Platform: Portable Computing Language
  Device: NVIDIA Tegra X1
    Driver version  : 1.3 (Linux ARM64)
    Compute units   : 1
    Clock frequency : 921 MHz

    Global memory bandwidth (GBPS)
      float   : 17.95
      float2  : 20.21
      float4  : 20.92
      float8  : 19.82
      float16 : 15.14

    Single-precision compute (GFLOPS)
      float   : 214.09
      float2  : 229.80
      float4  : 230.95
      float8  : 229.31
      float16 : 228.80

    Half-precision compute (GFLOPS)
      half   : 212.93
      half2  : 228.95
      half4  : 228.69
      half8  : 245.39
      half16 : 238.39

    Double-precision compute (GFLOPS)
      double   : 7.32
      double2  : 7.31
      double4  : 7.30
      double8  : 7.27
      double16 : 7.21

    Integer compute (GIOPS)
      int   : 70.95
      int2  : 74.95
      int4  : 76.43
      int8  : 76.62
      int16 : 76.78

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer         : 2.94
      enqueueReadBuffer          : 0.69
      enqueueMapBuffer(for read) : 2487.73
        memcpy from mapped ptr   : 0.70
      enqueueUnmap(after write)  : 0.68
        memcpy to mapped ptr     : 3.68

    Kernel launch latency : 32.77 us

Note via POCL 1.3