MIOpen
2.18.0

Contents:

  • MIOpen Release notes
    • 10/05/2022 [2.18.0]
    • 04/12/2022 [2.16.0]
    • 11/02/2021 [2.14.0]
    • 08/03/2021 [2.12.0]
    • 05/17/2021 [2.11.0]
    • 03/25/2021 [2.10.0]
    • 12/01/2020 [ 2.9.0 ]
    • 10/28/2020 [ 2.8.0 ]
    • 09/18/2020 [ 2.7.0 ]
    • 08/20/2020 [ 2.6.0 ]
    • 07/10/2020 [ 2.5.0 ]
    • 06/02/2020 [ 2.4.0 ]
    • 03/01/2020 [ 2.3.0 ]
    • 01/24/2020 [ 2.2.1 ]
    • 12/19/2019 [ 2.2.0 ]
    • 09/25/2019 [ 2.1.0 ]
    • 08/13/2019 [ 2.0.1 ]
    • 07/08/2019 [ 2.0.0 ]
    • 05/03/2019 [ 1.8.1 ]
    • 04/11/2019 [ 1.8.0 ]
    • 02/06/2019 [ 1.7.1 ]
    • 12/19/2018 [ 1.7.0 ]
    • 11/18/2018 [ 1.6.0 ]
    • 09/14/2018 [ 1.5.0 ]
    • 07/30/2018 [ 1.4.2 ]
    • 07/19/2018 [ 1.4.1 ]
    • 07/06/2018 [ 1.4.0 ]
    • 03/30/2018 [ 1.3.0 ]
    • 03/08/2018 [ 1.2.1 ]
    • 12/15/2017 [ 1.2.0 ]
    • 09/08/2017 [ 1.1.0 ]
    • 08/27/2017 [ 1.0.2 ]
    • 07/26/2017 [ 1.0.1 ]
    • 06/30/2017 [ 1.0.0 ] Initial release
  • Citing MIOpen
    • Citation BibTeX
  • Build and Install Instructions
    • Prerequisites
    • Installing MIOpen with pre-built packages
    • Installing MIOpen kernels package
    • Installing the dependencies
    • Building MIOpen from source
      • Configuring with cmake
      • For the HIP backend (ROCm 3.5 and later), run:
      • For OpenCL, run:
      • Setting Up Locations
      • System Performance Database and User Database
      • Persistent Program Cache
        • For MIOpen version 2.3 and earlier
        • For MIOpen version 2.4 and later
      • Changing the cmake configuration
    • Building the library
    • Building the driver
    • Running the tests
    • Building the documentation
    • Formatting the code
    • Installing the dependencies manually
    • Using docker
    • Citing MIOpen
      • Citation BibTeX
    • Porting from cuDNN to MIOpen
  • Building MIOpen for Embedded Systems
    • Install dependencies
    • Configuring for an embedded build
    • Embedding Find-Db and Performance database:
    • Embedding the precompiled kernels package:
    • Embedding precompiled package
      • Using a package install
      • Using the URL to a kernels binary
    • Full configuration line:
  • MIOpenDriver
    • Building the Driver
    • Base Arguments
    • Executing MIOpenDriver
  • Debugging and Logging
    • Logging
    • Layer Filtering
      • Filtering by algorithm
      • Filtering by build method
      • Filtering out all Solutions except one
      • Filtering the Solutions on individual basis
    • rocBlas Logging and Behavior
    • Numerical Checking
    • Controlling Parallel Compilation
    • Experimental controls
      • Code Object (CO) version selection (EXPERIMENTAL)
      • Winograd Multi-pass Maximum Workspace throttling
  • Kernel Cache
    • Clear the cache
    • Disabling the cache
    • Updating MIOpen and removing the cache
    • Installing pre-compiled kernels
  • Performance Database
    • Auto-tuning the kernels.
      • MIOPEN_FIND_ENFORCE
      • Updating MIOpen and the User Db
  • Find-Db Database
    • Populating the User Find-Db
    • Updating MIOpen and the User Find-Db
    • Disabling Find-Db
  • Find and Immediate Mode
    • Find API
    • Immediate Mode API
    • Immediate Mode Fall Back
    • Limitations of Immediate Mode
      • Architectual Limitations
      • Backend Limitations
      • Find Modes
  • Fusion API: Getting Started
    • Introduction
      • Intended Audience
    • Create a Fusion Plan
    • Create and add Operators
    • Compile the Fusion Plan
    • Set the runtime arguments
    • Execute a Fusion Plan
    • Cleanup
    • Supported Fusions
    • Performance Comparison to Non-Fused Kernels
  • MI200 MFMA Behavior Specifics
  • MIOpen Porting Guide
    • The key differences between MIOpen and cuDNN:
    • Helpful MIOpen Environment Variables
  • API Reference
    • Datatypes
    • Handle
      • miopenStatus_t
      • miopenCreate
      • miopenCreateWithStream
      • miopenDestroy
      • miopenSetStream
      • miopenGetStream
      • miopenGetKernelTime
      • miopenEnableProfiling
    • Tensors
      • miopenDataType_t
      • miopenTensorOp_t
      • miopenCreateTensorDescriptor
      • miopenSet4dTensorDescriptor
      • miopenGet4dTensorDescriptor
      • miopenSetTensorDescriptor
      • miopenGetTensorDescriptorSize
      • miopenGetTensorDescriptor
      • miopenDestroyTensorDescriptor
      • miopenOpTensor
      • miopenSetTensor
      • miopenScaleTensor
    • Activation Layers
      • miopenActivationMode_t
      • miopenCreateActivationDescriptor
      • miopenSetActivationDescriptor
      • miopenGetActivationDescriptor
      • miopenActivationForward
      • miopenActivationBackward
      • miopenDestroyActivationDescriptor
    • Convolutional Layer
      • miopenConvolutionMode_t
      • miopenConvFwdAlgorithm_t
      • miopenConvBwdWeightsAlgorithm_t
      • miopenConvBwdDataAlgorithm_t
      • miopenConvAlgoPerf_t
      • miopenConvSolution_t
      • miopenCreateConvolutionDescriptor
      • miopenInitConvolutionDescriptor
      • miopenInitConvolutionNdDescriptor
      • miopenGetConvolutionDescriptor
      • miopenGetConvolutionNdDescriptor
      • miopenSetConvolutionGroupCount
      • miopenSetTransposeConvOutputPadding
      • miopenSetTransposeConvNdOutputPadding
      • miopenGetConvolutionForwardOutputDim
      • miopenGetConvolutionNdForwardOutputDim
      • miopenConvolutionForwardGetWorkSpaceSize
      • miopenFindConvolutionForwardAlgorithm
      • miopenConvolutionForward
      • miopenConvolutionForwardBias
      • miopenFindConvolutionBackwardDataAlgorithm
      • miopenConvolutionBackwardData
      • miopenConvolutionBackwardDataGetWorkSpaceSize
      • miopenConvolutionBackwardWeightsGetWorkSpaceSize
      • miopenFindConvolutionBackwardWeightsAlgorithm
      • miopenConvolutionBackwardWeights
      • miopenConvolutionBackwardBias
      • miopenDestroyConvolutionDescriptor
    • Recurrent Neural Networks
      • miopenRNNMode_t
      • miopenRNNInputMode_t
      • miopenRNNAlgo_t
      • miopenRNNDirectionMode_t
      • miopenRNNBiasMode_t
      • miopenRNNGEMMalgoMode_t
      • miopenCreateRNNDescriptor
      • miopenGetRNNDescriptor
      • miopenGetRNNDescriptor_V2
      • miopenDestroyRNNDescriptor
      • miopenSetRNNDescriptor
      • miopenSetRNNDescriptor_V2
      • miopenGetRNNWorkspaceSize
      • miopenGetRNNTrainingReserveSize
      • miopenGetRNNParamsSize
      • miopenGetRNNParamsDescriptor
      • miopenGetRNNInputTensorSize
      • miopenGetRNNHiddenTensorSize
      • miopenGetRNNLayerParamSize
      • miopenGetRNNLayerBiasSize
      • miopenGetRNNLayerParam
      • miopenGetRNNLayerBias
      • miopenSetRNNLayerParam
      • miopenSetRNNLayerBias
      • miopenGetRNNLayerParamOffset
      • miopenGetRNNLayerBiasOffset
      • miopenRNNForwardTraining
      • miopenRNNBackwardData
      • miopenRNNBackwardWeights
      • miopenRNNForwardInference
    • Batch Normalization Layer
      • miopenBatchNormMode_t
      • miopenDeriveBNTensorDescriptor
      • miopenBatchNormalizationForwardTraining
      • miopenBatchNormalizationForwardInference
      • miopenBatchNormalizationBackward
    • Local Response Normalization Layer
      • miopenLRNMode_t
      • miopenCreateLRNDescriptor
      • miopenSetLRNDescriptor
      • miopenGetLRNDescriptor
      • miopenLRNGetWorkSpaceSize
      • miopenLRNForward
      • miopenLRNBackward
      • miopenDestroyLRNDescriptor
    • Pooling Layer
      • miopenPoolingMode_t
      • miopenIndexType_t
      • miopenCreatePoolingDescriptor
      • miopenSet2dPoolingDescriptor
      • miopenSetNdPoolingDescriptor
      • miopenSetPoolingIndexType
      • miopenGetPoolingIndexType
      • miopenGet2dPoolingDescriptor
      • miopenGetNdPoolingDescriptor
      • miopenGetPoolingForwardOutputDim
      • miopenGetPoolingNdForwardOutputDim
      • miopenPoolingGetWorkSpaceSize
      • miopenPoolingGetWorkSpaceSizeV2
      • miopenPoolingForward
      • miopenPoolingBackward
      • miopenDestroyPoolingDescriptor
    • Softmax Layer
      • miopenSoftmaxAlgorithm_t
      • miopenSoftmaxMode_t
      • miopenSoftmaxForward
      • miopenSoftmaxBackward
      • miopenSoftmaxForward_V2
      • miopenSoftmaxBackward_V2
    • Layer Fusion
      • miopenFusionDirection_t
      • miopenCreateFusionPlan
      • miopenDestroyFusionPlan
      • miopenCompileFusionPlan
      • miopenFusionPlanGetOp
      • miopenFusionPlanGetWorkSpaceSize
      • miopenFusionPlanConvolutionGetAlgo
      • miopenCreateOpConvForward
      • miopenCreateOpActivationForward
      • miopenCreateOpBiasForward
      • miopenCreateOpBatchNormInference
      • miopenCreateOperatorArgs
      • miopenDestroyOperatorArgs
      • miopenSetOpArgsConvForward
      • miopenSetOpArgsBatchNormInference
      • miopenSetOpArgsBiasForward
      • miopenExecuteFusionPlan
    • Loss Function Layer
      • miopenCTCLossAlgo_t
      • miopenCreateCTCLossDescriptor
      • miopenGetCTCLossDescriptor
      • miopenDestroyCTCLossDescriptor
      • miopenSetCTCLossDescriptor
      • miopenGetCTCLossWorkspaceSize
      • miopenCTCLoss
    • Dropout Layer
      • miopenRNGType_t
      • miopenCreateDropoutDescriptor
      • miopenGetDropoutDescriptor
      • miopenRestoreDropoutDescriptor
      • miopenDestroyDropoutDescriptor
      • miopenSetDropoutDescriptor
      • miopenDropoutGetReserveSpaceSize
      • miopenDropoutGetStatesSize
      • miopenDropoutForward
      • miopenDropoutBackward
    • Reduction Layer
      • miopenReduceTensorOp_t
      • miopenNanPropagation_t
      • miopenReduceTensorIndices_t
      • miopenIndicesType_t
      • miopenCreateReduceTensorDescriptor
      • miopenDestroyReduceTensorDescriptor
      • miopenSetReduceTensorDescriptor
      • miopenGetReduceTensorDescriptor
      • miopenGetReductionIndicesSize
      • miopenGetReductionWorkspaceSize
      • miopenReduceTensor
MIOpen
  • »
  • Search

© Copyright 2022, Advanced Micro Devices, Inc. All rights reserved.

Built with Sphinx using a theme provided by Read the Docs.