From 974b65fea888cc000a5164d5b56d9ed016391151 Mon Sep 17 00:00:00 2001 From: Nattapat Chaimanowong Date: Mon, 15 Oct 2018 15:07:34 +0100 Subject: IVGCVSW-1951 Remove type templating from NeonConvolution2dWorkload Change-Id: Id3d8137d60b14e93863ad5b6db752a7f6ef9b7fb --- src/backends/neon/NeonLayerSupport.cpp | 2 +- src/backends/neon/NeonWorkloadFactory.cpp | 4 +- src/backends/neon/backend.mk | 4 +- src/backends/neon/test/NeonCreateWorkloadTests.cpp | 12 +- src/backends/neon/workloads/CMakeLists.txt | 8 +- .../workloads/NeonConvolution2dBaseWorkload.cpp | 127 ------------------- .../workloads/NeonConvolution2dBaseWorkload.hpp | 47 -------- .../workloads/NeonConvolution2dFloatWorkload.cpp | 39 ------ .../workloads/NeonConvolution2dFloatWorkload.hpp | 29 ----- .../workloads/NeonConvolution2dUint8Workload.cpp | 35 ------ .../workloads/NeonConvolution2dUint8Workload.hpp | 29 ----- .../neon/workloads/NeonConvolution2dWorkload.cpp | 134 +++++++++++++++++++++ .../neon/workloads/NeonConvolution2dWorkload.hpp | 47 ++++++++ src/backends/neon/workloads/NeonWorkloads.hpp | 4 +- 14 files changed, 194 insertions(+), 327 deletions(-) delete mode 100644 src/backends/neon/workloads/NeonConvolution2dBaseWorkload.cpp delete mode 100644 src/backends/neon/workloads/NeonConvolution2dBaseWorkload.hpp delete mode 100644 src/backends/neon/workloads/NeonConvolution2dFloatWorkload.cpp delete mode 100644 src/backends/neon/workloads/NeonConvolution2dFloatWorkload.hpp delete mode 100644 src/backends/neon/workloads/NeonConvolution2dUint8Workload.cpp delete mode 100644 src/backends/neon/workloads/NeonConvolution2dUint8Workload.hpp create mode 100644 src/backends/neon/workloads/NeonConvolution2dWorkload.cpp create mode 100644 src/backends/neon/workloads/NeonConvolution2dWorkload.hpp diff --git a/src/backends/neon/NeonLayerSupport.cpp b/src/backends/neon/NeonLayerSupport.cpp index 99e227897f..e28c4f5add 100644 --- a/src/backends/neon/NeonLayerSupport.cpp +++ b/src/backends/neon/NeonLayerSupport.cpp @@ -17,7 +17,7 @@ #include "workloads/NeonAdditionFloatWorkload.hpp" #include "workloads/NeonActivationWorkload.hpp" #include "workloads/NeonBatchNormalizationFloatWorkload.hpp" -#include "workloads/NeonConvolution2dBaseWorkload.hpp" +#include "workloads/NeonConvolution2dWorkload.hpp" #include "workloads/NeonDepthwiseConvolutionBaseWorkload.hpp" #include "workloads/NeonL2NormalizationFloatWorkload.hpp" #include "workloads/NeonMultiplicationFloatWorkload.hpp" diff --git a/src/backends/neon/NeonWorkloadFactory.cpp b/src/backends/neon/NeonWorkloadFactory.cpp index a43722728d..f73f3aac1a 100644 --- a/src/backends/neon/NeonWorkloadFactory.cpp +++ b/src/backends/neon/NeonWorkloadFactory.cpp @@ -135,8 +135,8 @@ std::unique_ptr NeonWorkloadFactory::CreatePooling2d(const Poo std::unique_ptr NeonWorkloadFactory::CreateConvolution2d( const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info) const { - return MakeWorkload(descriptor, info, - m_MemoryManager.GetIntraLayerManager()); + return std::make_unique(descriptor, info, + m_MemoryManager.GetIntraLayerManager()); } std::unique_ptr NeonWorkloadFactory::CreateDepthwiseConvolution2d( diff --git a/src/backends/neon/backend.mk b/src/backends/neon/backend.mk index 6dfd951957..97105ec17e 100644 --- a/src/backends/neon/backend.mk +++ b/src/backends/neon/backend.mk @@ -19,9 +19,7 @@ BACKEND_SOURCES := \ workloads/NeonConstantWorkload.cpp \ workloads/NeonConvertFp16ToFp32Workload.cpp \ workloads/NeonConvertFp32ToFp16Workload.cpp \ - workloads/NeonConvolution2dBaseWorkload.cpp \ - workloads/NeonConvolution2dFloatWorkload.cpp \ - workloads/NeonConvolution2dUint8Workload.cpp \ + workloads/NeonConvolution2dWorkload.cpp \ workloads/NeonDepthwiseConvolutionBaseWorkload.cpp \ workloads/NeonDepthwiseConvolutionFloatWorkload.cpp \ workloads/NeonDepthwiseConvolutionUint8Workload.cpp \ diff --git a/src/backends/neon/test/NeonCreateWorkloadTests.cpp b/src/backends/neon/test/NeonCreateWorkloadTests.cpp index ec8fe803a1..63b3d41500 100644 --- a/src/backends/neon/test/NeonCreateWorkloadTests.cpp +++ b/src/backends/neon/test/NeonCreateWorkloadTests.cpp @@ -179,12 +179,12 @@ BOOST_AUTO_TEST_CASE(CreateBatchNormalizationFloatWorkload) NeonCreateBatchNormalizationWorkloadTest(); } -template +template static void NeonCreateConvolution2dWorkloadTest(DataLayout dataLayout = DataLayout::NCHW) { Graph graph; NeonWorkloadFactory factory; - auto workload = CreateConvolution2dWorkloadTest(factory, graph, dataLayout); TensorShape inputShape = (dataLayout == DataLayout::NCHW) ? TensorShape{2, 3, 8, 16} : TensorShape{2, 8, 16, 3}; @@ -201,23 +201,23 @@ static void NeonCreateConvolution2dWorkloadTest(DataLayout dataLayout = DataLayo #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC BOOST_AUTO_TEST_CASE(CreateConvolution2dFloat16NchwWorkload) { - NeonCreateConvolution2dWorkloadTest(); + NeonCreateConvolution2dWorkloadTest(); } BOOST_AUTO_TEST_CASE(CreateConvolution2dFloat16NhwcWorkload) { - NeonCreateConvolution2dWorkloadTest(DataLayout::NHWC); + NeonCreateConvolution2dWorkloadTest(DataLayout::NHWC); } #endif BOOST_AUTO_TEST_CASE(CreateConvolution2dFloatNchwWorkload) { - NeonCreateConvolution2dWorkloadTest(); + NeonCreateConvolution2dWorkloadTest(); } BOOST_AUTO_TEST_CASE(CreateConvolution2dFloatNhwcWorkload) { - NeonCreateConvolution2dWorkloadTest(DataLayout::NHWC); + NeonCreateConvolution2dWorkloadTest(DataLayout::NHWC); } template diff --git a/src/backends/neon/workloads/CMakeLists.txt b/src/backends/neon/workloads/CMakeLists.txt index d847df70fd..9c57dffd23 100644 --- a/src/backends/neon/workloads/CMakeLists.txt +++ b/src/backends/neon/workloads/CMakeLists.txt @@ -16,12 +16,8 @@ list(APPEND armnnNeonBackendWorkloads_sources NeonConvertFp16ToFp32Workload.hpp NeonConvertFp32ToFp16Workload.cpp NeonConvertFp32ToFp16Workload.hpp - NeonConvolution2dBaseWorkload.cpp - NeonConvolution2dBaseWorkload.hpp - NeonConvolution2dFloatWorkload.cpp - NeonConvolution2dFloatWorkload.hpp - NeonConvolution2dUint8Workload.cpp - NeonConvolution2dUint8Workload.hpp + NeonConvolution2dWorkload.cpp + NeonConvolution2dWorkload.hpp NeonDepthwiseConvolutionBaseWorkload.cpp NeonDepthwiseConvolutionBaseWorkload.hpp NeonDepthwiseConvolutionFloatWorkload.cpp diff --git a/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.cpp b/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.cpp deleted file mode 100644 index b11d10fd2f..0000000000 --- a/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.cpp +++ /dev/null @@ -1,127 +0,0 @@ -// -// Copyright © 2017 Arm Ltd. All rights reserved. -// SPDX-License-Identifier: MIT -// - -#include -#include -#include - -#include "NeonConvolution2dBaseWorkload.hpp" - -#include -#include - -namespace armnn -{ - -using namespace armcomputetensorutils; - -arm_compute::Status NeonConvolution2dWorkloadValidate(const TensorInfo& input, - const TensorInfo& output, - const Convolution2dDescriptor& descriptor, - const TensorInfo& weights, - const Optional& 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.has_value()); - - aclBiasesInfo = BuildArmComputeTensorInfo(biases.value(), descriptor.m_DataLayout); - optionalAclBiasesInfo = &aclBiasesInfo; - } - - arm_compute::PadStrideInfo layerInfo = BuildArmComputePadStrideInfo(descriptor); - - return arm_compute::NEConvolutionLayer::validate(&aclInputInfo, - &aclWeightsInfo, - optionalAclBiasesInfo, - &aclOutputInfo, - layerInfo); -} - -template -NeonConvolution2dBaseWorkload::NeonConvolution2dBaseWorkload( - const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, - std::shared_ptr& memoryManager) - : TypedWorkload(descriptor, info) -{ - using arm_compute::NEDirectConvolutionLayer; - - ValidateData(); - - // todo: check tensor shapes match. - - arm_compute::ITensor& input = boost::polymorphic_downcast(m_Data.m_Inputs[0])->GetTensor(); - arm_compute::ITensor& output = boost::polymorphic_downcast(m_Data.m_Outputs[0])->GetTensor(); - - arm_compute::DataLayout aclDataLayout = ConvertDataLayout(m_Data.m_Parameters.m_DataLayout); - input.info()->set_data_layout(aclDataLayout); - output.info()->set_data_layout(aclDataLayout); - - m_KernelTensor = std::make_unique(); - BuildArmComputeTensor(*m_KernelTensor, m_Data.m_Weight->GetTensorInfo(), m_Data.m_Parameters.m_DataLayout); - - if (m_Data.m_Parameters.m_BiasEnabled) - { - m_BiasTensor = std::make_unique(); - BuildArmComputeTensor(*m_BiasTensor, m_Data.m_Bias->GetTensorInfo(), m_Data.m_Parameters.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(memoryManager); - directConvolutionLayer->configure(&input, - m_KernelTensor.get(), - m_BiasTensor.get(), - &output, - padStrideInfo); - m_ConvolutionLayer.reset(directConvolutionLayer.release()); - } - else - { - auto convolutionLayer = std::make_unique(memoryManager); - convolutionLayer->configure(&input, - m_KernelTensor.get(), - m_BiasTensor.get(), - &output, - padStrideInfo); - m_ConvolutionLayer.reset(convolutionLayer.release()); - } - BOOST_ASSERT(m_ConvolutionLayer); - - InitializeArmComputeTensorData(*m_KernelTensor, m_Data.m_Weight); - -} - -template -void NeonConvolution2dBaseWorkload::FreeUnusedTensors() -{ - FreeTensorIfUnused(m_KernelTensor); - FreeTensorIfUnused(m_BiasTensor); -} - -// Generates known implementations for linker. -template class NeonConvolution2dBaseWorkload; -template class NeonConvolution2dBaseWorkload; - -} //namespace armnn diff --git a/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.hpp b/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.hpp deleted file mode 100644 index 1cd30c70f9..0000000000 --- a/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.hpp +++ /dev/null @@ -1,47 +0,0 @@ -// -// Copyright © 2017 Arm Ltd. All rights reserved. -// SPDX-License-Identifier: MIT -// - -#pragma once - -#include -#include -#include -#include -#include - -#include - -#include - -namespace armnn -{ - -arm_compute::Status NeonConvolution2dWorkloadValidate(const TensorInfo& input, - const TensorInfo& output, - const Convolution2dDescriptor& descriptor, - const TensorInfo& weights, - const Optional& biases); - -template -class NeonConvolution2dBaseWorkload : public TypedWorkload -{ -public: - using TypedWorkload::m_Data; - - NeonConvolution2dBaseWorkload(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, - std::shared_ptr& memoryManager); - - virtual void ValidateData() const {}; - -protected: - std::unique_ptr m_ConvolutionLayer; - - std::unique_ptr m_KernelTensor; - std::unique_ptr m_BiasTensor; - - void FreeUnusedTensors(); -}; - -} //namespace armnn diff --git a/src/backends/neon/workloads/NeonConvolution2dFloatWorkload.cpp b/src/backends/neon/workloads/NeonConvolution2dFloatWorkload.cpp deleted file mode 100644 index 9969154421..0000000000 --- a/src/backends/neon/workloads/NeonConvolution2dFloatWorkload.cpp +++ /dev/null @@ -1,39 +0,0 @@ -// -// Copyright © 2017 Arm Ltd. All rights reserved. -// SPDX-License-Identifier: MIT -// - -#include "NeonConvolution2dFloatWorkload.hpp" -#include -#include -#include - -namespace armnn -{ -using namespace armcomputetensorutils; - -NeonConvolution2dFloatWorkload::NeonConvolution2dFloatWorkload(const Convolution2dQueueDescriptor& descriptor, - const WorkloadInfo& info, std::shared_ptr& memoryManager) - : NeonConvolution2dBaseWorkload(descriptor, info, memoryManager) -{ - if (m_Data.m_Parameters.m_BiasEnabled) - { - InitializeArmComputeTensorData(*m_BiasTensor, m_Data.m_Bias); - } - - m_ConvolutionLayer->prepare(); - FreeUnusedTensors(); -} - -void NeonConvolution2dFloatWorkload::Execute() const -{ - ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonConvolution2dFloatWorkload_Execute"); - m_ConvolutionLayer->run(); -} - -void NeonConvolution2dFloatWorkload::ValidateData() const -{ - m_Data.ValidateInputsOutputs("NeonConvolution2dFloatWorkload", 1, 1); -} - -} //namespace armnn diff --git a/src/backends/neon/workloads/NeonConvolution2dFloatWorkload.hpp b/src/backends/neon/workloads/NeonConvolution2dFloatWorkload.hpp deleted file mode 100644 index 14c77c8bd0..0000000000 --- a/src/backends/neon/workloads/NeonConvolution2dFloatWorkload.hpp +++ /dev/null @@ -1,29 +0,0 @@ -// -// Copyright © 2017 Arm Ltd. All rights reserved. -// SPDX-License-Identifier: MIT -// - -#pragma once - -#include "NeonConvolution2dBaseWorkload.hpp" -#include - -#include "arm_compute/runtime/MemoryManagerOnDemand.h" - -#include - -namespace armnn -{ - -class NeonConvolution2dFloatWorkload : public NeonConvolution2dBaseWorkload -{ -public: - NeonConvolution2dFloatWorkload(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, - std::shared_ptr& memoryManager); - - void Execute() const override; - void ValidateData() const override; -}; - -} //namespace armnn - diff --git a/src/backends/neon/workloads/NeonConvolution2dUint8Workload.cpp b/src/backends/neon/workloads/NeonConvolution2dUint8Workload.cpp deleted file mode 100644 index 8572cbfb08..0000000000 --- a/src/backends/neon/workloads/NeonConvolution2dUint8Workload.cpp +++ /dev/null @@ -1,35 +0,0 @@ -// -// Copyright © 2017 Arm Ltd. All rights reserved. -// SPDX-License-Identifier: MIT -// - -#include "NeonConvolution2dUint8Workload.hpp" - -namespace armnn -{ - -NeonConvolution2dUint8Workload::NeonConvolution2dUint8Workload(const Convolution2dQueueDescriptor& descriptor, - const WorkloadInfo& info, std::shared_ptr& memoryManager) - : NeonConvolution2dBaseWorkload(descriptor, info, memoryManager) -{ - if (m_Data.m_Parameters.m_BiasEnabled) - { - InitializeArmComputeTensorData(*m_BiasTensor, m_Data.m_Bias); - } - - m_ConvolutionLayer->prepare(); - FreeUnusedTensors(); -} - -void NeonConvolution2dUint8Workload::Execute() const -{ - ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonConvolution2dUint8Workload_Execute"); - m_ConvolutionLayer->run(); -} - -void NeonConvolution2dUint8Workload::ValidateData() const -{ - m_Data.ValidateInputsOutputs("NeonConvolution2dUint8Workload", 1, 1); -} - -} //namespace armnn diff --git a/src/backends/neon/workloads/NeonConvolution2dUint8Workload.hpp b/src/backends/neon/workloads/NeonConvolution2dUint8Workload.hpp deleted file mode 100644 index ef60fc3e84..0000000000 --- a/src/backends/neon/workloads/NeonConvolution2dUint8Workload.hpp +++ /dev/null @@ -1,29 +0,0 @@ -// -// Copyright © 2017 Arm Ltd. All rights reserved. -// SPDX-License-Identifier: MIT -// - -#pragma once - -#include "NeonConvolution2dBaseWorkload.hpp" - -#include "arm_compute/runtime/MemoryManagerOnDemand.h" - -#include - -namespace armnn -{ - -class NeonConvolution2dUint8Workload : public NeonConvolution2dBaseWorkload -{ -public: - NeonConvolution2dUint8Workload(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, - std::shared_ptr& memoryManager); - - virtual void ValidateData() const override; - virtual void Execute() const override; -private: -}; - -} //namespace armnnn - diff --git a/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp b/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp new file mode 100644 index 0000000000..c26cdea92b --- /dev/null +++ b/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp @@ -0,0 +1,134 @@ +// +// Copyright © 2017 Arm Ltd. All rights reserved. +// SPDX-License-Identifier: MIT +// + +#include "NeonConvolution2dWorkload.hpp" + +#include +#include +#include + +#include +#include + +namespace armnn +{ + +using namespace armcomputetensorutils; + +arm_compute::Status NeonConvolution2dWorkloadValidate(const TensorInfo& input, + const TensorInfo& output, + const Convolution2dDescriptor& descriptor, + const TensorInfo& weights, + const Optional& 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.has_value()); + + aclBiasesInfo = BuildArmComputeTensorInfo(biases.value(), descriptor.m_DataLayout); + optionalAclBiasesInfo = &aclBiasesInfo; + } + + arm_compute::PadStrideInfo layerInfo = BuildArmComputePadStrideInfo(descriptor); + + return arm_compute::NEConvolutionLayer::validate(&aclInputInfo, + &aclWeightsInfo, + optionalAclBiasesInfo, + &aclOutputInfo, + layerInfo); +} + +NeonConvolution2dWorkload::NeonConvolution2dWorkload( + const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, + std::shared_ptr& memoryManager) + : BaseWorkload(descriptor, info) +{ + using arm_compute::NEDirectConvolutionLayer; + + m_Data.ValidateInputsOutputs("NeonConvolution2dWorkload", 1, 1); + + // todo: check tensor shapes match. + + arm_compute::ITensor& input = boost::polymorphic_downcast(m_Data.m_Inputs[0])->GetTensor(); + arm_compute::ITensor& output = boost::polymorphic_downcast(m_Data.m_Outputs[0])->GetTensor(); + + arm_compute::DataLayout aclDataLayout = ConvertDataLayout(m_Data.m_Parameters.m_DataLayout); + input.info()->set_data_layout(aclDataLayout); + output.info()->set_data_layout(aclDataLayout); + + m_KernelTensor = std::make_unique(); + BuildArmComputeTensor(*m_KernelTensor, m_Data.m_Weight->GetTensorInfo(), m_Data.m_Parameters.m_DataLayout); + + if (m_Data.m_Parameters.m_BiasEnabled) + { + m_BiasTensor = std::make_unique(); + BuildArmComputeTensor(*m_BiasTensor, m_Data.m_Bias->GetTensorInfo(), m_Data.m_Parameters.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(memoryManager); + directConvolutionLayer->configure(&input, + m_KernelTensor.get(), + m_BiasTensor.get(), + &output, + padStrideInfo); + m_ConvolutionLayer.reset(directConvolutionLayer.release()); + } + else + { + auto convolutionLayer = std::make_unique(memoryManager); + convolutionLayer->configure(&input, + m_KernelTensor.get(), + m_BiasTensor.get(), + &output, + padStrideInfo); + m_ConvolutionLayer.reset(convolutionLayer.release()); + } + BOOST_ASSERT(m_ConvolutionLayer); + + InitializeArmComputeTensorData(*m_KernelTensor, m_Data.m_Weight); + + if (m_Data.m_Parameters.m_BiasEnabled) + { + InitializeArmComputeTensorData(*m_BiasTensor, m_Data.m_Bias); + } + + m_ConvolutionLayer->prepare(); + FreeUnusedTensors(); +} + +void NeonConvolution2dWorkload::Execute() const +{ + ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonConvolution2dWorkload_Execute"); + m_ConvolutionLayer->run(); +} + +void NeonConvolution2dWorkload::FreeUnusedTensors() +{ + FreeTensorIfUnused(m_KernelTensor); + FreeTensorIfUnused(m_BiasTensor); +} + +} //namespace armnn diff --git a/src/backends/neon/workloads/NeonConvolution2dWorkload.hpp b/src/backends/neon/workloads/NeonConvolution2dWorkload.hpp new file mode 100644 index 0000000000..573bb1a3e2 --- /dev/null +++ b/src/backends/neon/workloads/NeonConvolution2dWorkload.hpp @@ -0,0 +1,47 @@ +// +// Copyright © 2017 Arm Ltd. All rights reserved. +// SPDX-License-Identifier: MIT +// + +#pragma once + +#include +#include +#include +#include +#include + +#include + +#include + +namespace armnn +{ + +arm_compute::Status NeonConvolution2dWorkloadValidate(const TensorInfo& input, + const TensorInfo& output, + const Convolution2dDescriptor& descriptor, + const TensorInfo& weights, + const Optional& biases); + +class NeonConvolution2dWorkload : public BaseWorkload +{ +public: + using BaseWorkload::m_Data; + + NeonConvolution2dWorkload(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, + std::shared_ptr& memoryManager); + + void Execute() const override; + +private: + std::unique_ptr m_ConvolutionLayer; + + std::unique_ptr m_KernelTensor; + std::unique_ptr m_BiasTensor; + + void FreeUnusedTensors(); + +}; + +} //namespace armnn diff --git a/src/backends/neon/workloads/NeonWorkloads.hpp b/src/backends/neon/workloads/NeonWorkloads.hpp index 702ddb5c82..688b1f5828 100644 --- a/src/backends/neon/workloads/NeonWorkloads.hpp +++ b/src/backends/neon/workloads/NeonWorkloads.hpp @@ -10,9 +10,7 @@ #include "NeonConstantWorkload.hpp" #include "NeonConvertFp16ToFp32Workload.hpp" #include "NeonConvertFp32ToFp16Workload.hpp" -#include "NeonConvolution2dBaseWorkload.hpp" -#include "NeonConvolution2dFloatWorkload.hpp" -#include "NeonConvolution2dUint8Workload.hpp" +#include "NeonConvolution2dWorkload.hpp" #include "NeonDepthwiseConvolutionFloatWorkload.hpp" #include "NeonDepthwiseConvolutionUint8Workload.hpp" #include "NeonFloorFloatWorkload.hpp" -- cgit v1.2.1