From 9e53a35b66b1ec7ceee7c712380a13596175b83b Mon Sep 17 00:00:00 2001 From: arovir01 Date: Fri, 31 Aug 2018 15:26:35 +0100 Subject: IVGCVSW-1784: Rename float32 workloads for ACL Change-Id: I98bdfe9cb12c663d1d5cfa456e2cc967d70ab22b --- .../ClWorkloads/ClConvolution2dFloatWorkload.cpp | 79 ++++++++++++++++++++++ 1 file changed, 79 insertions(+) create mode 100644 src/armnn/backends/ClWorkloads/ClConvolution2dFloatWorkload.cpp (limited to 'src/armnn/backends/ClWorkloads/ClConvolution2dFloatWorkload.cpp') diff --git a/src/armnn/backends/ClWorkloads/ClConvolution2dFloatWorkload.cpp b/src/armnn/backends/ClWorkloads/ClConvolution2dFloatWorkload.cpp new file mode 100644 index 0000000000..375ace277a --- /dev/null +++ b/src/armnn/backends/ClWorkloads/ClConvolution2dFloatWorkload.cpp @@ -0,0 +1,79 @@ +// +// Copyright © 2017 Arm Ltd. All rights reserved. +// See LICENSE file in the project root for full license information. +// + +#include "ClConvolution2dFloatWorkload.hpp" +#include "backends/ClTensorHandle.hpp" +#include "backends/CpuTensorHandle.hpp" +#include "backends/ArmComputeTensorUtils.hpp" +#include "backends/ClLayerSupport.hpp" + +namespace armnn +{ +using namespace armcomputetensorutils; + +ClConvolution2dFloatWorkload::ClConvolution2dFloatWorkload(const Convolution2dQueueDescriptor& descriptor, + const WorkloadInfo& info, std::shared_ptr& memoryManager) + : FloatWorkload(descriptor, info) + , m_ConvolutionLayer(memoryManager) +{ + + // todo: check tensor shapes match. + const TensorInfo& weightInfo = m_Data.m_Weight->GetTensorInfo(); + + m_KernelTensor = std::make_unique(); + BuildArmComputeTensor(*m_KernelTensor, weightInfo); + + arm_compute::PadStrideInfo padStrideInfo(m_Data.m_Parameters.m_StrideX, + m_Data.m_Parameters.m_StrideY, + m_Data.m_Parameters.m_PadLeft, + m_Data.m_Parameters.m_PadRight, + m_Data.m_Parameters.m_PadTop, + m_Data.m_Parameters.m_PadBottom, + arm_compute::DimensionRoundingType::FLOOR); + + if (m_Data.m_Parameters.m_BiasEnabled) + { + m_BiasTensor = std::make_unique(); + BuildArmComputeTensor(*m_BiasTensor, m_Data.m_Bias->GetTensorInfo()); + } + + m_Data.ValidateInputsOutputs("ClConvolution2dFloat32Workload", 1, 1); + + arm_compute::ICLTensor& input = static_cast(m_Data.m_Inputs[0])->GetTensor(); + arm_compute::ICLTensor& output = static_cast(m_Data.m_Outputs[0])->GetTensor(); + + m_ConvolutionLayer.configure(&input, + m_KernelTensor.get(), + m_BiasTensor.get(), + &output, + padStrideInfo); + + InitializeArmComputeClTensorDataForFloatTypes(*m_KernelTensor, m_Data.m_Weight); + + if (m_BiasTensor) + { + InitializeArmComputeClTensorDataForFloatTypes(*m_BiasTensor, m_Data.m_Bias); + } + + // Force Compute Library to perform the necessary copying and reshaping, after which + // delete all the input tensors that will no longer be needed + m_ConvolutionLayer.prepare(); + FreeUnusedTensors(); +} + +void ClConvolution2dFloatWorkload::Execute() const +{ + ARMNN_SCOPED_PROFILING_EVENT_CL("ClConvolution2dFloat32Workload_Execute"); + + m_ConvolutionLayer.run(); +} + +void ClConvolution2dFloatWorkload::FreeUnusedTensors() +{ + FreeTensorIfUnused(m_KernelTensor); + FreeTensorIfUnused(m_BiasTensor); +} + +} //namespace armnn -- cgit v1.2.1