Merge branch 'gpu_flops' into marenz.update_ci

include/firestarter/Cuda/Cuda.hpp

@@ -69,11 +69,14 @@ class Cuda {
 
 public:
   /// Initilize the cuda class. This will start a thread running the Cuda::initGpus function and wait until all gpus
-  /// are inititialized. \arg LoadVar A reference to the variable that controlls the current load of Firestarter. \arg
-  /// UseFloat Set to true if we want to stress using single precision floating points. \arg UseDouble Set to true if
-  /// we want to stress using double precision floating points. If neither UseFloat or UseDouble is set the precision
-  /// will be choosen automatically. \arg MatrixSize Set to a specific matrix size which will be choosen for the gemm
-  /// operation or set to 0 for automatic selection. \arg Gpus Select the number of gpus to stress or -1 for all.
+  /// are inititialized.
+  /// \arg LoadVar A reference to the variable that controlls the current load of Firestarter.
+  /// \arg UseFloat Set to true if we want to stress using single precision floating points.
+  /// \arg UseDouble Set to true if we want to stress using double precision floating points. If neither UseFloat or
+  /// UseDouble is set the precision will be choosen automatically.
+  /// \arg MatrixSize Set to a specific matrix size which will be choosen for the gemm operation or set to 0 for
+  /// automatic selection.
+  /// \arg Gpus Select the number of gpus to stress or -1 for all.
   Cuda(const volatile firestarter::LoadThreadWorkType& LoadVar, bool UseFloat, bool UseDouble, unsigned MatrixSize,
        int Gpus)
 #if defined(FIRESTARTER_BUILD_CUDA) || defined(FIRESTARTER_BUILD_HIP)

src/firestarter/Cuda/Cuda.cpp

@@ -53,9 +53,10 @@ template <std::size_t Multiple> auto roundUp(int NumToRound) -> int {
 /// Convert the UseDouble input (0 -> single precision, 1 -> double precision, 2 -> automatic) to either 0 or 1 for
 /// float or double respectively. For CUDART_VERSION at least equal 8000 and automatic selection we check if the card
 /// a singleToDoublePrecisionPerfRatio bigger than 3 and select float in this case otherwise double. In all other
-/// cases automatic results in double. \arg UseDouble The input that specifies either single precision, double
-/// precision or automatic selection. \arg Properties The device properties. \return The selected precision, either 0
-/// or 1 for float or double respectively.
+/// cases automatic results in double.
+/// \arg UseDouble The input that specifies either single precision, double precision or automatic selection.
+/// \arg Properties The device properties.
+/// \return The selected precision, either 0 or 1 for float or double respectively.
 auto getPrecision(int UseDouble, const compat::DeviceProperties& Properties) -> int {
 #if (CUDART_VERSION >= 8000)
   // read precision ratio (dp/sp) of GPU to choose the right variant for maximum