Platform: NVIDIA CUDA
  Device: NVIDIA GeForce GTX 1660 Ti
    Driver version  : 565.57.01 (Linux x64)
    Compute units   : 24
    Clock frequency : 1590 MHz

    Global memory bandwidth (GBPS)
      float   : 235.92
      float2  : 247.28
      float4  : 260.64
      float8  : 254.10
      float16 : 217.35

    Single-precision compute (GFLOPS)
      float   : 5692.43
      float2  : 5705.85
      float4  : 5697.71
      float8  : 5497.52
      float16 : 4822.71

    No half precision support! Skipped

    Double-precision compute (GFLOPS)
      double   : 166.56
      double2  : 169.71
      double4  : 151.43
      double8  : 152.88
      double16 : 163.43

    Integer compute (GIOPS)
      int   : 5009.23
      int2  : 5025.67
      int4  : 4511.78
      int8  : 4535.21
      int16 : 4828.46

    Integer compute Fast 24bit (GIOPS)
      int   : 5030.41
      int2  : 5000.83
      int4  : 5002.84
      int8  : 4461.20
      int16 : 4415.56

    Integer char (8bit) compute (GIOPS)
      char   : 4137.69
      char2  : 4238.38
      char4  : 4174.55
      char8  : 4234.00
      char16 : 3432.68

    Integer short (16bit) compute (GIOPS)
      short   : 4185.20
      short2  : 4014.07
      short4  : 4125.94
      short8  : 3622.42
      short16 : 3496.44

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer              : 6.85
      enqueueReadBuffer               : 6.92
      enqueueWriteBuffer non-blocking : 6.14
      enqueueReadBuffer non-blocking  : 6.08
      enqueueMapBuffer(for read)      : 9.77
        memcpy from mapped ptr        : 11.68
      enqueueUnmap(after write)       : 12.33
        memcpy to mapped ptr          : 11.99

    Kernel launch latency : 4.14 us