| /external/tensorflow/tensorflow/stream_executor/rocm/ |
| D | rocm_dnn.h | 45 dnn::PoolingDescriptor op;
50 bool IsSame(const dnn::BatchDescriptor& input_dimensions,
51 const dnn::BatchDescriptor& output_dimensions,
52 const dnn::PoolingDescriptor& pooling_dimensions, int _type);
61 bool find(const void* p, const dnn::BatchDescriptor& input_dimensions,
62 const dnn::BatchDescriptor& output_dimensions,
63 const dnn::PoolingDescriptor& pooling_dimensions, int _type,
65 void insert(const void* p, const dnn::BatchDescriptor& input_dimensions,
66 const dnn::BatchDescriptor& output_dimensions,
67 const dnn::PoolingDescriptor& pooling_dimensions, int _type,
[all …]
|
| D | rocm_dnn.cc | 63 using dnn::AlgorithmDesc;
64 using dnn::BatchDescriptor;
65 using dnn::ConvolutionDescriptor;
66 using dnn::FilterDescriptor;
67 using dnn::NormalizeDescriptor;
68 using dnn::PoolingDescriptor;
504 dnn::ProfileResult GetProfileResultFromConvSolution( in GetProfileResultFromConvSolution()
506 dnn::ProfileResult profile_result; in GetProfileResultFromConvSolution()
514 dnn::ProfileResult GetProfileResultFromConvAlgoPerf( in GetProfileResultFromConvAlgoPerf()
515 dnn::ConvolutionKind kind, miopenConvAlgoPerf_t algorithm) { in GetProfileResultFromConvAlgoPerf()
[all …]
|
| /external/tensorflow/tensorflow/stream_executor/cuda/ |
| D | cuda_dnn.h | 43 class CudnnSupport : public dnn::DnnSupport {
48 port::StatusOr<perftools::gputools::dnn::VersionInfo> GetVersion() override;
50 port::StatusOr<std::unique_ptr<dnn::RnnDescriptor>> createRnnDescriptor(
52 int batch_size, dnn::RnnInputMode input_mode,
53 dnn::RnnDirectionMode direction_mode, dnn::RnnMode rnn_mode,
54 dnn::DataType data_type, const dnn::AlgorithmConfig& algorithm_config,
58 port::StatusOr<std::unique_ptr<dnn::RnnSequenceTensorDescriptor>>
61 dnn::DataType data_type) override;
63 port::StatusOr<std::unique_ptr<dnn::RnnSequenceTensorDescriptor>>
68 dnn::DataType data_type) override;
[all …]
|
| D | cuda_dnn.cc | 205 cudnnDataType_t GetRnnComputeType(dnn::DataType data_type);
207 cudnnConvolutionFwdAlgo_t ToConvForwardAlgo(dnn::AlgorithmDesc algorithm) { in ToConvForwardAlgo()
227 dnn::AlgorithmDesc algorithm) { in ToConvBackwardDataAlgo()
246 dnn::AlgorithmDesc algorithm) { in ToConvBackwardFilterAlgo()
272 cudnnRNNAlgo_t ToCudnnRNNAlgo(absl::optional<dnn::AlgorithmDesc> algorithm) { in ToCudnnRNNAlgo()
357 port::StatusOr<perftools::gputools::dnn::VersionInfo>
361 return perftools::gputools::dnn::VersionInfo( in GetVersion()
517 CudnnTensorDescriptor(const dnn::BatchDescriptor& batch_descriptor, in CudnnTensorDescriptor()
521 case dnn::DataLayout::kBatchYXDepth: in CudnnTensorDescriptor()
522 case dnn::DataLayout::kBatchDepthYX: { in CudnnTensorDescriptor()
[all …]
|
| /external/tensorflow/tensorflow/stream_executor/ |
| D | dnn.h | 52 namespace dnn {
336 port::ArraySlice<dnn::BatchDescriptor> inputs);
469 std::ostream& operator<<(std::ostream& str, dnn::PadAlignment alignment);
1016 const DeviceMemory<float>& side_input, const dnn::BatchDescriptor& x_desc, in DoBatchNormalizationForward()
1017 const dnn::BatchDescriptor& scale_offset_desc, const double epsilon, in DoBatchNormalizationForward()
1019 dnn::ActivationMode activation_mode, DeviceMemory<float>* y, in DoBatchNormalizationForward()
1035 const DeviceMemory<float>& side_input, const dnn::BatchDescriptor& x_desc, in DoBatchNormalizationForward()
1036 const dnn::BatchDescriptor& scale_offset_desc, const double epsilon, in DoBatchNormalizationForward()
1038 dnn::ActivationMode activation_mode, DeviceMemory<Eigen::half>* y, in DoBatchNormalizationForward()
1068 const dnn::BatchDescriptor& x_desc, in DoBatchNormalizationBackward()
[all …]
|
| D | stream.h | 67 namespace dnn {
251 const DeviceMemory<float> &side_input, const dnn::BatchDescriptor &x_desc,
252 const dnn::BatchDescriptor &scale_offset_desc, const double epsilon,
254 dnn::ActivationMode activation_mode, DeviceMemory<float> *y,
264 const DeviceMemory<float> &inv_var, const dnn::BatchDescriptor &x_desc,
265 const dnn::BatchDescriptor &scale_offset_desc, const double epsilon,
276 const DeviceMemory<float> &side_input, const dnn::BatchDescriptor &x_desc,
277 const dnn::BatchDescriptor &scale_offset_desc, const double epsilon,
279 dnn::ActivationMode activation_mode, DeviceMemory<Eigen::half> *y,
290 const dnn::BatchDescriptor &x_desc,
[all …]
|
| D | stream.cc | 37 std::string ToVlogString(const dnn::BatchDescriptor &descriptor) { in ToVlogString()
41 std::string ToVlogString(const dnn::FilterDescriptor &descriptor) { in ToVlogString()
45 std::string ToVlogString(const dnn::ConvolutionDescriptor &descriptor) { in ToVlogString()
49 std::string ToVlogString(const dnn::PoolingDescriptor &descriptor) { in ToVlogString()
53 std::string ToVlogString(const dnn::NormalizeDescriptor &descriptor) { in ToVlogString()
57 std::string ToVlogString(dnn::ActivationMode mode) { in ToVlogString()
58 return dnn::ActivationModeString(mode); in ToVlogString()
61 std::string ToVlogString(const dnn::AlgorithmConfig &algo_config) { in ToVlogString()
65 std::string ToVlogString(dnn::ElementwiseOperation op) { in ToVlogString()
66 return dnn::ElementwiseOperationString(op); in ToVlogString()
[all …]
|
| D | stream_executor_pimpl.h | 366 std::vector<dnn::AlgorithmDesc> *out_algorithms);
371 dnn::ConvolutionKind kind, dnn::DataType element_type, Stream *stream,
372 const dnn::BatchDescriptor &input_descriptor, DeviceMemoryBase input_data,
373 const dnn::FilterDescriptor &filter_descriptor,
375 const dnn::BatchDescriptor &output_descriptor,
377 const dnn::ConvolutionDescriptor &convolution_descriptor,
379 std::vector<dnn::ProfileResult> *out_algorithms);
382 bool GetRnnAlgorithms(std::vector<dnn::AlgorithmDesc> *out_algorithms);
387 std::vector<dnn::AlgorithmDesc> *out_algorithms);
393 std::vector<dnn::AlgorithmDesc> *out_algorithms);
[all …]
|
| D | stream_executor_pimpl.cc | 266 std::vector<dnn::AlgorithmDesc> *out_algorithms) { in GetConvolveAlgorithms()
267 dnn::DnnSupport *dnn_support = AsDnn(); in GetConvolveAlgorithms()
278 dnn::ConvolutionKind kind, dnn::DataType element_type, Stream *stream, in GetMIOpenConvolveAlgorithms()
279 const dnn::BatchDescriptor &input_descriptor, DeviceMemoryBase input_data, in GetMIOpenConvolveAlgorithms()
280 const dnn::FilterDescriptor &filter_descriptor, in GetMIOpenConvolveAlgorithms()
281 DeviceMemoryBase filter_data, const dnn::BatchDescriptor &output_descriptor, in GetMIOpenConvolveAlgorithms()
283 const dnn::ConvolutionDescriptor &convolution_descriptor, in GetMIOpenConvolveAlgorithms()
285 std::vector<dnn::ProfileResult> *out_algorithms) { in GetMIOpenConvolveAlgorithms()
286 dnn::DnnSupport *dnn_support = AsDnn(); in GetMIOpenConvolveAlgorithms()
297 std::vector<dnn::AlgorithmDesc> *out_algorithms) { in GetRnnAlgorithms()
[all …]
|
| D | dnn.cc | 23 namespace dnn { namespace
51 dnn::ConvolutionKind /*kind*/, dnn::DataType /*element_type*/, in GetMIOpenConvolveAlgorithms()
52 Stream* /*stream*/, const dnn::BatchDescriptor& /*input_descriptor*/, in GetMIOpenConvolveAlgorithms()
54 const dnn::FilterDescriptor& /*filter_descriptor*/, in GetMIOpenConvolveAlgorithms()
56 const dnn::BatchDescriptor& /*output_descriptor*/, in GetMIOpenConvolveAlgorithms()
58 const dnn::ConvolutionDescriptor& /*convolution_descriptor*/, in GetMIOpenConvolveAlgorithms()
82 case dnn::QuantizedActivationMode::k8Bit: in QuantizedActivationModeString()
84 case dnn::QuantizedActivationMode::k16Bit: in QuantizedActivationModeString()
86 case dnn::QuantizedActivationMode::k32Bit: in QuantizedActivationModeString()
177 std::ostream& operator<<(std::ostream& str, dnn::PadAlignment alignment) { in operator <<()
[all …]
|
| /external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
| D | stream_executor_util.h | 44 se::dnn::DataLayout input,
45 se::dnn::FilterLayout filter,
46 se::dnn::DataLayout output);
50 std::tuple<se::dnn::DataLayout, se::dnn::FilterLayout, se::dnn::DataLayout>>
90 StatusOr<se::dnn::ConvolutionKind> GetDNNConvKindFromCudnnConvKind(
92 StatusOr<se::dnn::DataType> GetDNNDataTypeFromPrimitiveType(PrimitiveType type);
|
| D | gpu_conv_runner.h | 36 se::dnn::ProfileResult* profile_result = nullptr;
39 absl::optional<se::dnn::AlgorithmDesc> algo_override;
60 se::dnn::ActivationMode mode;
67 se::dnn::AlgorithmConfig algorithm;
70 se::dnn::BatchDescriptor input_descriptor;
71 se::dnn::FilterDescriptor filter_descriptor;
72 se::dnn::BatchDescriptor output_descriptor;
73 se::dnn::ConvolutionDescriptor conv_desc;
|
| D | hlo_algorithm_denylist_test.cc | 56 EXPECT_EQ(stream_executor::dnn::AlgorithmDesc(0, false), list[0]); in TEST_F()
57 EXPECT_EQ(stream_executor::dnn::AlgorithmDesc(0, true), list[1]); in TEST_F()
58 EXPECT_EQ(stream_executor::dnn::AlgorithmDesc(1, false), list[2]); in TEST_F()
59 EXPECT_EQ(stream_executor::dnn::AlgorithmDesc(1, true), list[3]); in TEST_F()
|
| D | stream_executor_util.cc | 36 using se::dnn::DataLayout;
37 using se::dnn::DataLayoutString;
38 using se::dnn::FilterLayout;
39 using se::dnn::FilterLayoutString;
322 StatusOr<se::dnn::ConvolutionKind> GetDNNConvKindFromCudnnConvKind( in GetDNNConvKindFromCudnnConvKind()
326 return se::dnn::BACKWARD_FILTER; in GetDNNConvKindFromCudnnConvKind()
328 return se::dnn::BACKWARD_DATA; in GetDNNConvKindFromCudnnConvKind()
330 return se::dnn::FORWARD; in GetDNNConvKindFromCudnnConvKind()
337 StatusOr<se::dnn::DataType> GetDNNDataTypeFromPrimitiveType( in GetDNNDataTypeFromPrimitiveType()
341 return se::dnn::ToDataType<Eigen::half>::value; in GetDNNDataTypeFromPrimitiveType()
[all …]
|
| D | cudnn_batchnorm_runner.cc | 33 se::dnn::BatchDescriptor operand_desc;
34 se::dnn::BatchDescriptor scale_offset_desc;
66 se::dnn::BatchDescriptor input_desc;
67 se::dnn::BatchDescriptor scale_offset_desc;
98 batch_descs.input_desc.set_layout(se::dnn::DataLayout::kBatchDepthYX) in MakeBatchNormDescriptors()
104 batch_descs.scale_offset_desc.set_layout(se::dnn::DataLayout::kBatchDepthYX) in MakeBatchNormDescriptors()
149 se::dnn::ActivationMode::kNone, // in RunCudnnBatchNormForwardInferenceImpl()
178 se::dnn::ActivationMode::kNone, // in RunCudnnBatchNormForwardTrainingImpl()
|
| /external/tensorflow/tensorflow/core/kernels/ |
| D | gpu_utils.cc | 101 if (auto* dnn = stream_executor->AsDnn()) { in GetCudnnVersion() local
102 se::port::StatusOr<se::dnn::VersionInfo> version_or = dnn->GetVersion(); in GetCudnnVersion()
126 void LogConvAutotuneResults(se::dnn::ConvolutionKind kind, in LogConvAutotuneResults()
127 se::dnn::DataType element_type, in LogConvAutotuneResults()
131 const se::dnn::BatchDescriptor& input_desc, in LogConvAutotuneResults()
132 const se::dnn::FilterDescriptor& filter_desc, in LogConvAutotuneResults()
133 const se::dnn::BatchDescriptor& output_desc, in LogConvAutotuneResults()
134 const se::dnn::ConvolutionDescriptor& conv_desc, in LogConvAutotuneResults()
171 se::dnn::DataType element_type, se::DeviceMemoryBase input_buffer, in LogFusedConvForwardAutotuneResults()
174 const se::dnn::BatchDescriptor& input_desc, in LogFusedConvForwardAutotuneResults()
[all …]
|
| D | gpu_utils.h | 217 void LogConvAutotuneResults(se::dnn::ConvolutionKind kind,
218 se::dnn::DataType element_type,
222 const se::dnn::BatchDescriptor& input_desc,
223 const se::dnn::FilterDescriptor& filter_desc,
224 const se::dnn::BatchDescriptor& output_desc,
225 const se::dnn::ConvolutionDescriptor& conv_desc,
231 se::dnn::DataType element_type, se::DeviceMemoryBase input_buffer,
234 const se::dnn::BatchDescriptor& input_desc,
235 const se::dnn::FilterDescriptor& filter_desc,
236 const se::dnn::BatchDescriptor& output_desc,
[all …]
|
| D | cudnn_pooling_gpu.cc | 35 se::dnn::PoolingMode pooling_mode, in Compute()
71 se::dnn::PoolingDescriptor pooling_desc(3); in Compute()
73 se::dnn::BatchDescriptor input_desc(3); in Compute()
76 .set_layout(se::dnn::DataLayout::kBatchDepthYX); in Compute()
77 se::dnn::BatchDescriptor output_desc(3); in Compute()
80 .set_layout(se::dnn::DataLayout::kBatchDepthYX); in Compute()
82 const auto dim_i = static_cast<se::dnn::DimIndex>(i); in Compute()
134 OpKernelContext* context, se::dnn::PoolingMode pooling_mode, in Compute()
140 CHECK((pooling_mode != se::dnn::PoolingMode::kMaximum) || in Compute()
204 se::dnn::PoolingDescriptor pooling_desc(3); in Compute()
[all …]
|
| D | pooling_ops_common.cc | 214 se::dnn::PoolingMode pooling_mode, in Compute()
266 se::dnn::DataLayout data_layout = se::dnn::DataLayout::kBatchDepthYX; in Compute()
270 se::dnn::DataLayout data_layout; in Compute()
273 data_layout = se::dnn::DataLayout::kBatchYXDepth; in Compute()
276 data_layout = se::dnn::DataLayout::kBatchDepthYX; in Compute()
281 data_layout = se::dnn::DataLayout::kBatchYXDepth; in Compute()
352 se::dnn::PoolingDescriptor pooling_desc; in Compute()
362 se::dnn::BatchDescriptor input_desc; in Compute()
369 se::dnn::BatchDescriptor output_desc; in Compute()
442 OpKernelContext* context, se::dnn::PoolingMode pooling_mode, in Compute()
[all …]
|
| D | conv_ops_fused_impl.h | 310 const se::dnn::ActivationMode activation_mode)
322 std::tuple<ConvParameters::ParameterDataType, se::dnn::ActivationMode>;
329 se::dnn::ActivationMode activation_mode_;
349 se::dnn::AlgorithmConfig>;
369 se::dnn::AlgorithmConfig* algorithm_config) {
376 std::vector<se::dnn::AlgorithmDesc> algorithms;
403 se::dnn::ProfileResult profile_result;
406 launch(se::dnn::AlgorithmConfig(profile_algorithm), allocator_used,
563 se::dnn::ActivationMode dnn_activation_mode;
566 dnn_activation_mode = se::dnn::ActivationMode::kRelu;
[all …]
|
| D | conv_grad_input_ops.cc | 43 se::dnn::AlgorithmConfig>
72 using se::dnn::AlgorithmConfig; in operator ()()
73 using se::dnn::AlgorithmDesc; in operator ()()
74 using se::dnn::ProfileResult; in operator ()()
237 std::make_tuple(se::dnn::DataLayout::kBatchYXDepth, in operator ()()
238 se::dnn::FilterLayout::kOutputYXInput); in operator ()()
240 std::make_tuple(se::dnn::DataLayout::kBatchDepthYX, in operator ()()
241 se::dnn::FilterLayout::kOutputInputYX); in operator ()()
243 se::dnn::DataLayout compute_data_layout; in operator ()()
244 se::dnn::FilterLayout filter_layout; in operator ()()
[all …]
|
| D | conv_ops_3d.cc | 39 using stream_executor::dnn::DimIndex;
206 se::dnn::AlgorithmConfig>
382 std::make_tuple(se::dnn::DataLayout::kBatchYXDepth, in launch()
383 se::dnn::FilterLayout::kOutputYXInput); in launch()
385 std::make_tuple(se::dnn::DataLayout::kBatchDepthYX, in launch()
386 se::dnn::FilterLayout::kOutputInputYX); in launch()
388 se::dnn::DataLayout compute_data_layout; in launch()
389 se::dnn::FilterLayout filter_layout; in launch()
397 se::dnn::BatchDescriptor input_desc(3); in launch()
404 se::dnn::BatchDescriptor output_desc(3); in launch()
[all …]
|
| /external/tensorflow/tensorflow/core/protobuf/ |
| D | conv_autotuning.proto | 7 import "tensorflow/stream_executor/dnn.proto";
14 stream_executor.dnn.ConvolutionKind kind = 1;
15 stream_executor.dnn.TensorDescriptorProto input = 2;
16 stream_executor.dnn.TensorDescriptorProto filter = 3;
17 stream_executor.dnn.TensorDescriptorProto output = 4;
18 stream_executor.dnn.ConvolutionDescriptorProto conv_desc = 5;
26 stream_executor.dnn.ActivationMode activation = 8;
|
| /external/tensorflow/tensorflow/python/estimator/canned/ |
| D | dnn.py | 26 from tensorflow_estimator.python.estimator.canned import dnn
30 dnn.__all__ = [s for s in dir(dnn) if not s.startswith('__')]
32 from tensorflow_estimator.python.estimator.canned.dnn import *
|
| /external/tensorflow/tensorflow/compiler/mlir/xla/ |
| D | attribute_exporter.cc | 60 StatusOr<stream_executor::dnn::ActivationMode> ConvertConvActivationMode( in ConvertConvActivationMode()
70 return stream_executor::dnn::kNone; in ConvertConvActivationMode()
72 return stream_executor::dnn::kSigmoid; in ConvertConvActivationMode()
74 return stream_executor::dnn::kTanh; in ConvertConvActivationMode()
76 return stream_executor::dnn::kRelu; in ConvertConvActivationMode()
78 return stream_executor::dnn::kRelu6; in ConvertConvActivationMode()
80 return stream_executor::dnn::kReluX; in ConvertConvActivationMode()
82 return stream_executor::dnn::kBandPass; in ConvertConvActivationMode()
|