Platform: NVIDIA CUDA
  Device: Tesla K80
    Driver version  : 455.32.00 (Linux x64)
    Compute units   : 13
    Clock frequency : 823 MHz

    Global memory bandwidth (GBPS)
      float   : 147.19
      float2  : 148.99
      float4  : 152.33
      float8  : 141.67
      float16 : 68.77

    Single-precision compute (GFLOPS)
      float   : 2835.78
      float2  : 2834.16
      float4  : 3700.81
      float8  : 3518.41
      float16 : 3288.67

    No half precision support! Skipped

    Double-precision compute (GFLOPS)
      double   : 1400.02
      double2  : 1399.04
      double4  : 1394.24
      double8  : 1396.52
      double16 : 1386.00

    Integer compute (GIOPS)
      int   : 711.60
      int2  : 711.39
      int4  : 711.65
      int8  : 711.87
      int16 : 711.75

    Integer compute Fast 24bit (GIOPS)
      int   : 711.52
      int2  : 711.37
      int4  : 711.58
      int8  : 711.36
      int16 : 709.81

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer              : 9.35
      enqueueReadBuffer               : 11.89
      enqueueWriteBuffer non-blocking : 8.42
      enqueueReadBuffer non-blocking  : 11.13
      enqueueMapBuffer(for read)      : 9.99
        memcpy from mapped ptr        : 9.80
      enqueueUnmap(after write)       : 12.05
        memcpy to mapped ptr          : 9.57

    Kernel launch latency : 8.20 us