Ubuntu - GPU - OpenCL - Check OpenCL Performance

clpeak is a synthetic benchmarking tool to measure peak capabilities of opencl devices.

• It only measures the peak metrics that can be achieved using vector operations and does not represent a real-world use case.

clpeak

returns:

Platform: AMD Accelerated Parallel Processing
  Device: gfx1100
    Driver version  : 3513.0 (HSA1.1,LC) (Linux x64)
    Compute units   : 48
    Clock frequency : 3220 MHz
 
    Global memory bandwidth (GBPS)
      float   : 751.23
      float2  : 789.32
      float4  : 825.00
      float8  : 856.98
      float16 : 877.29
 
    Single-precision compute (GFLOPS)
      float   : 34524.21
      float2  : 34385.53
      float4  : 34120.14
      float8  : 34467.64
      float16 : 32858.38
 
    Half-precision compute (GFLOPS)
      half   : 34022.72
      half2  : 65991.11
      half4  : 66666.16
      half8  : 62676.74
      half16 : 64124.60
 
    Double-precision compute (GFLOPS)
      double   : 1155.41
      double2  : 1171.97
      double4  : 1174.89
      double8  : 1169.35
      double16 : 1160.33
 
    Integer compute (GIOPS)
      int   : 8587.24
      int2  : 8392.64
      int4  : 8364.48
      int8  : 8410.87
      int16 : 8421.63
 
    Transfer bandwidth (GBPS)
      enqueueWriteBuffer         : 19.29
      enqueueReadBuffer          : 19.92
      enqueueMapBuffer(for read) : 249707.39
        memcpy from mapped ptr   : 19.80
      enqueueUnmap(after write)  : 363980.28
        memcpy to mapped ptr     : 19.32
 
    Kernel launch latency : 15.75 us