Poplar and PopLibs API Reference

• 1. Using the libraries
  • 1.1. Adding PopLibs code to a graph
  • 1.2. Setting Options
    • 1.2.1. Environment variables
    • 1.2.2. Option values
    • 1.2.3. Unsupported options
  • 1.3. Linking
• 2. Poplar runtime API reference
  • 2.1. Vertex types
  • 2.2. Input & output types
  • 2.3. Field types
  • 2.4. Vector types
  • 2.5. VectorList types
  • 2.6. Half-float types
  • 2.7. Stack size definition
• 3. Poplar API reference
  • 3.1. Utility classes
    • 3.1.1. poplar/ArrayRef.hpp
    • 3.1.2. poplar/Interval.hpp
    • 3.1.3. poplar/OptionFlags.hpp
    • 3.1.4. poplar/RandomSeed.hpp
    • 3.1.5. poplar/ReplicatedStreamMode.hpp
    • 3.1.6. poplar/SerializationFormat.hpp
    • 3.1.7. poplar/StringRef.hpp
    • 3.1.8. poplar/SyncType.hpp
    • 3.1.9. poplar/TypeTraits.hpp
    • 3.1.10. poplar/CSRFunctions.hpp
  • 3.2. Exceptions
    • 3.2.1. poplar/exceptions.hpp
  • 3.3. Graph classes
    • 3.3.1. poplar/CodeletFileType.hpp
    • 3.3.2. poplar/DataStream.hpp
    • 3.3.3. poplar/DataStreamType.hpp
    • 3.3.4. poplar/Graph.hpp
    • 3.3.5. poplar/GraphElements.hpp
    • 3.3.6. poplar/LateInitCallback.hpp
    • 3.3.7. poplar/PerfEstimateFunc.hpp
    • 3.3.8. poplar/Tensor.hpp
    • 3.3.9. poplar/TensorCloneMethod.hpp
    • 3.3.10. poplar/TensorRearranger.hpp
    • 3.3.11. poplar/Type.hpp
    • 3.3.12. poplar/VariableMappingMethod.hpp
    • 3.3.13. poplar/VariableRef.hpp
    • 3.3.14. poplar/VectorLayout.hpp
    • 3.3.15. poplar/VertexIntrospector.hpp
  • 3.4. Control program classes
    • 3.4.1. poplar/Program.hpp
  • 3.5. Device management
    • 3.5.1. poplar/TargetType.hpp
    • 3.5.2. poplar/Target.hpp
    • 3.5.3. poplar/Device.hpp
    • 3.5.4. poplar/DeviceManager.hpp
    • 3.5.5. poplar/IpuLinkConfiguration.hpp
    • 3.5.6. poplar/IpuLinkTopology.hpp
  • 3.6. Graph execution
    • 3.6.1. poplar/Engine.hpp
    • 3.6.2. poplar/StreamCallback.hpp
  • 3.7. Serializing executable state
    • 3.7.1. poplar/Executable.hpp
  • 3.8. Profiling & debugging
    • 3.8.1. poplar/DebugContext.hpp
    • 3.8.2. poplar/ProfileValue.hpp
    • 3.8.3. poplar/IPUModel.hpp
    • 3.8.4. poplar/GlobalExchangeConstraints.hpp
    • 3.8.5. poplar/CycleCount.hpp
• 4. PopLibs API reference
  • 4.1. Utility functions (poputil)
    • 4.1.1. poputil/Broadcast.hpp
    • 4.1.2. poputil/GraphFunction.hpp
    • 4.1.3. poputil/Loop.hpp
    • 4.1.4. poputil/OptionParsing.hpp
    • 4.1.5. poputil/TensorMetaData.hpp
    • 4.1.6. poputil/TileMapping.hpp
    • 4.1.7. poputil/Util.hpp
    • 4.1.8. poputil/VarStructure.hpp
    • 4.1.9. poputil/VertexTemplates.hpp
  • 4.2. Tensor operations (popops)
    • 4.2.1. popops/AllTrue.hpp
    • 4.2.2. popops/Cast.hpp
    • 4.2.3. popops/CircBuf.hpp
    • 4.2.4. popops/CollectiveTypes.hpp
    • 4.2.5. popops/DynamicSlice.hpp
    • 4.2.6. popops/ElementWise.hpp
    • 4.2.7. popops/ElementWiseUtil.hpp
    • 4.2.8. popops/Encoding.hpp
    • 4.2.9. popops/EncodingConstants.hpp
    • 4.2.10. popops/Expr.hpp
    • 4.2.11. popops/ExprOp.hpp
    • 4.2.12. popops/Fill.hpp
    • 4.2.13. popops/Gather.hpp
    • 4.2.14. popops/GatherStatistics.hpp
    • 4.2.15. popops/HostSliceTensor.hpp
    • 4.2.16. popops/NaN.hpp
    • 4.2.17. popops/Operation.hpp
    • 4.2.18. popops/Pad.hpp
    • 4.2.19. popops/Rearrange.hpp
    • 4.2.20. popops/Reduce.hpp
    • 4.2.21. popops/ScaledAdd.hpp
    • 4.2.22. popops/Scatter.hpp
    • 4.2.23. popops/SelectScalarFromRows.hpp
    • 4.2.24. popops/SequenceSlice.hpp
    • 4.2.25. popops/Sort.hpp
    • 4.2.26. popops/SortOrder.hpp
    • 4.2.27. popops/TensorCollectives.hpp
    • 4.2.28. popops/TopK.hpp
    • 4.2.29. popops/UpdateScalarInRows.hpp
    • 4.2.30. popops/Zero.hpp
  • 4.3. Linear algebra functions (poplin)
    • 4.3.1. poplin/Cholesky.hpp
    • 4.3.2. poplin/ConvParams.hpp
    • 4.3.3. poplin/ConvUtil.hpp
    • 4.3.4. poplin/Convolution.hpp
    • 4.3.5. poplin/FullyConnected.hpp
    • 4.3.6. poplin/MatMul.hpp
    • 4.3.7. poplin/MeshGrid.hpp
    • 4.3.8. poplin/MultiConvolution.hpp
    • 4.3.9. poplin/Norms.hpp
    • 4.3.10. poplin/TriangularSolve.hpp
  • 4.4. Random number operations (poprand)
    • 4.4.1. poprand/RandomGen.hpp
  • 4.5. Sparse tensor operations (popsparse)
    • 4.5.1. popsparse/experimental/BlockSparse.hpp
    • 4.5.2. popsparse/experimental/BlockSparseMatMul.hpp
    • 4.5.3. popsparse/MatMul.hpp
    • 4.5.4. popsparse/MatMulParams.hpp
    • 4.5.5. popsparse/FullyConnected.hpp
    • 4.5.6. popsparse/FullyConnectedParams.hpp
    • 4.5.7. popsparse/SparsePartitioner.hpp
    • 4.5.8. popsparse/SparseStorageFormats.hpp
    • 4.5.9. popsparse/SparseTensor.hpp
    • 4.5.10. popsparse/SparsityParams.hpp
  • 4.6. Neural network functions (popnn)
    • 4.6.1. popnn/BatchNorm.hpp
    • 4.6.2. popnn/CTCLoss.hpp
    • 4.6.3. popnn/GroupNorm.hpp
    • 4.6.4. popnn/Gru.hpp
    • 4.6.5. popnn/GruDef.hpp
    • 4.6.6. popnn/InstanceNorm.hpp
    • 4.6.7. popnn/LayerNorm.hpp
    • 4.6.8. popnn/LogSoftmax.hpp
    • 4.6.9. popnn/Loss.hpp
    • 4.6.10. popnn/Lstm.hpp
    • 4.6.11. popnn/LstmDef.hpp
    • 4.6.12. popnn/NonLinearity.hpp
    • 4.6.13. popnn/NonLinearityDef.hpp
    • 4.6.14. popnn/NonLinearityDefUtil.hpp
    • 4.6.15. popnn/Norms.hpp
    • 4.6.16. popnn/Pooling.hpp
    • 4.6.17. popnn/PoolingDef.hpp
    • 4.6.18. popnn/Recurrent.hpp
    • 4.6.19. popnn/Rnn.hpp
    • 4.6.20. popnn/SpatialSoftMax.hpp
• 5. GCL API reference
  • 5.1. gcl/TileAllocation.hpp
  • 5.2. gcl/Collectives.hpp
  • 5.3. gcl/CollectiveBalancedReorder.hpp
• 6. PopVision analysis C++ API
  • 6.1. Compilation reports
  • 6.2. Execution reports
• 7. PopVision Analysis Python API
• 8. PopVision trace instrumentation library (PVTI)
• 9. Index
• 10. Trademarks & copyright