diff options
Diffstat (limited to 'src/backends/neon')
-rw-r--r-- | src/backends/neon/NeonLayerSupport.cpp | 2 | ||||
-rw-r--r-- | src/backends/neon/NeonWorkloadFactory.cpp | 4 | ||||
-rw-r--r-- | src/backends/neon/backend.mk | 4 | ||||
-rw-r--r-- | src/backends/neon/test/NeonCreateWorkloadTests.cpp | 12 | ||||
-rw-r--r-- | src/backends/neon/workloads/CMakeLists.txt | 8 | ||||
-rw-r--r-- | src/backends/neon/workloads/NeonConvolution2dFloatWorkload.cpp | 39 | ||||
-rw-r--r-- | src/backends/neon/workloads/NeonConvolution2dFloatWorkload.hpp | 29 | ||||
-rw-r--r-- | src/backends/neon/workloads/NeonConvolution2dUint8Workload.cpp | 35 | ||||
-rw-r--r-- | src/backends/neon/workloads/NeonConvolution2dUint8Workload.hpp | 29 | ||||
-rw-r--r-- | src/backends/neon/workloads/NeonConvolution2dWorkload.cpp (renamed from src/backends/neon/workloads/NeonConvolution2dBaseWorkload.cpp) | 31 | ||||
-rw-r--r-- | src/backends/neon/workloads/NeonConvolution2dWorkload.hpp (renamed from src/backends/neon/workloads/NeonConvolution2dBaseWorkload.hpp) | 14 | ||||
-rw-r--r-- | src/backends/neon/workloads/NeonWorkloads.hpp | 4 |
12 files changed, 39 insertions, 172 deletions
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<armnn::IWorkload> NeonWorkloadFactory::CreatePooling2d(const Poo std::unique_ptr<armnn::IWorkload> NeonWorkloadFactory::CreateConvolution2d( const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info) const { - return MakeWorkload<NeonConvolution2dFloatWorkload, NeonConvolution2dUint8Workload>(descriptor, info, - m_MemoryManager.GetIntraLayerManager()); + return std::make_unique<NeonConvolution2dWorkload>(descriptor, info, + m_MemoryManager.GetIntraLayerManager()); } std::unique_ptr<IWorkload> 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<NeonBatchNormalizationFloatWorkload, DataType::Float32>(); } -template <typename Convolution2dWorkloadType, typename armnn::DataType DataType> +template <typename armnn::DataType DataType> static void NeonCreateConvolution2dWorkloadTest(DataLayout dataLayout = DataLayout::NCHW) { Graph graph; NeonWorkloadFactory factory; - auto workload = CreateConvolution2dWorkloadTest<Convolution2dWorkloadType, + auto workload = CreateConvolution2dWorkloadTest<NeonConvolution2dWorkload, DataType>(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<NeonConvolution2dFloatWorkload, DataType::Float16>(); + NeonCreateConvolution2dWorkloadTest<DataType::Float16>(); } BOOST_AUTO_TEST_CASE(CreateConvolution2dFloat16NhwcWorkload) { - NeonCreateConvolution2dWorkloadTest<NeonConvolution2dFloatWorkload, DataType::Float16>(DataLayout::NHWC); + NeonCreateConvolution2dWorkloadTest<DataType::Float16>(DataLayout::NHWC); } #endif BOOST_AUTO_TEST_CASE(CreateConvolution2dFloatNchwWorkload) { - NeonCreateConvolution2dWorkloadTest<NeonConvolution2dFloatWorkload, DataType::Float32>(); + NeonCreateConvolution2dWorkloadTest<DataType::Float32>(); } BOOST_AUTO_TEST_CASE(CreateConvolution2dFloatNhwcWorkload) { - NeonCreateConvolution2dWorkloadTest<NeonConvolution2dFloatWorkload, DataType::Float32>(DataLayout::NHWC); + NeonCreateConvolution2dWorkloadTest<DataType::Float32>(DataLayout::NHWC); } template <typename DepthwiseConvolution2dFloat32WorkloadType, typename armnn::DataType DataType> 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/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 <backends/CpuTensorHandle.hpp> -#include <backends/aclCommon/ArmComputeTensorUtils.hpp> -#include <backends/neon/NeonLayerSupport.hpp> - -namespace armnn -{ -using namespace armcomputetensorutils; - -NeonConvolution2dFloatWorkload::NeonConvolution2dFloatWorkload(const Convolution2dQueueDescriptor& descriptor, - const WorkloadInfo& info, std::shared_ptr<arm_compute::MemoryManagerOnDemand>& 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 <backends/neon/workloads/NeonWorkloadUtils.hpp> - -#include "arm_compute/runtime/MemoryManagerOnDemand.h" - -#include <memory> - -namespace armnn -{ - -class NeonConvolution2dFloatWorkload : public NeonConvolution2dBaseWorkload<DataType::Float16, DataType::Float32> -{ -public: - NeonConvolution2dFloatWorkload(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, - std::shared_ptr<arm_compute::MemoryManagerOnDemand>& 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<arm_compute::MemoryManagerOnDemand>& 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 <memory> - -namespace armnn -{ - -class NeonConvolution2dUint8Workload : public NeonConvolution2dBaseWorkload<DataType::QuantisedAsymm8> -{ -public: - NeonConvolution2dUint8Workload(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, - std::shared_ptr<arm_compute::MemoryManagerOnDemand>& memoryManager); - - virtual void ValidateData() const override; - virtual void Execute() const override; -private: -}; - -} //namespace armnnn - diff --git a/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.cpp b/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp index b11d10fd2f..c26cdea92b 100644 --- a/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.cpp +++ b/src/backends/neon/workloads/NeonConvolution2dWorkload.cpp @@ -3,12 +3,12 @@ // SPDX-License-Identifier: MIT // +#include "NeonConvolution2dWorkload.hpp" + #include <backends/CpuTensorHandle.hpp> #include <backends/aclCommon/ArmComputeTensorUtils.hpp> #include <backends/neon/NeonLayerSupport.hpp> -#include "NeonConvolution2dBaseWorkload.hpp" - #include <armnn/Types.hpp> #include <armnnUtils/Half.hpp> @@ -47,15 +47,14 @@ arm_compute::Status NeonConvolution2dWorkloadValidate(const TensorInfo& input, layerInfo); } -template<armnn::DataType... dataTypes> -NeonConvolution2dBaseWorkload<dataTypes...>::NeonConvolution2dBaseWorkload( +NeonConvolution2dWorkload::NeonConvolution2dWorkload( const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, std::shared_ptr<arm_compute::MemoryManagerOnDemand>& memoryManager) - : TypedWorkload<Convolution2dQueueDescriptor, dataTypes...>(descriptor, info) + : BaseWorkload<Convolution2dQueueDescriptor>(descriptor, info) { using arm_compute::NEDirectConvolutionLayer; - ValidateData(); + m_Data.ValidateInputsOutputs("NeonConvolution2dWorkload", 1, 1); // todo: check tensor shapes match. @@ -111,17 +110,25 @@ NeonConvolution2dBaseWorkload<dataTypes...>::NeonConvolution2dBaseWorkload( 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(); } -template<armnn::DataType... dataTypes> -void NeonConvolution2dBaseWorkload<dataTypes...>::FreeUnusedTensors() +void NeonConvolution2dWorkload::Execute() const +{ + ARMNN_SCOPED_PROFILING_EVENT_NEON("NeonConvolution2dWorkload_Execute"); + m_ConvolutionLayer->run(); +} + +void NeonConvolution2dWorkload::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 diff --git a/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.hpp b/src/backends/neon/workloads/NeonConvolution2dWorkload.hpp index 1cd30c70f9..573bb1a3e2 100644 --- a/src/backends/neon/workloads/NeonConvolution2dBaseWorkload.hpp +++ b/src/backends/neon/workloads/NeonConvolution2dWorkload.hpp @@ -24,24 +24,24 @@ arm_compute::Status NeonConvolution2dWorkloadValidate(const TensorInfo& input, const TensorInfo& weights, const Optional<TensorInfo>& biases); -template<armnn::DataType... dataTypes> -class NeonConvolution2dBaseWorkload : public TypedWorkload<Convolution2dQueueDescriptor, dataTypes...> +class NeonConvolution2dWorkload : public BaseWorkload<Convolution2dQueueDescriptor> { public: - using TypedWorkload<Convolution2dQueueDescriptor, dataTypes...>::m_Data; + using BaseWorkload<Convolution2dQueueDescriptor>::m_Data; - NeonConvolution2dBaseWorkload(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, - std::shared_ptr<arm_compute::MemoryManagerOnDemand>& memoryManager); + NeonConvolution2dWorkload(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info, + std::shared_ptr<arm_compute::MemoryManagerOnDemand>& memoryManager); - virtual void ValidateData() const {}; + void Execute() const override; -protected: +private: std::unique_ptr<arm_compute::IFunction> m_ConvolutionLayer; std::unique_ptr<arm_compute::Tensor> m_KernelTensor; std::unique_ptr<arm_compute::Tensor> 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" |