28 #include <arm_compute/core/CL/CLKernelLibrary.h> 29 #include <arm_compute/runtime/CL/CLBufferAllocator.h> 30 #include <arm_compute/runtime/CL/CLScheduler.h> 45 std::string& outReasonIfUnsupported)
52 std::string& outReasonIfUnsupported,
68 if (modelOptions->SaveCachedNetwork())
72 if (filePath !=
"" && fs::exists(filePath) && fs::is_regular_file(filePath))
77 std::ofstream file(filePath, std::ios::out | std::ios::binary);
85 std::unique_ptr<IWorkload> ClWorkloadFactory::MakeWorkload(
const QueueDescriptorType& descriptor,
91 return MakeWorkloadHelper<FloatWorkload, Uint8Workload>(descriptor,
info, std::forward<Args>(args)...);
99 template <
typename Workload,
typename QueueDescriptorType,
typename... Args>
100 std::unique_ptr<IWorkload> ClWorkloadFactory::MakeWorkload(
const QueueDescriptorType& descriptor,
106 return std::make_unique<Workload>(descriptor,
info, std::forward<Args>(args)...);
114 void ClWorkloadFactory::InitializeCLCompileContext()
117 auto context = arm_compute::CLKernelLibrary::get().context();
118 auto device = arm_compute::CLKernelLibrary::get().get_device();
119 m_CLCompileContext = arm_compute::CLCompileContext(context, device);
121 if (m_ModelContextPtr)
127 && fs::exists(filePath)
128 && fs::is_regular_file(filePath)
129 && !(modelOptions->SaveCachedNetwork()))
133 deserializer.
Deserialize(m_CLCompileContext, context, device, filePath);
139 : m_MemoryManager(memoryManager), m_ModelContextPtr(
IBackendInternal::IBackendSpecificModelContextPtr{})
141 InitializeCLCompileContext();
146 : m_MemoryManager(memoryManager), m_ModelContextPtr(modelContextPtr)
148 InitializeCLCompileContext();
152 const bool IsMemoryManaged)
const 155 std::unique_ptr<ClTensorHandle> tensorHandle = std::make_unique<ClTensorHandle>(tensorInfo);
156 tensorHandle->SetMemoryGroup(m_MemoryManager->GetInterLayerMemoryGroup());
163 const bool IsMemoryManaged)
const 166 std::unique_ptr<ClTensorHandle> tensorHandle = std::make_unique<ClTensorHandle>(tensorInfo, dataLayout);
167 tensorHandle->SetMemoryGroup(m_MemoryManager->GetInterLayerMemoryGroup());
174 unsigned int const* subTensorOrigin)
const 177 arm_compute::TensorShape shape = armcomputetensorutils::BuildArmComputeTensorShape(subTensorShape);
184 coords.set(i, armnn::numeric_cast<int>(subTensorOrigin[revertedIndex]));
187 const arm_compute::TensorShape parentShape = armcomputetensorutils::BuildArmComputeTensorShape(parent.
GetShape());
188 if (!::arm_compute::error_on_invalid_subtensor(__func__, __FILE__, __LINE__, parentShape, coords, shape))
193 return std::make_unique<ClSubTensorHandle>(
194 PolymorphicDowncast<IClTensorHandle*>(&parent), shape, coords);
211 return MakeWorkload<ClActivationWorkload>(descriptor,
info, m_CLCompileContext);
217 return MakeWorkload<ClAdditionWorkload>(descriptor,
info, m_CLCompileContext);
223 return std::make_unique<ClArgMinMaxWorkload>(descriptor,
info, m_CLCompileContext);
230 return MakeWorkload<ClBatchNormalizationFloatWorkload, NullWorkload>(descriptor,
info, m_CLCompileContext);
236 return MakeWorkload<ClBatchToSpaceNdWorkload>(descriptor,
info, m_CLCompileContext);
242 return MakeWorkload<ClCastWorkload>(descriptor,
info, m_CLCompileContext);
248 return MakeWorkload<ClComparisonWorkload>(descriptor,
info, m_CLCompileContext);
254 return MakeWorkload<ClConcatWorkload>(descriptor,
info, m_CLCompileContext);
260 return MakeWorkload<ClConstantWorkload>(descriptor,
info, m_CLCompileContext);
267 return MakeWorkload<ClConvertFp16ToFp32Workload>(descriptor,
info, m_CLCompileContext);
274 return MakeWorkload<ClConvertFp32ToFp16Workload>(descriptor,
info, m_CLCompileContext);
280 bool isFastMathEnabled =
false;
281 if (m_ModelContextPtr)
283 if (m_ModelContextPtr.get() !=
nullptr)
292 return MakeWorkload<ClConvolution2dWorkload>(descriptor,
294 m_MemoryManager->GetIntraLayerManager(),
302 return MakeWorkload<NullWorkload, NullWorkload>(descriptor,
info, m_CLCompileContext);
308 return MakeWorkload<ClDepthToSpaceWorkload>(descriptor,
info, m_CLCompileContext);
315 return MakeWorkload<ClDepthwiseConvolutionWorkload>(descriptor,
info, m_CLCompileContext);
321 return MakeWorkload<ClDequantizeWorkload>(descriptor,
info, m_CLCompileContext);
328 return MakeWorkload<NullWorkload, NullWorkload>(descriptor,
info, m_CLCompileContext);
334 return std::make_unique<ClDivisionWorkload>(descriptor,
info, m_CLCompileContext);
348 return std::make_unique<ClAbsWorkload>(absQueueDescriptor,
info, m_CLCompileContext);
351 return std::make_unique<ClExpWorkload>(descriptor,
info, m_CLCompileContext);
353 return std::make_unique<ClNegWorkload>(descriptor,
info, m_CLCompileContext);
357 rsqrtQueueDescriptor.
m_Inputs = descriptor.m_Inputs;
358 rsqrtQueueDescriptor.
m_Outputs = descriptor.m_Outputs;
360 return std::make_unique<ClRsqrtWorkload>(rsqrtQueueDescriptor,
info, m_CLCompileContext);
363 return std::make_unique<ClLogicalNotWorkload>(descriptor,
info, m_CLCompileContext);
383 return std::make_unique<ClFillWorkload>(descriptor,
info, m_CLCompileContext);
389 return MakeWorkload<ClFloorFloatWorkload, NullWorkload>(descriptor,
info, m_CLCompileContext);
395 return MakeWorkload<ClFullyConnectedWorkload>(descriptor,
397 m_MemoryManager->GetIntraLayerManager(),
404 return MakeWorkload<ClGatherWorkload>(descriptor,
info, m_CLCompileContext);
421 return std::make_unique<CopyMemGenericWorkload>(descriptor,
info);
428 return MakeWorkload<ClInstanceNormalizationWorkload>(descriptor,
info, m_CLCompileContext);
434 return MakeWorkload<ClL2NormalizationFloatWorkload, NullWorkload>(descriptor,
info, m_CLCompileContext);
443 return std::make_unique<ClLogicalAndWorkload>(descriptor,
info, m_CLCompileContext);
445 return std::make_unique<ClLogicalOrWorkload>(descriptor,
info, m_CLCompileContext);
454 return MakeWorkload<ClLogSoftmaxWorkload>(descriptor,
456 m_MemoryManager->GetIntraLayerManager(),
463 return MakeWorkload<ClLstmFloatWorkload, NullWorkload>(descriptor,
info, m_CLCompileContext);
469 return MakeWorkload<ClMaximumWorkload>(descriptor,
info, m_CLCompileContext);
475 return MakeWorkload<ClMeanWorkload>(descriptor,
info, m_CLCompileContext);
486 return MakeWorkload<CopyMemGenericWorkload>(descriptor,
info);
497 return std::make_unique<ImportMemGenericWorkload>(descriptor,
info);
509 return MakeWorkload<ClMinimumWorkload>(descriptor,
info, m_CLCompileContext);
515 return MakeWorkload<ClMultiplicationWorkload>(descriptor,
info, m_CLCompileContext);
521 return MakeWorkload<ClNormalizationFloatWorkload, NullWorkload>(descriptor,
info, m_CLCompileContext);
527 return std::make_unique<CopyMemGenericWorkload>(descriptor,
info);
533 return MakeWorkload<ClPadWorkload>(descriptor,
info, m_CLCompileContext);
539 return MakeWorkload<ClPermuteWorkload>(descriptor,
info, m_CLCompileContext);
545 return MakeWorkload<ClPooling2dWorkload>(descriptor,
info, m_CLCompileContext);
551 return MakeWorkload<NullWorkload, NullWorkload>(descriptor,
info, m_CLCompileContext);
557 return MakeWorkload<ClPreluWorkload>(descriptor,
info, m_CLCompileContext);
563 return std::make_unique<ClQLstmWorkload>(descriptor,
info, m_CLCompileContext);
569 return MakeWorkload<ClQuantizeWorkload>(descriptor,
info, m_CLCompileContext);
575 return MakeWorkload<ClQuantizedLstmWorkload>(descriptor,
info, m_CLCompileContext);
581 return std::make_unique<ClRankWorkload>(descriptor,
info);
587 return std::make_unique<ClReduceWorkload>(descriptor,
info);
593 return MakeWorkload<ClReshapeWorkload>(descriptor,
info, m_CLCompileContext);
599 return MakeWorkload<ClResizeWorkload>(descriptor,
info, m_CLCompileContext);
631 return MakeWorkload<ClSliceWorkload>(descriptor,
info, m_CLCompileContext);
637 return std::make_unique<ClSoftmaxWorkload>(descriptor,
639 m_MemoryManager->GetIntraLayerManager(),
646 return MakeWorkload<ClSpaceToBatchNdWorkload>(descriptor,
info, m_CLCompileContext);
652 return MakeWorkload<ClSpaceToDepthWorkload>(descriptor,
info, m_CLCompileContext);
658 return MakeWorkload<ClSplitterWorkload>(descriptor,
info, m_CLCompileContext);
664 return MakeWorkload<ClStackWorkload>(descriptor,
info, m_CLCompileContext);
670 return MakeWorkload<ClStridedSliceWorkload>(descriptor,
info, m_CLCompileContext);
676 return MakeWorkload<ClSubtractionWorkload>(descriptor,
info, m_CLCompileContext);
682 return MakeWorkload<ClTransposeWorkload>(descriptor,
info, m_CLCompileContext);
689 return MakeWorkload<ClTransposeConvolution2dWorkload>(descriptor,
691 m_MemoryManager->GetIntraLayerManager(),
std::unique_ptr< IWorkload > CreateDetectionPostProcess(const DetectionPostProcessQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< ITensorHandle > CreateSubTensorHandle(ITensorHandle &parent, TensorShape const &subTensorShape, unsigned int const *subTensorOrigin) const override
std::unique_ptr< IWorkload > CreateComparison(const ComparisonQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateGather(const GatherQueueDescriptor &descriptor, const WorkloadInfo &info) const override
ClWorkloadFactory(const std::shared_ptr< ClMemoryManager > &memoryManager)
UnaryOperation m_Operation
Specifies the elementwiseUnary operation to execute.
Interface for a layer that is connectable to other layers via InputSlots and OutputSlots.
std::unique_ptr< IWorkload > CreateDebug(const DebugQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateReshape(const ReshapeQueueDescriptor &descriptor, const WorkloadInfo &info) const override
void AfterWorkloadsCreated() override
std::unique_ptr< IWorkload > CreateConvertFp32ToFp16(const ConvertFp32ToFp16QueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateSpaceToBatchNd(const SpaceToBatchNdQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateDivision(const DivisionQueueDescriptor &descriptor, const WorkloadInfo &info) const override
A ComparisonDescriptor for the ComparisonLayer.
uint32_t m_TargetWidth
Target width value.
DataLayout m_DataLayout
The data layout to be used (NCHW, NHWC).
constexpr const char * ClBackendId()
std::vector< BackendOptions > ModelOptions
std::unique_ptr< IWorkload > CreateStridedSlice(const StridedSliceQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateConvertFp16ToFp32(const ConvertFp16ToFp32QueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateQuantize(const QuantizeQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::array< unsigned int, MaxNumOfTensorDimensions > Coordinates
std::unique_ptr< IWorkload > CreateInput(const InputQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateStack(const StackQueueDescriptor &descriptor, const WorkloadInfo &info) const override
ResizeMethod m_Method
The Interpolation method to use (Bilinear, NearestNeighbor).
std::unique_ptr< IWorkload > CreateFullyConnected(const FullyConnectedQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateSlice(const SliceQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< ITensorHandle > CreateTensorHandle(const TensorInfo &tensorInfo, const bool IsMemoryManaged=true) const override
std::unique_ptr< IWorkload > CreateQuantizedLstm(const QuantizedLstmQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateRank(const RankQueueDescriptor &descriptor, const WorkloadInfo &info) const override
const BackendId & GetBackendId() const override
bool SaveSerializedToStream(std::ostream &stream)
Serializes the ClContext to the stream.
Copyright (c) 2021 ARM Limited and Contributors.
void IgnoreUnused(Ts &&...)
TypedWorkload< QueueDescriptor, armnn::DataType::Float16, armnn::DataType::Float32 > FloatWorkload
std::unique_ptr< IWorkload > CreateAbs(const AbsQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateLstm(const LstmQueueDescriptor &descriptor, const WorkloadInfo &info) const override
LayerDescriptor m_Parameters
LogicalBinaryOperation m_Operation
Specifies the logical operation to execute.
std::unique_ptr< IWorkload > CreateLogicalBinary(const LogicalBinaryQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateLogSoftmax(const LogSoftmaxQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateMean(const MeanQueueDescriptor &descriptor, const WorkloadInfo &Info) const override
std::unique_ptr< IWorkload > CreateReduce(const ReduceQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateMinimum(const MinimumQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateTransposeConvolution2d(const TransposeConvolution2dQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateFill(const FillQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateCast(const CastQueueDescriptor &descriptor, const WorkloadInfo &info) const override
static bool IsLayerSupported(const Layer &layer, Optional< DataType > dataType, std::string &outReasonIfUnsupported)
std::unique_ptr< IWorkload > CreateQLstm(const QLstmQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateDepthwiseConvolution2d(const DepthwiseConvolution2dQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateDepthToSpace(const DepthToSpaceQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::shared_ptr< IBackendModelContext > IBackendSpecificModelContextPtr
std::unique_ptr< IWorkload > CreateResize(const ResizeQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreatePermute(const PermuteQueueDescriptor &descriptor, const WorkloadInfo &info) const override
void Serialize(const arm_compute::CLCompileContext &clCompileContext)
Serializes the CLCompileContext built-in programs.
std::unique_ptr< IWorkload > CreateMemCopy(const MemCopyQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateBatchToSpaceNd(const BatchToSpaceNdQueueDescriptor &descriptor, const WorkloadInfo &info) const override
uint32_t m_TargetWidth
Target width value.
std::unique_ptr< IWorkload > CreateSoftmax(const SoftmaxQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateMaximum(const MaximumQueueDescriptor &descriptor, const WorkloadInfo &info) const override
RuntimeException WrapClError(const cl::Error &clError, const CheckLocation &location)
std::unique_ptr< IWorkload > CreateEqual(const EqualQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateElementwiseUnary(const ElementwiseUnaryQueueDescriptor &descriptor, const WorkloadInfo &info) const override
bool IsFastMathEnabled() const
std::unique_ptr< IWorkload > CreateFloor(const FloorQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreatePreCompiled(const PreCompiledQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateSubtraction(const SubtractionQueueDescriptor &descriptor, const WorkloadInfo &info) const override
static bool IsLayerSupported(const BackendId &backendId, const IConnectableLayer &layer, Optional< DataType > dataType, std::string &outReasonIfUnsupported)
std::unique_ptr< IWorkload > CreateMultiplication(const MultiplicationQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateAddition(const AdditionQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreatePooling2d(const Pooling2dQueueDescriptor &descriptor, const WorkloadInfo &info) const override
virtual TensorShape GetShape() const =0
Get the number of elements for each dimension ordered from slowest iterating dimension to fastest ite...
uint32_t m_TargetHeight
Target height value.
uint32_t m_TargetHeight
Target height value.
std::unique_ptr< IWorkload > CreateMerger(const MergerQueueDescriptor &descriptor, const WorkloadInfo &info) const override
void Deserialize(arm_compute::CLCompileContext &clCompileContext, cl::Context &context, cl::Device &device, const std::string &filePath)
Deserializes the CLCompileContext built-in programs from a binary file.
TypedWorkload< QueueDescriptor, armnn::DataType::QAsymmU8 > Uint8Workload
std::unique_ptr< IWorkload > CreateRsqrt(const RsqrtQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateSplitter(const SplitterQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateGreater(const GreaterQueueDescriptor &descriptor, const WorkloadInfo &info) const override
A ElementwiseUnaryDescriptor for the ElementwiseUnaryLayer.
std::unique_ptr< IWorkload > CreateL2Normalization(const L2NormalizationQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateActivation(const ActivationQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::vector< ITensorHandle * > m_Outputs
std::unique_ptr< IWorkload > CreatePad(const PadQueueDescriptor &descriptor, const WorkloadInfo &info) const override
unsigned int GetNumDimensions() const
Function that returns the tensor rank.
std::unique_ptr< IWorkload > CreateSpaceToDepth(const SpaceToDepthQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateNormalization(const NormalizationQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateInstanceNormalization(const InstanceNormalizationQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateTranspose(const TransposeQueueDescriptor &descriptor, const WorkloadInfo &info) const override
Contains information about inputs and outputs to a layer.
std::vector< ITensorHandle * > m_Inputs
std::unique_ptr< IWorkload > CreateBatchNormalization(const BatchNormalizationQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateConcat(const ConcatQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateMemImport(const MemImportQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateResizeBilinear(const ResizeBilinearQueueDescriptor &descriptor, const WorkloadInfo &info) const override
The ClBackendModelContext is used to pass in CL specific backend ModelOptions.
std::unique_ptr< IWorkload > CreatePrelu(const PreluQueueDescriptor &descriptor, const WorkloadInfo &info) const override
DataLayout m_DataLayout
The data layout to be used (NCHW, NHWC).
std::unique_ptr< IWorkload > CreateDequantize(const DequantizeQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateConstant(const ConstantQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateConvolution2d(const Convolution2dQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::unique_ptr< IWorkload > CreateArgMinMax(const ArgMinMaxQueueDescriptor &descriptor, const WorkloadInfo &info) const override
std::string GetCachedNetworkFilePath() const
std::unique_ptr< IWorkload > CreateOutput(const OutputQueueDescriptor &descriptor, const WorkloadInfo &info) const override