H A D | cuda_dnn.cc | 1426 const DeviceMemory<T>& input_data, 1536 const DeviceMemory<T>& input_data, 1552 rnn_desc, input_desc, input_data, input_h_desc, input_h_data, 1610 input_desc.handles() /*xDesc*/, input_data.opaque() /*x*/, 1628 input_desc.handles() /*xDesc*/, input_data.opaque() /*x*/, 1653 const DeviceMemory<T>& input_data, 1676 rnn_desc, input_desc, input_data, input_h_desc, input_h_data, 1733 input_desc.handles() /*xDesc*/, input_data.opaque() /*x*/, 1831 const DeviceMemory<Eigen::half>& input_data, 1862 stream, cudnn_rnn_desc, cudnn_input_desc, input_data, cudnn_input_h_des 1423 ExtractAndCheckRnnForward( const CudnnRnnDescriptor& rnn_desc, const CudnnRnnSequenceTensorDescriptor& input_desc, const DeviceMemory<T>& input_data, const CudnnRnnStateTensorDescriptor& input_h_desc, const DeviceMemory<T>& input_h_data, const CudnnRnnStateTensorDescriptor& input_c_desc, const DeviceMemory<T>& input_c_data, const DeviceMemory<T>& params, const CudnnRnnSequenceTensorDescriptor& output_desc, const DeviceMemory<T>& output_data, const CudnnRnnStateTensorDescriptor& output_h_desc, const DeviceMemory<T>& output_h_data, const CudnnRnnStateTensorDescriptor& output_c_desc, const DeviceMemory<T>& output_c_data, RnnModelDims* model_dims) argument 1533 DoRnnForwardImpl( Stream* stream, const CudnnRnnDescriptor& rnn_desc, const CudnnRnnSequenceTensorDescriptor& input_desc, const DeviceMemory<T>& input_data, const CudnnRnnStateTensorDescriptor& input_h_desc, const DeviceMemory<T>& input_h_data, const CudnnRnnStateTensorDescriptor& input_c_desc, const DeviceMemory<T>& input_c_data, const DeviceMemory<T>& params, const CudnnRnnSequenceTensorDescriptor& output_desc, DeviceMemory<T>* output_data, const CudnnRnnStateTensorDescriptor& output_h_desc, DeviceMemory<T>* output_h_data, const CudnnRnnStateTensorDescriptor& output_c_desc, DeviceMemory<T>* output_c_data, bool is_training, ScratchAllocator* reserve_space_allocator, ScratchAllocator* workspace_allocator) argument 1650 DoRnnBackwardImpl( Stream* stream, const CudnnRnnDescriptor& rnn_desc, const CudnnRnnSequenceTensorDescriptor& input_desc, const DeviceMemory<T>& input_data, const CudnnRnnStateTensorDescriptor& input_h_desc, const DeviceMemory<T>& input_h_data, const CudnnRnnStateTensorDescriptor& input_c_desc, const DeviceMemory<T>& input_c_data, const DeviceMemory<T>& params, const CudnnRnnSequenceTensorDescriptor& output_desc, const DeviceMemory<T>& output_data, const CudnnRnnStateTensorDescriptor& output_h_desc, const DeviceMemory<T>& output_h_data, const CudnnRnnStateTensorDescriptor& output_c_desc, const DeviceMemory<T>& output_c_data, const DeviceMemory<T>& output_backprop_data, const DeviceMemory<T>& output_h_backprop_data, const DeviceMemory<T>& output_c_backprop_data, DeviceMemory<T>* input_backprop_data, DeviceMemory<T>* input_h_backprop_data, DeviceMemory<T>* input_c_backprop_data, DeviceMemory<T>* params_backprop_data, DeviceMemory<uint8>* reserve_space_data, ScratchAllocator* workspace_allocator) argument 1828 DoRnnForward( Stream* stream, const dnn::RnnDescriptor& rnn_desc, const dnn::RnnSequenceTensorDescriptor& input_desc, const DeviceMemory<Eigen::half>& input_data, const dnn::RnnStateTensorDescriptor& input_h_desc, const DeviceMemory<Eigen::half>& input_h_data, const dnn::RnnStateTensorDescriptor& input_c_desc, const DeviceMemory<Eigen::half>& input_c_data, const DeviceMemory<Eigen::half>& params, const dnn::RnnSequenceTensorDescriptor& output_desc, DeviceMemory<Eigen::half>* output_data, const dnn::RnnStateTensorDescriptor& output_h_desc, DeviceMemory<Eigen::half>* output_h_data, const dnn::RnnStateTensorDescriptor& output_c_desc, DeviceMemory<Eigen::half>* output_c_data, bool is_training, ScratchAllocator* reserve_space_allocator, ScratchAllocator* workspace_allocator) argument 1871 DoRnnForward( Stream* stream, const dnn::RnnDescriptor& rnn_desc, const dnn::RnnSequenceTensorDescriptor& input_desc, const DeviceMemory<float>& input_data, const dnn::RnnStateTensorDescriptor& input_h_desc, const DeviceMemory<float>& input_h_data, const dnn::RnnStateTensorDescriptor& input_c_desc, const DeviceMemory<float>& input_c_data, const DeviceMemory<float>& params, const dnn::RnnSequenceTensorDescriptor& output_desc, DeviceMemory<float>* output_data, const dnn::RnnStateTensorDescriptor& output_h_desc, DeviceMemory<float>* output_h_data, const dnn::RnnStateTensorDescriptor& output_c_desc, DeviceMemory<float>* output_c_data, bool is_training, ScratchAllocator* reserve_space_allocator, ScratchAllocator* workspace_allocator) argument 1913 DoRnnForward( Stream* stream, const dnn::RnnDescriptor& rnn_desc, const dnn::RnnSequenceTensorDescriptor& input_desc, const DeviceMemory<double>& input_data, const dnn::RnnStateTensorDescriptor& input_h_desc, const DeviceMemory<double>& input_h_data, const dnn::RnnStateTensorDescriptor& input_c_desc, const DeviceMemory<double>& input_c_data, const DeviceMemory<double>& params, const dnn::RnnSequenceTensorDescriptor& output_desc, DeviceMemory<double>* output_data, const dnn::RnnStateTensorDescriptor& output_h_desc, DeviceMemory<double>* output_h_data, const dnn::RnnStateTensorDescriptor& output_c_desc, DeviceMemory<double>* output_c_data, bool is_training, ScratchAllocator* reserve_space_allocator, ScratchAllocator* workspace_allocator) argument 1956 DoRnnBackward( Stream* stream, const dnn::RnnDescriptor& rnn_desc, const dnn::RnnSequenceTensorDescriptor& input_desc, const DeviceMemory<Eigen::half>& input_data, const dnn::RnnStateTensorDescriptor& input_h_desc, const DeviceMemory<Eigen::half>& input_h_data, const dnn::RnnStateTensorDescriptor& input_c_desc, const DeviceMemory<Eigen::half>& input_c_data, const DeviceMemory<Eigen::half>& params, const dnn::RnnSequenceTensorDescriptor& output_desc, const DeviceMemory<Eigen::half>& output_data, const dnn::RnnStateTensorDescriptor& output_h_desc, const DeviceMemory<Eigen::half>& output_h_data, const dnn::RnnStateTensorDescriptor& output_c_desc, const DeviceMemory<Eigen::half>& output_c_data, const DeviceMemory<Eigen::half>& output_backprop_data, const DeviceMemory<Eigen::half>& output_h_backprop_data, const DeviceMemory<Eigen::half>& output_c_backprop_data, DeviceMemory<Eigen::half>* input_backprop_data, DeviceMemory<Eigen::half>* input_h_backprop_data, DeviceMemory<Eigen::half>* input_c_backprop_data, DeviceMemory<Eigen::half>* params_backprop_data, DeviceMemory<uint8>* reserve_space_data, ScratchAllocator* workspace_allocator) argument 2009 DoRnnBackward( Stream* stream, const dnn::RnnDescriptor& rnn_desc, const dnn::RnnSequenceTensorDescriptor& input_desc, const DeviceMemory<float>& input_data, const dnn::RnnStateTensorDescriptor& input_h_desc, const DeviceMemory<float>& input_h_data, const dnn::RnnStateTensorDescriptor& input_c_desc, const DeviceMemory<float>& input_c_data, const DeviceMemory<float>& params, const dnn::RnnSequenceTensorDescriptor& output_desc, const DeviceMemory<float>& output_data, const dnn::RnnStateTensorDescriptor& output_h_desc, const DeviceMemory<float>& output_h_data, const dnn::RnnStateTensorDescriptor& output_c_desc, const DeviceMemory<float>& output_c_data, const DeviceMemory<float>& output_backprop_data, const DeviceMemory<float>& output_h_backprop_data, const DeviceMemory<float>& output_c_backprop_data, DeviceMemory<float>* input_backprop_data, DeviceMemory<float>* input_h_backprop_data, DeviceMemory<float>* input_c_backprop_data, DeviceMemory<float>* params_backprop_data, DeviceMemory<uint8>* reserve_space_data, ScratchAllocator* workspace_allocator) argument 2061 DoRnnBackward( Stream* stream, const dnn::RnnDescriptor& rnn_desc, const dnn::RnnSequenceTensorDescriptor& input_desc, const DeviceMemory<double>& input_data, const dnn::RnnStateTensorDescriptor& input_h_desc, const DeviceMemory<double>& input_h_data, const dnn::RnnStateTensorDescriptor& input_c_desc, const DeviceMemory<double>& input_c_data, const DeviceMemory<double>& params, const dnn::RnnSequenceTensorDescriptor& output_desc, const DeviceMemory<double>& output_data, const dnn::RnnStateTensorDescriptor& output_h_desc, const DeviceMemory<double>& output_h_data, const dnn::RnnStateTensorDescriptor& output_c_desc, const DeviceMemory<double>& output_c_data, const DeviceMemory<double>& output_backprop_data, const DeviceMemory<double>& output_h_backprop_data, const DeviceMemory<double>& output_c_backprop_data, DeviceMemory<double>* input_backprop_data, DeviceMemory<double>* input_h_backprop_data, DeviceMemory<double>* input_c_backprop_data, DeviceMemory<double>* params_backprop_data, DeviceMemory<uint8>* reserve_space_data, ScratchAllocator* workspace_allocator) argument 2302 DoConvolveImpl( Stream* stream, const BatchDescriptor& batch_descriptor, const DeviceMemory<T>& input_data, const FilterDescriptor& filter_descriptor, const DeviceMemory<T>& filter_data, const ConvolutionDescriptor& convolution_descriptor, const BatchDescriptor& output_descriptor, DeviceMemory<T>* output_data, ScratchAllocator* scratch_allocator, const dnn::AlgorithmConfig& algorithm_config, dnn::ProfileResult* output_profile_result) argument 2924 DoConvolve( Stream* stream, const BatchDescriptor& batch_descriptor, const DeviceMemory<float>& input_data, const FilterDescriptor& filter_descriptor, const DeviceMemory<float>& filter_data, const ConvolutionDescriptor& convolution_descriptor, const BatchDescriptor& output_descriptor, DeviceMemory<float>* output_data, ScratchAllocator* scratch_allocator, const dnn::AlgorithmConfig& algorithm_config, dnn::ProfileResult* output_profile_result) argument 2940 DoConvolve( Stream* stream, const BatchDescriptor& batch_descriptor, const DeviceMemory<double>& input_data, const FilterDescriptor& filter_descriptor, const DeviceMemory<double>& filter_data, const ConvolutionDescriptor& convolution_descriptor, const BatchDescriptor& output_descriptor, DeviceMemory<double>* output_data) argument 2952 DoConvolve( Stream* stream, const BatchDescriptor& batch_descriptor, const DeviceMemory<Eigen::half>& input_data, const FilterDescriptor& filter_descriptor, const DeviceMemory<Eigen::half>& filter_data, const ConvolutionDescriptor& convolution_descriptor, const BatchDescriptor& output_descriptor, DeviceMemory<Eigen::half>* output_data, ScratchAllocator* scratch_allocator, const dnn::AlgorithmConfig& algorithm_config, dnn::ProfileResult* output_profile_result) argument 3107 DoTransformTensor(Stream* stream, const dnn::BatchDescriptor& input_desc, dnn::DataType input_type, const DeviceMemoryBase& input_data, const dnn::BatchDescriptor& output_desc, dnn::DataType output_type, float scale, DeviceMemoryBase* output_data) argument 3384 DoConvolveBackwardFilterImpl( Stream* stream, const dnn::BatchDescriptor& input_descriptor, const DeviceMemory<T>& input_data, const dnn::BatchDescriptor& output_descriptor_in, DeviceMemory<T> backward_output_data, const dnn::ConvolutionDescriptor& convolution_descriptor, const dnn::FilterDescriptor& filter_descriptor, DeviceMemory<T>* backward_filter_data, ScratchAllocator* scratch_allocator, const dnn::AlgorithmConfig& algorithm_config, dnn::ProfileResult* output_profile_result) argument 3595 DoConvolveBackwardFilter( Stream* stream, const dnn::BatchDescriptor& input_descriptor, const DeviceMemory<float>& input_data, const dnn::BatchDescriptor& output_descriptor_in, DeviceMemory<float> backward_output_data, const dnn::ConvolutionDescriptor& convolution_descriptor, const dnn::FilterDescriptor& filter_descriptor, DeviceMemory<float>* backward_filter_data, ScratchAllocator* scratch_allocator, const dnn::AlgorithmConfig& algorithm_config, dnn::ProfileResult* output_profile_result) argument 3613 DoConvolveBackwardFilter( Stream* stream, const dnn::BatchDescriptor& input_descriptor, const DeviceMemory<Eigen::half>& input_data, const dnn::BatchDescriptor& output_descriptor_in, DeviceMemory<Eigen::half> backward_output_data, const dnn::ConvolutionDescriptor& convolution_descriptor, const dnn::FilterDescriptor& filter_descriptor, DeviceMemory<Eigen::half>* backward_filter_data, ScratchAllocator* scratch_allocator, const dnn::AlgorithmConfig& algorithm_config, dnn::ProfileResult* output_profile_result) argument 3632 DoConvolveBackwardBiasImpl( Stream* stream, const dnn::BatchDescriptor& input_descriptor, const DeviceMemory<T>& input_data, const dnn::BatchDescriptor& bias_descriptor, DeviceMemory<T>* backward_bias_data) argument 3665 DoConvolveBackwardBias( Stream* stream, const BatchDescriptor& input_descriptor, const DeviceMemory<double>& input_data, const BatchDescriptor& bias_descriptor, DeviceMemory<double>* backward_bias_data) argument 3674 DoConvolveBackwardBias( Stream* stream, const BatchDescriptor& input_descriptor, const DeviceMemory<float>& input_data, const BatchDescriptor& bias_descriptor, DeviceMemory<float>* backward_bias_data) argument 3683 DoConvolveBackwardBias( Stream* stream, const BatchDescriptor& input_descriptor, const DeviceMemory<Eigen::half>& input_data, const BatchDescriptor& bias_descriptor, DeviceMemory<Eigen::half>* backward_bias_data) argument 3692 DoMatMul(Stream* stream, const DeviceMemory<float>& input_data, const DeviceMemory<float>& weights, const dnn::BatchDescriptor& input_dimensions, const dnn::BatchDescriptor& output_dimensions, DeviceMemory<float>* output_data) argument 3827 DoBiasAdd(Stream* stream, const DeviceMemory<float>& input_data, const DeviceMemory<float>& biases, const dnn::BatchDescriptor& dimensions, DeviceMemory<float>* output_data) argument 3885 DoActivate(Stream* stream, dnn::ActivationMode activation_mode, const dnn::BatchDescriptor& dimensions, const DeviceMemory<float>& input_data, DeviceMemory<float>* output_data, uint64 options) argument 3954 DoPoolForward( Stream* stream, const dnn::PoolingDescriptor& pooling_dimensions, const dnn::BatchDescriptor& input_dimensions, const DeviceMemory<double>& input_data, const dnn::BatchDescriptor& output_dimensions, DeviceMemory<double>* output_data) argument 3989 DoPoolForward( Stream* stream, const dnn::PoolingDescriptor& pooling_dimensions, const dnn::BatchDescriptor& input_dimensions, const DeviceMemory<float>& input_data, const dnn::BatchDescriptor& output_dimensions, DeviceMemory<float>* output_data) argument 4024 DoPoolForward( Stream* stream, const dnn::PoolingDescriptor& pooling_dimensions, const dnn::BatchDescriptor& input_dimensions, const DeviceMemory<Eigen::half>& input_data, const dnn::BatchDescriptor& output_dimensions, DeviceMemory<Eigen::half>* output_data) argument 4058 DoPoolBackward( Stream* stream, const dnn::PoolingDescriptor& pooling_dimensions, const dnn::BatchDescriptor& input_dimensions, const DeviceMemory<double>& input_data, const dnn::BatchDescriptor& output_dimensions, const DeviceMemory<double>& output_data, const DeviceMemory<double>& input_diff_data, DeviceMemory<double>* output_diff_data) argument 4096 DoPoolBackward( Stream* stream, const dnn::PoolingDescriptor& pooling_dimensions, const dnn::BatchDescriptor& input_dimensions, const DeviceMemory<float>& input_data, const dnn::BatchDescriptor& output_dimensions, const DeviceMemory<float>& output_data, const DeviceMemory<float>& input_diff_data, DeviceMemory<float>* output_diff_data) argument 4134 DoPoolBackward( Stream* stream, const dnn::PoolingDescriptor& pooling_dimensions, const dnn::BatchDescriptor& input_dimensions, const DeviceMemory<Eigen::half>& input_data, const dnn::BatchDescriptor& output_dimensions, const DeviceMemory<Eigen::half>& output_data, const DeviceMemory<Eigen::half>& input_diff_data, DeviceMemory<Eigen::half>* output_diff_data) argument 4171 DoNormalize( Stream* stream, const dnn::NormalizeDescriptor& normalize_descriptor, const DeviceMemory<float>& input_data, DeviceMemory<float>* output_data) argument 4178 DoNormalizeWithDimensions( Stream* stream, const dnn::NormalizeDescriptor& normalize_descriptor, const dnn::BatchDescriptor& dimensions, const DeviceMemory<float>& input_data, DeviceMemory<float>* output_data) argument 4263 DoDepthConcatenate( Stream* stream, port::ArraySlice<dnn::BatchDescriptor> input_dimensions, port::ArraySlice<const DeviceMemory<float>*> input_data, DeviceMemory<float>* output_data) argument 4320 DoElementwiseOperate( Stream* stream, dnn::ElementwiseOperation operation, port::ArraySlice<dnn::BatchDescriptor> input_dimensions, port::ArraySlice<const DeviceMemory<float>*> input_data, const dnn::BatchDescriptor& output_dimensions, DeviceMemory<float>* output_data) argument 4330 DoXYPad(Stream* stream, const dnn::BatchDescriptor& dimensions, const DeviceMemory<float>& input_data, int64 left_pad, int64 right_pad, int64 top_pad, int64 bottom_pad, DeviceMemory<float>* output_data) argument 4339 DoXYSlice(Stream* stream, const dnn::BatchDescriptor& dimensions, const DeviceMemory<float>& input_data, int64 left_trim, int64 right_trim, int64 top_trim, int64 bottom_trim, DeviceMemory<float>* output_data) argument [all...] |