Skip to content

Commit

Permalink
[HIPIFY][ROCm#584][DNN][MIOpen] cuDNN -> MIOpen - Part 13 - cuDNN Bat…
Browse files Browse the repository at this point in the history 
…ch Normalization functions

+ Continued supporting hipification to MIOpen based on `miopen.h`
+ Updated the synthetic test `cudnn2miopen.cu` accordingly
  • Loading branch information
@emankov
emankov committed Apr 4, 2023
1 parent 3483f9b commit 0b48b16
Show file tree
Hide file tree
Showing 2 changed files with 51 additions and 6 deletions.
8 changes: 4 additions & 4 deletions src/CUDA2HIP_DNN_API_functions.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -172,11 +172,11 @@ const std::map<llvm::StringRef, hipCounter> CUDA_DNN_FUNCTION_MAP {
{"cudnnDivisiveNormalizationBackward", {"hipdnnDivisiveNormalizationBackward", "", CONV_LIB_FUNC, API_DNN, 2, HIP_UNSUPPORTED}},

// cuDNN Batch Normalization functions
{"cudnnDeriveBNTensorDescriptor", {"hipdnnDeriveBNTensorDescriptor", "", CONV_LIB_FUNC, API_DNN, 2}},
{"cudnnBatchNormalizationForwardTraining", {"hipdnnBatchNormalizationForwardTraining", "", CONV_LIB_FUNC, API_DNN, 2}},
{"cudnnDeriveBNTensorDescriptor", {"hipdnnDeriveBNTensorDescriptor", "miopenDeriveBNTensorDescriptor", CONV_LIB_FUNC, API_DNN, 2}},
{"cudnnBatchNormalizationForwardTraining", {"hipdnnBatchNormalizationForwardTraining", "miopenBatchNormalizationForwardTraining", CONV_LIB_FUNC, API_DNN, 2}},
{"cudnnBatchNormalizationForwardTrainingEx", {"hipdnnBatchNormalizationForwardTrainingEx", "", CONV_LIB_FUNC, API_DNN, 2, HIP_UNSUPPORTED}},
{"cudnnBatchNormalizationForwardInference", {"hipdnnBatchNormalizationForwardInference", "", CONV_LIB_FUNC, API_DNN, 2}},
{"cudnnBatchNormalizationBackward", {"hipdnnBatchNormalizationBackward", "", CONV_LIB_FUNC, API_DNN, 2}},
{"cudnnBatchNormalizationForwardInference", {"hipdnnBatchNormalizationForwardInference", "miopenBatchNormalizationForwardInference", CONV_LIB_FUNC, API_DNN, 2}},
{"cudnnBatchNormalizationBackward", {"hipdnnBatchNormalizationBackward", "miopenBatchNormalizationBackward", CONV_LIB_FUNC, API_DNN, 2}},
{"cudnnBatchNormalizationBackwardEx", {"hipdnnBatchNormalizationBackwardEx", "", CONV_LIB_FUNC, API_DNN, 2, HIP_UNSUPPORTED}},
{"cudnnGetBatchNormalizationForwardTrainingExWorkspaceSize", {"hipdnnGetBatchNormalizationForwardTrainingExWorkspaceSize", "", CONV_LIB_FUNC, API_DNN, 2, HIP_UNSUPPORTED}},
{"cudnnGetBatchNormalizationBackwardExWorkspaceSize", {"hipdnnGetBatchNormalizationBackwardExWorkspaceSize", "", CONV_LIB_FUNC, API_DNN, 2, HIP_UNSUPPORTED}},
Expand Down
49 changes: 47 additions & 2 deletions tests/unit_tests/synthetic/libraries/cudnn2miopen.cu
View file
Original file line number Diff line number Diff line change
Expand Up @@ -272,7 +272,6 @@ int main() {
// CHECK: status = miopenDestroyTensorDescriptor(tensorDescriptor);
status = cudnnDestroyTensorDescriptor(tensorDescriptor);


// CHECK: miopenTensorDescriptor_t aD;
// CHECK-NEXT: miopenTensorDescriptor_t bD;
// CHECK-NEXT: miopenTensorDescriptor_t cD;
Expand All @@ -281,6 +280,7 @@ int main() {
// CHECK-NEXT: miopenTensorDescriptor_t wD;
// CHECK-NEXT: miopenTensorDescriptor_t inputD;
// CHECK-NEXT: miopenTensorDescriptor_t dbD;
// CHECK-NEXT: miopenTensorDescriptor_t dxD;
cudnnTensorDescriptor_t aD;
cudnnTensorDescriptor_t bD;
cudnnTensorDescriptor_t cD;
Expand All @@ -289,6 +289,7 @@ int main() {
cudnnTensorDescriptor_t wD;
cudnnTensorDescriptor_t inputD;
cudnnTensorDescriptor_t dbD;
cudnnTensorDescriptor_t dxD;
void* A = nullptr;
void* B = nullptr;
void* C = nullptr;
Expand Down Expand Up @@ -386,6 +387,7 @@ int main() {
// CHECK: status = miopenCreatePoolingDescriptor(&poolingDescriptor);
status = cudnnCreatePoolingDescriptor(&poolingDescriptor);

// CHECK: miopenNanPropagation_t maxpoolingNanOpt;
cudnnNanPropagation_t maxpoolingNanOpt;
int wH = 0;
int wW = 0;
Expand Down Expand Up @@ -452,7 +454,7 @@ int main() {
status = cudnnSetLRNDescriptor(LRNDescriptor, lrnN, lrnAlpha, lrnBeta, lrnK);

// TODO: add a referrence to miopenLRNMode_t as a 2nd arg
// TODO: [feature] Add a new type of transformation with declaring a var before the function call to add that var referrence as an arg to the below function call
// TODO: [feature] Add a new type of transformation by declaring a var before the function call to add that var reference as an arg to the below function call
// CUDA: cudnnStatus_t CUDNNWINAPI cudnnGetLRNDescriptor(cudnnLRNDescriptor_t normDesc, unsigned* lrnN, double* lrnAlpha, double* lrnBeta, double* lrnK);
// MIOPEN: MIOPEN_EXPORT miopenStatus_t miopenGetLRNDescriptor(const miopenLRNDescriptor_t lrnDesc, miopenLRNMode_t* mode, unsigned int* lrnN, double* lrnAlpha, double* lrnBeta, double* lrnK);
// CHECK: status = miopenGetLRNDescriptor(LRNDescriptor, &lrnN, &lrnAlpha, &lrnBeta, &lrnK);
Expand All @@ -463,5 +465,48 @@ int main() {
// CHECK: status = miopenDestroyLRNDescriptor(LRNDescriptor);
status = cudnnDestroyLRNDescriptor(LRNDescriptor);

// CHECK: miopenTensorDescriptor_t bnScaleBiasMeanVarDesc;
// CHECK: miopenTensorDescriptor_t bnScaleBiasDiffDesc;
cudnnTensorDescriptor_t bnScaleBiasMeanVarDesc;
cudnnTensorDescriptor_t bnScaleBiasDiffDesc;
void *bnScale = nullptr;
void *bnBias = nullptr;
double expAvgFactor = 0.0f;
void *resultRunningMean = nullptr;
void *resultRunningVariance = nullptr;
double epsilon = 0.0f;
void *resultSaveMean = nullptr;
void *resultSaveInvVariance = nullptr;
void *estimatedMean = nullptr;
void *estimatedVariance = nullptr;
void *alphaDataDiff = nullptr;
void *betaDataDiff = nullptr;
void *alphaParamDiff = nullptr;
void *betaParamDiff = nullptr;
void *resultBnScaleDiff = nullptr;
void *resultBnBiasDiff = nullptr;
void *savedMean = nullptr;
void *savedInvVariance = nullptr;

// CUDA: cudnnStatus_t CUDNNWINAPI cudnnDeriveBNTensorDescriptor(cudnnTensorDescriptor_t derivedBnDesc, const cudnnTensorDescriptor_t xDesc, cudnnBatchNormMode_t mode);
// MIOPEN: MIOPEN_EXPORT miopenStatus_t miopenDeriveBNTensorDescriptor(miopenTensorDescriptor_t derivedBnDesc, const miopenTensorDescriptor_t xDesc, miopenBatchNormMode_t bn_mode);
// CHECK: status = miopenDeriveBNTensorDescriptor(tensorDescriptor, xD, batchNormMode);
status = cudnnDeriveBNTensorDescriptor(tensorDescriptor, xD, batchNormMode);

// CUDA: cudnnStatus_t CUDNNWINAPI cudnnBatchNormalizationForwardTraining(cudnnHandle_t handle, cudnnBatchNormMode_t mode, const void* alpha, const void* beta, const cudnnTensorDescriptor_t xDesc, const void* x, const cudnnTensorDescriptor_t yDesc, void* y, const cudnnTensorDescriptor_t bnScaleBiasMeanVarDesc, const void* bnScale, const void* bnBias, double exponentialAverageFactor, void* resultRunningMean, double epsilon, void* resultSaveMean, void* resultSaveInvVariance);
// MIOPEN: MIOPEN_EXPORT miopenStatus_t miopenBatchNormalizationForwardTraining(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void* alpha, void* beta, const miopenTensorDescriptor_t xDesc, const void* x, const miopenTensorDescriptor_t yDesc, void* y, const miopenTensorDescriptor_t bnScaleBiasMeanVarDesc, void* bnScale, void* bnBias, double expAvgFactor, void* resultRunningMean, void* resultRunningVariance, double epsilon, void* resultSaveMean, void* resultSaveInvVariance);
// CHECK: status = miopenBatchNormalizationForwardTraining(handle, batchNormMode, alpha, beta, xD, x, yD, y, bnScaleBiasMeanVarDesc, bnScale, bnBias, expAvgFactor, resultRunningMean, resultRunningVariance, epsilon, resultSaveMean, resultSaveInvVariance);
status = cudnnBatchNormalizationForwardTraining(handle, batchNormMode, alpha, beta, xD, x, yD, y, bnScaleBiasMeanVarDesc, bnScale, bnBias, expAvgFactor, resultRunningMean, resultRunningVariance, epsilon, resultSaveMean, resultSaveInvVariance);

// CUDA: cudnnStatus_t CUDNNWINAPI cudnnBatchNormalizationForwardInference(cudnnHandle_t handle, cudnnBatchNormMode_t mode, const void* alpha, const void* beta, const cudnnTensorDescriptor_t xDesc, const void* x, const cudnnTensorDescriptor_t yDesc, void* y, const cudnnTensorDescriptor_t bnScaleBiasMeanVarDesc, const void* bnScale, const void* bnBias, const void* estimatedMean, const void* estimatedVariance, double epsilon);
// MIOPEN: miopenBatchNormalizationForwardInference(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, void* alpha, void* beta, const miopenTensorDescriptor_t xDesc, const void* x, const miopenTensorDescriptor_t yDesc, void* y, const miopenTensorDescriptor_t bnScaleBiasMeanVarDesc, void* bnScale, void* bnBias, void* estimatedMean, void* estimatedVariance, double epsilon);
// CHECK: status = miopenBatchNormalizationForwardInference(handle, batchNormMode, alpha, beta, xD, x, yD, y, bnScaleBiasMeanVarDesc, bnScale, bnBias, estimatedMean, estimatedVariance, epsilon);
status = cudnnBatchNormalizationForwardInference(handle, batchNormMode, alpha, beta, xD, x, yD, y, bnScaleBiasMeanVarDesc, bnScale, bnBias, estimatedMean, estimatedVariance, epsilon);

// CUDA: cudnnStatus_t CUDNNWINAPI cudnnBatchNormalizationBackward(cudnnHandle_t handle, cudnnBatchNormMode_t mode, const void* alphaDataDiff, const void* betaDataDiff, const void* alphaParamDiff, const void* betaParamDiff, const cudnnTensorDescriptor_t xDesc, const void* x, const cudnnTensorDescriptor_t dyDesc, const void* dy, const cudnnTensorDescriptor_t dxDesc, void* dx, const cudnnTensorDescriptor_t dBnScaleBiasDesc, const void* bnScale, void* dBnScaleResult, void* dBnBiasResult, double epsilon, const void* savedMean, const void* savedInvVariance);
// MIOPEN: MIOPEN_EXPORT miopenStatus_t miopenBatchNormalizationBackward(miopenHandle_t handle, miopenBatchNormMode_t bn_mode, const void* alphaDataDiff, const void* betaDataDiff, const void* alphaParamDiff, const void* betaParamDiff, const miopenTensorDescriptor_t xDesc, const void* x, const miopenTensorDescriptor_t dyDesc, const void* dy, const miopenTensorDescriptor_t dxDesc, void* dx, const miopenTensorDescriptor_t bnScaleBiasDiffDesc, const void* bnScale, void* resultBnScaleDiff, void* resultBnBiasDiff, double epsilon, const void* savedMean, const void* savedInvVariance);
// CHECK: status = miopenBatchNormalizationBackward(handle, batchNormMode, alphaDataDiff, betaDataDiff, alphaParamDiff, betaParamDiff, xD, x, yD, y, dxD, dx, bnScaleBiasDiffDesc, bnScale, resultBnScaleDiff, resultBnBiasDiff, epsilon, savedMean, savedInvVariance);
status = cudnnBatchNormalizationBackward(handle, batchNormMode, alphaDataDiff, betaDataDiff, alphaParamDiff, betaParamDiff, xD, x, yD, y, dxD, dx, bnScaleBiasDiffDesc, bnScale, resultBnScaleDiff, resultBnBiasDiff, epsilon, savedMean, savedInvVariance);

return 0;
}

0 comments on commit 0b48b16

Please sign in to comment.