Fixing tnlLongVectorCUDA :: setNewSize. (ecd167e8) · Commits · TNL / tnl-dev

src/Makefile.am

+4 −2

Original line number	Diff line number	Diff line
		@@ -72,5 +72,7 @@ endif



		TESTS = tnl-unit-tests \
		tnl-benchmarks
		No newline at end of file
		TESTS = tnl-unit-tests

		##tnl-unit-tests \
		## tnl-benchmarks
		No newline at end of file

src/core/tnl-cuda-kernels.h

+6 −6

Original line number	Diff line number	Diff line
		@@ -403,7 +403,7 @@ bool tnlCUDAReduction( const int size,
		if( ! deviceAux )
		{
		int sizeAlloc = :: Max( 1, size / desBlockSize );
		if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
		if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
		return false;
		deviceAux = deviceAuxVct. Data();
		}
		@@ -648,7 +648,7 @@ bool tnlCUDASimpleReduction5( const int size,
		if( ! deviceAux )
		{
		int sizeAlloc = :: Max( 1, size / desBlockSize );
		if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
		if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
		return false;
		deviceAux = deviceAuxVct. Data();
		}
		@@ -874,7 +874,7 @@ bool tnlCUDASimpleReduction4( const int size,
		if( ! deviceAux )
		{
		int sizeAlloc = :: Max( 1, size / desBlockSize );
		if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
		if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
		return false;
		deviceAux = deviceAuxVct. Data();
		}
		@@ -1085,7 +1085,7 @@ bool tnlCUDASimpleReduction3( const int size,
		if( ! deviceAux )
		{
		int sizeAlloc = :: Max( 1, size / desBlockSize );
		if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
		if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
		return false;
		deviceAux = deviceAuxVct. Data();
		}
		@@ -1261,7 +1261,7 @@ bool tnlCUDASimpleReduction2( const int size,
		if( ! deviceAux )
		{
		int sizeAlloc = :: Max( 1, size / desBlockSize );
		if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
		if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
		return false;
		deviceAux = deviceAuxVct. Data();
		}
		@@ -1420,7 +1420,7 @@ bool tnlCUDASimpleReduction1( const int size,
		if( ! deviceAux )
		{
		int sizeAlloc = :: Max( 1, size / desBlockSize );
		if( ! deviceAuxVct. SetNewSize( sizeAlloc ) )
		if( ! deviceAuxVct. setNewSize( sizeAlloc ) )
		return false;
		deviceAux = deviceAuxVct. Data();
		}

src/core/tnlCUDAKernelsTester.h

+1 −1

Original line number	Diff line number	Diff line
		@@ -106,7 +106,7 @@ template< class T > class tnlCUDAKernelsTester : public CppUnit :: TestCase
		{
		const int size = host_input. GetSize();
		tnlLongVectorCUDA< T > device_input;
		if( ! device_input. SetNewSize( size ) )
		if( ! device_input. setNewSize( size ) )
		return false;
		device_input. copyFrom( host_input );

src/core/tnlFieldCUDA2D.h

+6 −1

Original line number	Diff line number	Diff line
		@@ -70,7 +70,7 @@ template< class T > class tnlFieldCUDA2D : public tnlLongVectorCUDA< T >
		{
		x_size = new_x_size;
		y_size = new_y_size;
		return tnlLongVectorCUDA< T > :: SetNewSize( x_size * y_size );
		return tnlLongVectorCUDA< T > :: setNewSize( x_size * y_size );
		};

		bool SetNewDimensions( const tnlFieldCUDA2D< T >& f )
		@@ -78,6 +78,11 @@ template< class T > class tnlFieldCUDA2D : public tnlLongVectorCUDA< T >
		return SetNewDimensions( f. GetXSize(), f. GetYSize() );
		};

		bool SetNewDimensions( const tnlField2D< T >& f )
		{
		return SetNewDimensions( f. GetXSize(), f. GetYSize() );
		};

		void SetSharedData( T* _data, const int _x_size, const int _y_size )
		{
		tnlLongVectorCUDA< T > :: SetSharedData( _data, _x_size * _y_size );

src/core/tnlLongVectorCUDA.h

+22 −41

Original line number	Diff line number	Diff line
		@@ -44,6 +44,7 @@ using namespace std;
		#include <core/tnlObject.h>
		#include <core/param-types.h>
		#include <core/tnlLongVector.h>
		#include <debug/tnlDebug.h>

		template< class T > class tnlLongVectorCUDA : public tnlObject
		{
		@@ -51,17 +52,9 @@ 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 ), data( NULL ), shared_data( false )
		: tnlObject( name ), size( 0 ), data( NULL ), shared_data( false )
		{
		if( size && cudaMalloc( ( void** ) &data, ( size ) * sizeof( T ) ) != cudaSuccess )
		{
		cerr << "Unable to allocate new long vector '" << GetName() << "' with size "
		<< size << " on CUDA device for "
		<< GetName() << "." << endl;
		data = NULL;
		abort();
		}
		//data ++;
		if( _size ) setNewSize( _size );
		};
		#else
		{
		@@ -72,16 +65,10 @@ 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 ), data( NULL ), shared_data( false )
		: tnlObject( v ), size( 0 ), data( NULL ), shared_data( false )
		{
		if( size )
		if( cudaMalloc( ( void** ) &data, size * sizeof( T ) ) != cudaSuccess )
		{
		cerr << "Unable to allocate new long vector '" << GetName() << "' with size " << size << " on CUDA device." << endl;
		data = NULL;
		abort();
		}
		//data ++;
		if( v. GetSize() )
		setNewSize( v. size );
		};
		#else
		{
		@@ -92,15 +79,9 @@ 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() ), data( NULL ), shared_data( false )
		{
		if( size && cudaMalloc( ( void** ) &data, size * sizeof( T ) ) != cudaSuccess )
		: tnlObject( v ), size( 0 ), data( NULL ), shared_data( false )
		{
		cerr << "Unable to allocate new long vector with size " << size << " on CUDA device." << endl;
		data = NULL;
		abort();
		}
		//data ++;
		if( v. GetSize() ) setNewSize( v. GetSize() );
		};
		#else
		{
		@@ -116,19 +97,21 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
		}


		bool SetNewSize( int _size )
		bool setNewSize( int _size )
		#ifdef HAVE_CUDA
		{
		if( debug )
		cout << "Setting new size to " << _size << " for " << GetName() << endl;
		dbgFunctionName( "tnlLongVectorCUDA", "setNewSize" );
		dbgCout( "Setting new size to " << _size << " for " << GetName() );
		if( size == _size ) return true;
		if( data && ! shared_data )
		{
		dbgCout( "Freeing allocated memory on CUDA device of " << GetName() );
		cudaFree( data );
		data = NULL;
		}
		size = _size;
		shared_data = false;
		cerr << "###########" << endl;
		if( cudaMalloc( ( void** ) &data, size * sizeof( T ) ) != cudaSuccess )
		{
		cerr << "Unable to allocate new long vector with size "
		@@ -147,9 +130,14 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
		};
		#endif

		bool SetNewSize( const tnlLongVectorCUDA< T >& v )
		bool setNewSize( const tnlLongVectorCUDA< T >& v )
		{
		return setNewSize( v. GetSize() );
		};

		bool setNewSize( const tnlLongVector< T >& v )
		{
		return SetNewSize( v. GetSize() );
		return setNewSize( v. GetSize() );
		};

		void SetSharedData( T* _data, const int _size )
		@@ -239,11 +227,11 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
		virtual
		~tnlLongVectorCUDA()
		{
		dbgFunctionName( "tnlLongVectorCUDA", "~tnlLOngVectorCUDA" );
		#ifdef HAVE_CUDA
		if( data && ! shared_data )
		{
		if( debug )
		cout << "Freeing allocated memory of " << GetName() << " on CUDA device." << endl;
		dbgCout( "Freeing allocated memory of " << GetName() << " on CUDA device." );
		if( cudaFree( data ) != cudaSuccess )
		{
		cerr << "Unable to free alocated memory on CUDA device of " << GetName() << "." << endl;
		@@ -254,11 +242,6 @@ template< class T > class tnlLongVectorCUDA : public tnlObject
		#endif
		};

		static void setDebug( bool _debug )
		{
		debug = _debug;
		};

		private:

		int size;
		@@ -290,6 +273,4 @@ template< class T > bool tnlLongVector< T > :: copyFrom( const tnlLongVectorCUDA

		}

		template< class T > bool tnlLongVectorCUDA< T > :: debug;

		#endif /* TNLLONGVECTORCUDA_H_ */