/external/tensorflow/tensorflow/core/grappler/clusters/ |
virtual_cluster_test.cc | 89 EXPECT_EQ(1500, node.compute_cost()); 91 EXPECT_EQ(2500, node.compute_cost());
|
single_machine_test.cc | 95 EXPECT_LE(0, node.compute_cost()); 96 EXPECT_GE(run_duration_micros, node.compute_cost());
|
/external/tensorflow/tensorflow/core/grappler/costs/ |
op_performance_data.proto | 81 int64 compute_cost = 3;
|
op_level_cost_estimator.cc | 420 Costs::NanoSeconds compute_cost(std::ceil(operations / device_perf.gigaops)); 422 << " Execution Time (ns):" << compute_cost.count(); 437 costs.compute_time = compute_cost; 440 costs.execution_time = std::max(compute_cost, memory_cost); 442 costs.execution_time = compute_cost + memory_cost; [all...] |
utils.cc | 278 // Note that CostGraphDef::Node::compute_cost is microseconds, while 279 // OpPerformance.compute_cost is nanoseconds. 280 perf->set_compute_cost(cost_node->compute_cost() * 1000);
|
/external/eigen/unsupported/Eigen/CXX11/src/Tensor/ |
TensorBroadcasting.h | 353 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);
|
TensorCostModel.h | 84 double load_cost, double store_cost, double compute_cost) const { 86 compute_cost * compute_cycles_;
|
TensorReverse.h | 213 double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() + local 218 compute_cost += 2 * TensorOpCost::AddCost<Index>(); 222 TensorOpCost(0, 0, compute_cost, false /* vectorized */, PacketSize);
|
TensorStriding.h | 212 double compute_cost = (NumDims - 1) * (TensorOpCost::AddCost<Index>() + local 217 compute_cost *= 2; // packet() computes two indices 222 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
|
TensorArgMax.h | 259 const double compute_cost = 1.0 + local 262 m_impl.costPerCoeff(vectorized) + TensorOpCost(0, 0, compute_cost);
|
TensorInflation.h | 204 const double compute_cost = NumDims * (3 * TensorOpCost::DivCost<Index>() + local 213 compute_cost, vectorized, PacketSize);
|
TensorShuffling.h | 181 const double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() + local 185 TensorOpCost(0, 0, compute_cost, false /* vectorized */, PacketSize);
|
TensorPatch.h | 249 const double compute_cost = NumDims * (TensorOpCost::DivCost<Index>() + local 253 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
|
TensorConcatenation.h | 265 const double compute_cost = NumDims * (2 * TensorOpCost::AddCost<Index>() + local 275 TensorOpCost(0, 0, compute_cost);
|
TensorImagePatch.h | 441 const double compute_cost = 3 * TensorOpCost::DivCost<Index>() + local 445 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
|
TensorVolumePatch.h | 499 const double compute_cost = local 502 return TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
|
TensorReduction.h | 659 const double compute_cost = num_values_to_reduce * internal::functor_traits<Op>::Cost; local 661 TensorOpCost(0, 0, compute_cost, vectorized, PacketSize); [all...] |
/external/tensorflow/tensorflow/python/grappler/ |
cost_analyzer.cc | 103 double measured_cost = perf->compute_cost(); 141 ops[op_name].time += op_perf.compute_cost();
|
/external/tensorflow/tensorflow/core/common_runtime/ |
direct_session_with_tracking_alloc_test.cc | 298 EXPECT_LE(0, node.compute_cost()); 299 EXPECT_GE(run_duration_micros, node.compute_cost());
|
/external/tensorflow/tensorflow/core/kernels/ |
mirror_pad_op.h | 240 const double compute_cost = Dims * (7 * TensorOpCost::AddCost<Index>() + 244 TensorOpCost(1, 0, compute_cost, vectorized, kPacketSize);
|
eigen_volume_patch.h | 434 const double compute_cost = 10 * TensorOpCost::DivCost<Index>() + local 437 return TensorOpCost(0, 0, compute_cost, vectorized, PacketSize);
|