src/backends/NeonWorkloads/NeonConvolution2dBaseWorkload.cpp

//
// Copyright © 2017 Arm Ltd. All rights reserved.
// SPDX-License-Identifier: MIT
//

#include <backends/CpuTensorHandle.hpp>
#include <backends/aclCommon/ArmComputeTensorUtils.hpp>
#include <backends/NeonLayerSupport.hpp>

#include "NeonConvolution2dBaseWorkload.hpp"

#include <armnn/Types.hpp>
#include <Half.hpp>

namespace armnn
{

using namespace armcomputetensorutils;

arm_compute::Status NeonConvolution2dWorkloadValidate(const TensorInfo& input,
    const TensorInfo& output,
    const Convolution2dDescriptor& descriptor,
    const TensorInfo& weights,
    const boost::optional<TensorInfo>& biases)
{
    const arm_compute::TensorInfo aclInputInfo = BuildArmComputeTensorInfo(input, descriptor.m_DataLayout);
    const arm_compute::TensorInfo aclOutputInfo = BuildArmComputeTensorInfo(output, descriptor.m_DataLayout);
    const arm_compute::TensorInfo aclWeightsInfo = BuildArmComputeTensorInfo(weights, descriptor.m_DataLayout);

    arm_compute::TensorInfo aclBiasesInfo;
    arm_compute::TensorInfo *optionalAclBiasesInfo = nullptr;

    if (descriptor.m_BiasEnabled)
    {
        BOOST_ASSERT(biases.is_initialized());

        aclBiasesInfo = BuildArmComputeTensorInfo(biases.get(), descriptor.m_DataLayout);
        optionalAclBiasesInfo = &aclBiasesInfo;
    }

    arm_compute::PadStrideInfo layerInfo = BuildArmComputePadStrideInfo(descriptor);

    return arm_compute::NEConvolutionLayer::validate(&aclInputInfo,
                                                     &aclWeightsInfo,
                                                     optionalAclBiasesInfo,
                                                     &aclOutputInfo,
                                                     layerInfo);
}

template<armnn::DataType... dataTypes>
NeonConvolution2dBaseWorkload<dataTypes...>::NeonConvolution2dBaseWorkload(
    const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info,
    std::shared_ptr<arm_compute::MemoryManagerOnDemand>& memoryManager)
    : TypedWorkload<Convolution2dQueueDescriptor, dataTypes...>(descriptor, info)
{
    using arm_compute::NEDirectConvolutionLayer;

    ValidateData();

    // todo: check tensor shapes match.

    arm_compute::ITensor& input = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Inputs[0])->GetTensor();
    arm_compute::ITensor& output = boost::polymorphic_downcast<INeonTensorHandle*>(m_Data.m_Outputs[0])->GetTensor();

    m_KernelTensor = std::make_unique<arm_compute::Tensor>();
    BuildArmComputeTensor(*m_KernelTensor, m_Data.m_Weight->GetTensorInfo(), descriptor.m_DataLayout);

    if (m_Data.m_Parameters.m_BiasEnabled)
    {
        m_BiasTensor = std::make_unique<arm_compute::Tensor>();
        BuildArmComputeTensor(*m_BiasTensor, m_Data.m_Bias->GetTensorInfo(), descriptor.m_DataLayout);
    }

    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);

    const bool preferDirectConvolution =
        IsNeonDirectConvolutionPreferred(m_Data.m_Weight->GetTensorInfo(),
                                         m_Data.m_Parameters);

    if (preferDirectConvolution)
    {
        auto directConvolutionLayer = std::make_unique<arm_compute::NEDirectConvolutionLayer>(memoryManager);
        directConvolutionLayer->configure(&input,
                                          m_KernelTensor.get(),
                                          m_BiasTensor.get(),
                                          &output,
                                          padStrideInfo);
        m_ConvolutionLayer.reset(directConvolutionLayer.release());
    }
    else
    {
        auto convolutionLayer = std::make_unique<arm_compute::NEConvolutionLayer>(memoryManager);
        convolutionLayer->configure(&input,
                                    m_KernelTensor.get(),
                                    m_BiasTensor.get(),
                                    &output,
                                    padStrideInfo);
        m_ConvolutionLayer.reset(convolutionLayer.release());
    }
    BOOST_ASSERT(m_ConvolutionLayer);

    armnn::DataType dataType = m_Data.m_Weight->GetTensorInfo().GetDataType();

    switch (dataType)
    {
        case DataType::Float16:
        {
            InitialiseArmComputeTensorData(*m_KernelTensor, m_Data.m_Weight->template GetConstTensor<Half>());
            break;
        }
        case DataType::Float32:
        {
            InitialiseArmComputeTensorData(*m_KernelTensor, m_Data.m_Weight->template GetConstTensor<float>());
            break;
        }
        case DataType::QuantisedAsymm8:
        {
            InitialiseArmComputeTensorData(*m_KernelTensor, m_Data.m_Weight->template GetConstTensor<uint8_t>());
            break;
        }
        default:
        {
            BOOST_ASSERT_MSG(false, "Unknown DataType.");
        }
    }
}

template<armnn::DataType... dataTypes>
void NeonConvolution2dBaseWorkload<dataTypes...>::FreeUnusedTensors()
{
    FreeTensorIfUnused(m_KernelTensor);
    FreeTensorIfUnused(m_BiasTensor);
}

// Generates known implementations for linker.
template class NeonConvolution2dBaseWorkload<armnn::DataType::Float16, armnn::DataType::Float32>;
template class NeonConvolution2dBaseWorkload<armnn::DataType::QuantisedAsymm8>;

} //namespace armnn