Platform: ARM Platform
  Device: Mali-G31 r0p0
    Driver version  : 3.0 (Linux ARM64)
    Compute units   : 1
    Clock frequency : 5 MHz # <= probably wrong

    Global memory bandwidth (GBPS)
      float   : 3.53
      float2  : 4.30
      float4  : 4.17
      float8  : 3.48
      float16 : 1.87

    Single-precision compute (GFLOPS)
      float   : 13.47
      float2  : 13.45
      float4  : 13.43
      float8  : 13.39
      float16 : 12.54

    Half-precision compute (GFLOPS)
      half   : 13.46
      half2  : 26.72
      half4  : 26.66
      half8  : 26.55
      half16 : 26.42

    No double precision support! Skipped

    Integer compute (GIOPS)
      int   : 12.61
      int2  : 12.97
      int4  : 13.11
      int8  : 13.16
      int16 : 11.86

    Integer compute Fast 24bit (GIOPS)
      int   : 12.62
      int2  : 12.98
      int4  : 13.11
      int8  : 13.16
      int16 : 11.87

    Integer char (8bit) compute (GIOPS)
      char   : 11.92
      char2  : 23.59
      char4  : 45.74
      char8  : 47.25
      char16 : 47.77

    Integer short (16bit) compute (GIOPS)
      short   : 11.92
      short2  : 23.56
      short4  : 24.70
      short8  : 24.98
      short16 : 25.28

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer              : 1.70
      enqueueReadBuffer               : 1.85
      enqueueWriteBuffer non-blocking : 1.71
      enqueueReadBuffer non-blocking  : 1.86
      enqueueMapBuffer(for read)      : 22.21
        memcpy from mapped ptr        : 2.04
      enqueueUnmap(after write)       : 22.19
        memcpy to mapped ptr          : 2.03

    Kernel launch latency : 100.09 us