Added constexpr to some Cuda methods

src/TNL/Devices/Cuda.cpp

@@ -26,11 +26,6 @@ String Cuda::getDeviceType()
    return String( "Cuda" );
 }
 
-int Cuda::getGPUTransferBufferSize()
-{
-   return 1 << 20;
-}
-
 int Cuda::getNumberOfBlocks( const int threads,
                              const int blockSize )
 {

src/TNL/Devices/Cuda.h

@@ -38,11 +38,15 @@ class Cuda
 
    static String getDeviceType();
 
-   __cuda_callable__ static inline int getMaxGridSize();
+   __cuda_callable__ static inline constexpr int getMaxGridSize();
 
-   __cuda_callable__ static inline int getMaxBlockSize();
+   __cuda_callable__ static inline constexpr int getMaxBlockSize();
 
-   __cuda_callable__ static inline int getWarpSize();
+   __cuda_callable__ static inline constexpr int getWarpSize();
+
+   __cuda_callable__ static inline constexpr int getNumberOfSharedMemoryBanks();
+
+   static inline constexpr int getGPUTransferBufferSize();
 
 #ifdef HAVE_CUDA
    static int getDeviceId();
@@ -51,10 +55,6 @@ class Cuda
    __device__ static Index getGlobalThreadIdx( const Index gridIdx = 0 );
 #endif
 
-   __cuda_callable__ static inline int getNumberOfSharedMemoryBanks();
-
-   static int getGPUTransferBufferSize();
-
    static int getNumberOfBlocks( const int threads,
                                  const int blockSize );
 

src/TNL/Devices/Cuda_impl.h

@@ -16,26 +16,38 @@ namespace TNL {
 namespace Devices {   
 
 __cuda_callable__ 
-inline int Cuda::getMaxGridSize()
+inline constexpr int Cuda::getMaxGridSize()
 {
    // TODO: make it preprocessor macro constant defined in tnlConfig
    return 65535;
 };
 
 __cuda_callable__
-inline int Cuda::getMaxBlockSize()
+inline constexpr int Cuda::getMaxBlockSize()
 {
    // TODO: make it preprocessor macro constant defined in tnlConfig
    return 1024;
 };
 
 __cuda_callable__ 
-inline int Cuda::getWarpSize()
+inline constexpr int Cuda::getWarpSize()
 {
    // TODO: make it preprocessor macro constant defined in tnlConfig
    return 32;
 }
 
+__cuda_callable__
+inline constexpr int Cuda::getNumberOfSharedMemoryBanks()
+{
+   // TODO: make it preprocessor macro constant defined in tnlConfig
+   return 32;
+}
+
+inline constexpr int Cuda::getGPUTransferBufferSize()
+{
+   return 1 << 20;
+}
+
 #ifdef HAVE_CUDA
 template< typename Index >
 __device__ Index Cuda::getGlobalThreadIdx( const Index gridIdx )
@@ -45,13 +57,6 @@ __device__ Index Cuda::getGlobalThreadIdx( const Index gridIdx )
 #endif
 
 
-__cuda_callable__ 
-inline int Cuda::getNumberOfSharedMemoryBanks()
-{
-   // TODO: make it preprocessor macro constant defined in tnlConfig
-   return 32;
-}
-
 template< typename ObjectType >
 ObjectType* Cuda::passToDevice( const ObjectType& object )
 {