Commit 47da23bd authored by Tomáš Oberhuber's avatar Tomáš Oberhuber
Browse files

Fixing tnlLongVectorCUDA and constructors of CUDA grids and fields.

parent 03235b52

src/core/tnl-cuda-kernels.h

+6 −6
Original line number Diff line number Diff line
@@ -399,7 +399,7 @@ bool tnlCUDAReduction( const int size, 
    * If one calls the CUDA reduction more then once then one can provide

    * auxiliary array by passing it via the parameter deviceAux.

    */

   tnlLongVectorCUDA< T > deviceAuxVct;

   tnlLongVectorCUDA< T > deviceAuxVct( "tnlCUDAReduction:deviceAuxVct" );

   if( ! deviceAux )

   {

      int sizeAlloc = :: Max( 1, size / desBlockSize );
@@ -644,7 +644,7 @@ bool tnlCUDASimpleReduction5( const int size, 
    * If one calls the CUDA reduction more then once then one can provide

    * auxiliary array by passing it via the parameter deviceAux.

    */

   tnlLongVectorCUDA< T > deviceAuxVct;

   tnlLongVectorCUDA< T > deviceAuxVct( "tnlCUDAReduction:deviceAuxVct" );

   if( ! deviceAux )

   {

      int sizeAlloc = :: Max( 1, size / desBlockSize );
@@ -870,7 +870,7 @@ bool tnlCUDASimpleReduction4( const int size, 
    * If one calls the CUDA reduction more then once then one can provide

    * auxiliary array by passing it via the parameter deviceAux.

    */

   tnlLongVectorCUDA< T > deviceAuxVct;

   tnlLongVectorCUDA< T > deviceAuxVct( "tnlCUDAReduction:deviceAuxVct" );

   if( ! deviceAux )

   {

      int sizeAlloc = :: Max( 1, size / desBlockSize );
@@ -1081,7 +1081,7 @@ bool tnlCUDASimpleReduction3( const int size, 
    * If one calls the CUDA reduction more then once then one can provide

    * auxiliary array by passing it via the parameter deviceAux.

    */

   tnlLongVectorCUDA< T > deviceAuxVct;

   tnlLongVectorCUDA< T > deviceAuxVct( "tnlCUDAReduction:deviceAuxVct" );

   if( ! deviceAux )

   {

      int sizeAlloc = :: Max( 1, size / desBlockSize );
@@ -1257,7 +1257,7 @@ bool tnlCUDASimpleReduction2( const int size, 
    * If one calls the CUDA reduction more then once then one can provide

    * auxiliary array by passing it via the parameter deviceAux.

    */

   tnlLongVectorCUDA< T > deviceAuxVct;

   tnlLongVectorCUDA< T > deviceAuxVct( "tnlCUDAReduction:deviceAuxVct" );

   if( ! deviceAux )

   {

      int sizeAlloc = :: Max( 1, size / desBlockSize );
@@ -1416,7 +1416,7 @@ bool tnlCUDASimpleReduction1( const int size, 
    * If one calls the CUDA reduction more then once then one can provide

    * auxiliary array by passing it via the parameter deviceAux.

    */

   tnlLongVectorCUDA< T > deviceAuxVct;

   tnlLongVectorCUDA< T > deviceAuxVct( "tnlCUDAReduction:deviceAuxVct" );

   if( ! deviceAux )

   {

      int sizeAlloc = :: Max( 1, size / desBlockSize );

src/core/tnlFieldCUDA2D.h

+1 −3
Original line number Diff line number Diff line
@@ -28,10 +28,8 @@ template< class T > class tnlFieldCUDA2D : public tnlLongVectorCUDA< T > 
   public:



   tnlFieldCUDA2D( const char* name = 0 )

   : tnlLongVectorCUDA< T >( 0 )

   : tnlLongVectorCUDA< T >( name )

   {

      if( name )

         tnlLongVectorCUDA< T > :: SetName( name );

   };



   tnlFieldCUDA2D( const char* name,

src/core/tnlFieldCUDA3D.h

+2 −2
Original line number Diff line number Diff line
@@ -26,8 +26,8 @@ template< class T > class tnlFieldCUDA3D : public tnlLongVectorCUDA< T > 
{

   public:



   tnlFieldCUDA3D()

   : tnlLongVectorCUDA< T >( 0 )

   tnlFieldCUDA3D( const char* name = 0 )

   : tnlLongVectorCUDA< T >( name )

   { };



   tnlFieldCUDA3D( int _x_size,

src/core/tnlLongVectorCUDA.h

+14 −19
Original line number Diff line number Diff line
@@ -51,11 +51,11 @@ template< class T > class tnlLongVectorCUDA : public tnlObject 
   //! Constructor with given size

   tnlLongVectorCUDA( const char* name = 0, int _size = 0 )

#ifdef HAVE_CUDA

    : tnlObject( name ), size( _size ), shared_data( false )

    : tnlObject( name ), size( _size ), data( NULL ), shared_data( false )

   {

      if( cudaMalloc( ( void** ) &data, ( size + 1 ) * sizeof( T ) ) != cudaSuccess  )

      if( size && cudaMalloc( ( void** ) &data, ( size ) * sizeof( T ) ) != cudaSuccess  )

      {

         cerr << "Unable to allocate new long vector with size "

         cerr << "Unable to allocate new long vector '" << GetName() << "' with size "

              << size << " on CUDA device for "

              << GetName() << "." << endl;

         data = NULL;
@@ -72,12 +72,12 @@ template< class T > class tnlLongVectorCUDA : public tnlObject 
   //! Constructor with another long vector as template

   tnlLongVectorCUDA( const tnlLongVectorCUDA& v )

#ifdef HAVE_CUDA

    : tnlObject( v ), size( v. size ), shared_data( false )

    : tnlObject( v ), size( v. size ), data( NULL ), shared_data( false )

   {



      if( cudaMalloc( ( void** ) &data, ( size + 1 ) * sizeof( T ) ) != cudaSuccess )

	  if( size )

		  if( cudaMalloc( ( void** ) &data, size * sizeof( T ) ) != cudaSuccess )

		  {

         cerr << "Unable to allocate new long vector with size " << size << " on CUDA device." << endl;

			  cerr << "Unable to allocate new long vector '" << GetName() << "' with size " << size << " on CUDA device." << endl;

			  data = NULL;

			  abort();

		  }
@@ -92,10 +92,9 @@ template< class T > class tnlLongVectorCUDA : public tnlObject 
   //! Constructor with another long vector as template

   tnlLongVectorCUDA( const tnlLongVector< T >& v )

#ifdef HAVE_CUDA

    : tnlObject( v ), size( v. GetSize() ), shared_data( false )

    : tnlObject( v ), size( v. GetSize() ), data( NULL ), shared_data( false )

   {



      if( cudaMalloc( ( void** ) &data, ( size + 1 ) * sizeof( T ) ) != cudaSuccess )

      if( size && cudaMalloc( ( void** ) &data, size * sizeof( T ) ) != cudaSuccess )

      {

         cerr << "Unable to allocate new long vector with size " << size << " on CUDA device." << endl;

         data = NULL;
@@ -123,7 +122,7 @@ template< class T > class tnlLongVectorCUDA : public tnlObject 
      if( debug )

	   cout << "Setting new size to " << _size << " for " << GetName() << endl;

      if( size == _size ) return true;

      if( ! shared_data )

      if( data && ! shared_data )

      {

         cudaFree( data );

         data = NULL;
@@ -178,16 +177,12 @@ template< class T > class tnlLongVectorCUDA : public tnlObject 
    */

   const T* Data() const

   {

	   if( ! size )

		   return NULL;

      return data;

   };



   //! Returns pointer to data

   T* Data()

   {

	  if( ! size )

		  return NULL;

      return data;

   }

src/diff/tnlGridCUDA2D.h

+1 −2
Original line number Diff line number Diff line
@@ -27,9 +27,8 @@ template< class T = double > class tnlGridCUDA2D : 
   public:



   tnlGridCUDA2D( const char* name = 0 )

   : tnlFieldCUDA2D< T >( name )

   {

      if( name )

         tnlFieldCUDA2D< T > :: SetName( name );

   };



   //! Constructor with the grid and the domain dimensions