1 // This file is part of Eigen, a lightweight C++ template library 2 // for linear algebra. 3 // 4 // Copyright (C) 2014 Benoit Steiner <benoit.steiner.goog@gmail.com> 5 // 6 // This Source Code Form is subject to the terms of the Mozilla 7 // Public License v. 2.0. If a copy of the MPL was not distributed 8 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/. 9 10 #ifndef EIGEN_CXX11_TENSOR_TENSOR_LAYOUT_SWAP_H 11 #define EIGEN_CXX11_TENSOR_TENSOR_LAYOUT_SWAP_H 12 13 namespace Eigen { 14 15 /** \class TensorLayoutSwap 16 * \ingroup CXX11_Tensor_Module 17 * 18 * \brief Swap the layout from col-major to row-major, or row-major 19 * to col-major, and invert the order of the dimensions. 20 * 21 * Beware: the dimensions are reversed by this operation. If you want to 22 * preserve the ordering of the dimensions, you need to combine this 23 * operation with a shuffle. 24 * 25 * \example: 26 * Tensor<float, 2, ColMajor> input(2, 4); 27 * Tensor<float, 2, RowMajor> output = input.swap_layout(); 28 * eigen_assert(output.dimension(0) == 4); 29 * eigen_assert(output.dimension(1) == 2); 30 * 31 * array<int, 2> shuffle(1, 0); 32 * output = input.swap_layout().shuffle(shuffle); 33 * eigen_assert(output.dimension(0) == 2); 34 * eigen_assert(output.dimension(1) == 4); 35 * 36 */ 37 namespace internal { 38 template<typename XprType> 39 struct traits<TensorLayoutSwapOp<XprType> > : public traits<XprType> 40 { 41 typedef typename XprType::Scalar Scalar; 42 typedef traits<XprType> XprTraits; 43 typedef typename XprTraits::StorageKind StorageKind; 44 typedef typename XprTraits::Index Index; 45 typedef typename XprType::Nested Nested; 46 typedef typename remove_reference<Nested>::type _Nested; 47 static const int NumDimensions = traits<XprType>::NumDimensions; 48 static const int Layout = (traits<XprType>::Layout == ColMajor) ? RowMajor : ColMajor; 49 }; 50 51 template<typename XprType> 52 struct eval<TensorLayoutSwapOp<XprType>, Eigen::Dense> 53 { 54 typedef const TensorLayoutSwapOp<XprType>& type; 55 }; 56 57 template<typename XprType> 58 struct nested<TensorLayoutSwapOp<XprType>, 1, typename eval<TensorLayoutSwapOp<XprType> >::type> 59 { 60 typedef TensorLayoutSwapOp<XprType> type; 61 }; 62 63 } // end namespace internal 64 65 66 67 template<typename XprType> 68 class TensorLayoutSwapOp : public TensorBase<TensorLayoutSwapOp<XprType>, WriteAccessors> 69 { 70 public: 71 typedef typename Eigen::internal::traits<TensorLayoutSwapOp>::Scalar Scalar; 72 typedef typename Eigen::NumTraits<Scalar>::Real RealScalar; 73 typedef typename internal::remove_const<typename XprType::CoeffReturnType>::type CoeffReturnType; 74 typedef typename Eigen::internal::nested<TensorLayoutSwapOp>::type Nested; 75 typedef typename Eigen::internal::traits<TensorLayoutSwapOp>::StorageKind StorageKind; 76 typedef typename Eigen::internal::traits<TensorLayoutSwapOp>::Index Index; 77 78 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorLayoutSwapOp(const XprType& expr) 79 : m_xpr(expr) {} 80 81 EIGEN_DEVICE_FUNC 82 const typename internal::remove_all<typename XprType::Nested>::type& 83 expression() const { return m_xpr; } 84 85 EIGEN_DEVICE_FUNC 86 EIGEN_STRONG_INLINE TensorLayoutSwapOp& operator = (const TensorLayoutSwapOp& other) 87 { 88 typedef TensorAssignOp<TensorLayoutSwapOp, const TensorLayoutSwapOp> Assign; 89 Assign assign(*this, other); 90 internal::TensorExecutor<const Assign, DefaultDevice>::run(assign, DefaultDevice()); 91 return *this; 92 } 93 94 template<typename OtherDerived> 95 EIGEN_DEVICE_FUNC 96 EIGEN_STRONG_INLINE TensorLayoutSwapOp& operator = (const OtherDerived& other) 97 { 98 typedef TensorAssignOp<TensorLayoutSwapOp, const OtherDerived> Assign; 99 Assign assign(*this, other); 100 internal::TensorExecutor<const Assign, DefaultDevice>::run(assign, DefaultDevice()); 101 return *this; 102 } 103 104 protected: 105 typename XprType::Nested m_xpr; 106 }; 107 108 109 // Eval as rvalue 110 template<typename ArgType, typename Device> 111 struct TensorEvaluator<const TensorLayoutSwapOp<ArgType>, Device> 112 { 113 typedef TensorLayoutSwapOp<ArgType> XprType; 114 typedef typename XprType::Index Index; 115 static const int NumDims = internal::array_size<typename TensorEvaluator<ArgType, Device>::Dimensions>::value; 116 typedef DSizes<Index, NumDims> Dimensions; 117 118 enum { 119 IsAligned = TensorEvaluator<ArgType, Device>::IsAligned, 120 PacketAccess = TensorEvaluator<ArgType, Device>::PacketAccess, 121 Layout = (static_cast<int>(TensorEvaluator<ArgType, Device>::Layout) == static_cast<int>(ColMajor)) ? RowMajor : ColMajor, 122 CoordAccess = false, // to be implemented 123 RawAccess = TensorEvaluator<ArgType, Device>::RawAccess 124 }; 125 126 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorEvaluator(const XprType& op, const Device& device) 127 : m_impl(op.expression(), device) 128 { 129 for(int i = 0; i < NumDims; ++i) { 130 m_dimensions[i] = m_impl.dimensions()[NumDims-1-i]; 131 } 132 } 133 134 typedef typename XprType::Scalar Scalar; 135 typedef typename XprType::CoeffReturnType CoeffReturnType; 136 typedef typename PacketType<CoeffReturnType, Device>::type PacketReturnType; 137 138 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Dimensions& dimensions() const { return m_dimensions; } 139 140 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE bool evalSubExprsIfNeeded(CoeffReturnType* data) { 141 return m_impl.evalSubExprsIfNeeded(data); 142 } 143 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void cleanup() { 144 m_impl.cleanup(); 145 } 146 147 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE CoeffReturnType coeff(Index index) const 148 { 149 return m_impl.coeff(index); 150 } 151 152 template<int LoadMode> 153 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE PacketReturnType packet(Index index) const 154 { 155 return m_impl.template packet<LoadMode>(index); 156 } 157 158 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 159 return m_impl.costPerCoeff(vectorized); 160 } 161 162 EIGEN_DEVICE_FUNC Scalar* data() const { return m_impl.data(); } 163 164 const TensorEvaluator<ArgType, Device>& impl() const { return m_impl; } 165 166 protected: 167 TensorEvaluator<ArgType, Device> m_impl; 168 Dimensions m_dimensions; 169 }; 170 171 172 // Eval as lvalue 173 template<typename ArgType, typename Device> 174 struct TensorEvaluator<TensorLayoutSwapOp<ArgType>, Device> 175 : public TensorEvaluator<const TensorLayoutSwapOp<ArgType>, Device> 176 { 177 typedef TensorEvaluator<const TensorLayoutSwapOp<ArgType>, Device> Base; 178 typedef TensorLayoutSwapOp<ArgType> XprType; 179 180 enum { 181 IsAligned = TensorEvaluator<ArgType, Device>::IsAligned, 182 PacketAccess = TensorEvaluator<ArgType, Device>::PacketAccess, 183 Layout = (static_cast<int>(TensorEvaluator<ArgType, Device>::Layout) == static_cast<int>(ColMajor)) ? RowMajor : ColMajor, 184 CoordAccess = false // to be implemented 185 }; 186 187 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorEvaluator(const XprType& op, const Device& device) 188 : Base(op, device) 189 { } 190 191 typedef typename XprType::Index Index; 192 typedef typename XprType::Scalar Scalar; 193 typedef typename XprType::CoeffReturnType CoeffReturnType; 194 typedef typename PacketType<CoeffReturnType, Device>::type PacketReturnType; 195 196 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE CoeffReturnType& coeffRef(Index index) 197 { 198 return this->m_impl.coeffRef(index); 199 } 200 template <int StoreMode> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE 201 void writePacket(Index index, const PacketReturnType& x) 202 { 203 this->m_impl.template writePacket<StoreMode>(index, x); 204 } 205 }; 206 207 } // end namespace Eigen 208 209 #endif // EIGEN_CXX11_TENSOR_TENSOR_LAYOUT_SWAP_H 210