/external/eigen/unsupported/Eigen/CXX11/src/Tensor/ |
D | TensorCostModel.h | 25 class TensorOpCost { 55 TensorOpCost() : bytes_loaded_(0), bytes_stored_(0), compute_cycles_(0) {} in TensorOpCost() function 57 TensorOpCost(double bytes_loaded, double bytes_stored, double compute_cycles) in TensorOpCost() function 63 TensorOpCost(double bytes_loaded, double bytes_stored, double compute_cycles, in TensorOpCost() function 97 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost cwiseMin( in cwiseMin() 98 const TensorOpCost& rhs) const { in cwiseMin() 102 return TensorOpCost(bytes_loaded, bytes_stored, compute_cycles); in cwiseMin() 106 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost cwiseMax( in cwiseMax() 107 const TensorOpCost& rhs) const { in cwiseMax() 111 return TensorOpCost(bytes_loaded, bytes_stored, compute_cycles); in cwiseMax() [all …]
|
D | TensorBroadcasting.h | 351 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 353 double compute_cost = TensorOpCost::AddCost<Index>(); 356 compute_cost += TensorOpCost::DivCost<Index>(); 359 TensorOpCost::MulCost<Index>() + TensorOpCost::AddCost<Index>(); 362 compute_cost += TensorOpCost::MulCost<Index>() + 363 TensorOpCost::ModCost<Index>() + 364 TensorOpCost::AddCost<Index>(); 368 TensorOpCost::MulCost<Index>() + TensorOpCost::AddCost<Index>(); 372 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
|
D | TensorPadding.h | 189 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 190 TensorOpCost cost = m_impl.costPerCoeff(vectorized); 238 void updateCostPerDimension(TensorOpCost& cost, int i, bool first) const { 246 cost += TensorOpCost(0, 0, 2 * TensorOpCost::AddCost<Index>() + 247 reduction * (1 * TensorOpCost::AddCost<Index>())); 249 cost += TensorOpCost(0, 0, 2 * TensorOpCost::AddCost<Index>() + 250 2 * TensorOpCost::MulCost<Index>() + 251 reduction * (2 * TensorOpCost::MulCost<Index>() + 252 1 * TensorOpCost::DivCost<Index>()));
|
D | TensorInflation.h | 203 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 204 const double compute_cost = NumDims * (3 * TensorOpCost::DivCost<Index>() + 205 3 * TensorOpCost::MulCost<Index>() + 206 2 * TensorOpCost::AddCost<Index>()); 210 return TensorOpCost(); 212 TensorOpCost(sizeof(CoeffReturnType) * input_size / output_size, 0,
|
D | TensorReverse.h | 212 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 213 double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() + 214 2 * TensorOpCost::MulCost<Index>() + 215 TensorOpCost::DivCost<Index>()); 218 compute_cost += 2 * TensorOpCost::AddCost<Index>(); 222 TensorOpCost(0, 0, compute_cost, false /* vectorized */, PacketSize);
|
D | TensorEvaluator.h | 106 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { in costPerCoeff() 107 return TensorOpCost(sizeof(CoeffReturnType), 0, 0, vectorized, in costPerCoeff() 205 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 206 return TensorOpCost(sizeof(CoeffReturnType), 0, 0, vectorized, 268 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 270 return TensorOpCost(sizeof(CoeffReturnType), 0, 0, vectorized, 338 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 341 TensorOpCost(0, 0, functor_cost, vectorized, PacketSize); 416 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 421 TensorOpCost(0, 0, functor_cost, vectorized, PacketSize); [all …]
|
D | TensorArgMax.h | 115 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 117 return m_impl.costPerCoeff(vectorized) + TensorOpCost(0, 0, 1); 257 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 260 … (m_return_dim < 0 ? 0.0 : (TensorOpCost::ModCost<Index>() + TensorOpCost::DivCost<Index>())); 262 m_impl.costPerCoeff(vectorized) + TensorOpCost(0, 0, compute_cost);
|
D | TensorChipping.h | 244 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 251 cost += TensorOpCost::MulCost<Index>() + TensorOpCost::AddCost<Index>(); 256 cost += TensorOpCost::AddCost<Index>(); 258 cost += 3 * TensorOpCost::MulCost<Index>() + TensorOpCost::DivCost<Index>() + 259 3 * TensorOpCost::AddCost<Index>(); 263 TensorOpCost(0, 0, cost, vectorized, PacketSize);
|
D | TensorGenerator.h | 148 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 152 return TensorOpCost(0, 0, TensorOpCost::AddCost<Scalar>() + 153 TensorOpCost::MulCost<Scalar>());
|
D | TensorShuffling.h | 180 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 181 const double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() + 182 2 * TensorOpCost::MulCost<Index>() + 183 TensorOpCost::DivCost<Index>()); 185 TensorOpCost(0, 0, compute_cost, false /* vectorized */, PacketSize);
|
D | TensorAssign.h | 153 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 158 TensorOpCost left = m_leftImpl.costPerCoeff(vectorized); 160 TensorOpCost( 163 TensorOpCost(0, sizeof(CoeffReturnType), 0, vectorized, PacketSize);
|
D | TensorPatch.h | 248 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 249 const double compute_cost = NumDims * (TensorOpCost::DivCost<Index>() + 250 TensorOpCost::MulCost<Index>() + 251 2 * TensorOpCost::AddCost<Index>()); 253 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
|
D | TensorStriding.h | 211 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 212 double compute_cost = (NumDims - 1) * (TensorOpCost::AddCost<Index>() + 213 TensorOpCost::MulCost<Index>() + 214 TensorOpCost::DivCost<Index>()) + 215 TensorOpCost::MulCost<Index>(); 222 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
|
D | TensorConvolution.h | 451 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 456 TensorOpCost::AddCost<Scalar>() + TensorOpCost::MulCost<Scalar>(); 459 (2 * TensorOpCost::AddCost<Index>() + 2 * TensorOpCost::MulCost<Index>() + 460 TensorOpCost::DivCost<Index>()); 461 return TensorOpCost(0, 0, firstIndex_compute_cost, vectorized, PacketSize) + 464 TensorOpCost(0, 0, convolve_compute_cost, vectorized, 1065 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 1072 TensorOpCost::AddCost<Scalar>() + TensorOpCost::MulCost<Scalar>(); 1075 (2 * TensorOpCost::AddCost<Index>() + 2 * TensorOpCost::MulCost<Index>() + 1076 TensorOpCost::DivCost<Index>()); [all …]
|
D | TensorConcatenation.h | 263 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 265 const double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() + 266 2 * TensorOpCost::MulCost<Index>() + 267 TensorOpCost::DivCost<Index>() + 268 TensorOpCost::ModCost<Index>()); 275 TensorOpCost(0, 0, compute_cost);
|
D | TensorCustomOp.h | 136 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 138 return TensorOpCost(sizeof(CoeffReturnType), 0, 0, vectorized, PacketSize); 291 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost costPerCoeff(bool vectorized) const { 293 return TensorOpCost(sizeof(CoeffReturnType), 0, 0, vectorized, PacketSize);
|
D | TensorConversion.h | 232 EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE TensorOpCost 234 const double cast_cost = TensorOpCost::CastCost<SrcType, TargetType>(); 241 TensorOpCost(0, 0, TgtCoeffRatio * (cast_cost / PacketSize)); 243 return m_impl.costPerCoeff(vectorized) + TensorOpCost(0, 0, cast_cost);
|
/external/tensorflow/tensorflow/core/kernels/ |
D | parameterized_truncated_normal_op.cc | 279 (Eigen::TensorOpCost::AddCost<T>() + in operator ()() 280 Eigen::TensorOpCost::MulCost<T>()) * in operator ()() 283 + Eigen::TensorOpCost::AddCost<T>() + in operator ()() 284 Eigen::TensorOpCost::MulCost<T>() + in operator ()() 288 + Eigen::TensorOpCost::MulCost<T>() * 4 + in operator ()() 291 + Eigen::TensorOpCost::AddCost<T>(); in operator ()() 297 uniformSampleCost + Eigen::TensorOpCost::MulCost<T>() + in operator ()() 298 Eigen::TensorOpCost::AddCost<T>() + in operator ()() 299 Eigen::TensorOpCost::MulCost<T>() * 2 + in operator ()() 300 Eigen::TensorOpCost::AddCost<T>() + uniformSampleCost + in operator ()() [all …]
|
D | transpose_functor_cpu.cc | 60 (conjugate ? 1 : 0) + ndims * (Eigen::TensorOpCost::DivCost<int64>() + in TransposeSimple() 61 2 * Eigen::TensorOpCost::MulCost<int64>() + in TransposeSimple() 62 2 * Eigen::TensorOpCost::AddCost<int64>()); in TransposeSimple() 63 Eigen::TensorOpCost cost(/*bytes_loaded=*/sizeof(T), in TransposeSimple()
|
D | dense_update_functor.cc | 44 Eigen::TensorOpCost(.1, // chosen to force large chunks in operator ()() 66 Eigen::TensorOpCost(estimated_string_size, estimated_string_size, 0), in operator ()()
|
D | population_count_op.cc | 111 const double total_cost = (Eigen::TensorOpCost::CastCost<T, uint8>() + in operator ()() 112 Eigen::TensorOpCost::CastCost<int64, uint8>()); in operator ()()
|
/external/tensorflow/tensorflow/core/kernels/image/ |
D | crop_and_resize_op.cc | 328 depth * (Eigen::TensorOpCost::AddCost<float>() * 6 + in operator ()() 329 Eigen::TensorOpCost::MulCost<float>() * 3 + in operator ()() 330 Eigen::TensorOpCost::CastCost<T, float>() * 4) + in operator ()() 331 (Eigen::TensorOpCost::AddCost<float>() * 2 + in operator ()() 332 Eigen::TensorOpCost::AddCost<float>() * 3); in operator ()() 334 cost_per_pixel = depth * Eigen::TensorOpCost::CastCost<T, float>() + in operator ()() 335 Eigen::TensorOpCost::AddCost<float>() * 4 + in operator ()() 336 Eigen::TensorOpCost::MulCost<float>() * 4; in operator ()() 536 ? depth * (Eigen::TensorOpCost::AddCost<float>() * 7 + in operator ()() 537 Eigen::TensorOpCost::MulCost<float>() * 6 + in operator ()() [all …]
|
D | non_max_suppression_op.cc | 487 int compute_cycles = Eigen::TensorOpCost::AddCost<int>() * num_boxes * 14 + in BatchedNonMaxSuppressionOp() 488 Eigen::TensorOpCost::MulCost<int>() * num_boxes * 9 + in BatchedNonMaxSuppressionOp() 489 Eigen::TensorOpCost::MulCost<float>() * num_boxes * 9 + in BatchedNonMaxSuppressionOp() 490 Eigen::TensorOpCost::AddCost<float>() * num_boxes * 8; in BatchedNonMaxSuppressionOp() 493 const Eigen::TensorOpCost cost(input_bytes, output_bytes, compute_cycles); in BatchedNonMaxSuppressionOp() 526 compute_cycles = Eigen::TensorOpCost::AddCost<int>() * num_boxes * 5 + in BatchedNonMaxSuppressionOp() 527 Eigen::TensorOpCost::AddCost<float>() * num_boxes * 5; in BatchedNonMaxSuppressionOp() 530 const Eigen::TensorOpCost cost_result(input_bytes, output_bytes, in BatchedNonMaxSuppressionOp() 566 compute_cycles = Eigen::TensorOpCost::AddCost<int>() * 2 + in BatchedNonMaxSuppressionOp() 567 Eigen::TensorOpCost::MulCost<int>() * 2 + in BatchedNonMaxSuppressionOp() [all …]
|
/external/tensorflow/tensorflow/core/kernels/linalg/ |
D | tridiagonal_matmul_op.cc | 71 const double add_cost = Eigen::TensorOpCost::AddCost<Scalar>(); in GetCostPerUnit() 72 const double mult_cost = Eigen::TensorOpCost::MulCost<Scalar>(); in GetCostPerUnit()
|
D | tridiagonal_solve_op.cc | 81 const double add_cost = Eigen::TensorOpCost::AddCost<Scalar>(); in GetCostPerUnit() 82 const double mult_cost = Eigen::TensorOpCost::MulCost<Scalar>(); in GetCostPerUnit() 83 const double div_cost = Eigen::TensorOpCost::DivCost<Scalar>(); in GetCostPerUnit()
|