Per-operator build dependencies
Creates a list of operators their respective dependencies.
Alters the build system to walk-through them resolve the dependencies
and build Compute Library.
Removes the following unused kernels/functions:
-[NE|CL]MinMaxLayerKernel
-CLFillBorder
Resolves: COMPMID-4695,COMPMID-4696
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Change-Id: I35ebeef38dac25ec5459cfe9c5f7c9a708621124
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/357914
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Michele DiGiorgio <michele.digiorgio@arm.com>
Comments-Addressed: bsgcomp <bsgcomp@arm.com>
Signed-off-by: Freddie Liardet <frederick.liardet@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/6295
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
diff --git a/filelist.json b/filelist.json
index 5171f39..4b85408 100644
--- a/filelist.json
+++ b/filelist.json
@@ -7,8 +7,80 @@
"src/common/AllocatorWrapper.cpp",
"src/common/ITensorV2.cpp",
"src/common/TensorPack.cpp",
- "src/common/IOperator.cpp"
+ "src/common/IOperator.cpp",
+ "src/core/AccessWindowAutoPadding.cpp",
+ "src/core/AccessWindowStatic.cpp",
+ "src/core/AccessWindowTranspose.cpp",
+ "src/core/Error.cpp",
+ "src/core/GPUTarget.cpp",
+ "src/core/Helpers.cpp",
+ "src/core/IAccessWindow.cpp",
+ "src/core/IKernel.cpp",
+ "src/core/ITensor.cpp",
+ "src/core/ITensorPack.cpp",
+ "src/core/Rounding.cpp",
+ "src/core/Size2D.cpp",
+ "src/core/SubTensorInfo.cpp",
+ "src/core/TensorInfo.cpp",
+ "src/core/Utils.cpp",
+ "src/core/Validate.cpp",
+ "src/core/Version.cpp",
+ "src/core/helpers/SoftmaxHelpers.cpp",
+ "src/core/helpers/WindowHelpers.cpp",
+ "src/core/utils/AssemblyUtils.cpp",
+ "src/core/utils/ScaleUtils.cpp",
+ "src/core/utils/helpers/fft.cpp",
+ "src/core/utils/helpers/tensor_transform.cpp",
+ "src/core/utils/io/FileHandler.cpp",
+ "src/core/utils/misc/MMappedFile.cpp",
+ "src/core/utils/quantization/AsymmHelpers.cpp",
+ "src/core/CPP/CPPTypes.cpp",
+ "src/core/CPP/kernels/CPPBoxWithNonMaximaSuppressionLimitKernel.cpp",
+ "src/core/CPP/kernels/CPPNonMaximumSuppressionKernel.cpp",
+ "src/core/CPP/kernels/CPPPermuteKernel.cpp",
+ "src/core/CPP/kernels/CPPTopKVKernel.cpp",
+ "src/core/CPP/kernels/CPPUpsampleKernel.cpp",
+ "src/runtime/Allocator.cpp",
+ "src/runtime/BlobLifetimeManager.cpp",
+ "src/runtime/BlobMemoryPool.cpp",
+ "src/runtime/ISimpleLifetimeManager.cpp",
+ "src/runtime/ITensorAllocator.cpp",
+ "src/runtime/IWeightsManager.cpp",
+ "src/runtime/IScheduler.cpp",
+ "src/runtime/Memory.cpp",
+ "src/runtime/MemoryManagerOnDemand.cpp",
+ "src/runtime/OffsetLifetimeManager.cpp",
+ "src/runtime/OffsetMemoryPool.cpp",
+ "src/runtime/OperatorTensor.cpp",
+ "src/runtime/PoolManager.cpp",
+ "src/runtime/RuntimeContext.cpp",
+ "src/runtime/Scheduler.cpp",
+ "src/runtime/SchedulerFactory.cpp",
+ "src/runtime/SchedulerUtils.cpp",
+ "src/runtime/SubTensor.cpp",
+ "src/runtime/Tensor.cpp",
+ "src/runtime/TensorAllocator.cpp",
+ "src/runtime/Utils.cpp",
+ "src/runtime/CPP/ICPPSimpleFunction.cpp",
+ "src/runtime/CPP/functions/CPPBoxWithNonMaximaSuppressionLimit.cpp",
+ "src/runtime/CPP/functions/CPPDetectionOutputLayer.cpp",
+ "src/runtime/CPP/functions/CPPDetectionPostProcessLayer.cpp",
+ "src/runtime/CPP/functions/CPPNonMaximumSuppression.cpp",
+ "src/runtime/CPP/functions/CPPPermute.cpp",
+ "src/runtime/CPP/functions/CPPTopKV.cpp",
+ "src/runtime/CPP/functions/CPPUpsample.cpp"
],
+ "logging": [
+ "src/core/utils/logging/FilePrinter.cpp",
+ "src/core/utils/logging/Helpers.cpp",
+ "src/core/utils/logging/Logger.cpp",
+ "src/core/utils/logging/LoggerRegistry.cpp"
+ ],
+ "scheduler": {
+ "single": [ "src/runtime/CPP/SingleThreadScheduler.cpp" ],
+ "threads": [ "src/runtime/CPP/CPPScheduler.cpp" ],
+ "omp": [ "src/runtime/OMP/OMPScheduler.cpp"]
+ },
"c_api": {
"common": [
"src/c/AclContext.cpp",
@@ -28,6 +100,14 @@
"src/c/operators/AclActivation.cpp"
]
},
+ "high_priority": [
+ "Activation",
+ "DepthwiseConv2d",
+ "Conv2d",
+ "Permute",
+ "Pool2d",
+ "Reshape"
+ ],
"gpu": {
"common": [
"src/core/CL/CLCompileContext.cpp",
@@ -41,19 +121,11 @@
"src/core/CL/ICLSimpleKernel.cpp",
"src/core/CL/ICLTensor.cpp",
"src/core/CL/OpenCL.cpp",
- "src/gpu/cl/ClKernelLibrary.cpp",
- "src/gpu/cl/kernels/gemm/ClGemmHelpers.cpp",
- "src/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeBifrost.cpp",
- "src/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeMidgard.cpp",
- "src/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeValhall.cpp",
- "src/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedBifrost.cpp",
- "src/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedValhall.cpp",
- "src/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyBifrost.cpp",
- "src/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp",
- "src/core/CL/kernels/CLFillBorderKernel.cpp",
"src/gpu/cl/ClContext.cpp",
+ "src/gpu/cl/ClKernelLibrary.cpp",
"src/gpu/cl/ClQueue.cpp",
"src/gpu/cl/ClTensor.cpp",
+ "src/core/CL/kernels/CLFillBorderKernel.cpp",
"src/runtime/CL/CLBufferAllocator.cpp",
"src/runtime/CL/CLGEMMHeuristicsHandle.cpp",
"src/runtime/CL/CLHelpers.cpp",
@@ -68,888 +140,1022 @@
"src/runtime/CL/CLTuner.cpp",
"src/runtime/CL/ICLSimpleFunction.cpp",
"src/runtime/CL/Utils.cpp",
- "src/runtime/CL/gemm/CLGEMMDefaultTypeBifrost.cpp",
- "src/runtime/CL/gemm/CLGEMMDefaultTypeMidgard.cpp",
- "src/runtime/CL/gemm/CLGEMMDefaultTypeValhall.cpp",
- "src/runtime/CL/gemm_auto_heuristics/CLGEMMAutoHeuristics.cpp",
"src/runtime/CL/mlgo/HeuristicTree.cpp",
"src/runtime/CL/mlgo/MLGOHeuristics.cpp",
"src/runtime/CL/mlgo/MLGOParser.cpp",
"src/runtime/CL/mlgo/Utils.cpp",
"src/runtime/CL/tuners/CLTuningParametersList.cpp"
],
- "high_priority": [
- "Activation",
- "DepthwiseConv2d",
- "DirectConv2d",
- "Permute",
- "Pool2d",
- "Reshape"
- ],
"operators": {
- "Activation": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClActivation.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClActivationKernel.cpp"
- ]
- }
- },
- "Add": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClAdd.cpp"
- ]
- }
- },
- "Cast": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClCast.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClCastKernel.cpp"
- ]
- }
- },
- "Concatenate": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClConcatenate.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.cpp",
- "src/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.cpp",
- "src/gpu/cl/kernels/ClWidthConcatenateKernel.cpp",
- "src/gpu/cl/kernels/ClHeightConcatenateKernel.cpp",
- "src/gpu/cl/kernels/ClDepthConcatenateKernel.cpp",
- "src/gpu/cl/kernels/ClBatchConcatenateKernel.cpp"
- ]
- }
- },
- "DirectConv2d": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClDirectConv2d.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClDirectConv2dKernel.cpp"
- ]
- }
- },
- "FullyConnected": {
- "deps": [
- "ClFlatten",
- "ClConvertFullyConnectedWeights",
- "ClGemm",
- "ClGemmLowpMatrixMultiplyCore",
- "ClTranspose"
- ],
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClFullyConnected.cpp"
- ]
- }
- },
- "ConvertFullyConnectedWeights": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClConvertFullyConnectedWeights.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClConvertFullyConnectedWeightsKernel.cpp"
- ]
- }
- },
- "Permute": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClPermute.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClPermuteKernel.cpp"
- ]
- }
- },
- "Pool2d": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClPool2d.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClPool2dKernel.cpp"
- ]
- }
- },
- "Conv2d": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClConv2d.cpp"
- ]
- }
- },
- "PRelu": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClPRelu.cpp"
- ]
- }
- },
- "Reshape": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClReshape.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClReshapeKernel.cpp"
- ]
- }
- },
- "Copy": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClCopy.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClCopyKernel.cpp"
- ]
- }
- },
- "Crop": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClCrop.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClCropKernel.cpp"
- ]
- }
- },
- "Dequantize": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClDequantize.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClDequantizeKernel.cpp"
- ]
- }
- },
- "Elementwise": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClElementwiseOperations.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClElementwiseKernel.cpp"
- ]
- }
- },
- "ElementwiseUnary": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClElementwiseUnary.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClElementwiseUnaryKernel.cpp"
- ]
- }
- },
- "Fill": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClFill.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClFillKernel.cpp"
- ]
- }
- },
- "Flatten": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClFlatten.cpp"
- ]
- }
- },
- "Floor": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClFloor.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClFloorKernel.cpp"
- ]
- }
- },
- "GEMM": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClGemm.cpp",
- "src/gpu/cl/operators/ClGemmConv2d.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClGemmMatrixMultiplyNativeKernel.cpp",
- "src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.cpp",
- "src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsKernel.cpp",
- "src/gpu/cl/kernels/ClGemmMatrixMultiplyKernel.cpp",
- "src/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.cpp",
- "src/gpu/cl/kernels/ClGemmReshapeLhsMatrixKernel.cpp"
- ]
- }
- },
- "GEMMLowp": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClGemmLowpMatrixMultiplyCore.cpp",
- "src/gpu/cl/operators/ClGemmLowpOutputStage.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClGemmLowpMatrixMultiplyNativeKernel.cpp",
- "src/gpu/cl/kernels/ClGemmLowpMatrixMultiplyReshapedKernel.cpp",
- "src/gpu/cl/kernels/ClGemmLowpMatrixMultiplyReshapedOnlyRhsKernel.cpp",
- "src/gpu/cl/kernels/ClGemmLowpOffsetContributionKernel.cpp",
- "src/gpu/cl/kernels/ClGemmLowpOffsetContributionOutputStageKernel.cpp",
- "src/gpu/cl/kernels/ClGemmLowpQuantizeDownInt32ScaleByFixedPointKernel.cpp",
- "src/gpu/cl/kernels/ClGemmLowpQuantizeDownInt32ScaleByFloatKernel.cpp",
- "src/gpu/cl/kernels/ClGemmLowpQuantizeDownInt32ScaleKernel.cpp",
- "src/gpu/cl/kernels/ClGemmLowpReductionKernel.cpp"
- ]
- }
- },
- "Mul": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClMul.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClMulKernel.cpp"
- ]
- }
- },
- "Quantize": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClQuantize.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClQuantizeKernel.cpp"
- ]
- }
- },
- "Scale": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClScale.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClScaleKernel.cpp"
- ]
- }
- },
- "Softmax": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClSoftmax.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClSoftmaxKernel.cpp"
- ]
- }
- },
- "Sub": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClSub.cpp"
- ]
- }
- },
- "Transpose": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClTranspose.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClTransposeKernel.cpp"
- ]
- }
- },
- "GenerateProposals": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLGenerateProposalsLayerKernel.cpp"
- ]
- }
- },
- "ArgMinMax": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLArgMinMaxLayerKernel.cpp"
- ]
- }
- },
- "BatchNormalization": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLBatchNormalizationLayerKernel.cpp"
- ]
- }
- },
- "BatchToSpace": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLBatchToSpaceLayerKernel.cpp"
- ]
- }
- },
- "Bitwise": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLBitwiseKernel.cpp"
- ]
- }
- },
- "BoundingBoxTransform": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLBoundingBoxTransformKernel.cpp"
- ]
- }
- },
- "ChannelShuffleLayer": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLChannelShuffleLayerKernel.cpp"
- ]
- }
- },
- "GEMMConv2d": {
- "files": {
- "kernel": [
- "src/gpu/cl/kernels/ClCol2ImKernel.cpp",
- "src/gpu/cl/kernels/ClIm2ColKernel.cpp"
- ]
- }
- },
- "Comparison": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLComparisonKernel.cpp"
- ]
- }
- },
- "DeconvolutionLayerUpsample": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLDeconvolutionLayerUpsampleKernel.cpp"
- ]
- }
- },
- "DeconvolutionReshapeOutput": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLDeconvolutionReshapeOutputKernel.cpp"
- ]
- }
- },
- "DepthToSpace": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLDepthToSpaceLayerKernel.cpp"
- ]
- }
- },
- "DepthwiseConvolutionLayerNative": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLDepthwiseConvolutionLayerNativeKernel.cpp"
- ]
- }
- },
- "FFTDigitReverse": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLFFTDigitReverseKernel.cpp"
- ]
- }
- },
- "FFTRadixStage": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLFFTRadixStageKernel.cpp"
- ]
- }
- },
- "FFTScale": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLFFTScaleKernel.cpp"
- ]
- }
- },
- "FuseBatchNormalization": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLFuseBatchNormalizationKernel.cpp"
- ]
- }
- },
- "Gather": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLGatherKernel.cpp"
- ]
- }
- },
- "InstanceNormalization": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLInstanceNormalizationLayerKernel.cpp"
- ]
- }
- },
- "L2Normalize": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLL2NormalizeLayerKernel.cpp"
- ]
- }
- },
- "LogicalNot": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClLogicalNot.cpp"
- ]
- }
- },
- "MaxUnpooling": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLMaxUnpoolingLayerKernel.cpp"
- ]
- }
- },
- "MeanStdDevNormalization": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLMeanStdDevNormalizationKernel.cpp"
- ]
- }
- },
- "MinMax": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLMinMaxLayerKernel.cpp"
- ]
- }
- },
- "Normalization": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLNormalizationLayerKernel.cpp"
- ]
- }
- },
- "NormalizePlanarYUV": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.cpp"
- ]
- }
- },
- "Pad": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLPadLayerKernel.cpp"
- ]
- }
- },
- "PriorBox": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLPriorBoxLayerKernel.cpp"
- ]
- }
- },
- "QLSTMLayerNormalization": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLQLSTMLayerNormalizationKernel.cpp"
- ]
- }
- },
- "Range": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLRangeKernel.cpp"
- ]
- }
- },
- "ReductionOperation": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLReductionOperationKernel.cpp"
- ]
- }
- },
- "Remap": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLRemapKernel.cpp"
- ]
- }
- },
- "Reorg": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLReorgLayerKernel.cpp"
- ]
- }
- },
- "Reverse": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLReverseKernel.cpp"
- ]
- }
- },
- "ROIAlign": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLROIAlignLayerKernel.cpp"
- ]
- }
- },
- "ROIPooling": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLROIPoolingLayerKernel.cpp"
- ]
- }
- },
- "Select": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLSelectKernel.cpp"
- ]
- }
- },
- "SpaceToBatch": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp"
- ]
- }
- },
- "SpaceToDepth": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLSpaceToDepthLayerKernel.cpp"
- ]
- }
- },
- "Stack": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLStackLayerKernel.cpp"
- ]
- }
- },
- "StridedSlice": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLStridedSliceKernel.cpp"
- ]
- }
- },
- "Tile": {
- "files": {
- "kernel": [
- "src/core/CL/kernels/CLTileKernel.cpp"
- ]
- }
- },
- "WeightsReshape": {
- "files": {
- "kernel": [
- "src/gpu/cl/kernels/ClWeightsReshapeKernel.cpp"
- ]
- }
- },
- "WinogradConv2d": {
- "files": {
- "operator": [
- "src/gpu/cl/operators/ClWinogradConv2d.cpp"
- ],
- "kernel": [
- "src/gpu/cl/kernels/ClWinogradFilterTransformKernel.cpp",
- "src/gpu/cl/kernels/ClWinogradInputTransformKernel.cpp",
- "src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp"
- ]
- }
+ "Activation":{
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClActivationKernel.cpp",
+ "src/gpu/cl/operators/ClActivation.cpp",
+ "src/runtime/CL/functions/CLActivationLayer.cpp"
+ ]
+ }
+ },
+ "ArgMinMax": {
+ "deps": [ "Reshape" ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLArgMinMaxLayerKernel.cpp",
+ "src/runtime/CL/functions/CLArgMinMaxLayer.cpp"
+ ]
+ }
+ },
+ "Add": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClElementwiseKernel.cpp",
+ "src/gpu/cl/operators/ClAdd.cpp"
+ ]
+ }
+ },
+ "BatchNormalization": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLBatchNormalizationLayerKernel.cpp",
+ "src/core/CL/kernels/CLFuseBatchNormalizationKernel.cpp",
+ "src/runtime/CL/functions/CLBatchNormalizationLayer.cpp",
+ "src/runtime/CL/functions/CLFuseBatchNormalization.cpp"
+ ]
+ }
+ },
+ "BatchToSpace": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLBatchToSpaceLayerKernel.cpp",
+ "src/runtime/CL/functions/CLBatchToSpaceLayer.cpp"
+ ]
+ }
+ },
+ "Bitwise": {
+ "files": {
+ "common": [ "src/core/CL/kernels/CLBitwiseKernel.cpp" ]
+ }
+ },
+ "BitwiseAnd": {
+ "deps": [ "Bitwise" ],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLBitwiseAnd.cpp" ]
+ }
+ },
+ "BitwiseNot": {
+ "deps": [ "Bitwise" ],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLBitwiseNot.cpp" ]
+ }
+ },
+ "BitwiseOr": {
+ "deps": [ "Bitwise" ],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLBitwiseOr.cpp" ]
+ }
+ },
+ "BitwiseXor": {
+ "deps": [ "Bitwise" ],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLBitwiseXor.cpp" ]
+ }
+ },
+ "BoundingBoxTransform": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLBoundingBoxTransformKernel.cpp",
+ "src/runtime/CL/functions/CLBoundingBoxTransform.cpp"
+ ]
+ }
+ },
+ "Cast": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClCastKernel.cpp",
+ "src/gpu/cl/operators/ClCast.cpp",
+ "src/runtime/CL/functions/CLCast.cpp"
+ ]
+ }
+ },
+ "ChannelShuffle": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLChannelShuffleLayerKernel.cpp",
+ "src/runtime/CL/functions/CLChannelShuffleLayer.cpp"
+ ]
+ }
+ },
+ "Comparison": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLComparisonKernel.cpp",
+ "src/runtime/CL/functions/CLComparison.cpp"
+ ]
+ }
+ },
+ "Concatenate": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClBatchConcatenateKernel.cpp",
+ "src/gpu/cl/kernels/ClDepthConcatenateKernel.cpp",
+ "src/gpu/cl/kernels/ClHeightConcatenateKernel.cpp",
+ "src/gpu/cl/kernels/ClWidthConcatenate2TensorsKernel.cpp",
+ "src/gpu/cl/kernels/ClWidthConcatenate4TensorsKernel.cpp",
+ "src/gpu/cl/kernels/ClWidthConcatenateKernel.cpp",
+ "src/gpu/cl/operators/ClConcatenate.cpp",
+ "src/runtime/CL/functions/CLConcatenateLayer.cpp"
+ ]
+ }
+ },
+ "Conv2d": {
+ "deps": [
+ "Activation",
+ "ElementwiseBinary",
+ "FFT2D",
+ "Gemm",
+ "Mul",
+ "Pad",
+ "Permute",
+ "Reduction",
+ "Reshape",
+ "Reverse",
+ "Slice"
+ ],
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClDirectConv2dKernel.cpp",
+ "src/gpu/cl/kernels/ClWinogradFilterTransformKernel.cpp",
+ "src/gpu/cl/kernels/ClWinogradInputTransformKernel.cpp",
+ "src/gpu/cl/kernels/ClWinogradOutputTransformKernel.cpp",
+ "src/gpu/cl/kernels/ClIm2ColKernel.cpp",
+ "src/gpu/cl/kernels/ClCol2ImKernel.cpp",
+ "src/gpu/cl/operators/ClConv2d.cpp",
+ "src/gpu/cl/operators/ClDirectConv2d.cpp",
+ "src/gpu/cl/operators/ClGemmConv2d.cpp",
+ "src/gpu/cl/operators/ClWinogradConv2d.cpp",
+ "src/gpu/cl/kernels/ClWeightsReshapeKernel.cpp",
+ "src/runtime/CL/functions/CLConvolutionLayer.cpp",
+ "src/runtime/CL/functions/CLDirectConvolutionLayer.cpp",
+ "src/runtime/CL/functions/CLFFTConvolutionLayer.cpp",
+ "src/runtime/CL/functions/CLGEMMConvolutionLayer.cpp",
+ "src/runtime/CL/functions/CLWinogradConvolutionLayer.cpp"
+ ]
+ }
+ },
+ "Copy": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClCopyKernel.cpp",
+ "src/gpu/cl/operators/ClCopy.cpp",
+ "src/runtime/CL/functions/CLCopy.cpp"
+ ]
+ }
+ },
+ "CropResize": {
+ "deps": [ "Copy", "Fill", "Scale" ],
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClCropKernel.cpp",
+ "src/gpu/cl/operators/ClCrop.cpp",
+ "src/runtime/CL/functions/CLCrop.cpp",
+ "src/runtime/CL/functions/CLCropResize.cpp"
+ ]
+ }
+ },
+ "Deconv2d": {
+ "deps": [ "Conv2d", "Reverse", "Transpose"],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLDeconvolutionLayerUpsampleKernel.cpp",
+ "src/core/CL/kernels/CLDeconvolutionReshapeOutputKernel.cpp",
+ "src/runtime/CL/functions/CLDeconvolutionLayer.cpp",
+ "src/runtime/CL/functions/CLDeconvolutionLayerUpsample.cpp",
+ "src/runtime/CL/functions/CLDirectDeconvolutionLayer.cpp",
+ "src/runtime/CL/functions/CLGEMMDeconvolutionLayer.cpp"
+ ]
+ }
+ },
+ "DepthConvert": {
+ "deps": [ "Cast"],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLDepthConvertLayer.cpp" ]
+ }
+ },
+ "DepthToSpace": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLDepthToSpaceLayerKernel.cpp",
+ "src/runtime/CL/functions/CLDepthToSpaceLayer.cpp"
+ ]
+ }
+ },
+ "DepthwiseConv2d": {
+ "deps": [ "Permute" ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLDepthwiseConvolutionLayerNativeKernel.cpp",
+ "src/runtime/CL/functions/CLDepthwiseConvolutionLayer.cpp"
+ ]
+ }
+ },
+ "Dequantize": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClDequantizeKernel.cpp",
+ "src/gpu/cl/operators/ClDequantize.cpp",
+ "src/runtime/CL/functions/CLDequantizationLayer.cpp"
+ ]
+ }
+ },
+ "ElementwiseBinary": {
+ "deps": ["Add", "Sub"],
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClElementwiseKernel.cpp",
+ "src/gpu/cl/operators/ClElementwiseOperations.cpp",
+ "src/runtime/CL/functions/CLElementwiseOperations.cpp"
+ ]
+ }
+ },
+ "ElementwiseUnary":{
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClElementwiseUnaryKernel.cpp",
+ "src/gpu/cl/operators/ClElementwiseUnary.cpp",
+ "src/runtime/CL/functions/CLElementwiseUnaryLayer.cpp"
+ ]
+ }
+ },
+ "FFT1D": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLFFTDigitReverseKernel.cpp",
+ "src/core/CL/kernels/CLFFTRadixStageKernel.cpp",
+ "src/core/CL/kernels/CLFFTScaleKernel.cpp",
+ "src/runtime/CL/functions/CLFFT1D.cpp"
+ ]
+ }
+ },
+ "FFT2D": {
+ "deps": [ "FFT1D" ],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLFFT2D.cpp" ]
+ }
+ },
+ "Fill": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClFillKernel.cpp",
+ "src/gpu/cl/operators/ClFill.cpp",
+ "src/runtime/CL/functions/CLFill.cpp"
+ ]
+ }
+ },
+ "Flatten": {
+ "files": {
+ "common": [
+ "src/gpu/cl/operators/ClFlatten.cpp",
+ "src/runtime/CL/functions/CLFlattenLayer.cpp"
+ ]
+ }
+ },
+ "Floor": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClFloorKernel.cpp",
+ "src/gpu/cl/operators/ClFloor.cpp",
+ "src/runtime/CL/functions/CLFloor.cpp"
+ ]
+ }
+ },
+ "FullyConnected": {
+ "deps": [ "Flatten", "Gemm", "Transpose"],
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClConvertFullyConnectedWeightsKernel.cpp",
+ "src/gpu/cl/operators/ClConvertFullyConnectedWeights.cpp",
+ "src/gpu/cl/operators/ClFullyConnected.cpp",
+ "src/runtime/CL/functions/CLConvertFullyConnectedWeights.cpp",
+ "src/runtime/CL/functions/CLFullyConnectedLayer.cpp"
+ ]
+ }
+ },
+ "Gather": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLGatherKernel.cpp",
+ "src/runtime/CL/functions/CLGather.cpp"]
+ }
+ },
+ "Gemm": {
+ "deps": [ "Cast" ],
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/gemm/ClGemmHelpers.cpp",
+ "src/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeBifrost.cpp",
+ "src/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeMidgard.cpp",
+ "src/gpu/cl/kernels/gemm/native/ClGemmDefaultConfigNativeValhall.cpp",
+ "src/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedBifrost.cpp",
+ "src/gpu/cl/kernels/gemm/reshaped/ClGemmDefaultConfigReshapedValhall.cpp",
+ "src/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyBifrost.cpp",
+ "src/gpu/cl/kernels/gemm/reshaped_only_rhs/ClGemmDefaultConfigReshapedRhsOnlyValhall.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpMatrixMultiplyNativeKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpMatrixMultiplyReshapedKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpMatrixMultiplyReshapedOnlyRhsKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpOffsetContributionKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpOffsetContributionOutputStageKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpQuantizeDownInt32ScaleByFixedPointKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpQuantizeDownInt32ScaleByFloatKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpQuantizeDownInt32ScaleKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmMatrixMultiplyKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmMatrixMultiplyNativeKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmMatrixMultiplyReshapedOnlyRhsKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmReshapeLhsMatrixKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmReshapeRhsMatrixKernel.cpp",
+ "src/gpu/cl/kernels/ClGemmLowpReductionKernel.cpp",
+ "src/gpu/cl/operators/ClGemm.cpp",
+ "src/gpu/cl/operators/ClGemmLowpMatrixMultiplyCore.cpp",
+ "src/gpu/cl/operators/ClGemmLowpOutputStage.cpp",
+ "src/runtime/CL/gemm/CLGEMMDefaultTypeBifrost.cpp",
+ "src/runtime/CL/gemm/CLGEMMDefaultTypeMidgard.cpp",
+ "src/runtime/CL/gemm/CLGEMMDefaultTypeValhall.cpp",
+ "src/runtime/CL/gemm_auto_heuristics/CLGEMMAutoHeuristics.cpp",
+ "src/runtime/CL/functions/CLGEMM.cpp",
+ "src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp",
+ "src/runtime/CL/functions/CLGEMMLowpOutputStage.cpp"
+ ]
+ }
+ },
+ "GenerateProposals": {
+ "deps": [ "BoundingBoxTransform", "Dequantize", "Pad", "Permute", "Quantize", "Reshape" ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLGenerateProposalsLayerKernel.cpp",
+ "src/runtime/CL/functions/CLGenerateProposalsLayer.cpp"
+ ]
+ }
+ },
+ "InstanceNormalize": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLInstanceNormalizationLayerKernel.cpp",
+ "src/runtime/CL/functions/CLInstanceNormalizationLayer.cpp"
+ ]
+ }
+ },
+ "L2Normalize": {
+ "deps": [ "Reduction" ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLL2NormalizeLayerKernel.cpp",
+ "src/runtime/CL/functions/CLL2NormalizeLayer.cpp"
+ ]
+ }
+ },
+ "Logical": {
+ "files": {
+ "common": [
+ "src/gpu/cl/operators/ClLogicalNot.cpp",
+ "src/runtime/CL/functions/CLLogicalAnd.cpp",
+ "src/runtime/CL/functions/CLLogicalNot.cpp",
+ "src/runtime/CL/functions/CLLogicalOr.cpp"
+ ]
+ }
+ },
+ "LSTM": {
+ "deps": [
+ "Activation",
+ "Concatenate",
+ "Copy",
+ "Dequantize",
+ "ElementwiseBinary",
+ "Fill",
+ "FullyConnected",
+ "Gemm",
+ "MeanStdDevNormalize",
+ "Mul",
+ "Quantize",
+ "Slice",
+ "Transpose"
+ ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLQLSTMLayerNormalizationKernel.cpp",
+ "src/runtime/CL/functions/CLQLSTMLayer.cpp",
+ "src/runtime/CL/functions/CLLSTMLayer.cpp",
+ "src/runtime/CL/functions/CLLSTMLayerQuantized.cpp"
+ ]
+ }
+ },
+ "MaxUnpool2d": {
+ "deps": [ "Fill" ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLMaxUnpoolingLayerKernel.cpp",
+ "src/runtime/CL/functions/CLMaxUnpoolingLayer.cpp"
+ ]
+ }
+ },
+ "MeanStdDevNormalize": {
+ "deps": [ "Reduction" ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLMeanStdDevNormalizationKernel.cpp",
+ "src/runtime/CL/functions/CLMeanStdDevNormalizationLayer.cpp",
+ "src/runtime/CL/functions/CLReduceMean.cpp"
+ ]
+ }
+ },
+ "Mul": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClMulKernel.cpp",
+ "src/gpu/cl/operators/ClMul.cpp",
+ "src/runtime/CL/functions/CLPixelWiseMultiplication.cpp"
+ ]
+ }
+ },
+ "Normalize": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLNormalizationLayerKernel.cpp",
+ "src/runtime/CL/functions/CLNormalizationLayer.cpp"
+ ]
+ }
+ },
+ "Pad": {
+ "deps": [ "Copy" ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLPadLayerKernel.cpp",
+ "src/runtime/CL/functions/CLPadLayer.cpp"
+ ]
+ }
+ },
+ "Permute": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClPermuteKernel.cpp",
+ "src/gpu/cl/operators/ClPermute.cpp",
+ "src/runtime/CL/functions/CLPermute.cpp"
+ ]
+ }
+ },
+ "Pool2d": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClPool2dKernel.cpp",
+ "src/gpu/cl/operators/ClPool2d.cpp",
+ "src/runtime/CL/functions/CLPoolingLayer.cpp"
+ ]
+ }
+ },
+ "PRelu": {
+ "deps": [ "ElementwiseBinary" ],
+ "files": {
+ "common": [
+ "src/gpu/cl/operators/ClPRelu.cpp",
+ "src/runtime/CL/functions/CLPReluLayer.cpp"
+ ]
+ }
+ },
+ "PriorBox": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLPriorBoxLayerKernel.cpp",
+ "src/runtime/CL/functions/CLPriorBoxLayer.cpp"
+ ]
+ }
+ },
+ "Quantize": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClQuantizeKernel.cpp",
+ "src/gpu/cl/operators/ClQuantize.cpp",
+ "src/runtime/CL/functions/CLQuantizationLayer.cpp"
+ ]
+ }
+ },
+ "Range": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLRangeKernel.cpp",
+ "src/runtime/CL/functions/CLRange.cpp"
+ ]
+ }
+ },
+ "Reduction": {
+ "deps": [ "Reshape" ],
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLReductionOperationKernel.cpp",
+ "src/runtime/CL/functions/CLReductionOperation.cpp"
+ ]
+ }
+ },
+ "Remap": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLRemapKernel.cpp",
+ "src/runtime/CL/functions/CLRemap.cpp"]
+ }
+ },
+ "Reorg": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLReorgLayerKernel.cpp",
+ "src/runtime/CL/functions/CLReorgLayer.cpp"
+ ]
+ }
+ },
+ "Reshape": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClReshapeKernel.cpp",
+ "src/gpu/cl/operators/ClReshape.cpp",
+ "src/runtime/CL/functions/CLReshapeLayer.cpp"
+ ]
+ }
+ },
+ "Reverse": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLReverseKernel.cpp",
+ "src/runtime/CL/functions/CLReverse.cpp"
+ ]
+ }
+ },
+ "RNN": {
+ "deps": [ "Activation", "Cast", "ElementwiseBinary", "FullyConnected", "Gemm"],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLRNNLayer.cpp" ]
+ }
+ },
+ "ROIAlign": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLROIAlignLayerKernel.cpp",
+ "src/runtime/CL/functions/CLROIAlignLayer.cpp"
+ ]
+ }
+ },
+ "ROIPool2d": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLROIPoolingLayerKernel.cpp",
+ "src/runtime/CL/functions/CLROIPoolingLayer.cpp"
+ ]
+ }
+ },
+ "Scale": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClScaleKernel.cpp",
+ "src/gpu/cl/operators/ClScale.cpp",
+ "src/runtime/CL/functions/CLScale.cpp"
+ ]
+ }
+ },
+ "Select": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLSelectKernel.cpp",
+ "src/runtime/CL/functions/CLSelect.cpp"
+ ]
+ }
+ },
+ "Slice": {
+ "deps": [ "StridedSlice" ],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLSlice.cpp" ]
+ }
+ },
+ "Softmax": {
+ "deps": [ "Permute" ],
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClSoftmaxKernel.cpp",
+ "src/gpu/cl/operators/ClSoftmax.cpp",
+ "src/runtime/CL/functions/CLSoftmaxLayer.cpp"
+ ]
+ }
+ },
+ "SpaceToBatch": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLSpaceToBatchLayerKernel.cpp",
+ "src/runtime/CL/functions/CLSpaceToBatchLayer.cpp"
+ ]
+ }
+ },
+ "SpaceToDepth": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLSpaceToDepthLayerKernel.cpp",
+ "src/runtime/CL/functions/CLSpaceToDepthLayer.cpp"
+ ]
+ }
+ },
+ "Split": {
+ "deps": [ "StridedSlice" ],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLSplit.cpp" ]
+ }
+ },
+ "Stack": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLStackLayerKernel.cpp",
+ "src/runtime/CL/functions/CLStackLayer.cpp"
+ ]
+ }
+ },
+ "StridedSlice": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLStridedSliceKernel.cpp",
+ "src/runtime/CL/functions/CLStridedSlice.cpp"
+ ]
+ }
+ },
+ "Sub": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClElementwiseKernel.cpp",
+ "src/gpu/cl/operators/ClSub.cpp"
+ ]
+ }
+ },
+ "Tile": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLTileKernel.cpp",
+ "src/runtime/CL/functions/CLTile.cpp"
+ ]
+ }
+ },
+ "Transpose": {
+ "files": {
+ "common": [
+ "src/gpu/cl/kernels/ClTransposeKernel.cpp",
+ "src/gpu/cl/operators/ClTranspose.cpp",
+ "src/runtime/CL/functions/CLTranspose.cpp"
+ ]
+ }
+ },
+ "Unstack": {
+ "deps": [ "StridedSlice" ],
+ "files": {
+ "common": [ "src/runtime/CL/functions/CLUnstack.cpp" ]
+ }
+ },
+ "YUVNormalize": {
+ "files": {
+ "common": [
+ "src/core/CL/kernels/CLNormalizePlanarYUVLayerKernel.cpp",
+ "src/runtime/CL/functions/CLNormalizePlanarYUVLayer.cpp"
+ ]
}
}
- },
+ }
+},
"cpu": {
"common": [
"src/cpu/CpuContext.cpp",
"src/cpu/CpuQueue.cpp",
- "src/cpu/CpuTensor.cpp"
- ],
- "high_priority": [
- "Activation",
- "DepthwiseConv2d",
- "DirectConv2d",
- "Permute",
- "Pool2d",
- "Reshape",
- "FillBorder"
+ "src/cpu/CpuTensor.cpp",
+ "src/core/NEON/kernels/NEFillBorderKernel.cpp",
+ "src/runtime/NEON/INEOperator.cpp",
+ "src/runtime/NEON/INESimpleFunction.cpp",
+ "src/runtime/NEON/INESimpleFunctionNoBorder.cpp"
],
"operators": {
"Activation": {
"files": {
- "operator": [
- "src/cpu/operators/CpuActivation.cpp"
+ "common": [
+ "src/cpu/operators/CpuActivation.cpp",
+ "src/cpu/kernels/CpuActivationKernel.cpp",
+ "src/runtime/NEON/functions/NEActivationLayer.cpp"
],
- "kernel": [
- "src/cpu/kernels/CpuActivationKernel.cpp"
- ],
- "sve": {
- "fp32": [
- "src/cpu/kernels/activation/sve/fp32.cpp"
- ],
- "fp16": [
- "src/cpu/kernels/activation/sve/fp16.cpp"
- ],
- "qsymm16": [
- "src/cpu/kernels/activation/sve/qsymm16.cpp"
- ],
- "qasymm8": [
- "src/cpu/kernels/activation/sve/qasymm8.cpp"
- ],
- "qasymm8_signed": [
- "src/cpu/kernels/activation/sve/qasymm8_signed.cpp"
- ]
- },
"neon": {
- "fp32": [
- "src/cpu/kernels/activation/neon/fp32.cpp"
- ],
- "fp16": [
- "src/cpu/kernels/activation/neon/fp16.cpp"
- ],
- "qsymm16": [
- "src/cpu/kernels/activation/neon/qsymm16.cpp"
- ],
- "qasymm8": [
- "src/cpu/kernels/activation/neon/qasymm8.cpp"
- ],
- "qasymm8_signed": [
- "src/cpu/kernels/activation/neon/qasymm8_signed.cpp"
- ]
+ "fp16": [ "src/cpu/kernels/activation/neon/fp16.cpp" ],
+ "fp32": [ "src/cpu/kernels/activation/neon/fp32.cpp" ],
+ "qasymm8": [ "src/cpu/kernels/activation/neon/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/activation/neon/qasymm8_signed.cpp" ],
+ "qsymm16": [ "src/cpu/kernels/activation/neon/qsymm16.cpp" ]
+ },
+ "sve": {
+ "fp16": [ "src/cpu/kernels/activation/sve/fp16.cpp" ],
+ "fp32": [ "src/cpu/kernels/activation/sve/fp32.cpp" ],
+ "qasymm8": [ "src/cpu/kernels/activation/neon/qasymm8.cpp", "src/cpu/kernels/activation/sve/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/activation/neon/qasymm8_signed.cpp", "src/cpu/kernels/activation/sve/qasymm8_signed.cpp" ],
+ "qsymm16": [ "src/cpu/kernels/activation/neon/qsymm16.cpp", "src/cpu/kernels/activation/sve/qsymm16.cpp" ]
}
}
},
+ "ArgMinMax": {
+ "deps": [ "Reduction" ],
+ "files": {
+ "common": [ "src/runtime/NEON/functions/NEArgMinMaxLayer.cpp" ]
+ }
+ },
"Add": {
"files": {
- "operator": [
- "src/cpu/operators/CpuAdd.cpp"
+ "common": [
+ "src/cpu/operators/CpuAdd.cpp",
+ "src/cpu/kernels/CpuAddKernel.cpp",
+ "src/runtime/NEON/functions/NEArithmeticAddition.cpp"
],
- "kernel": [
- "src/cpu/kernels/CpuAddKernel.cpp"
- ],
- "sve": {
- "all": [
- "src/cpu/kernels/add/sve/impl.cpp"
- ],
- "qsymm16": [
- "src/cpu/kernels/add/sve/qsymm16.cpp"
- ],
- "qasymm8": [
- "src/cpu/kernels/add/sve/qasymm8.cpp"
- ],
- "qasymm8_signed": [
- "src/cpu/kernels/add/sve/qasymm8_signed.cpp"
- ]
- },
"neon": {
- "qsymm16": [
- "src/cpu/kernels/add/neon/qsymm16.cpp"
- ],
- "qasymm8": [
- "src/cpu/kernels/add/neon/qasymm8.cpp"
- ],
- "qasymm8_signed": [
- "src/cpu/kernels/add/neon/qasymm8_signed.cpp"
- ]
+ "qasymm8": [ "src/cpu/kernels/add/neon/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/add/neon/qasymm8_signed.cpp" ],
+ "qsymm16": [ "src/cpu/kernels/add/neon/qsymm16.cpp" ]
+ },
+ "sve": {
+ "common": [ "src/cpu/kernels/add/sve/impl.cpp" ],
+ "qasymm8": [ "src/cpu/kernels/add/neon/qasymm8.cpp", "src/cpu/kernels/add/sve/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/add/neon/qasymm8_signed.cpp", "src/cpu/kernels/add/sve/qasymm8_signed.cpp" ],
+ "qsymm16": [ "src/cpu/kernels/add/neon/qsymm16.cpp", "src/cpu/kernels/add/sve/qsymm16.cpp" ]
}
}
},
- "BatchNorm": {
+ "BatchNormalize": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEBatchNormalizationLayerKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEBatchNormalizationLayerKernel.cpp",
+ "src/core/NEON/kernels/NEFuseBatchNormalizationKernel.cpp",
+ "src/runtime/NEON/functions/NEBatchNormalizationLayer.cpp",
+ "src/runtime/NEON/functions/NEFuseBatchNormalization.cpp"
],
- "sve": {
- "fp32": [
- "src/core/NEON/kernels/batchnormalization/impl/SVE/fp32.cpp"
- ],
- "fp16": [
- "src/core/NEON/kernels/batchnormalization/impl/SVE/fp16.cpp"
- ]
- },
"neon": {
- "fp32": [
- "src/core/NEON/kernels/batchnormalization/impl/NEON/fp32.cpp"
- ],
- "fp16": [
- "src/core/NEON/kernels/batchnormalization/impl/NEON/fp16.cpp"
- ]
+ "fp16": [ "src/core/NEON/kernels/batchnormalization/impl/NEON/fp16.cpp" ],
+ "fp32": [ "src/core/NEON/kernels/batchnormalization/impl/NEON/fp32.cpp" ]
+ },
+ "sve": {
+ "fp16": [ "src/core/NEON/kernels/batchnormalization/impl/SVE/fp16.cpp" ],
+ "fp32": [ "src/core/NEON/kernels/batchnormalization/impl/SVE/fp32.cpp" ]
}
}
},
"BatchToSpace": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEBatchToSpaceLayerKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEBatchToSpaceLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEBatchToSpaceLayer.cpp"
]
}
},
"BitwiseAnd": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEBitwiseAndKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEBitwiseAndKernel.cpp",
+ "src/runtime/NEON/functions/NEBitwiseAnd.cpp"
]
}
},
"BitwiseNot": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEBitwiseNotKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEBitwiseNotKernel.cpp",
+ "src/runtime/NEON/functions/NEBitwiseNot.cpp"
]
}
},
"BitwiseOr": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEBitwiseOrKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEBitwiseOrKernel.cpp",
+ "src/runtime/NEON/functions/NEBitwiseOr.cpp"
]
}
},
"BitwiseXor": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEBitwiseXorKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEBitwiseXorKernel.cpp",
+ "src/runtime/NEON/functions/NEBitwiseXor.cpp"
]
}
},
"BoundingBoxTransform": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEBoundingBoxTransformKernel.cpp"
- ]
- }
- },
- "ChannelShuffleLayer": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEChannelShuffleLayerKernel.cpp"
- ]
- }
- },
- "Col2Im": {
- "files": {
- "kernel": [
- "src/cpu/kernels/CpuCol2ImKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEBoundingBoxTransformKernel.cpp",
+ "src/runtime/NEON/functions/NEBoundingBoxTransform.cpp"
]
}
},
"Cast": {
"files": {
- "operator": [
- "src/cpu/operators/CpuCast.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuCastKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuCast.cpp",
+ "src/cpu/kernels/CpuCastKernel.cpp",
+ "src/runtime/NEON/functions/NECast.cpp"
+ ]
+ }
+ },
+ "ChannelShuffle": {
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NEChannelShuffleLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEChannelShuffleLayer.cpp"
]
}
},
"Concatenate": {
"files": {
- "operator": [
- "src/cpu/operators/CpuConcatenate.cpp"
- ],
- "kernel": [
+ "common": [
+ "src/cpu/operators/CpuConcatenate.cpp",
"src/cpu/kernels/CpuConcatenateWidthKernel.cpp",
"src/cpu/kernels/CpuConcatenateBatchKernel.cpp",
"src/cpu/kernels/CpuConcatenateDepthKernel.cpp",
- "src/cpu/kernels/CpuConcatenateHeightKernel.cpp"
+ "src/cpu/kernels/CpuConcatenateHeightKernel.cpp",
+ "src/runtime/NEON/functions/NEConcatenateLayer.cpp"
]
}
},
- "ConvertFullyConnectedWeights": {
+ "Conv2d": {
+ "deps": [
+ "Activation",
+ "ElementwiseBinary",
+ "FFT2D",
+ "Gemm",
+ "Mul",
+ "Pad",
+ "Permute",
+ "Reshape",
+ "Reverse",
+ "Slice"
+ ],
"files": {
- "operator": [
- "src/cpu/operators/CpuConvertFullyConnectedWeights.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuConvertFullyConnectedWeightsKernel.cpp"
- ]
- }
- },
- "ConvertQuantizedSignedness": {
- "files": {
- "kernel": [
- "src/cpu/kernels/CpuConvertQuantizedSignednessKernel.cpp"
- ]
- }
- },
- "Convolution": {
- "files": {
- "operator": [
- "src/cpu/operators/CpuConv2d.cpp"
+ "common": [
+ "src/cpu/operators/CpuConv2d.cpp",
+ "src/cpu/operators/CpuDirectConv2d.cpp",
+ "src/cpu/operators/CpuGemmDirectConv2d.cpp",
+ "src/cpu/operators/CpuGemmConv2d.cpp",
+ "src/cpu/operators/CpuWinogradConv2d.cpp",
+ "src/cpu/operators/internal/CpuGemmAssemblyDispatch.cpp",
+ "src/cpu/kernels/CpuDirectConv2dKernel.cpp",
+ "src/cpu/kernels/CpuDirectConv2dOutputStageKernel.cpp",
+ "src/cpu/kernels/CpuWinogradConv2dKernel.cpp",
+ "src/cpu/kernels/CpuCol2ImKernel.cpp",
+ "src/cpu/kernels/CpuIm2ColKernel.cpp",
+ "src/cpu/kernels/CpuWeightsReshapeKernel.cpp",
+ "src/core/NEON/kernels/convolution/common/padding.cpp",
+ "src/core/NEON/kernels/convolution/common/qasymm8.cpp",
+ "src/core/NEON/kernels/convolution/common/qsymm8.cpp",
+ "src/core/NEON/kernels/convolution/common/utils.cpp",
+ "src/core/NEON/kernels/convolution/winograd/padding.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_1x8_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_4x4_fp16_fp16_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_4x4_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_6x6_fp16_fp16_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_6x6_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_2_7_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_2x2_3x3_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_2x2_5x5_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_4_5_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_4x4_3x3_fp16_fp16_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_4x4_3x3_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_6_3_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_2_7_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_2x2_3x3_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_2x2_5x5_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_4_5_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_4x4_3x3_fp16_fp16_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_4x4_3x3_fp32_fp32_integers.cpp",
+ "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_6_3_fp32_fp32_integers.cpp",
+ "src/runtime/NEON/functions/NEConvolutionLayer.cpp",
+ "src/runtime/NEON/functions/NEDirectConvolutionLayer.cpp",
+ "src/runtime/NEON/functions/NEFFTConvolutionLayer.cpp",
+ "src/runtime/NEON/functions/NEGEMMConv2d.cpp",
+ "src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp",
+ "src/runtime/NEON/functions/NEWinogradConvolutionLayer.cpp"
]
}
},
"Copy": {
"files": {
- "operator": [
- "src/cpu/operators/CpuCopy.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuCopyKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuCopy.cpp",
+ "src/cpu/kernels/CpuCopyKernel.cpp",
+ "src/runtime/NEON/functions/NECopy.cpp"
]
}
},
- "Crop": {
+ "CropResize": {
+ "deps": [ "Scale" ],
"files": {
- "kernel": [
- "src/core/NEON/kernels/NECropKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NECropKernel.cpp",
+ "src/runtime/NEON/functions/NECropResize.cpp"
+ ]
+ }
+ },
+ "Deconv2d": {
+ "deps": [ "Conv2d", "Reverse", "Transpose"],
+ "files": {
+ "common": [
+ "src/runtime/NEON/functions/NEDeconvolutionLayer.cpp"
+ ]
+ }
+ },
+ "DepthConvert": {
+ "deps": [ "Cast"],
+ "files": {
+ "common": [
+ "src/runtime/NEON/functions/NEDepthConvertLayer.cpp"
+ ]
+ }
+ },
+ "DepthToSpace": {
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NEDepthToSpaceLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEDepthToSpaceLayer.cpp"
]
}
},
"DepthwiseConv2d": {
- "deps": [
- "Activation",
- "Permute"
- ],
+ "deps": [ "Activation", "Permute" ],
"files": {
- "operator": [
+ "common": [
"src/cpu/operators/CpuDepthwiseConv2d.cpp",
"src/cpu/operators/CpuDepthwiseConv2dAssemblyDispatch.cpp",
- "src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp"
- ],
- "kernel": [
+ "src/cpu/kernels/CpuDepthwiseConv2dNativeKernel.cpp",
+ "src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp",
"src/core/NEON/kernels/convolution/common/padding.cpp",
"src/core/NEON/kernels/convolution/common/qasymm8.cpp",
"src/core/NEON/kernels/convolution/common/qsymm8.cpp",
"src/core/NEON/kernels/convolution/common/utils.cpp",
- "src/cpu/kernels/CpuDepthwiseConv2dNativeKernel.cpp"
+ "src/core/NEON/kernels/arm_conv/depthwise/depthwise_fp16.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/depthwise_u8q.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/depthwise_u8s8u8q.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/interleaves/8b_mla.cpp",
+ "src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp"
],
+ "neon": {
+ "estate64": [
+ "src/core/NEON/kernels/arm_conv/depthwise/depthwise_s8q.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/depthwise_fp32.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/interleaves/a64_s8q_3x3_dot.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/interleaves/a64_u8q_3x3_dot.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s1_output4x4_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s1_output4x4_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s2_output2x2_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s2_output2x2_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_5x5_s1_output2x2_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_5x5_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_generic_output9_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_packed_to_nhwc_generic_with_multiplier_output2x8_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_3x3_s1_output3x3_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_3x3_s1_output4x4_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_3x3_s1_output4x4_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_3x3_s2_output2x2_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_3x3_s2_output2x2_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_5x5_s1_output2x2_mla_depthfirst/generic_direct.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_5x5_s1_output2x2_mla_depthfirst/generic_indirect.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_nhwc_generic_output9_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_packed_to_nhwc_3x3_s2_with_multiplier_output3x3_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_packed_to_nhwc_5x5_s1_with_multiplier_output2x4_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp32_packed_to_nhwc_generic_with_multiplier_output2x8_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8q_nhwc_3x3_s1_output2x2_dot_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8q_nhwc_3x3_s1_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8q_nhwc_3x3_s2_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8q_nhwc_5x5_s1_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8q_nhwc_generic_output9_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8q_packed_to_nhwc_3x3_s2_with_multiplier_output2x4_dot_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8q_packed_to_nhwc_5x5_s1_with_multiplier_output4x2_dot_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8q_packed_to_nhwc_generic_with_multiplier_output2x8_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_s8qs_nhwc_3x3_s1_output2x2_dot_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8q_nhwc_3x3_s1_output2x2_dot_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8q_nhwc_3x3_s1_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8q_nhwc_3x3_s2_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8q_nhwc_5x5_s1_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8q_nhwc_generic_output9_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8q_packed_to_nhwc_3x3_s2_with_multiplier_output2x4_dot_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8q_packed_to_nhwc_5x5_s1_with_multiplier_output4x2_dot_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8q_packed_to_nhwc_generic_with_multiplier_output2x8_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8s8u8q_nhwc_3x3_s1_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8s8u8q_nhwc_3x3_s2_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8s8u8q_nhwc_5x5_s1_output2x2_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8s8u8q_nhwc_generic_output9_mla_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_u8s8u8q_packed_to_nhwc_generic_with_multiplier_output2x8_mla_depthfirst/generic.cpp"
+ ]
+ },
"sve": {
- "all": [
+ "common": [
+ "src/core/NEON/kernels/arm_conv/depthwise/depthwise_s8q.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/depthwise_fp32.cpp",
"src/core/NEON/kernels/arm_conv/depthwise/interleaves/sve_8b_mla.cpp",
"src/core/NEON/kernels/arm_conv/depthwise/interleaves/sve_s8q_3x3_dot.cpp",
"src/core/NEON/kernels/arm_conv/depthwise/interleaves/sve_u8q_3x3_dot.cpp",
@@ -994,17 +1200,7 @@
"src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_u8q_packed_to_nhwc_5x5_s1_with_multiplier_output4x2_dot_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_u8s8u8q_nhwc_3x3_s1_output2x2_mla_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_u8s8u8q_nhwc_3x3_s2_output2x2_mla_depthfirst/generic.cpp",
- "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_u8s8u8q_nhwc_5x5_s1_output2x2_mla_depthfirst/generic.cpp"
- ]
- },
- "neon": {
- "estate64": [
- "src/core/NEON/kernels/arm_conv/depthwise/depthwise_fp16.cpp",
- "src/core/NEON/kernels/arm_conv/depthwise/depthwise_fp32.cpp",
- "src/core/NEON/kernels/arm_conv/depthwise/depthwise_s8q.cpp",
- "src/core/NEON/kernels/arm_conv/depthwise/depthwise_u8q.cpp",
- "src/core/NEON/kernels/arm_conv/depthwise/depthwise_u8s8u8q.cpp",
- "src/core/NEON/kernels/arm_conv/depthwise/interleaves/8b_mla.cpp",
+ "src/core/NEON/kernels/arm_conv/depthwise/kernels/sve_u8s8u8q_nhwc_5x5_s1_output2x2_mla_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/depthwise/interleaves/a64_s8q_3x3_dot.cpp",
"src/core/NEON/kernels/arm_conv/depthwise/interleaves/a64_u8q_3x3_dot.cpp",
"src/core/NEON/kernels/arm_conv/depthwise/kernels/a64_fp16_nhwc_3x3_s1_output2x2_mla_depthfirst/generic_direct.cpp",
@@ -1059,166 +1255,122 @@
}
}
},
- "DepthToSpaceLayer": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEDepthToSpaceLayerKernel.cpp"
- ]
- }
- },
"Dequantize": {
"files": {
- "operator": [
- "src/cpu/operators/CpuDequantize.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuDequantizeKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuDequantize.cpp",
+ "src/cpu/kernels/CpuDequantizeKernel.cpp",
+ "src/runtime/NEON/functions/NEDequantizationLayer.cpp"
]
}
},
- "DirectConv2d": {
- "deps": [
- "Activation",
- "FillBorder"
- ],
+ "DetectionPostProcess": {
+ "deps": [ "Dequantize" ],
"files": {
- "operator": [
- "src/cpu/operators/CpuDirectConv2d.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuDirectConv2dKernel.cpp",
- "src/cpu/kernels/CpuDirectConv2dOutputStageKernel.cpp"
- ]
+ "common" : [ "src/runtime/NEON/functions/NEDetectionPostProcessLayer.cpp" ]
}
},
- "Elementwise": {
+ "ElementwiseBinary": {
"files": {
- "operator": [
- "src/cpu/operators/CpuElementwise.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuElementwiseKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuElementwise.cpp",
+ "src/cpu/kernels/CpuElementwiseKernel.cpp",
+ "src/runtime/NEON/functions/NEElementwiseOperations.cpp"
],
"sve": {
- "all": [
- "src/cpu/kernels/elementwise/sve/elementwise.cpp"
- ]
+ "common": [ "src/cpu/kernels/elementwise/sve/elementwise.cpp" ]
}
}
},
- "ElementwiseUnary": {
+ "ElementwiseUnary":{
"files": {
- "operator": [
- "src/cpu/operators/CpuElementwiseUnary.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuElementwiseUnaryKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuElementwiseUnary.cpp",
+ "src/cpu/kernels/CpuElementwiseUnaryKernel.cpp",
+ "src/runtime/NEON/functions/NEElementwiseUnaryLayer.cpp"
],
"sve": {
- "all": [
- "src/cpu/kernels/elementwise/sve/elementwise_unary.cpp"
- ]
+ "common": [ "src/cpu/kernels/elementwise/sve/elementwise_unary.cpp" ]
}
}
},
"FFT1D": {
"files": {
- "kernel": [
+ "common": [
"src/core/NEON/kernels/NEFFTDigitReverseKernel.cpp",
"src/core/NEON/kernels/NEFFTRadixStageKernel.cpp",
- "src/core/NEON/kernels/NEFFTScaleKernel.cpp"
+ "src/core/NEON/kernels/NEFFTScaleKernel.cpp",
+ "src/runtime/NEON/functions/NEFFT1D.cpp"
]
}
},
- "FillBorder": {
+ "FFT2D": {
+ "deps": [ "FFT1D" ],
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEFillBorderKernel.cpp"
- ]
- }
- },
- "Flatten": {
- "deps: ": [
- "Reshape"
- ],
- "files": {
- "operator": [
- "src/cpu/operators/CpuFlatten.cpp"
+ "common": [
+ "src/runtime/NEON/functions/NEFFT2D.cpp"
]
}
},
"Fill": {
"files": {
- "operator": [
- "src/cpu/operators/CpuFill.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuFillKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuFill.cpp",
+ "src/cpu/kernels/CpuFillKernel.cpp",
+ "src/runtime/NEON/functions/NEFill.cpp"
+ ]
+ }
+ },
+ "Flatten": {
+ "deps: ": [ "Reshape" ],
+ "files": {
+ "common": [
+ "src/cpu/operators/CpuFlatten.cpp",
+ "src/runtime/NEON/functions/NEFlattenLayer.cpp"
]
}
},
"Floor": {
"files": {
- "operator": [
- "src/cpu/operators/CpuFloor.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuFloorKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuFloor.cpp",
+ "src/cpu/kernels/CpuFloorKernel.cpp",
+ "src/runtime/NEON/functions/NEFloor.cpp"
],
"neon": {
- "fp32": [
- "src/cpu/kernels/floor/neon/fp32.cpp"
- ],
- "fp16": [
- "src/cpu/kernels/floor/neon/fp16.cpp"
- ]
+ "fp32": [ "src/cpu/kernels/floor/neon/fp32.cpp" ],
+ "fp16": [ "src/cpu/kernels/floor/neon/fp16.cpp" ]
}
}
},
"FullyConnected": {
- "deps": [
- "CpuFlatten",
- "CpuConvertFullyConnectedWeights",
- "CpuGemm",
- "CpuGemmLowpMatrixMultiplyCore"
- ],
+ "deps": [ "Flatten", "Gemm", "Transpose"],
"files": {
- "operator": [
- "src/cpu/operators/CpuFullyConnected.cpp"
- ]
- },
- "kernel": [
- "CpuTransposeKernel"
- ]
- },
- "FuseBatchNormalization": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEFuseBatchNormalizationKernel.cpp"
+ "common": [
+ "src/cpu/kernels/CpuConvertFullyConnectedWeightsKernel.cpp",
+ "src/cpu/operators/CpuConvertFullyConnectedWeights.cpp",
+ "src/cpu/operators/CpuFullyConnected.cpp",
+ "src/runtime/NEON/functions/NEConvertFullyConnectedWeights.cpp",
+ "src/runtime/NEON/functions/NEFullyConnectedLayer.cpp"
]
}
},
- "GEMM": {
+ "Gather": {
"files": {
- "operator" : ["src/cpu/operators/CpuGemm.cpp"],
- "kernel": [
+ "common": [
+ "src/core/NEON/kernels/NEGatherKernel.cpp",
+ "src/runtime/NEON/functions/NEGather.cpp"
+ ]
+ }
+ },
+ "Gemm": {
+ "files": {
+ "common": [
+ "src/cpu/kernels/CpuConvertQuantizedSignednessKernel.cpp",
"src/cpu/kernels/CpuGemmMatrixAdditionKernel.cpp",
"src/cpu/kernels/CpuGemmMatrixMultiplyKernel.cpp",
"src/cpu/kernels/CpuGemmTranspose1xWKernel.cpp",
- "src/cpu/kernels/CpuGemmInterleave4x4Kernel.cpp"
- ]
- }
- },
- "GEMMLowp": {
- "deps": [
- "GemmAssemblyDispatch"
- ],
- "files": {
- "operator" : [
- "src/cpu/operators/CpuGemmLowpOutputStage.cpp",
- "src/cpu/operators/CpuGemmLowpMatrixMultiplyCore.cpp"
- ],
- "kernel": [
+ "src/cpu/kernels/CpuGemmInterleave4x4Kernel.cpp",
"src/cpu/kernels/CpuGemmLowpQuantizeDownInt32ScaleKernel.cpp",
"src/cpu/kernels/CpuGemmLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.cpp",
"src/cpu/kernels/CpuGemmLowpQuantizeDownInt32ToInt8ScaleByFixedPointKernel.cpp",
@@ -1226,36 +1378,12 @@
"src/cpu/kernels/CpuGemmLowpMatrixMultiplyKernel.cpp",
"src/cpu/kernels/CpuGemmLowpMatrixReductionKernel.cpp",
"src/cpu/kernels/CpuGemmLowpOffsetContributionOutputStageKernel.cpp",
- "src/cpu/kernels/CpuGemmLowpOffsetContributionKernel.cpp"
- ]
- }
- },
- "GEMMConvolution": {
- "deps": [
- "Activation",
- "Col2Im",
- "Reshape",
- "Im2Col",
- "GEMMLowpOffsetContributionOutputStage",
- "ConvertQuantizedSignedness"
- ],
- "files": {
- "operator": [
- "src/cpu/operators/CpuGemmConv2d.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuWeightsReshapeKernel.cpp"
- ]
- }
- },
- "GemmAssemblyDispatch": {
- "files": {
- "operator": [
- "src/cpu/operators/internal/CpuGemmAssemblyDispatch.cpp"
- ],
- "kernel": [
- "src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp",
+ "src/cpu/kernels/CpuGemmLowpOffsetContributionKernel.cpp",
+ "src/cpu/operators/CpuGemm.cpp",
+ "src/cpu/operators/CpuGemmLowpOutputStage.cpp",
+ "src/cpu/operators/CpuGemmLowpMatrixMultiplyCore.cpp",
"src/core/NEON/kernels/arm_gemm/gemm_bf16.cpp",
+ "src/core/NEON/kernels/arm_gemm/gemm_fp16.cpp",
"src/core/NEON/kernels/arm_gemm/gemm_fp32.cpp",
"src/core/NEON/kernels/arm_gemm/gemm_int16.cpp",
"src/core/NEON/kernels/arm_gemm/gemm_int8.cpp",
@@ -1263,14 +1391,17 @@
"src/core/NEON/kernels/arm_gemm/gemm_quint8.cpp",
"src/core/NEON/kernels/arm_gemm/gemm_uint16.cpp",
"src/core/NEON/kernels/arm_gemm/gemm_uint8.cpp",
+ "src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp",
"src/core/NEON/kernels/arm_gemm/mergeresults-fp16.cpp",
"src/core/NEON/kernels/arm_gemm/mergeresults.cpp",
- "src/core/NEON/kernels/arm_gemm/interleave_indirect.cpp",
"src/core/NEON/kernels/arm_gemm/misc.cpp",
"src/core/NEON/kernels/arm_gemm/quantized.cpp",
"src/core/NEON/kernels/arm_gemm/rowsum_indirect_s8.cpp",
"src/core/NEON/kernels/arm_gemm/rowsum_indirect_u8.cpp",
- "src/core/NEON/kernels/arm_gemm/transform.cpp"
+ "src/core/NEON/kernels/arm_gemm/transform.cpp",
+ "src/runtime/NEON/functions/NEGEMM.cpp",
+ "src/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.cpp",
+ "src/runtime/NEON/functions/NEGEMMLowpOutputStage.cpp"
],
"neon": {
"estate32": [
@@ -1344,7 +1475,7 @@
]
},
"sve": {
- "all": [
+ "common": [
"src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_bf16fp32_dot_6x4VL/generic.cpp",
"src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_bf16fp32_mmla_6x4VL/generic.cpp",
"src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_fp16_mla_6x4VL/a64fx.cpp",
@@ -1384,152 +1515,196 @@
"src/core/NEON/kernels/arm_gemm/kernels/sve_smallK_hybrid_s8s32_dot_8x1VL/generic.cpp",
"src/core/NEON/kernels/arm_gemm/kernels/sve_smallK_hybrid_u8u32_dot_8x1VL/generic.cpp",
"src/core/NEON/kernels/arm_gemm/mergeresults-sve.cpp",
- "src/core/NEON/kernels/arm_gemm/transform-sve.cpp"
+ "src/core/NEON/kernels/arm_gemm/transform-sve.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s16_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s8_4x4/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s8_8x12/a55r1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s8_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s8_8x12/x1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u16_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_8x12/a55r1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_8x12/x1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_8x24/a55r1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_8x24/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_8x24/x1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_bf16fp32_dot_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_bf16fp32_mmla_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp16_mla_6x32/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp16_mla_6x32/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_4x24/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_4x24/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_6x16/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_8x4/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_8x4/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32bf16fp32_mmla_4x24/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32bf16fp32_mmla_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8qa_dot_4x16/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8qa_dot_4x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8qa_mmla_4x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8qs_dot_6x16/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8qs_dot_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8qs_mmla_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8s32_dot_6x16/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8s32_dot_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8s32_mmla_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8qa_dot_4x16/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8qa_dot_4x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8qa_mmla_4x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8u32_dot_6x16/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8u32_dot_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8u32_mmla_6x16/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_dot_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_dot_8x12/x1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_mmla_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_u8u32_mmla_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_8x12/a53.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_8x12/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_8x12/a55r1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_8x12/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_8x12/x1.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_8x6/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemv_pretransposed/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_fp32_mla_6x4/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_fp32_mla_8x4/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_s8s32_dot_6x4/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_s8s32_dot_6x4/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_s8s32_dot_8x4/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_s8s32_dot_8x4/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_u8u32_dot_6x4/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_u8u32_dot_6x4/generic.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_u8u32_dot_8x4/a55.cpp",
+ "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_u8u32_dot_8x4/generic.cpp"
]
}
}
},
- "GemmDirectConv2d": {
- "deps": [
- "Activation",
- "GemmAssemblyDispatch",
- "Permute"
- ],
+ "GenerateProposals": {
+ "deps": [ "BoundingBoxTransform", "Dequantize", "Pad", "Permute", "Quantize", "Reshape" ],
"files": {
- "operator": [
- "src/cpu/operators/CpuGemmDirectConv2d.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEGenerateProposalsLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEGenerateProposalsLayer.cpp"
]
}
},
- "Mul": {
+ "InstanceNormalize": {
+ "deps": [ "Permute", "Reduction" ],
"files": {
- "operator": [
- "src/cpu/operators/CpuMul.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuMulKernel.cpp"
- ]
- }
- },
- "Quantize": {
- "files": {
- "operator": [
- "src/cpu/operators/CpuQuantize.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuQuantizeKernel.cpp"
- ]
- }
- },
- "Reshape": {
- "files": {
- "operator": [
- "src/cpu/operators/CpuReshape.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuReshapeKernel.cpp"
- ]
- }
- },
- "Gather": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEGatherKernel.cpp"
- ]
- }
- },
- "GenerateProposalsLayer": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEGenerateProposalsLayerKernel.cpp"
- ]
- }
- },
- "Im2Col": {
- "files": {
- "kernel": [
- "src/cpu/kernels/CpuIm2ColKernel.cpp"
- ]
- }
- },
- "InstanceNormalization": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEInstanceNormalizationLayerKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEInstanceNormalizationLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEInstanceNormalizationLayer.cpp"
]
}
},
"L2Normalize": {
- "deps": [
- "Reduction"
- ],
+ "deps": [ "Reduction" ],
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEL2NormalizeLayerKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEL2NormalizeLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEL2NormalizeLayer.cpp"
]
}
},
"Logical": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NELogicalKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NELogicalKernel.cpp",
+ "src/runtime/NEON/functions/NELogical.cpp"
]
}
},
- "MaxUnpooling": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEMaxUnpoolingLayerKernel.cpp"
- ]
- }
- },
- "MeanStdDevNormalization": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEMeanStdDevNormalizationKernel.cpp"
- ]
- }
- },
- "MinMax": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEMinMaxLayerKernel.cpp"
- ]
- }
- },
- "Normalization": {
+ "LSTM": {
"deps": [
- "PixelWiseMultiplication"
+ "Activation",
+ "Concatenate",
+ "Copy",
+ "Dequantize",
+ "ElementwiseBinary",
+ "Fill",
+ "FullyConnected",
+ "Gemm",
+ "MeanStdDevNormalize",
+ "Mul",
+ "Quantize",
+ "Slice",
+ "Transpose"
],
"files": {
- "kernel": [
- "src/core/NEON/kernels/NENormalizationLayerKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEQLSTMLayerNormalizationKernel.cpp",
+ "src/runtime/NEON/functions/NELSTMLayer.cpp",
+ "src/runtime/NEON/functions/NELSTMLayerQuantized.cpp",
+ "src/runtime/NEON/functions/NEQLSTMLayer.cpp"
+ ]
+ }
+ },
+ "MaxUnpool2d": {
+ "deps": [ "Fill" ],
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NEMaxUnpoolingLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEMaxUnpoolingLayer.cpp"
+ ]
+ }
+ },
+ "Mean": {
+ "deps" : [ "Reduction" ],
+ "files": {
+ "common": [ "src/runtime/NEON/functions/NEReduceMean.cpp" ]
+ }
+ },
+ "MeanStdDevNormalize": {
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NEMeanStdDevNormalizationKernel.cpp",
+ "src/runtime/NEON/functions/NEMeanStdDevNormalizationLayer.cpp"
+ ]
+ }
+ },
+ "Mul": {
+ "files": {
+ "common": [
+ "src/cpu/operators/CpuMul.cpp",
+ "src/cpu/kernels/CpuMulKernel.cpp",
+ "src/runtime/NEON/functions/NEPixelWiseMultiplication.cpp"
+ ]
+ }
+ },
+ "Normalize": {
+ "deps": [ "Mul" ],
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NENormalizationLayerKernel.cpp",
+ "src/runtime/NEON/functions/NENormalizationLayer.cpp"
]
}
},
"Pad": {
+ "deps": [ "Concatenate", "Copy", "StridedSlice" ],
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEPadLayerKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEPadLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEPadLayer.cpp"
]
}
},
"Permute": {
"files": {
- "operator": [
- "src/cpu/operators/CpuPermute.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuPermuteKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuPermute.cpp",
+ "src/cpu/kernels/CpuPermuteKernel.cpp",
+ "src/runtime/NEON/functions/NEPermute.cpp"
]
}
},
"Pool2d": {
"files": {
- "operator": [
- "src/cpu/operators/CpuPool2d.cpp"
- ],
- "kernel": [
+ "common": [
+ "src/cpu/operators/CpuPool2d.cpp",
"src/cpu/kernels/CpuPool2dKernel.cpp",
"src/cpu/kernels/internal/CpuPool2dAssemblyWrapperKernel.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/cpp_nhwc_1x1_stride_any_depthfirst/generic.cpp",
@@ -1538,24 +1713,15 @@
"src/core/NEON/kernels/arm_conv/pooling/pooling_s8.cpp",
"src/core/NEON/kernels/arm_conv/pooling/pooling_s8q.cpp",
"src/core/NEON/kernels/arm_conv/pooling/pooling_u8.cpp",
- "src/core/NEON/kernels/arm_conv/pooling/pooling_u8q.cpp"
+ "src/core/NEON/kernels/arm_conv/pooling/pooling_u8q.cpp",
+ "src/runtime/NEON/functions/NEPoolingLayer.cpp"
],
"neon": {
- "nchw": [
- "src/cpu/kernels/pool2d/neon/nchw/all.cpp"
- ],
- "fp32": [
- "src/cpu/kernels/pool2d/neon/fp32.cpp"
- ],
- "fp16": [
- "src/cpu/kernels/pool2d/neon/fp16.cpp"
- ],
- "qasymm8": [
- "src/cpu/kernels/pool2d/neon/qasymm8.cpp"
- ],
- "qasymm8_signed": [
- "src/cpu/kernels/pool2d/neon/qasymm8_signed.cpp"
- ],
+ "nchw": [ "src/cpu/kernels/pool2d/neon/nchw/all.cpp" ],
+ "fp16": [ "src/cpu/kernels/pool2d/neon/fp16.cpp" ],
+ "fp32": [ "src/cpu/kernels/pool2d/neon/fp32.cpp" ],
+ "qasymm8": [ "src/cpu/kernels/pool2d/neon/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/pool2d/neon/qasymm8_signed.cpp" ],
"estate64": [
"src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_avg_3x3_s1_output2x2_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_avg_generic_depthfirst/generic.cpp",
@@ -1578,15 +1744,17 @@
]
},
"sve": {
- "all": [
- "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_avg_3x3_s1_output2x2_depthfirst/generic.cpp",
- "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_avg_generic_depthfirst/generic.cpp",
- "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
- "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_generic_depthfirst/generic.cpp",
+ "qasymm8": [ "src/cpu/kernels/pool2d/neon/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/pool2d/neon/qasymm8_signed.cpp" ],
+ "common": [
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_avg_3x3_s1_output2x2_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_avg_generic_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp16_nhwc_max_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_avg_3x3_s1_output2x2_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_avg_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_generic_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_s8_nhwc_avg_generic_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_s8_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_s8_nhwc_max_generic_depthfirst/generic.cpp",
@@ -1596,240 +1764,259 @@
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_u8_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_u8_nhwc_max_generic_depthfirst/generic.cpp",
"src/core/NEON/kernels/arm_conv/pooling/kernels/sve_u8q_nhwc_avg_generic_depthfirst/generic.cpp",
- "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_u8q_nhwc_max_generic_depthfirst/generic.cpp"
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/sve_u8q_nhwc_max_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_avg_3x3_s1_output2x2_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_avg_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp16_nhwc_max_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp32_nhwc_avg_3x3_s1_output2x2_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp32_nhwc_avg_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_fp32_nhwc_max_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_s8_nhwc_avg_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_s8_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_s8_nhwc_max_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_s8q_nhwc_avg_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_s8q_nhwc_max_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_u8_nhwc_avg_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_u8_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_u8_nhwc_max_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_u8q_nhwc_avg_generic_depthfirst/generic.cpp",
+ "src/core/NEON/kernels/arm_conv/pooling/kernels/a64_u8q_nhwc_max_generic_depthfirst/generic.cpp"
]
}
}
},
- "PriorBox": {
+ "PRelu": {
+ "deps": [ "ElementwiseBinary" ],
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEPriorBoxLayerKernel.cpp"
+ "common": [
+ "src/runtime/NEON/functions/NEPReluLayer.cpp"
]
}
},
- "QLSTMLayerNormalization": {
+ "PriorBox": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEQLSTMLayerNormalizationKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEPriorBoxLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEPriorBoxLayer.cpp"
+ ]
+ }
+ },
+ "Quantize": {
+ "files": {
+ "common": [
+ "src/cpu/operators/CpuQuantize.cpp",
+ "src/cpu/kernels/CpuQuantizeKernel.cpp",
+ "src/runtime/NEON/functions/NEQuantizationLayer.cpp"
]
}
},
"Range": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NERangeKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NERangeKernel.cpp",
+ "src/runtime/NEON/functions/NERange.cpp"
]
}
},
- "ReductionOperation": {
+ "Reduction":{
+ "deps": [ "Reshape" ],
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEReductionOperationKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEReductionOperationKernel.cpp",
+ "src/runtime/NEON/functions/NEReductionOperation.cpp"
]
}
},
"Remap": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NERemapKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NERemapKernel.cpp",
+ "src/runtime/NEON/functions/NERemap.cpp"
]
}
},
"Reorg": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEReorgLayerKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEReorgLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEReorgLayer.cpp"
+ ]
+ }
+ },
+ "Reshape": {
+ "files": {
+ "common": [
+ "src/cpu/operators/CpuReshape.cpp",
+ "src/cpu/kernels/CpuReshapeKernel.cpp",
+ "src/runtime/NEON/functions/NEReshapeLayer.cpp"
]
}
},
"Reverse": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEReverseKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEReverseKernel.cpp",
+ "src/runtime/NEON/functions/NEReverse.cpp"
]
}
},
+ "RNN": {
+ "deps": [ "Activation", "Add", "FullyConnected", "Gemm"],
+ "files": {
+ "common": [ "src/runtime/NEON/functions/NERNNLayer.cpp" ]
+ }
+ },
"ROIAlign": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEROIAlignLayerKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEROIAlignLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEROIAlignLayer.cpp"
]
}
},
- "ROIPooling": {
+ "ROIPool2d": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NEROIPoolingLayerKernel.cpp"
- ]
- }
- },
- "Select": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NESelectKernel.cpp"
- ]
- }
- },
- "SpaceToBatch": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NESpaceToBatchLayerKernel.cpp"
- ]
- }
- },
- "SpaceToDepth": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NESpaceToDepthLayerKernel.cpp"
- ]
- }
- },
- "Stack": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEStackLayerKernel.cpp"
- ]
- }
- },
- "StridedSlice": {
- "files": {
- "kernel": [
- "src/core/NEON/kernels/NEStridedSliceKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NEROIPoolingLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEROIPoolingLayer.cpp"
]
}
},
"Scale": {
"files": {
- "operator": [
- "src/cpu/operators/CpuScale.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuScaleKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuScale.cpp",
+ "src/cpu/kernels/CpuScaleKernel.cpp",
+ "src/runtime/NEON/functions/NEScale.cpp"
],
"sve": {
- "fp32": [
- "src/cpu/kernels/scale/sve/fp32.cpp"
- ],
- "fp16": [
- "src/cpu/kernels/scale/sve/fp16.cpp"
- ],
- "qasymm8": [
- "src/cpu/kernels/scale/sve/qasymm8.cpp"
- ],
- "qasymm8_signed": [
- "src/cpu/kernels/scale/sve/qasymm8_signed.cpp"
- ],
- "integer": [
- "src/cpu/kernels/scale/sve/integer.cpp"
- ]
+ "fp16": [ "src/cpu/kernels/scale/sve/fp16.cpp" ],
+ "fp32": [ "src/cpu/kernels/scale/sve/fp32.cpp" ],
+ "integer": [ "src/cpu/kernels/scale/sve/integer.cpp" ],
+ "qasymm8": [ "src/cpu/kernels/scale/sve/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/scale/sve/qasymm8_signed.cpp" ]
+
},
"neon": {
- "fp16": [
- "src/cpu/kernels/scale/neon/fp16.cpp"
- ],
- "qasymm8": [
- "src/cpu/kernels/scale/neon/qasymm8.cpp"
- ],
- "qasymm8_signed": [
- "src/cpu/kernels/scale/neon/qasymm8_signed.cpp"
- ],
- "integer": [
- "src/cpu/kernels/scale/neon/integer.cpp"
- ]
+ "fp16": [ "src/cpu/kernels/scale/neon/fp16.cpp" ],
+ "integer": [ "src/cpu/kernels/scale/neon/integer.cpp" ],
+ "qasymm8": [ "src/cpu/kernels/scale/neon/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/scale/neon/qasymm8_signed.cpp" ]
}
}
},
+ "Select": {
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NESelectKernel.cpp",
+ "src/runtime/NEON/functions/NESelect.cpp"
+ ]
+ }
+ },
+ "Slice": {
+ "deps": [ "StridedSlice" ],
+ "files": {
+ "common": [ "src/runtime/NEON/functions/NESlice.cpp" ]
+ }
+ },
"Softmax": {
"deps": [
"Permute"
],
"files": {
- "operator": [
- "src/cpu/operators/CpuSoftmax.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuSoftmaxKernel.cpp"
+ "common": [
+ "src/cpu/operators/CpuSoftmax.cpp",
+ "src/cpu/kernels/CpuSoftmaxKernel.cpp",
+ "src/runtime/NEON/functions/NESoftmaxLayer.cpp"
],
"sve": {
- "all": [
- "src/cpu/kernels/softmax/impl/sve/impl.cpp"
- ]
+ "common": [ "src/cpu/kernels/softmax/impl/sve/impl.cpp" ]
}
}
},
+ "SpaceToBatch": {
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NESpaceToBatchLayerKernel.cpp",
+ "src/runtime/NEON/functions/NESpaceToBatchLayer.cpp"
+ ]
+ }
+ },
+ "SpaceToDepth": {
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NESpaceToDepthLayerKernel.cpp",
+ "src/runtime/NEON/functions/NESpaceToDepthLayer.cpp"
+ ]
+ }
+ },
+ "Split": {
+ "deps": [ "StridedSlice" ],
+ "files": {
+ "common": [
+ "src/runtime/NEON/functions/NESplit.cpp"
+ ]
+ }
+ },
+ "Stack": {
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NEStackLayerKernel.cpp",
+ "src/runtime/NEON/functions/NEStackLayer.cpp"
+ ]
+ }
+ },
+ "StridedSlice": {
+ "files": {
+ "common": [
+ "src/core/NEON/kernels/NEStridedSliceKernel.cpp",
+ "src/runtime/NEON/functions/NEStridedSlice.cpp"
+ ]
+ }
+ },
"Sub": {
"files": {
- "operator": [
- "src/cpu/operators/CpuSub.cpp"
+ "common": [
+ "src/cpu/operators/CpuSub.cpp",
+ "src/cpu/kernels/CpuSubKernel.cpp",
+ "src/runtime/NEON/functions/NEArithmeticSubtraction.cpp"
],
- "kernel": [
- "src/cpu/kernels/CpuSubKernel.cpp"
- ],
+ "sve": {
+ "qasymm8": [ "src/cpu/kernels/sub/neon/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/sub/neon/qasymm8_signed.cpp" ],
+ "qsymm16": [ "src/cpu/kernels/sub/neon/qsymm16.cpp" ]
+ },
"neon": {
- "qsymm16": [
- "src/cpu/kernels/sub/neon/qsymm16.cpp"
- ],
- "qasymm8": [
- "src/cpu/kernels/sub/neon/qasymm8.cpp"
- ],
- "qasymm8_signed": [
- "src/cpu/kernels/sub/neon/qasymm8_signed.cpp"
- ]
+ "qasymm8": [ "src/cpu/kernels/sub/neon/qasymm8.cpp" ],
+ "qasymm8_signed": [ "src/cpu/kernels/sub/neon/qasymm8_signed.cpp" ],
+ "qsymm16": [ "src/cpu/kernels/sub/neon/qsymm16.cpp" ]
}
}
},
- "Transpose": {
- "files": {
- "operator": [
- "src/cpu/operators/CpuTranspose.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuTransposeKernel.cpp"
- ]
- }
- },
"Tile": {
"files": {
- "kernel": [
- "src/core/NEON/kernels/NETileKernel.cpp"
+ "common": [
+ "src/core/NEON/kernels/NETileKernel.cpp",
+ "src/runtime/NEON/functions/NETile.cpp"
]
}
},
- "WinogradConvolution": {
- "deps": [
- "Activation",
- "Permute"
- ],
+ "Transpose": {
"files": {
- "operator": [
- "src/cpu/operators/CpuWinogradConv2d.cpp"
- ],
- "kernel": [
- "src/cpu/kernels/CpuWinogradConv2dKernel.cpp",
- "src/core/NEON/kernels/convolution/winograd/padding.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_1x8_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_4x4_fp16_fp16_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_4x4_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_6x6_fp16_fp16_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/input_6x6_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_2_7_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_2x2_3x3_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_2x2_5x5_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_4_5_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_4x4_3x3_fp16_fp16_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_4x4_3x3_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/output_6_3_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_2_7_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_2x2_3x3_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_2x2_5x5_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_4_5_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_4x4_3x3_fp16_fp16_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_4x4_3x3_fp32_fp32_integers.cpp",
- "src/core/NEON/kernels/convolution/winograd/winograd_transforms/weights_6_3_fp32_fp32_integers.cpp"
+ "common": [
+ "src/cpu/kernels/CpuTransposeKernel.cpp",
+ "src/cpu/operators/CpuTranspose.cpp",
+ "src/runtime/NEON/functions/NETranspose.cpp"
]
}
+ },
+ "Unstack": {
+ "deps": [ "StridedSlice" ],
+ "files": {
+ "common": [ "src/runtime/NEON/functions/NEUnstack.cpp" ]
+ }
}
}
}