diff options
author | telsoa01 <telmo.soares@arm.com> | 2018-03-09 14:13:49 +0000 |
---|---|---|
committer | telsoa01 <telmo.soares@arm.com> | 2018-03-09 14:13:49 +0000 |
commit | 4fcda0101ec3d110c1d6d7bee5c83416b645528a (patch) | |
tree | c9a70aeb2887006160c1b3d265c27efadb7bdbae /src/armnn/backends/ClWorkloads/ClFullyConnectedFloat32Workload.cpp | |
download | armnn-4fcda0101ec3d110c1d6d7bee5c83416b645528a.tar.gz |
Release 18.02
Change-Id: Id3c11dc5ee94ef664374a988fcc6901e9a232fa6
Diffstat (limited to 'src/armnn/backends/ClWorkloads/ClFullyConnectedFloat32Workload.cpp')
-rw-r--r-- | src/armnn/backends/ClWorkloads/ClFullyConnectedFloat32Workload.cpp | 52 |
1 files changed, 52 insertions, 0 deletions
diff --git a/src/armnn/backends/ClWorkloads/ClFullyConnectedFloat32Workload.cpp b/src/armnn/backends/ClWorkloads/ClFullyConnectedFloat32Workload.cpp new file mode 100644 index 0000000000..96596b9d9c --- /dev/null +++ b/src/armnn/backends/ClWorkloads/ClFullyConnectedFloat32Workload.cpp @@ -0,0 +1,52 @@ +// +// Copyright © 2017 Arm Ltd. All rights reserved. +// See LICENSE file in the project root for full license information. +// + +#include "ClFullyConnectedFloat32Workload.hpp" +#include "backends/ClTensorHandle.hpp" +#include "backends/CpuTensorHandle.hpp" +#include "backends/ArmComputeTensorUtils.hpp" + +namespace armnn +{ +using namespace armcomputetensorutils; + +ClFullyConnectedFloat32Workload::ClFullyConnectedFloat32Workload(const FullyConnectedQueueDescriptor& descriptor, + const WorkloadInfo& info) + : Float32Workload<FullyConnectedQueueDescriptor>(descriptor, info) +{ + + BuildArmComputeTensor(m_WeightsTensor, m_Data.m_Weight->GetTensorInfo()); + + arm_compute::CLTensor* optionalBiasTensor = nullptr; + if (m_Data.m_Parameters.m_BiasEnabled) + { + BuildArmComputeTensor(m_BiasesTensor, m_Data.m_Bias->GetTensorInfo()); + optionalBiasTensor = &m_BiasesTensor; + } + + m_Data.ValidateInputsOutputs("ClFullyConnectedFloat32Workload", 1, 1); + + arm_compute::ICLTensor& input = static_cast<IClTensorHandle*>(m_Data.m_Inputs[0])->GetTensor(); + arm_compute::ICLTensor& output = static_cast<IClTensorHandle*>(m_Data.m_Outputs[0])->GetTensor(); + // Construct + m_FullyConnected.configure( + &input, &m_WeightsTensor, optionalBiasTensor, &output, m_Data.m_Parameters.m_TransposeWeightMatrix); + + // Allocate + InitialiseArmComputeClTensorData(m_WeightsTensor, m_Data.m_Weight->GetConstTensor<float>()); + + if (optionalBiasTensor) + { + InitialiseArmComputeClTensorData(*optionalBiasTensor, m_Data.m_Bias->GetConstTensor<float>()); + } +} + +void ClFullyConnectedFloat32Workload::Execute() const +{ + ARMNN_SCOPED_PROFILING_EVENT(Compute::GpuAcc, "ClFullyConnectedFloat32Workload_Execute"); + m_FullyConnected.run(); +} + +} //namespace armnn
\ No newline at end of file |