Replaced custom lambda functions with instances of STL types where possible (0a57393f) · Commits · TNL / tnl-dev

src/Benchmarks/BLAS/CommonVectorOperations.hpp

+9 −18

Original line number	Diff line number	Diff line
		@@ -97,8 +97,7 @@ getVectorL1Norm( const Vector& v )

		const auto* data = v.getData();
		auto fetch = [=] __cuda_callable__ ( IndexType i ) { return TNL::abs( data[ i ] ); };
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v.getSize(), reduction, fetch, ( ResultType ) 0 );
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 );
		}

		template< typename Device >
		@@ -114,8 +113,7 @@ getVectorL2Norm( const Vector& v )

		const auto* data = v.getData();
		auto fetch = [=] __cuda_callable__ ( IndexType i ) { return data[ i ] * data[ i ]; };
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return std::sqrt( Containers::Algorithms::Reduction< DeviceType >::reduce( v.getSize(), reduction, fetch, ( ResultType ) 0 ) );
		return std::sqrt( Containers::Algorithms::Reduction< DeviceType >::reduce( v.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 ) );
		}

		template< typename Device >
		@@ -138,8 +136,7 @@ getVectorLpNorm( const Vector& v,

		const auto* data = v.getData();
		auto fetch = [=] __cuda_callable__ ( IndexType i ) { return TNL::pow( TNL::abs( data[ i ] ), p ); };
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return std::pow( Containers::Algorithms::Reduction< DeviceType >::reduce( v.getSize(), reduction, fetch, ( ResultType ) 0 ), 1.0 / p );
		return std::pow( Containers::Algorithms::Reduction< DeviceType >::reduce( v.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 ), 1.0 / p );
		}

		template< typename Device >
		@@ -158,8 +155,7 @@ getVectorSum( const Vector& v )

		const auto* data = v.getData();
		auto fetch = [=] __cuda_callable__ ( IndexType i ) -> ResultType { return data[ i ]; };
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v.getSize(), reduction, fetch, ( ResultType ) 0 );
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 );
		}

		template< typename Device >
		@@ -258,8 +254,7 @@ getVectorDifferenceL1Norm( const Vector1& v1,
		const auto* data1 = v1.getData();
		const auto* data2 = v2.getData();
		auto fetch = [=] __cuda_callable__ ( IndexType i ) { return TNL::abs( data1[ i ] - data2[ i ] ); };
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), reduction, fetch, ( ResultType ) 0 );
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 );
		}

		template< typename Device >
		@@ -281,8 +276,7 @@ getVectorDifferenceL2Norm( const Vector1& v1,
		auto diff = data1[ i ] - data2[ i ];
		return diff * diff;
		};
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return std::sqrt( Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), reduction, fetch, ( ResultType ) 0 ) );
		return std::sqrt( Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 ) );
		}

		template< typename Device >
		@@ -308,8 +302,7 @@ getVectorDifferenceLpNorm( const Vector1& v1,
		const auto* data1 = v1.getData();
		const auto* data2 = v2.getData();
		auto fetch = [=] __cuda_callable__ ( IndexType i ) { return TNL::pow( TNL::abs( data1[ i ] - data2[ i ] ), p ); };
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return std::pow( Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), reduction, fetch, ( ResultType ) 0 ), 1.0 / p );
		return std::pow( Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 ), 1.0 / p );
		}

		template< typename Device >
		@@ -328,8 +321,7 @@ getVectorDifferenceSum( const Vector1& v1,
		const auto* data1 = v1.getData();
		const auto* data2 = v2.getData();
		auto fetch = [=] __cuda_callable__ ( IndexType i ) { return data1[ i ] - data2[ i ]; };
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), reduction, fetch, ( ResultType ) 0 );
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 );
		}

		template< typename Device >
		@@ -348,8 +340,7 @@ getScalarProduct( const Vector1& v1,
		const auto* data1 = v1.getData();
		const auto* data2 = v2.getData();
		auto fetch = [=] __cuda_callable__ ( IndexType i ) { return data1[ i ] * data2[ i ]; };
		auto reduction = [] __cuda_callable__ ( const ResultType& a, const ResultType& b ) { return a + b; };
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), reduction, fetch, ( ResultType ) 0 );
		return Containers::Algorithms::Reduction< DeviceType >::reduce( v1.getSize(), std::plus<>{}, fetch, ( ResultType ) 0 );
		}

		} // namespace Benchmarks

src/TNL/Containers/Algorithms/ArrayOperationsCuda.hpp

+3 −6

Original line number	Diff line number	Diff line
		@@ -133,8 +133,7 @@ compare( const Element1* destination,
		TNL_ASSERT_TRUE( source, "Attempted to compare data through a nullptr." );

		auto fetch = [=] __cuda_callable__ ( Index i ) -> bool { return destination[ i ] == source[ i ]; };
		auto reduction = [] __cuda_callable__ ( bool a, bool b ) { return a && b; };
		return Reduction< Devices::Cuda >::reduce( size, reduction, fetch, true );
		return Reduction< Devices::Cuda >::reduce( size, std::logical_and<>{}, fetch, true );
		}

		template< typename Element,
		@@ -150,8 +149,7 @@ containsValue( const Element* data,
		TNL_ASSERT_GE( size, (Index) 0, "" );

		auto fetch = [=] __cuda_callable__ ( Index i ) -> bool { return data[ i ] == value; };
		auto reduction = [] __cuda_callable__ ( bool a, bool b ) { return a \|\| b; };
		return Reduction< Devices::Cuda >::reduce( size, reduction, fetch, false );
		return Reduction< Devices::Cuda >::reduce( size, std::logical_or<>{}, fetch, false );
		}

		template< typename Element,
		@@ -167,8 +165,7 @@ containsOnlyValue( const Element* data,
		TNL_ASSERT_GE( size, 0, "" );

		auto fetch = [=] __cuda_callable__ ( Index i ) -> bool { return data[ i ] == value; };
		auto reduction = [] __cuda_callable__ ( bool a, bool b ) { return a && b; };
		return Reduction< Devices::Cuda >::reduce( size, reduction, fetch, true );
		return Reduction< Devices::Cuda >::reduce( size, std::logical_and<>{}, fetch, true );
		}

src/TNL/Containers/Algorithms/Multireduction.h

+6 −0

Original line number	Diff line number	Diff line
		@@ -12,6 +12,8 @@

		#pragma once

		#include <functional> // reduction functions like std::plus, std::logical_and, std::logical_or etc.

		#include <TNL/Devices/Host.h>
		#include <TNL/Devices/Cuda.h>

		@@ -32,6 +34,8 @@ struct Multireduction< Devices::Host >
		* the i-th value to be reduced from the j-th dataset
		* (i = 0,...,size-1; j = 0,...,n-1)
		* reduction: callable object representing the reduction operation
		* for example, it can be an instance of std::plus, std::logical_and,
		* std::logical_or etc.
		* size: the size of each dataset
		* n: number of datasets to be reduced
		* result: output array of size = n
		@@ -59,6 +63,8 @@ struct Multireduction< Devices::Cuda >
		* the i-th value to be reduced from the j-th dataset
		* (i = 0,...,size-1; j = 0,...,n-1)
		* reduction: callable object representing the reduction operation
		* for example, it can be an instance of std::plus, std::logical_and,
		* std::logical_or etc.
		* size: the size of each dataset
		* n: number of datasets to be reduced
		* hostResult: output array of size = n

src/TNL/Containers/Algorithms/Reduction.h

+5 −4

Original line number	Diff line number	Diff line
		@@ -13,6 +13,7 @@
		#pragma once

		#include <utility> // std::pair
		#include <functional> // reduction functions like std::plus, std::logical_and, std::logical_or etc.

		#include <TNL/Devices/Host.h>
		#include <TNL/Devices/Cuda.h>
		@@ -33,7 +34,7 @@ struct Reduction< Devices::Host >
		typename DataFetcher >
		static Result
		reduce( const Index size,
		ReductionOperation& reduction,
		const ReductionOperation& reduction,
		DataFetcher& dataFetcher,
		const Result& zero );

		@@ -43,7 +44,7 @@ struct Reduction< Devices::Host >
		typename DataFetcher >
		static std::pair< Index, Result >
		reduceWithArgument( const Index size,
		ReductionOperation& reduction,
		const ReductionOperation& reduction,
		DataFetcher& dataFetcher,
		const Result& zero );
		};
		@@ -57,7 +58,7 @@ struct Reduction< Devices::Cuda >
		typename DataFetcher >
		static Result
		reduce( const Index size,
		ReductionOperation& reduction,
		const ReductionOperation& reduction,
		DataFetcher& dataFetcher,
		const Result& zero );

		@@ -67,7 +68,7 @@ struct Reduction< Devices::Cuda >
		typename DataFetcher >
		static std::pair< Index, Result >
		reduceWithArgument( const Index size,
		ReductionOperation& reduction,
		const ReductionOperation& reduction,
		DataFetcher& dataFetcher,
		const Result& zero );
		};

src/TNL/Containers/Algorithms/Reduction.hpp

+4 −4

Original line number	Diff line number	Diff line
		@@ -45,7 +45,7 @@ template< typename Index,
		Result
		Reduction< Devices::Host >::
		reduce( const Index size,
		ReductionOperation& reduction,
		const ReductionOperation& reduction,
		DataFetcher& dataFetcher,
		const Result& zero )
		{
		@@ -137,7 +137,7 @@ template< typename Index,
		std::pair< Index, Result >
		Reduction< Devices::Host >::
		reduceWithArgument( const Index size,
		ReductionOperation& reduction,
		const ReductionOperation& reduction,
		DataFetcher& dataFetcher,
		const Result& zero )
		{
		@@ -260,7 +260,7 @@ template< typename Index,
		Result
		Reduction< Devices::Cuda >::
		reduce( const Index size,
		ReductionOperation& reduction,
		const ReductionOperation& reduction,
		DataFetcher& dataFetcher,
		const Result& zero )
		{
		@@ -336,7 +336,7 @@ template< typename Index,
		std::pair< Index, Result >
		Reduction< Devices::Cuda >::
		reduceWithArgument( const Index size,
		ReductionOperation& reduction,
		const ReductionOperation& reduction,
		DataFetcher& dataFetcher,
		const Result& zero )
		{