Home
last modified time | relevance | path

Searched refs:compute_cost (Results 1 – 23 of 23) sorted by relevance

/external/eigen/unsupported/Eigen/CXX11/src/Tensor/
DTensorBroadcasting.h353 double compute_cost = TensorOpCost::AddCost<Index>();
356 compute_cost += TensorOpCost::DivCost<Index>();
358 compute_cost +=
362 compute_cost += TensorOpCost::MulCost<Index>() +
367 compute_cost +=
372 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
DTensorCostModel.h84 double load_cost, double store_cost, double compute_cost) const { in total_cost() argument
86 compute_cost * compute_cycles_; in total_cost()
DTensorReverse.h213 double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() +
218 compute_cost += 2 * TensorOpCost::AddCost<Index>();
222 TensorOpCost(0, 0, compute_cost, false /* vectorized */, PacketSize);
DTensorStriding.h212 double compute_cost = (NumDims - 1) * (TensorOpCost::AddCost<Index>() +
217 compute_cost *= 2; // packet() computes two indices
222 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
DTensorInflation.h204 const double compute_cost = NumDims * (3 * TensorOpCost::DivCost<Index>() +
213 compute_cost, vectorized, PacketSize);
DTensorShuffling.h181 const double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() +
185 TensorOpCost(0, 0, compute_cost, false /* vectorized */, PacketSize);
DTensorArgMax.h259 const double compute_cost = 1.0 +
262 m_impl.costPerCoeff(vectorized) + TensorOpCost(0, 0, compute_cost);
DTensorPatch.h249 const double compute_cost = NumDims * (TensorOpCost::DivCost<Index>() +
253 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
DTensorConcatenation.h265 const double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() +
275 TensorOpCost(0, 0, compute_cost);
DTensorImagePatch.h441 const double compute_cost = 3 * TensorOpCost::DivCost<Index>() +
445 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
DTensorVolumePatch.h499 const double compute_cost =
502 return TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
DTensorReduction.h659 const double compute_cost = num_values_to_reduce * internal::functor_traits<Op>::Cost;
661 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
/external/tensorflow/tensorflow/core/grappler/clusters/
Dvirtual_cluster_test.cc90 EXPECT_EQ(1500, node.compute_cost()); in TEST_F()
92 EXPECT_EQ(2500, node.compute_cost()); in TEST_F()
Dsingle_machine_test.cc106 EXPECT_LE(0, node.compute_cost()); in TEST_F()
107 EXPECT_GE(run_duration_micros, node.compute_cost()); in TEST_F()
/external/tensorflow/tensorflow/python/grappler/
Dcost_analyzer.cc109 double measured_cost = perf->compute_cost(); in PreprocessCosts()
146 ops[op_name].time += op_perf.compute_cost(); in AnalyzeCosts()
255 os << std::setw(width_wide) << perf.compute_cost() << ","; in PrintAnalysis()
/external/tensorflow/tensorflow/core/framework/
Dcost_graph.proto58 int64 compute_cost = 9; field
/external/tensorflow/tensorflow/core/common_runtime/
Ddirect_session_with_tracking_alloc_test.cc305 EXPECT_LE(0, node.compute_cost()); in TEST()
306 EXPECT_GE(run_duration_micros, node.compute_cost()); in TEST()
/external/tensorflow/tensorflow/core/grappler/costs/
Dop_performance_data.proto85 int64 compute_cost = 3; field
Dvirtual_scheduler.cc1030 const auto& compute_cost = op_cost_pair.second.compute_time.count(); in Summary() local
1038 compute_cost, memory_cost, in Summary()
1128 const auto& compute_cost = op_cost_pair.second.compute_time.count(); in Summary() local
1151 (is_op_cost_accurate ? ' ' : '~'), cost, compute_cost, in Summary()
Dutils.cc323 perf->set_compute_cost(cost_node->compute_cost() * 1000); in CostGraphToOpPerformanceData()
Dop_level_cost_estimator.cc840 Costs::NanoSeconds compute_cost(std::ceil(operations / device_info.gigaops)); in PredictOpCountBasedCost() local
842 << " Compute Time (ns):" << compute_cost.count(); in PredictOpCountBasedCost()
871 costs.compute_time = compute_cost; in PredictOpCountBasedCost()
/external/tensorflow/tensorflow/core/kernels/image/
Dmirror_pad_op.h247 const double compute_cost = Dims * (7 * TensorOpCost::AddCost<Index>() +
251 TensorOpCost(1, 0, compute_cost, vectorized, kPacketSize);
/external/tensorflow/tensorflow/core/kernels/
Deigen_volume_patch.h435 const double compute_cost = 10 * TensorOpCost::DivCost<Index>() + in costPerCoeff() local
438 return TensorOpCost(0, 0, compute_cost, vectorized, PacketSize); in costPerCoeff()