// // Copyright © 2017 Arm Ltd. All rights reserved. // See LICENSE file in the project root for full license information. // #pragma once #include "WorkloadFactory.hpp" #include "OutputHandler.hpp" #include "armnn/IRuntime.hpp" #ifdef ARMCOMPUTECL_ENABLED #include #endif namespace cl { class Context; class CommandQueue; class Device; } namespace armnn { class IClTunedParameters; class ClTunedParameters; // ARM Compute OpenCL workload factory class ClWorkloadFactory : public IWorkloadFactory { public: ClWorkloadFactory(IClTunedParameters* clTunedParameters = nullptr); virtual ~ClWorkloadFactory(); virtual Compute GetCompute() const override { return Compute::GpuAcc; } static bool IsLayerSupported(const Layer& layer, DataType dataType, std::string& outReasonIfUnsupported); void LoadOpenClRuntime(); virtual bool SupportsSubTensors() const override { return true; } virtual std::unique_ptr CreateSubTensorHandle(ITensorHandle& parent, TensorShape const& subTensorShape, unsigned int const* subTensorOrigin) const override; virtual std::unique_ptr CreateTensorHandle(const TensorInfo& tensorInfo) const override; virtual std::unique_ptr CreateInput(const InputQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateOutput(const OutputQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateActivation(const ActivationQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateSoftmax(const SoftmaxQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateSplitter(const SplitterQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateMerger(const MergerQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateFullyConnected(const FullyConnectedQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreatePermute(const PermuteQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreatePooling2d(const Pooling2dQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateConvolution2d(const Convolution2dQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateDepthwiseConvolution2d( const DepthwiseConvolution2dQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateNormalization(const NormalizationQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateAddition(const AdditionQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateMultiplication(const MultiplicationQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateBatchNormalization(const BatchNormalizationQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateMemCopy(const MemCopyQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateResizeBilinear(const ResizeBilinearQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateFakeQuantization(const FakeQuantizationQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateL2Normalization(const L2NormalizationQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateConstant(const ConstantQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateReshape(const ReshapeQueueDescriptor& descriptor, const WorkloadInfo& info) const override; virtual std::unique_ptr CreateFloor(const FloorQueueDescriptor& descriptor, const WorkloadInfo& info) const override; private: ClTunedParameters* m_clTunedParameters; }; class ClTunedParameters : public IClTunedParameters { public: ClTunedParameters(armnn::IClTunedParameters::Mode mode); virtual void Load(const char* filename); virtual void Save(const char* filename) const; Mode m_Mode; #ifdef ARMCOMPUTECL_ENABLED arm_compute::CLTuner m_Tuner; #endif }; } // namespace armnn