.20 516.99 518.88 519.22
     5 425.42 520.41 515.32 519.90 519.05 1596.83 518.07 521.19
     6 427.54 517.65 515.59 517.49 517.84 518.53 1588.71 518.19
     7 427.34 519.37 516.10 517.83 518.19 519.74 519.38 1592.76
P2P=Disabled Latency Matrix (us)
   GPU     0      1      2      3      4      5      6      7
     0   2.87  18.16  23.59  21.82  23.63  23.43  23.61  23.61
     1  23.50   2.08  21.71  21.56  23.54  17.95  23.53  23.37
     2  23.49  20.39   2.25  21.63  23.55  23.61  23.62  23.46
     3  22.10  21.55  22.09   2.08  23.57  23.44  23.49  23.60
     4  23.56  22.89  23.44  23.16   2.19  23.37  23.47  23.47
     5  23.57  23.56  23.58  23.46  21.14   2.16  18.30  19.88
     6  23.59  23.39  23.57  23.47  21.97  21.14   2.47  21.42
     7  23.68  23.43  23.48  23.31  23.26  15.81  21.05   2.31

   CPU     0      1      2      3      4      5      6      7
     0   3.57   5.39   5.09   5.18   5.63   5.51   5.51   5.52
     1   5.22   3.40   5.16   5.04   5.52   5.48   5.43   5.38
     2   4.88   4.78   3.30   4.94   5.36   5.36   5.42   5.37
     3   4.89   4.73   4.81   2.62   5.35   5.53   5.41   5.35
     4   5.26   5.19   5.21   5.18   2.77   5.79   5.80   5.78
     5   5.16   5.05   5.06   5.16   5.56   2.79   5.75   5.70
     6   5.20   5.11   5.16   5.19   5.69   5.72   2.82   5.75
     7   5.77   5.25   5.24   5.40   5.67   5.65   5.77   3.57
P2P=Enabled Latency (P2P Writes) Matrix (us)
   GPU     0      1      2      3      4      5      6      7
     0   2.86   3.24   3.22   3.33   3.22   3.22   3.26   3.23
     1   2.72   2.08   2.78   2.79   2.86   2.79   2.78   2.78
     2   2.82   2.81   2.25   2.80   2.81   2.82   2.87   2.81
     3   2.91   2.86   2.80   2.07   2.85   2.79   2.79   2.92
     4   2.86   2.90   2.86   2.86   2.18   2.87   2.86   2.93
     5   2.83   2.85   2.78   2.79   2.79   2.16   2.84   2.80
     6   2.82   2.81   2.88   2.81   2.86   2.88   2.46   2.82
     7   2.86   2.86   2.92   2.87   2.86   2.93   2.85   2.30

   CPU     0      1      2      3      4      5      6      7
     0   3.54   1.92   1.84   1.84   1.92   1.92   1.84   1.98
     1   1.99   3.41   1.85   1.89   1.84   1.82   1.99   1.87
     2   1.77   1.72   3.56   1.71   1.71   1.71   1.68   1.69
     3   1.77   1.77   1.73   2.70   1.69   1.72   1.73   1.69
     4   1.91   1.86   1.88   1.92   2.96   1.92   1.84   1.83
     5   1.97   1.89   1.89   1.86   1.89   2.89   1.85   1.95
     6   1.89   1.86   1.85   1.84   1.90   1.83   2.91   1.82
     7   2.13   2.05   2.06   2.04   1.97   2.05   2.08   3.69

NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.


./p2pBandwidthLatencyTest --sm_copy

[P2P (Peer-to-Peer) GPU Bandwidth Latency Test]
Device: 0, NVIDIA A100-SXM4-80GB, pciBusID: 65, pciDeviceID: 1, pciDomainID:0
Device: 1, NVIDIA A100-SXM4-80GB, pciBusID: 65, pciDeviceID: 2, pciDomainID:0
Device: 2, NVIDIA A100-SXM4-80GB, pciBusID: 67, pciDeviceID: 1, pciDomainID:0
Device: 3, NVIDIA A100-SXM4-80GB, pciBusID: 67, pciDeviceID: 2, pciDomainID:0
Device: 4, NVIDIA A100-SXM4-80GB, pciBusID: 69, pciDeviceID: 1, pciDomainID:0
Device: 5, NVIDIA A100-SXM4-80GB, pciBusID: 69, pciDeviceID: 2, pciDomainID:0
Device: 6, NVIDIA A100-SXM4-80GB, pciBusID: 6b, pciDeviceID: 1, pciDomainID:0
Device: 7, NVIDIA A100-SXM4-80GB, pciBusID: 6b, pciDeviceID: 2, pciDomainID:0
Device=0 CAN Access Peer Device=1
Device=0 CAN Access Peer Device=2
Device=0 CAN Access Peer Device=3
Device=0 CAN Access Peer Device=4
Device=0 CAN Access Peer Device=5
Device=0 CAN Access Peer Device=6
Device=0 CAN Access Peer Device=7
Device=1 CAN Access Peer Device=0
Device=1 CAN Access Peer Device=2
Device=1 CAN Access Peer Device=3
Device=1 CAN Access Peer Device=4
Device=1 CAN Access Peer Device=5
Device=1 CAN Access Peer Device=6
Device=1 CAN Access Peer Device=7
Device=2 CAN Access Peer Device=0
Device=2 CAN Access Peer Device=1
Device=2 CAN Access Peer Device=3
Device=2 CAN Access Peer Device=4
Device=2 CAN Access Peer Device=5
Device=2 CAN Access Peer Device=6
Device=2 CAN Access Peer Device=7
Device=3 CAN Access Peer Device=0
Device=3 CAN Access Peer Device=1
Device=3 CAN Access Peer Device=2
Device=3 CAN Access Peer Device=4
Device=3 CAN Access Peer Device=5
Device=3 CAN Access Peer Device=6
Device=3 CAN Access Peer Device=7
Device=4 CAN Access Peer Device=0
Device=4 CAN Access Peer Device=1
Device=4 CAN Access Peer Device=2
Device=4 CAN Access Peer Device=3
Device=4 CAN Access Peer Device=5
Device=4 CAN Access Peer Device=6
Device=4 CAN Access Peer Device=7
Device=5 CAN Access Peer Device=0
Device=5 CAN Access Peer Device=1
Device=5 CAN Access Peer Device=2
Device=5 CAN Access Peer Device=3
Device=5 CAN Access Peer Device=4
Device=5 CAN Access Peer Device=6
Device=5 CAN Access Peer Device=7
Device=6 CAN Access Peer Device=0
Device=6 CAN Access Peer Device=1
Device=6 CAN Access Peer Device=2
Device=6 CAN Access Peer Device=3
Device=6 CAN Access Peer Device=4
Device=6 CAN Access Peer Device=5
Device=6 CAN Access Peer Device=7
Device=7 CAN Access Peer Device=0
Device=7 CAN Access Peer Device=1
Device=7 CAN Access Peer Device=2
Device=7 CAN Access Peer Device=3
Device=7 CAN Access Peer Device=4
Device=7 CAN Access Peer Device=5
Device=7 CAN Access Peer Device=6

***NOTE: In case a device doesn't have P2P access to other one, it falls back to normal memcopy procedure.
So you can see lesser Bandwidth (GB/s) and unstable Latency (us) in those cases.

P2P Connectivity Matrix
     D\D     0     1     2     3     4     5     6     7
     0	     1     1     1     1     1     1     1     1
     1	     1     1     1     1     1     1     1     1
     2	     1     1     1     1     1     1     1     1
     3	     1     1     1     1     1     1     1     1
     4	     1     1     1     1     1     1     1     1
     5	     1     1     1     1     1     1     1     1
     6	     1     1     1     1     1     1     1     1
     7	     1     1     1     1     1     1     1     1
Unidirectional P2P=Disabled Bandwidth Matrix (GB/s)
   D\D     0      1      2      3      4      5      6      7
     0 1521.42  16.15  19.60  19.99  20.35  20.32  20.40  13.78
     1  16.39 1534.87  19.39  19.52  20.32  20.31  20.33  13.72
     2  19.48  19.55 1531.86  15.02  20.29  20.31  20.31  13.74
     3  19.65  19.57  16.33 1543.97  19.63  19.69  20.35  13.72
     4  19.69  19.59  19.89  20.52 1568.78  15.51  20.60  12.66
     5  19.68  19.59  19.68  20.33  15.67 1573.51  20.68  12.76
     6  19.81  19.54  20.47  20.53  20.74  20.69 1565.63   8.31
     7  19.78  19.75  19.67  20.49  20.78  20.66  15.75 1567.20
Unidirectional P2P=Enabled Bandwidth (P2P Writes) Matrix (GB/s)
   D\D     0      1      2      3      4      5      6      7
     0 1530.36 205.00 205.97 205.05 205.27 205.81 205.54 204.94
     1 204.94 1545.50 205.75 205.59 202.97 205.70 205.32 205.48
     2 220.07 221.00 1584.69 225.79 200.58 200.53 200.94 201.09
     3 221.69 221.95 221.88 1586.29 226.58 226.25 226.19 225.79
     4 221.51 222.45 222.01 225.21 1589.52 226.19 226.78 225.60
     5 224.76 225.40 225.34 231.41 228.90 1579.88 231.28 230.39
     6 226.06 225.66 225.47 231.21 231.34 231.21 1587.91 230.32
     7 221.76 222.26 222.20 226.38 225.60 226.12 226.84 1583.08
Bidirectional P2P=Disabled Bandwidth Matrix (GB/s)
   D\D     0      1      2      3      4      5      6      7
     0 1551.64  15.83  27.28  27.45  28.68  28.62  28.94  20.49
     1  15.84 1560.94  26.86  27.51  27.74  27.85  28.83  20.54
     2  27.67  27.52 1592.76  15.88  27.82  27.98  27.91  20.94
     3  27.93  27.80  15.46 1600.92  28.67  28.73  27.80  20.95
     4  28.64  28.50  28.54  28.55 1599.28  15.69  27.97  19.71
     5  27.73  28.59  28.55  28.58  15.80 1598.47  27.86  19.70
     6  27.20  27.23  28.70  28.76  29.19  29.06 1596.02  12.18
     7  20.03  20.06  20.07  20.30  18.04  18.05  12.27 1595.20
Bidirectional P2P=Enabled Bandwidth Matrix (GB/s)
   D\D     0      1      2      3      4      5      6      7
     0 1550.10 314.07 306.73 313.94 324.30 323.97 323.97 324.71
     1 314.96 1564.06 307.16 315.40 323.50 324.04 324.04 323.03
     2 307.34 308.25 1565.63 308.25 316.36 316.42 316.87 316.10
     3 314.70 315.27 306.01 1559.38 324.04 324.17 323.63 322.90
     4 323.23 322.83 315.47 322.83 1601.74 387.62 387.33 388.01
     5 322.96 323.30 315.40 323.23 387.72 1593.57 388.10 386.95
     6 322.96 322.90 315.21 323.16 387.43 388.10 1593.57 387.14
     7 322.50 322.03 314.96 322.10 387.33 386.37 386.95 1594.39
P2P=Disabled Latency Matrix (us)
   GPU     0      1      2      3      4      5      6      7
     0   2.95  23.60  23.59  23.56  23.60  23.58  23.51  23.59
     1  23.43   2.60  23.51  23.47  23.59  23.58  23.64  23.61
     2  23.48  23.47   2.57  23.49  23.54  23.58  23.58  23.64
     3  23.49  23.48  23.46   2.55  23.59  23.58  23.59  23.58
     4  23.60  23.58  23.52  23.58   2.07  21.96  21.67  23.49
     5  23.68  23.53  23.57  23.59  21.09   2.14  21.23  21.79
     6  23.50  23.53  23.63  23.57  21.62  22.48   2.27  23.48
     7  23.62  23.55  23.66  23.67  22.46  21.68  22.39   2.30

   CPU     0      1      2      3      4      5      6      7
     0   2.87   6.01   6.07   6.03   5.26   5.26   5.16   5.37
     1   6.32   2.83   5.88   5.97   5.26   5.24   5.16   5.36
     2   5.84   5.87   2.81   5.83   5.28   5.18   5.14   5.22
     3   5.85   5.78   5.80   2.82   5.32   5.13   5.31   5.21
     4   5.42   5.43   5.43   5.42   2.59   4.72   4.85   4.70
     5   5.46   5.27   5.34   5.35   4.74   2.62   4.80   4.78
     6   5.27   5.38   5.31   5.30   4.72   4.73   2.61   4.69
     7   5.55   5.41   5.57   5.36   4.84   4.62   4.78   2.61
P2P=Enabled Latency (P2P Writes) Matrix (us)
   GPU     0      1      2      3      4      5      6      7
     0   2.96   5.34   5.34   5.35   5.34   5.34   5.38   5.34
     1   5.22   2.59   5.20   5.12   5.21   5.20   5.17   5.19
     2   5.47   5.45   2.58   5.46   5.46   5.45   5.46   5.47
     3   5.24   5.26   5.25   2.53   5.25   5.23   5.24   5.25
     4   4.61   4.65   4.60   4.60   2.07   4.61   4.60   4.62
     5   4.54   4.57   4.53   4.53   4.53   2.13   4.53   4.54
     6   4.55   4.58   4.58   4.54   4.54   4.59   2.24   4.54
     7   4.67   4.68   4.70   4.66   4.67   4.67   4.68   2.29

   CPU     0      1      2      3      4      5      6      7
     0   3.05   2.16   2.16   2.18   2.37   2.28   2.15   2.15
     1   2.18   2.87   2.17   2.12   2.15   2.12   2.11   2.29
     2   2.18   2.19   2.93   2.16   2.23   2.22   2.14   2.14
     3   2.18   2.23   2.13   2.90   2.24   2.16   2.25   2.13
     4   1.99   1.94   1.93   1.99   2.78   1.97   1.99   1.93
     5   1.96   1.95   1.94   1.95   2.07   2.72   1.93   1.94
     6   1.99   1.96   1.92   1.96   2.06   1.95   2.73   1.94
     7   1.98   1.97   1.97   1.95   2.03   2.01   1.95   2.73

NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.
(torch1.12) root@GPU-12:/data4/docker_root/docker/tmp# history -c
(torch1.12) root@GPU-12:/data4/docker_root/docker/tmp# logout

nvmlc's People

Contributors

yanbc avatar

Watchers

avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.