diff options
Diffstat (limited to 'unsupported/Eigen/CXX11/src/Tensor/TensorReverse.h')
-rw-r--r-- | unsupported/Eigen/CXX11/src/Tensor/TensorReverse.h | 40 |
1 files changed, 26 insertions, 14 deletions
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorReverse.h b/unsupported/Eigen/CXX11/src/Tensor/TensorReverse.h index 96d92038c..1a59cc8f7 100644 --- a/unsupported/Eigen/CXX11/src/Tensor/TensorReverse.h +++ b/unsupported/Eigen/CXX11/src/Tensor/TensorReverse.h @@ -104,6 +104,10 @@ struct TensorEvaluator<const TensorReverseOp<ReverseDimensions, ArgType>, Device typedef typename XprType::Index Index; static const int NumDims = internal::array_size<ReverseDimensions>::value; typedef DSizes<Index, NumDims> Dimensions; + typedef typename XprType::Scalar Scalar; + typedef typename XprType::CoeffReturnType CoeffReturnType; + typedef typename PacketType<CoeffReturnType, Device>::type PacketReturnType; + static const int PacketSize = internal::unpacket_traits<PacketReturnType>::size; enum { IsAligned = false, @@ -135,10 +139,6 @@ struct TensorEvaluator<const TensorReverseOp<ReverseDimensions, ArgType>, Device } } - typedef typename XprType::Scalar Scalar; - typedef typename XprType::CoeffReturnType CoeffReturnType; - typedef typename PacketType<CoeffReturnType, Device>::type PacketReturnType; - EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Dimensions& dimensions() const { return m_dimensions; } @@ -195,21 +195,33 @@ struct TensorEvaluator<const TensorReverseOp<ReverseDimensions, ArgType>, Device EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE PacketReturnType packet(Index index) const { - const int packetSize = internal::unpacket_traits<PacketReturnType>::size; - EIGEN_STATIC_ASSERT(packetSize > 1, YOU_MADE_A_PROGRAMMING_MISTAKE) - eigen_assert(index+packetSize-1 < dimensions().TotalSize()); + EIGEN_STATIC_ASSERT(PacketSize > 1, YOU_MADE_A_PROGRAMMING_MISTAKE) + eigen_assert(index+PacketSize-1 < dimensions().TotalSize()); // TODO(ndjaitly): write a better packing routine that uses // local structure. EIGEN_ALIGN_MAX typename internal::remove_const<CoeffReturnType>::type - values[packetSize]; - for (int i = 0; i < packetSize; ++i) { + values[PacketSize]; + for (int i = 0; i < PacketSize; ++i) { values[i] = coeff(index+i); } PacketReturnType rslt = internal::pload<PacketReturnType>(values); return rslt; } + EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { + double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() + + 2 * TensorOpCost::MulCost<Index>() + + TensorOpCost::DivCost<Index>()); + for (int i = 0; i < NumDims; ++i) { + if (m_reverse[i]) { + compute_cost += 2 * TensorOpCost::AddCost<Index>(); + } + } + return m_impl.costPerCoeff(vectorized) + + TensorOpCost(0, 0, compute_cost, false /* vectorized */, PacketSize); + } + EIGEN_DEVICE_FUNC Scalar* data() const { return NULL; } protected: @@ -246,6 +258,7 @@ struct TensorEvaluator<TensorReverseOp<ReverseDimensions, ArgType>, Device> typedef typename XprType::Scalar Scalar; typedef typename XprType::CoeffReturnType CoeffReturnType; typedef typename PacketType<CoeffReturnType, Device>::type PacketReturnType; + static const int PacketSize = internal::unpacket_traits<PacketReturnType>::size; EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Dimensions& dimensions() const { return this->m_dimensions; } @@ -256,14 +269,13 @@ struct TensorEvaluator<TensorReverseOp<ReverseDimensions, ArgType>, Device> template <int StoreMode> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void writePacket(Index index, const PacketReturnType& x) { - const int packetSize = internal::unpacket_traits<PacketReturnType>::size; - EIGEN_STATIC_ASSERT(packetSize > 1, YOU_MADE_A_PROGRAMMING_MISTAKE) - eigen_assert(index+packetSize-1 < dimensions().TotalSize()); + EIGEN_STATIC_ASSERT(PacketSize > 1, YOU_MADE_A_PROGRAMMING_MISTAKE) + eigen_assert(index+PacketSize-1 < dimensions().TotalSize()); // This code is pilfered from TensorMorphing.h - EIGEN_ALIGN_MAX CoeffReturnType values[packetSize]; + EIGEN_ALIGN_MAX CoeffReturnType values[PacketSize]; internal::pstore<CoeffReturnType, PacketReturnType>(values, x); - for (int i = 0; i < packetSize; ++i) { + for (int i = 0; i < PacketSize; ++i) { this->coeffRef(index+i) = values[i]; } } |