Loading src/TNL/Containers/DistributedArray.h +71 −0 Original line number Diff line number Diff line Loading @@ -193,6 +193,67 @@ public: template< typename Array > bool operator!=( const Array& array ) const; /** * \brief Process the lambda function \e f for each array element in interval [ \e begin, \e end). * * The lambda function is supposed to be declared as * * ``` * f( IndexType elementIdx, ValueType& elementValue ) * ``` * * where * * - \e elementIdx is an index of the array element being currently processed * - \e elementValue is a value of the array element being currently processed * * This is performed at the same place where the array is allocated, * i.e. it is efficient even on GPU. * * \param begin The beginning of the array elements interval. * \param end The end of the array elements interval. * \param f The lambda function to be processed. * * \par Example * \include Containers/ArrayExample_forElements.cpp * \par Output * \include ArrayExample_forElements.out * */ template< typename Function > void forElements( IndexType begin, IndexType end, Function&& f ); /** * \brief Process the lambda function \e f for each array element in interval [ \e begin, \e end) for constant instances of the array. * * The lambda function is supposed to be declared as * * ``` * f( IndexType elementIdx, ValueType& elementValue ) * ``` * * where * * - \e elementIdx is an index of the array element being currently processed * - \e elementValue is a value of the array element being currently processed * * This is performed at the same place where the array is allocated, * i.e. it is efficient even on GPU. * * \param begin The beginning of the array elements interval. * \param end The end of the array elements interval. * \param f The lambda function to be processed. * * \par Example * \include Containers/ArrayExample_forElements.cpp * \par Output * \include ArrayExample_forElements.out * */ template< typename Function > void forElements( IndexType begin, IndexType end, Function&& f ) const; // Checks if there is an element with given value in this array bool containsValue( ValueType value ) const; Loading @@ -215,6 +276,16 @@ private: static void setSynchronizerHelper( ViewType& view, const Array& array ) {} }; template< typename Value, typename Device, typename Index, typename Allocator > std::ostream& operator<<( std::ostream& str, const DistributedArray< Value, Device, Index, Allocator >& array ) { return array.getConstView().print( str ); } } // namespace Containers } // namespace TNL Loading src/TNL/Containers/DistributedArray.hpp +24 −0 Original line number Diff line number Diff line Loading @@ -449,6 +449,30 @@ operator!=( const Array& array ) const return view != array; } template< typename Value, typename Device, typename Index, typename Allocator > template< typename Function > void DistributedArray< Value, Device, Index, Allocator >:: forElements( IndexType begin, IndexType end, Function&& f ) { this->view.forElements( begin, end, f ); } template< typename Value, typename Device, typename Index, typename Allocator > template< typename Function > void DistributedArray< Value, Device, Index, Allocator >:: forElements( IndexType begin, IndexType end, Function&& f ) const { this->view.forElements( begin, end, f ); } template< typename Value, typename Device, typename Index, Loading src/TNL/Containers/DistributedArrayView.h +71 −0 Original line number Diff line number Diff line Loading @@ -170,12 +170,73 @@ public: template< typename Array > bool operator!=( const Array& array ) const; /** * \brief Process the lambda function \e f for each array element in interval [ \e begin, \e end). * * The lambda function is supposed to be declared as * * ``` * f( IndexType elementIdx, ValueType& elementValue ) * ``` * * where * * - \e elementIdx is an index of the array element being currently processed * - \e elementValue is a value of the array element being currently processed * * This is performed at the same place where the array is allocated, * i.e. it is efficient even on GPU. * * \param begin The beginning of the array elements interval. * \param end The end of the array elements interval. * \param f The lambda function to be processed. * * \par Example * \include Containers/ArrayExample_forElements.cpp * \par Output * \include ArrayExample_forElements.out * */ template< typename Function > void forElements( IndexType begin, IndexType end, Function&& f ); /** * \brief Process the lambda function \e f for each array element in interval [ \e begin, \e end) for constant instances of the array. * * The lambda function is supposed to be declared as * * ``` * f( IndexType elementIdx, ValueType& elementValue ) * ``` * * where * * - \e elementIdx is an index of the array element being currently processed * - \e elementValue is a value of the array element being currently processed * * This is performed at the same place where the array is allocated, * i.e. it is efficient even on GPU. * * \param begin The beginning of the array elements interval. * \param end The end of the array elements interval. * \param f The lambda function to be processed. * * \par Example * \include Containers/ArrayExample_forElements.cpp * \par Output * \include ArrayExample_forElements.out * */ template< typename Function > void forElements( IndexType begin, IndexType end, Function&& f ) const; // Checks if there is an element with given value in this array bool containsValue( ValueType value ) const; // Checks if all elements in this array have the same given value bool containsOnlyValue( ValueType value ) const; std::ostream& print( std::ostream& str ) const; protected: LocalRangeType localRange; IndexType ghosts = 0; Loading @@ -187,6 +248,16 @@ protected: int valuesPerElement = 1; }; template< typename Value, typename Device = Devices::Host, typename Index = int > std::ostream& operator<<( std::ostream& str, const DistributedArrayView< Value, Device, Index >& view ) { return view.print( str ); } } // namespace Containers } // namespace TNL Loading src/TNL/Containers/DistributedArrayView.hpp +57 −0 Original line number Diff line number Diff line Loading @@ -435,6 +435,37 @@ operator!=( const Array& array ) const return ! (*this == array); } template< typename Value, typename Device, typename Index > template< typename Function > void DistributedArrayView< Value, Device, Index >:: forElements( IndexType begin, IndexType end, Function&& f ) { IndexType localBegin = max( begin, localRange.getBegin() ); IndexType localEnd = min( end, localRange.getEnd() ); auto local_f = [=] __cuda_callable__ ( const IndexType& idx, ValueType& value ) mutable { f( idx + localRange.getBegin(), value ); }; this->localData.forElements( localBegin - localRange.getBegin(), localEnd - localRange.getBegin(), local_f ); } template< typename Value, typename Device, typename Index > template< typename Function > void DistributedArrayView< Value, Device, Index >:: forElements( IndexType begin, IndexType end, Function&& f ) const { } template< typename Value, typename Device, typename Index > Loading Loading @@ -465,5 +496,31 @@ containsOnlyValue( ValueType value ) const return result; } template< typename Value, typename Device, typename Index > std::ostream& DistributedArrayView< Value, Device, Index >:: print( std::ostream& str ) const { // The following does not work properly /*if( MPI::GetRank( group ) == 0 ) { str << "[ "; for( IndexType i = 0; i < localData.getSize(); i++ ) str << ", " << localData.getElement( i ); for( int proc = 1; proc < MPI::GetSize( group ); proc++ ) { Array< std::remove_const_t< Value >, Device, Index > localArray; receive( localArray, proc, 0, group ); for( IndexType i = 0; i < localArray.getSize(); i++ ) str << ", " << localArray.getElement( i ); } str << " ]"; } else send( this->localData, 0, 0, this->group ); return str;*/ } } // namespace Containers } // namespace TNL Loading
src/TNL/Containers/DistributedArray.h +71 −0 Original line number Diff line number Diff line Loading @@ -193,6 +193,67 @@ public: template< typename Array > bool operator!=( const Array& array ) const; /** * \brief Process the lambda function \e f for each array element in interval [ \e begin, \e end). * * The lambda function is supposed to be declared as * * ``` * f( IndexType elementIdx, ValueType& elementValue ) * ``` * * where * * - \e elementIdx is an index of the array element being currently processed * - \e elementValue is a value of the array element being currently processed * * This is performed at the same place where the array is allocated, * i.e. it is efficient even on GPU. * * \param begin The beginning of the array elements interval. * \param end The end of the array elements interval. * \param f The lambda function to be processed. * * \par Example * \include Containers/ArrayExample_forElements.cpp * \par Output * \include ArrayExample_forElements.out * */ template< typename Function > void forElements( IndexType begin, IndexType end, Function&& f ); /** * \brief Process the lambda function \e f for each array element in interval [ \e begin, \e end) for constant instances of the array. * * The lambda function is supposed to be declared as * * ``` * f( IndexType elementIdx, ValueType& elementValue ) * ``` * * where * * - \e elementIdx is an index of the array element being currently processed * - \e elementValue is a value of the array element being currently processed * * This is performed at the same place where the array is allocated, * i.e. it is efficient even on GPU. * * \param begin The beginning of the array elements interval. * \param end The end of the array elements interval. * \param f The lambda function to be processed. * * \par Example * \include Containers/ArrayExample_forElements.cpp * \par Output * \include ArrayExample_forElements.out * */ template< typename Function > void forElements( IndexType begin, IndexType end, Function&& f ) const; // Checks if there is an element with given value in this array bool containsValue( ValueType value ) const; Loading @@ -215,6 +276,16 @@ private: static void setSynchronizerHelper( ViewType& view, const Array& array ) {} }; template< typename Value, typename Device, typename Index, typename Allocator > std::ostream& operator<<( std::ostream& str, const DistributedArray< Value, Device, Index, Allocator >& array ) { return array.getConstView().print( str ); } } // namespace Containers } // namespace TNL Loading
src/TNL/Containers/DistributedArray.hpp +24 −0 Original line number Diff line number Diff line Loading @@ -449,6 +449,30 @@ operator!=( const Array& array ) const return view != array; } template< typename Value, typename Device, typename Index, typename Allocator > template< typename Function > void DistributedArray< Value, Device, Index, Allocator >:: forElements( IndexType begin, IndexType end, Function&& f ) { this->view.forElements( begin, end, f ); } template< typename Value, typename Device, typename Index, typename Allocator > template< typename Function > void DistributedArray< Value, Device, Index, Allocator >:: forElements( IndexType begin, IndexType end, Function&& f ) const { this->view.forElements( begin, end, f ); } template< typename Value, typename Device, typename Index, Loading
src/TNL/Containers/DistributedArrayView.h +71 −0 Original line number Diff line number Diff line Loading @@ -170,12 +170,73 @@ public: template< typename Array > bool operator!=( const Array& array ) const; /** * \brief Process the lambda function \e f for each array element in interval [ \e begin, \e end). * * The lambda function is supposed to be declared as * * ``` * f( IndexType elementIdx, ValueType& elementValue ) * ``` * * where * * - \e elementIdx is an index of the array element being currently processed * - \e elementValue is a value of the array element being currently processed * * This is performed at the same place where the array is allocated, * i.e. it is efficient even on GPU. * * \param begin The beginning of the array elements interval. * \param end The end of the array elements interval. * \param f The lambda function to be processed. * * \par Example * \include Containers/ArrayExample_forElements.cpp * \par Output * \include ArrayExample_forElements.out * */ template< typename Function > void forElements( IndexType begin, IndexType end, Function&& f ); /** * \brief Process the lambda function \e f for each array element in interval [ \e begin, \e end) for constant instances of the array. * * The lambda function is supposed to be declared as * * ``` * f( IndexType elementIdx, ValueType& elementValue ) * ``` * * where * * - \e elementIdx is an index of the array element being currently processed * - \e elementValue is a value of the array element being currently processed * * This is performed at the same place where the array is allocated, * i.e. it is efficient even on GPU. * * \param begin The beginning of the array elements interval. * \param end The end of the array elements interval. * \param f The lambda function to be processed. * * \par Example * \include Containers/ArrayExample_forElements.cpp * \par Output * \include ArrayExample_forElements.out * */ template< typename Function > void forElements( IndexType begin, IndexType end, Function&& f ) const; // Checks if there is an element with given value in this array bool containsValue( ValueType value ) const; // Checks if all elements in this array have the same given value bool containsOnlyValue( ValueType value ) const; std::ostream& print( std::ostream& str ) const; protected: LocalRangeType localRange; IndexType ghosts = 0; Loading @@ -187,6 +248,16 @@ protected: int valuesPerElement = 1; }; template< typename Value, typename Device = Devices::Host, typename Index = int > std::ostream& operator<<( std::ostream& str, const DistributedArrayView< Value, Device, Index >& view ) { return view.print( str ); } } // namespace Containers } // namespace TNL Loading
src/TNL/Containers/DistributedArrayView.hpp +57 −0 Original line number Diff line number Diff line Loading @@ -435,6 +435,37 @@ operator!=( const Array& array ) const return ! (*this == array); } template< typename Value, typename Device, typename Index > template< typename Function > void DistributedArrayView< Value, Device, Index >:: forElements( IndexType begin, IndexType end, Function&& f ) { IndexType localBegin = max( begin, localRange.getBegin() ); IndexType localEnd = min( end, localRange.getEnd() ); auto local_f = [=] __cuda_callable__ ( const IndexType& idx, ValueType& value ) mutable { f( idx + localRange.getBegin(), value ); }; this->localData.forElements( localBegin - localRange.getBegin(), localEnd - localRange.getBegin(), local_f ); } template< typename Value, typename Device, typename Index > template< typename Function > void DistributedArrayView< Value, Device, Index >:: forElements( IndexType begin, IndexType end, Function&& f ) const { } template< typename Value, typename Device, typename Index > Loading Loading @@ -465,5 +496,31 @@ containsOnlyValue( ValueType value ) const return result; } template< typename Value, typename Device, typename Index > std::ostream& DistributedArrayView< Value, Device, Index >:: print( std::ostream& str ) const { // The following does not work properly /*if( MPI::GetRank( group ) == 0 ) { str << "[ "; for( IndexType i = 0; i < localData.getSize(); i++ ) str << ", " << localData.getElement( i ); for( int proc = 1; proc < MPI::GetSize( group ); proc++ ) { Array< std::remove_const_t< Value >, Device, Index > localArray; receive( localArray, proc, 0, group ); for( IndexType i = 0; i < localArray.getSize(); i++ ) str << ", " << localArray.getElement( i ); } str << " ]"; } else send( this->localData, 0, 0, this->group ); return str;*/ } } // namespace Containers } // namespace TNL
