Platform: Moore Threads OpenCL
  Device: MUSA GEN1-104
    Driver version  : 20241010 release kuae1.3.0_musa3.1.0 db329f8fb@20241009  (Linux x64)
    Compute units   : 32
    Clock frequency : 1799 MHz

    Global memory bandwidth (GBPS)
      float   : 273.11
      float2  : 374.98
      float4  : 387.00
      float8  : 391.09
      float16 : 399.77

    Single-precision compute (GFLOPS)
      float   : 14176.27
      float2  : 13376.75
      float4  : 13451.20
      float8  : 13425.36
      float16 : 13350.33

    Half-precision compute (GFLOPS)
      half   : 13296.25
      half2  : 13387.43
      half4  : 13450.06
      half8  : 13437.34
      half16 : 13368.42

    Double-precision compute (GFLOPS)
      double   : 35.60
      double2  : 30.19
      double4  : 21.89
      double8  : 13.00
      double16 : 7.04

    Integer compute (GIOPS)
      int   : 2094.24
      int2  : 2092.97
      int4  : 2096.44
      int8  : 2096.89
      int16 : 2099.12

    Integer compute Fast 24bit (GIOPS)
      int   : 2094.74
      int2  : 2092.97
      int4  : 2095.84
      int8  : 2097.56
      int16 : 2099.13

    Integer char (8bit) compute (GIOPS)
      char   : 2095.43
      char2  : 2093.63
      char4  : 2097.98
      char8  : 2097.77
      char16 : 2100.38

    Integer short (16bit) compute (GIOPS)
      short   : 2095.20
      short2  : 2093.82
      short4  : 2096.92
      short8  : 2098.04
      short16 : 2099.83

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer              : 5.84
      enqueueReadBuffer               : 5.92
      enqueueWriteBuffer non-blocking : 5.85
      enqueueReadBuffer non-blocking  : 5.91
      enqueueMapBuffer(for read)      : 5743.47
        memcpy from mapped ptr        : 0.02
      enqueueUnmap(after write)       : 6042.44
        memcpy to mapped ptr          : 5.03

    Kernel launch latency : 48.02 us