Platform: NVIDIA CUDA
  Device: GeForce GTX 1080
    Driver version  : 367.27 (Linux x64)
    Compute units   : 20
    Clock frequency : 1733 MHz

    Global memory bandwidth (GBPS)
      float   : 224.25
      float2  : 227.78
      float4  : 236.81
      float8  : 216.52
      float16 : 179.27

    Single-precision compute (GFLOPS)
      float   : 8549.33
      float2  : 9216.67
      float4  : 9262.55
      float8  : 9164.55
      float16 : 9158.85

    Double-precision compute (GFLOPS)
      double   : 303.79
      double2  : 303.89
      double4  : 303.46
      double8  : 302.27
      double16 : 299.86

    Integer compute (GIOPS)
      int   : 2458.08
      int2  : 2620.93
      int4  : 2582.49
      int8  : 2621.57
      int16 : 2602.94

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer         : 1.77
      enqueueReadBuffer          : 12.85
      enqueueMapBuffer(for read) : 10.95
        memcpy from mapped ptr   : 10.30
      enqueueUnmap(after write)  : 12.15
        memcpy to mapped ptr     : 10.21

    Kernel launch latency : 4.35 us