Platform: NVIDIA CUDA
  Device: Graphics Device
    Driver version  : 378.13 (Linux x64)
    Compute units   : 28
    Clock frequency : 1683 MHz

    Global memory bandwidth (GBPS)
      float   : 389.99
      float2  : 394.86
      float4  : 410.15
      float8  : 388.05
      float16 : 263.58

    Single-precision compute (GFLOPS)
      float   : 11675.87
      float2  : 13240.07
      float4  : 13317.21
      float8  : 13151.05
      float16 : 12939.08

    Double-precision compute (GFLOPS)
      double   : 425.21
      double2  : 432.63
      double4  : 425.45
      double8  : 420.62
      double16 : 409.39

    Integer compute (GIOPS)
      int   : 3507.68
      int2  : 3801.87
      int4  : 3772.84
      int8  : 3774.45
      int16 : 3748.59

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer         : 9.96
      enqueueReadBuffer          : 8.95
      enqueueMapBuffer(for read) : 11.11
        memcpy from mapped ptr   : 12.16
      enqueueUnmap(after write)  : 12.40
        memcpy to mapped ptr     : 12.48

    Kernel launch latency : 4.22 us