Platform: NVIDIA CUDA
  Device: Quadro K620
    Driver version  : 515.76 (Linux x64)
    Compute units   : 3
    Clock frequency : 1124 MHz

    Global memory bandwidth (GBPS)
      float   : 25.41
      float2  : 26.21
      float4  : 26.69
      float8  : 25.73
      float16 : 22.42

    Single-precision compute (GFLOPS)
      float   : 569.12
      float2  : 839.43
      float4  : 856.16
      float8  : 851.06
      float16 : 848.14

    No half precision support! Skipped

    Double-precision compute (GFLOPS)
      double   : 27.47
      double2  : 27.48
      double4  : 27.42
      double8  : 27.32
      double16 : 27.11

    Integer compute (GIOPS)
      int   : 258.66
      int2  : 287.30
      int4  : 289.72
      int8  : 275.19
      int16 : 264.43

    Integer compute Fast 24bit (GIOPS)
      int   : 258.66
      int2  : 287.06
      int4  : 289.57
      int8  : 287.89
      int16 : 286.48

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer              : 5.16
      enqueueReadBuffer               : 5.13
      enqueueWriteBuffer non-blocking : 2.17
      enqueueReadBuffer non-blocking  : 2.73
      enqueueMapBuffer(for read)      : 5.43
        memcpy from mapped ptr        : 4.48
      enqueueUnmap(after write)       : 6.14
        memcpy to mapped ptr          : 4.49

    Kernel launch latency : 6.87 us