| // Auto-generated, do not edit. |
| |
| extern "C" { |
| |
| size_t CUDNNWINAPI |
| cudnnGetVersion(void) { |
| using FuncPtr = size_t (CUDNNWINAPI *)(); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetVersion"); |
| if (!func_ptr) return 0; |
| return func_ptr(); |
| } |
| |
| size_t CUDNNWINAPI |
| cudnnGetCudartVersion(void) { |
| using FuncPtr = size_t (CUDNNWINAPI *)(); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetCudartVersion"); |
| if (!func_ptr) return 0; |
| return func_ptr(); |
| } |
| |
| const char *CUDNNWINAPI |
| cudnnGetErrorString(cudnnStatus_t status) { |
| using FuncPtr = const char * (CUDNNWINAPI *)(cudnnStatus_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetErrorString"); |
| if (!func_ptr) return "cudnnGetErrorString symbol not found."; |
| return func_ptr(status); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnQueryRuntimeError(cudnnHandle_t handle, cudnnStatus_t *rstatus, cudnnErrQueryMode_t mode, cudnnRuntimeTag_t *tag) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnStatus_t *, cudnnErrQueryMode_t, cudnnRuntimeTag_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnQueryRuntimeError"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rstatus, mode, tag); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetProperty(libraryPropertyType type, int *value) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(libraryPropertyType, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetProperty"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(type, value); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreate(cudnnHandle_t *handle) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreate"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroy(cudnnHandle_t handle) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroy"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetStream(cudnnHandle_t handle, cudaStream_t streamId) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudaStream_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetStream"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, streamId); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetStream(cudnnHandle_t handle, cudaStream_t *streamId) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudaStream_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetStream"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, streamId); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateTensorDescriptor(cudnnTensorDescriptor_t *tensorDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnTensorDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetTensor4dDescriptor(cudnnTensorDescriptor_t tensorDesc, |
| cudnnTensorFormat_t format, |
| cudnnDataType_t dataType, /* image data type */ |
| int n, /* number of inputs (batch size) */ |
| int c, /* number of input feature maps */ |
| int h, /* height of input section */ |
| int w) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnTensorDescriptor_t, cudnnTensorFormat_t, cudnnDataType_t, int, int, int, int); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetTensor4dDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc, format, dataType, n, c, h, w); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetTensor4dDescriptorEx(cudnnTensorDescriptor_t tensorDesc, |
| cudnnDataType_t dataType, /* image data type */ |
| int n, /* number of inputs (batch size) */ |
| int c, /* number of input feature maps */ |
| int h, /* height of input section */ |
| int w, /* width of input section */ |
| int nStride, |
| int cStride, |
| int hStride, |
| int wStride) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnTensorDescriptor_t, cudnnDataType_t, int, int, int, int, int, int, int, int); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetTensor4dDescriptorEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc, dataType, n, c, h, w, nStride, cStride, hStride, wStride); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetTensor4dDescriptor(const cudnnTensorDescriptor_t tensorDesc, |
| cudnnDataType_t *dataType, /* image data type */ |
| int *n, /* number of inputs (batch size) */ |
| int *c, /* number of input feature maps */ |
| int *h, /* height of input section */ |
| int *w, /* width of input section */ |
| int *nStride, |
| int *cStride, |
| int *hStride, |
| int *wStride) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnTensorDescriptor_t, cudnnDataType_t *, int *, int *, int *, int *, int *, int *, int *, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetTensor4dDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc, dataType, n, c, h, w, nStride, cStride, hStride, wStride); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetTensorNdDescriptor(cudnnTensorDescriptor_t tensorDesc, |
| cudnnDataType_t dataType, |
| int nbDims, |
| const int dimA[], |
| const int strideA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnTensorDescriptor_t, cudnnDataType_t, int, const int [], const int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetTensorNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc, dataType, nbDims, dimA, strideA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetTensorNdDescriptorEx(cudnnTensorDescriptor_t tensorDesc, |
| cudnnTensorFormat_t format, |
| cudnnDataType_t dataType, |
| int nbDims, |
| const int dimA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnTensorDescriptor_t, cudnnTensorFormat_t, cudnnDataType_t, int, const int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetTensorNdDescriptorEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc, format, dataType, nbDims, dimA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetTensorNdDescriptor(const cudnnTensorDescriptor_t tensorDesc, |
| int nbDimsRequested, |
| cudnnDataType_t *dataType, |
| int *nbDims, |
| int dimA[], |
| int strideA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnTensorDescriptor_t, int, cudnnDataType_t *, int *, int [], int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetTensorNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc, nbDimsRequested, dataType, nbDims, dimA, strideA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetTensorSizeInBytes(const cudnnTensorDescriptor_t tensorDesc, size_t *size) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnTensorDescriptor_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetTensorSizeInBytes"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc, size); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyTensorDescriptor(cudnnTensorDescriptor_t tensorDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnTensorDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(tensorDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnTransformTensor(cudnnHandle_t handle, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnTransformTensor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, alpha, xDesc, x, beta, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnAddTensor(cudnnHandle_t handle, |
| const void *alpha, |
| const cudnnTensorDescriptor_t aDesc, |
| const void *A, |
| const void *beta, |
| const cudnnTensorDescriptor_t cDesc, |
| void *C) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnAddTensor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, alpha, aDesc, A, beta, cDesc, C); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateOpTensorDescriptor(cudnnOpTensorDescriptor_t *opTensorDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnOpTensorDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateOpTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(opTensorDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetOpTensorDescriptor(cudnnOpTensorDescriptor_t opTensorDesc, |
| cudnnOpTensorOp_t opTensorOp, |
| cudnnDataType_t opTensorCompType, |
| cudnnNanPropagation_t opTensorNanOpt) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnOpTensorDescriptor_t, cudnnOpTensorOp_t, cudnnDataType_t, cudnnNanPropagation_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetOpTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(opTensorDesc, opTensorOp, opTensorCompType, opTensorNanOpt); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetOpTensorDescriptor(const cudnnOpTensorDescriptor_t opTensorDesc, |
| cudnnOpTensorOp_t *opTensorOp, |
| cudnnDataType_t *opTensorCompType, |
| cudnnNanPropagation_t *opTensorNanOpt) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnOpTensorDescriptor_t, cudnnOpTensorOp_t *, cudnnDataType_t *, cudnnNanPropagation_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetOpTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(opTensorDesc, opTensorOp, opTensorCompType, opTensorNanOpt); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyOpTensorDescriptor(cudnnOpTensorDescriptor_t opTensorDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnOpTensorDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyOpTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(opTensorDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnOpTensor(cudnnHandle_t handle, |
| const cudnnOpTensorDescriptor_t opTensorDesc, |
| const void *alpha1, |
| const cudnnTensorDescriptor_t aDesc, |
| const void *A, |
| const void *alpha2, |
| const cudnnTensorDescriptor_t bDesc, |
| const void *B, |
| const void *beta, |
| const cudnnTensorDescriptor_t cDesc, |
| void *C) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnOpTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnOpTensor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, opTensorDesc, alpha1, aDesc, A, alpha2, bDesc, B, beta, cDesc, C); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateReduceTensorDescriptor(cudnnReduceTensorDescriptor_t *reduceTensorDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnReduceTensorDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateReduceTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(reduceTensorDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetReduceTensorDescriptor(cudnnReduceTensorDescriptor_t reduceTensorDesc, |
| cudnnReduceTensorOp_t reduceTensorOp, |
| cudnnDataType_t reduceTensorCompType, |
| cudnnNanPropagation_t reduceTensorNanOpt, |
| cudnnReduceTensorIndices_t reduceTensorIndices, |
| cudnnIndicesType_t reduceTensorIndicesType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnReduceTensorDescriptor_t, cudnnReduceTensorOp_t, cudnnDataType_t, cudnnNanPropagation_t, cudnnReduceTensorIndices_t, cudnnIndicesType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetReduceTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(reduceTensorDesc, reduceTensorOp, reduceTensorCompType, reduceTensorNanOpt, reduceTensorIndices, reduceTensorIndicesType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetReduceTensorDescriptor(const cudnnReduceTensorDescriptor_t reduceTensorDesc, |
| cudnnReduceTensorOp_t *reduceTensorOp, |
| cudnnDataType_t *reduceTensorCompType, |
| cudnnNanPropagation_t *reduceTensorNanOpt, |
| cudnnReduceTensorIndices_t *reduceTensorIndices, |
| cudnnIndicesType_t *reduceTensorIndicesType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnReduceTensorDescriptor_t, cudnnReduceTensorOp_t *, cudnnDataType_t *, cudnnNanPropagation_t *, cudnnReduceTensorIndices_t *, cudnnIndicesType_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetReduceTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(reduceTensorDesc, reduceTensorOp, reduceTensorCompType, reduceTensorNanOpt, reduceTensorIndices, reduceTensorIndicesType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyReduceTensorDescriptor(cudnnReduceTensorDescriptor_t reduceTensorDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnReduceTensorDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyReduceTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(reduceTensorDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetReductionIndicesSize(cudnnHandle_t handle, |
| const cudnnReduceTensorDescriptor_t reduceTensorDesc, |
| const cudnnTensorDescriptor_t aDesc, |
| const cudnnTensorDescriptor_t cDesc, |
| size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnReduceTensorDescriptor_t, const cudnnTensorDescriptor_t, const cudnnTensorDescriptor_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetReductionIndicesSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, reduceTensorDesc, aDesc, cDesc, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetReductionWorkspaceSize(cudnnHandle_t handle, |
| const cudnnReduceTensorDescriptor_t reduceTensorDesc, |
| const cudnnTensorDescriptor_t aDesc, |
| const cudnnTensorDescriptor_t cDesc, |
| size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnReduceTensorDescriptor_t, const cudnnTensorDescriptor_t, const cudnnTensorDescriptor_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetReductionWorkspaceSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, reduceTensorDesc, aDesc, cDesc, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnReduceTensor(cudnnHandle_t handle, |
| const cudnnReduceTensorDescriptor_t reduceTensorDesc, |
| void *indices, |
| size_t indicesSizeInBytes, |
| void *workspace, |
| size_t workspaceSizeInBytes, |
| const void *alpha, |
| const cudnnTensorDescriptor_t aDesc, |
| const void *A, |
| const void *beta, |
| const cudnnTensorDescriptor_t cDesc, |
| void *C) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnReduceTensorDescriptor_t, void *, size_t, void *, size_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnReduceTensor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, reduceTensorDesc, indices, indicesSizeInBytes, workspace, workspaceSizeInBytes, alpha, aDesc, A, beta, cDesc, C); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetTensor(cudnnHandle_t handle, const cudnnTensorDescriptor_t yDesc, void *y, const void *valuePtr) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, void *, const void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetTensor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, yDesc, y, valuePtr); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnScaleTensor(cudnnHandle_t handle, const cudnnTensorDescriptor_t yDesc, void *y, const void *alpha) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, void *, const void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnScaleTensor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, yDesc, y, alpha); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateFilterDescriptor(cudnnFilterDescriptor_t *filterDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnFilterDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateFilterDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(filterDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetFilter4dDescriptor(cudnnFilterDescriptor_t filterDesc, |
| cudnnDataType_t dataType, /* image data type */ |
| cudnnTensorFormat_t format, |
| int k, /* number of output feature maps */ |
| int c, /* number of input feature maps */ |
| int h, /* height of each input filter */ |
| int w) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnFilterDescriptor_t, cudnnDataType_t, cudnnTensorFormat_t, int, int, int, int); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetFilter4dDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(filterDesc, dataType, format, k, c, h, w); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetFilter4dDescriptor(const cudnnFilterDescriptor_t filterDesc, |
| cudnnDataType_t *dataType, /* image data type */ |
| cudnnTensorFormat_t *format, |
| int *k, /* number of output feature maps */ |
| int *c, /* number of input feature maps */ |
| int *h, /* height of each input filter */ |
| int *w) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnFilterDescriptor_t, cudnnDataType_t *, cudnnTensorFormat_t *, int *, int *, int *, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetFilter4dDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(filterDesc, dataType, format, k, c, h, w); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetFilterNdDescriptor(cudnnFilterDescriptor_t filterDesc, |
| cudnnDataType_t dataType, /* image data type */ |
| cudnnTensorFormat_t format, |
| int nbDims, |
| const int filterDimA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnFilterDescriptor_t, cudnnDataType_t, cudnnTensorFormat_t, int, const int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetFilterNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(filterDesc, dataType, format, nbDims, filterDimA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetFilterNdDescriptor(const cudnnFilterDescriptor_t filterDesc, |
| int nbDimsRequested, |
| cudnnDataType_t *dataType, /* image data type */ |
| cudnnTensorFormat_t *format, |
| int *nbDims, |
| int filterDimA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnFilterDescriptor_t, int, cudnnDataType_t *, cudnnTensorFormat_t *, int *, int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetFilterNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(filterDesc, nbDimsRequested, dataType, format, nbDims, filterDimA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyFilterDescriptor(cudnnFilterDescriptor_t filterDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnFilterDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyFilterDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(filterDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateConvolutionDescriptor(cudnnConvolutionDescriptor_t *convDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnConvolutionDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateConvolutionDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetConvolutionMathType(cudnnConvolutionDescriptor_t convDesc, cudnnMathType_t mathType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnConvolutionDescriptor_t, cudnnMathType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetConvolutionMathType"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, mathType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionMathType(cudnnConvolutionDescriptor_t convDesc, cudnnMathType_t *mathType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnConvolutionDescriptor_t, cudnnMathType_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionMathType"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, mathType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetConvolutionGroupCount(cudnnConvolutionDescriptor_t convDesc, int groupCount) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnConvolutionDescriptor_t, int); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetConvolutionGroupCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, groupCount); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionGroupCount(cudnnConvolutionDescriptor_t convDesc, int *groupCount) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnConvolutionDescriptor_t, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionGroupCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, groupCount); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetConvolution2dDescriptor(cudnnConvolutionDescriptor_t convDesc, |
| int pad_h, /* zero-padding height */ |
| int pad_w, /* zero-padding width */ |
| int u, /* vertical filter stride */ |
| int v, /* horizontal filter stride */ |
| int dilation_h, /* filter dilation in the vertical dimension */ |
| int dilation_w, /* filter dilation in the horizontal dimension */ |
| cudnnConvolutionMode_t mode, |
| cudnnDataType_t computeType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnConvolutionDescriptor_t, int, int, int, int, int, int, cudnnConvolutionMode_t, cudnnDataType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetConvolution2dDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, pad_h, pad_w, u, v, dilation_h, dilation_w, mode, computeType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolution2dDescriptor(const cudnnConvolutionDescriptor_t convDesc, |
| int *pad_h, /* zero-padding height */ |
| int *pad_w, /* zero-padding width */ |
| int *u, /* vertical filter stride */ |
| int *v, /* horizontal filter stride */ |
| int *dilation_h, /* filter dilation in the vertical dimension */ |
| int *dilation_w, /* filter dilation in the horizontal dimension */ |
| cudnnConvolutionMode_t *mode, |
| cudnnDataType_t *computeType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnConvolutionDescriptor_t, int *, int *, int *, int *, int *, int *, cudnnConvolutionMode_t *, cudnnDataType_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolution2dDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, pad_h, pad_w, u, v, dilation_h, dilation_w, mode, computeType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolution2dForwardOutputDim(const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t inputTensorDesc, |
| const cudnnFilterDescriptor_t filterDesc, |
| int *n, |
| int *c, |
| int *h, |
| int *w) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, const cudnnFilterDescriptor_t, int *, int *, int *, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolution2dForwardOutputDim"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, inputTensorDesc, filterDesc, n, c, h, w); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetConvolutionNdDescriptor(cudnnConvolutionDescriptor_t convDesc, |
| int arrayLength, /* nbDims-2 size */ |
| const int padA[], |
| const int filterStrideA[], |
| const int dilationA[], |
| cudnnConvolutionMode_t mode, |
| cudnnDataType_t computeType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnConvolutionDescriptor_t, int, const int [], const int [], const int [], cudnnConvolutionMode_t, cudnnDataType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetConvolutionNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, arrayLength, padA, filterStrideA, dilationA, mode, computeType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionNdDescriptor(const cudnnConvolutionDescriptor_t convDesc, |
| int arrayLengthRequested, |
| int *arrayLength, |
| int padA[], |
| int strideA[], |
| int dilationA[], |
| cudnnConvolutionMode_t *mode, |
| cudnnDataType_t *computeType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnConvolutionDescriptor_t, int, int *, int [], int [], int [], cudnnConvolutionMode_t *, cudnnDataType_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, arrayLengthRequested, arrayLength, padA, strideA, dilationA, mode, computeType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionNdForwardOutputDim(const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t inputTensorDesc, |
| const cudnnFilterDescriptor_t filterDesc, |
| int nbDims, |
| int tensorOutputDimA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, const cudnnFilterDescriptor_t, int, int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionNdForwardOutputDim"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc, inputTensorDesc, filterDesc, nbDims, tensorOutputDimA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyConvolutionDescriptor(cudnnConvolutionDescriptor_t convDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnConvolutionDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyConvolutionDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(convDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionForwardAlgorithmMaxCount(cudnnHandle_t handle, int *count) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionForwardAlgorithmMaxCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, count); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindConvolutionForwardAlgorithm(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const cudnnFilterDescriptor_t wDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t yDesc, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionFwdAlgoPerf_t *perfResults) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnFilterDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, const int, int *, cudnnConvolutionFwdAlgoPerf_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindConvolutionForwardAlgorithm"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, wDesc, convDesc, yDesc, requestedAlgoCount, returnedAlgoCount, perfResults); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindConvolutionForwardAlgorithmEx(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionFwdAlgoPerf_t *perfResults, |
| void *workSpace, |
| size_t workSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, void *, const int, int *, cudnnConvolutionFwdAlgoPerf_t *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindConvolutionForwardAlgorithmEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, x, wDesc, w, convDesc, yDesc, y, requestedAlgoCount, returnedAlgoCount, perfResults, workSpace, workSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionForwardAlgorithm(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const cudnnFilterDescriptor_t wDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t yDesc, |
| cudnnConvolutionFwdPreference_t preference, |
| size_t memoryLimitInBytes, |
| cudnnConvolutionFwdAlgo_t *algo) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnFilterDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, cudnnConvolutionFwdPreference_t, size_t, cudnnConvolutionFwdAlgo_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionForwardAlgorithm"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, wDesc, convDesc, yDesc, preference, memoryLimitInBytes, algo); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionForwardAlgorithm_v7(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t srcDesc, |
| const cudnnFilterDescriptor_t filterDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t destDesc, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionFwdAlgoPerf_t *perfResults) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnFilterDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, const int, int *, cudnnConvolutionFwdAlgoPerf_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionForwardAlgorithm_v7"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, srcDesc, filterDesc, convDesc, destDesc, requestedAlgoCount, returnedAlgoCount, perfResults); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionForwardWorkspaceSize(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const cudnnFilterDescriptor_t wDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t yDesc, |
| cudnnConvolutionFwdAlgo_t algo, |
| size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnFilterDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, cudnnConvolutionFwdAlgo_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionForwardWorkspaceSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, wDesc, convDesc, yDesc, algo, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnConvolutionForward(cudnnHandle_t handle, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnConvolutionDescriptor_t convDesc, |
| cudnnConvolutionFwdAlgo_t algo, |
| void *workSpace, |
| size_t workSpaceSizeInBytes, |
| const void *beta, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnConvolutionDescriptor_t, cudnnConvolutionFwdAlgo_t, void *, size_t, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnConvolutionForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, alpha, xDesc, x, wDesc, w, convDesc, algo, workSpace, workSpaceSizeInBytes, beta, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnConvolutionBiasActivationForward(cudnnHandle_t handle, |
| const void *alpha1, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnConvolutionDescriptor_t convDesc, |
| cudnnConvolutionFwdAlgo_t algo, |
| void *workSpace, |
| size_t workSpaceSizeInBytes, |
| const void *alpha2, |
| const cudnnTensorDescriptor_t zDesc, |
| const void *z, |
| const cudnnTensorDescriptor_t biasDesc, |
| const void *bias, |
| const cudnnActivationDescriptor_t activationDesc, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnConvolutionDescriptor_t, cudnnConvolutionFwdAlgo_t, void *, size_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnActivationDescriptor_t, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnConvolutionBiasActivationForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, alpha1, xDesc, x, wDesc, w, convDesc, algo, workSpace, workSpaceSizeInBytes, alpha2, zDesc, z, biasDesc, bias, activationDesc, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnConvolutionBackwardBias(cudnnHandle_t handle, |
| const void *alpha, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const void *beta, |
| const cudnnTensorDescriptor_t dbDesc, |
| void *db) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnConvolutionBackwardBias"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, alpha, dyDesc, dy, beta, dbDesc, db); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionBackwardFilterAlgorithmMaxCount(cudnnHandle_t handle, int *count) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionBackwardFilterAlgorithmMaxCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, count); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindConvolutionBackwardFilterAlgorithm(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const cudnnTensorDescriptor_t dyDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnFilterDescriptor_t dwDesc, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionBwdFilterAlgoPerf_t *perfResults) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnTensorDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnFilterDescriptor_t, const int, int *, cudnnConvolutionBwdFilterAlgoPerf_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindConvolutionBackwardFilterAlgorithm"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, dyDesc, convDesc, dwDesc, requestedAlgoCount, returnedAlgoCount, perfResults); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindConvolutionBackwardFilterAlgorithmEx(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *y, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnFilterDescriptor_t dwDesc, |
| void *dw, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionBwdFilterAlgoPerf_t *perfResults, |
| void *workSpace, |
| size_t workSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnConvolutionDescriptor_t, const cudnnFilterDescriptor_t, void *, const int, int *, cudnnConvolutionBwdFilterAlgoPerf_t *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindConvolutionBackwardFilterAlgorithmEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, x, dyDesc, y, convDesc, dwDesc, dw, requestedAlgoCount, returnedAlgoCount, perfResults, workSpace, workSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionBackwardFilterAlgorithm(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const cudnnTensorDescriptor_t dyDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnFilterDescriptor_t dwDesc, |
| cudnnConvolutionBwdFilterPreference_t preference, |
| size_t memoryLimitInBytes, |
| cudnnConvolutionBwdFilterAlgo_t *algo) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnTensorDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnFilterDescriptor_t, cudnnConvolutionBwdFilterPreference_t, size_t, cudnnConvolutionBwdFilterAlgo_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionBackwardFilterAlgorithm"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, dyDesc, convDesc, dwDesc, preference, memoryLimitInBytes, algo); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionBackwardFilterAlgorithm_v7(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t srcDesc, |
| const cudnnTensorDescriptor_t diffDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnFilterDescriptor_t gradDesc, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionBwdFilterAlgoPerf_t *perfResults) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnTensorDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnFilterDescriptor_t, const int, int *, cudnnConvolutionBwdFilterAlgoPerf_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionBackwardFilterAlgorithm_v7"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, srcDesc, diffDesc, convDesc, gradDesc, requestedAlgoCount, returnedAlgoCount, perfResults); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionBackwardFilterWorkspaceSize(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const cudnnTensorDescriptor_t dyDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnFilterDescriptor_t gradDesc, |
| cudnnConvolutionBwdFilterAlgo_t algo, |
| size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnTensorDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnFilterDescriptor_t, cudnnConvolutionBwdFilterAlgo_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionBackwardFilterWorkspaceSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, dyDesc, convDesc, gradDesc, algo, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnConvolutionBackwardFilter(cudnnHandle_t handle, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const cudnnConvolutionDescriptor_t convDesc, |
| cudnnConvolutionBwdFilterAlgo_t algo, |
| void *workSpace, |
| size_t workSpaceSizeInBytes, |
| const void *beta, |
| const cudnnFilterDescriptor_t dwDesc, |
| void *dw) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnConvolutionDescriptor_t, cudnnConvolutionBwdFilterAlgo_t, void *, size_t, const void *, const cudnnFilterDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnConvolutionBackwardFilter"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, alpha, xDesc, x, dyDesc, dy, convDesc, algo, workSpace, workSpaceSizeInBytes, beta, dwDesc, dw); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionBackwardDataAlgorithmMaxCount(cudnnHandle_t handle, int *count) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionBackwardDataAlgorithmMaxCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, count); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindConvolutionBackwardDataAlgorithm(cudnnHandle_t handle, |
| const cudnnFilterDescriptor_t wDesc, |
| const cudnnTensorDescriptor_t dyDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t dxDesc, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionBwdDataAlgoPerf_t *perfResults) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnFilterDescriptor_t, const cudnnTensorDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, const int, int *, cudnnConvolutionBwdDataAlgoPerf_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindConvolutionBackwardDataAlgorithm"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, wDesc, dyDesc, convDesc, dxDesc, requestedAlgoCount, returnedAlgoCount, perfResults); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindConvolutionBackwardDataAlgorithmEx(cudnnHandle_t handle, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t dxDesc, |
| void *dx, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionBwdDataAlgoPerf_t *perfResults, |
| void *workSpace, |
| size_t workSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, void *, const int, int *, cudnnConvolutionBwdDataAlgoPerf_t *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindConvolutionBackwardDataAlgorithmEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, wDesc, w, dyDesc, dy, convDesc, dxDesc, dx, requestedAlgoCount, returnedAlgoCount, perfResults, workSpace, workSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionBackwardDataAlgorithm(cudnnHandle_t handle, |
| const cudnnFilterDescriptor_t wDesc, |
| const cudnnTensorDescriptor_t dyDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t dxDesc, |
| cudnnConvolutionBwdDataPreference_t preference, |
| size_t memoryLimitInBytes, |
| cudnnConvolutionBwdDataAlgo_t *algo) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnFilterDescriptor_t, const cudnnTensorDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, cudnnConvolutionBwdDataPreference_t, size_t, cudnnConvolutionBwdDataAlgo_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionBackwardDataAlgorithm"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, wDesc, dyDesc, convDesc, dxDesc, preference, memoryLimitInBytes, algo); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionBackwardDataAlgorithm_v7(cudnnHandle_t handle, |
| const cudnnFilterDescriptor_t filterDesc, |
| const cudnnTensorDescriptor_t diffDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t gradDesc, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnConvolutionBwdDataAlgoPerf_t *perfResults) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnFilterDescriptor_t, const cudnnTensorDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, const int, int *, cudnnConvolutionBwdDataAlgoPerf_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionBackwardDataAlgorithm_v7"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, filterDesc, diffDesc, convDesc, gradDesc, requestedAlgoCount, returnedAlgoCount, perfResults); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetConvolutionBackwardDataWorkspaceSize(cudnnHandle_t handle, |
| const cudnnFilterDescriptor_t wDesc, |
| const cudnnTensorDescriptor_t dyDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| const cudnnTensorDescriptor_t dxDesc, |
| cudnnConvolutionBwdDataAlgo_t algo, |
| size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnFilterDescriptor_t, const cudnnTensorDescriptor_t, const cudnnConvolutionDescriptor_t, const cudnnTensorDescriptor_t, cudnnConvolutionBwdDataAlgo_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetConvolutionBackwardDataWorkspaceSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, wDesc, dyDesc, convDesc, dxDesc, algo, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnConvolutionBackwardData(cudnnHandle_t handle, |
| const void *alpha, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const cudnnConvolutionDescriptor_t convDesc, |
| cudnnConvolutionBwdDataAlgo_t algo, |
| void *workSpace, |
| size_t workSpaceSizeInBytes, |
| const void *beta, |
| const cudnnTensorDescriptor_t dxDesc, |
| void *dx) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnConvolutionDescriptor_t, cudnnConvolutionBwdDataAlgo_t, void *, size_t, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnConvolutionBackwardData"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, alpha, wDesc, w, dyDesc, dy, convDesc, algo, workSpace, workSpaceSizeInBytes, beta, dxDesc, dx); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnIm2Col(cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const cudnnFilterDescriptor_t wDesc, |
| const cudnnConvolutionDescriptor_t convDesc, |
| void *colBuffer) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const cudnnConvolutionDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnIm2Col"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, xDesc, x, wDesc, convDesc, colBuffer); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSoftmaxForward(cudnnHandle_t handle, |
| cudnnSoftmaxAlgorithm_t algo, |
| cudnnSoftmaxMode_t mode, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnSoftmaxAlgorithm_t, cudnnSoftmaxMode_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSoftmaxForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, algo, mode, alpha, xDesc, x, beta, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSoftmaxBackward(cudnnHandle_t handle, |
| cudnnSoftmaxAlgorithm_t algo, |
| cudnnSoftmaxMode_t mode, |
| const void *alpha, |
| const cudnnTensorDescriptor_t yDesc, |
| const void *y, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const void *beta, |
| const cudnnTensorDescriptor_t dxDesc, |
| void *dx) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnSoftmaxAlgorithm_t, cudnnSoftmaxMode_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSoftmaxBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, algo, mode, alpha, yDesc, y, dyDesc, dy, beta, dxDesc, dx); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreatePoolingDescriptor(cudnnPoolingDescriptor_t *poolingDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnPoolingDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreatePoolingDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(poolingDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetPooling2dDescriptor(cudnnPoolingDescriptor_t poolingDesc, |
| cudnnPoolingMode_t mode, |
| cudnnNanPropagation_t maxpoolingNanOpt, |
| int windowHeight, |
| int windowWidth, |
| int verticalPadding, |
| int horizontalPadding, |
| int verticalStride, |
| int horizontalStride) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnPoolingDescriptor_t, cudnnPoolingMode_t, cudnnNanPropagation_t, int, int, int, int, int, int); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetPooling2dDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(poolingDesc, mode, maxpoolingNanOpt, windowHeight, windowWidth, verticalPadding, horizontalPadding, verticalStride, horizontalStride); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetPooling2dDescriptor(const cudnnPoolingDescriptor_t poolingDesc, |
| cudnnPoolingMode_t *mode, |
| cudnnNanPropagation_t *maxpoolingNanOpt, |
| int *windowHeight, |
| int *windowWidth, |
| int *verticalPadding, |
| int *horizontalPadding, |
| int *verticalStride, |
| int *horizontalStride) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnPoolingDescriptor_t, cudnnPoolingMode_t *, cudnnNanPropagation_t *, int *, int *, int *, int *, int *, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetPooling2dDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(poolingDesc, mode, maxpoolingNanOpt, windowHeight, windowWidth, verticalPadding, horizontalPadding, verticalStride, horizontalStride); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetPoolingNdDescriptor(cudnnPoolingDescriptor_t poolingDesc, |
| const cudnnPoolingMode_t mode, |
| const cudnnNanPropagation_t maxpoolingNanOpt, |
| int nbDims, |
| const int windowDimA[], |
| const int paddingA[], |
| const int strideA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnPoolingDescriptor_t, const cudnnPoolingMode_t, const cudnnNanPropagation_t, int, const int [], const int [], const int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetPoolingNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(poolingDesc, mode, maxpoolingNanOpt, nbDims, windowDimA, paddingA, strideA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetPoolingNdDescriptor(const cudnnPoolingDescriptor_t poolingDesc, |
| int nbDimsRequested, |
| cudnnPoolingMode_t *mode, |
| cudnnNanPropagation_t *maxpoolingNanOpt, |
| int *nbDims, |
| int windowDimA[], |
| int paddingA[], |
| int strideA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnPoolingDescriptor_t, int, cudnnPoolingMode_t *, cudnnNanPropagation_t *, int *, int [], int [], int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetPoolingNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(poolingDesc, nbDimsRequested, mode, maxpoolingNanOpt, nbDims, windowDimA, paddingA, strideA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetPoolingNdForwardOutputDim(const cudnnPoolingDescriptor_t poolingDesc, |
| const cudnnTensorDescriptor_t inputTensorDesc, |
| int nbDims, |
| int outputTensorDimA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnPoolingDescriptor_t, const cudnnTensorDescriptor_t, int, int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetPoolingNdForwardOutputDim"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(poolingDesc, inputTensorDesc, nbDims, outputTensorDimA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetPooling2dForwardOutputDim(const cudnnPoolingDescriptor_t poolingDesc, |
| const cudnnTensorDescriptor_t inputTensorDesc, |
| int *n, |
| int *c, |
| int *h, |
| int *w) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnPoolingDescriptor_t, const cudnnTensorDescriptor_t, int *, int *, int *, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetPooling2dForwardOutputDim"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(poolingDesc, inputTensorDesc, n, c, h, w); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyPoolingDescriptor(cudnnPoolingDescriptor_t poolingDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnPoolingDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyPoolingDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(poolingDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnPoolingForward(cudnnHandle_t handle, |
| const cudnnPoolingDescriptor_t poolingDesc, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnPoolingDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnPoolingForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, poolingDesc, alpha, xDesc, x, beta, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnPoolingBackward(cudnnHandle_t handle, |
| const cudnnPoolingDescriptor_t poolingDesc, |
| const void *alpha, |
| const cudnnTensorDescriptor_t yDesc, |
| const void *y, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t dxDesc, |
| void *dx) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnPoolingDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnPoolingBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, poolingDesc, alpha, yDesc, y, dyDesc, dy, xDesc, x, beta, dxDesc, dx); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateActivationDescriptor(cudnnActivationDescriptor_t *activationDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnActivationDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateActivationDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(activationDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetActivationDescriptor(cudnnActivationDescriptor_t activationDesc, |
| cudnnActivationMode_t mode, |
| cudnnNanPropagation_t reluNanOpt, |
| double coef) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnActivationDescriptor_t, cudnnActivationMode_t, cudnnNanPropagation_t, double); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetActivationDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(activationDesc, mode, reluNanOpt, coef); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetActivationDescriptor(const cudnnActivationDescriptor_t activationDesc, |
| cudnnActivationMode_t *mode, |
| cudnnNanPropagation_t *reluNanOpt, |
| double *coef) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnActivationDescriptor_t, cudnnActivationMode_t *, cudnnNanPropagation_t *, double *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetActivationDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(activationDesc, mode, reluNanOpt, coef); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyActivationDescriptor(cudnnActivationDescriptor_t activationDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnActivationDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyActivationDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(activationDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnActivationForward(cudnnHandle_t handle, |
| cudnnActivationDescriptor_t activationDesc, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnActivationDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnActivationForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, activationDesc, alpha, xDesc, x, beta, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnActivationBackward(cudnnHandle_t handle, |
| cudnnActivationDescriptor_t activationDesc, |
| const void *alpha, |
| const cudnnTensorDescriptor_t yDesc, |
| const void *y, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t dxDesc, |
| void *dx) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnActivationDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnActivationBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, activationDesc, alpha, yDesc, y, dyDesc, dy, xDesc, x, beta, dxDesc, dx); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateLRNDescriptor(cudnnLRNDescriptor_t *normDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnLRNDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateLRNDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(normDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetLRNDescriptor(cudnnLRNDescriptor_t normDesc, unsigned lrnN, double lrnAlpha, double lrnBeta, double lrnK) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnLRNDescriptor_t, unsigned int, double, double, double); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetLRNDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(normDesc, lrnN, lrnAlpha, lrnBeta, lrnK); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetLRNDescriptor(cudnnLRNDescriptor_t normDesc, unsigned *lrnN, double *lrnAlpha, double *lrnBeta, double *lrnK) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnLRNDescriptor_t, unsigned int *, double *, double *, double *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetLRNDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(normDesc, lrnN, lrnAlpha, lrnBeta, lrnK); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyLRNDescriptor(cudnnLRNDescriptor_t lrnDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnLRNDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyLRNDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(lrnDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnLRNCrossChannelForward(cudnnHandle_t handle, |
| cudnnLRNDescriptor_t normDesc, |
| cudnnLRNMode_t lrnMode, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnLRNDescriptor_t, cudnnLRNMode_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnLRNCrossChannelForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, normDesc, lrnMode, alpha, xDesc, x, beta, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnLRNCrossChannelBackward(cudnnHandle_t handle, |
| cudnnLRNDescriptor_t normDesc, |
| cudnnLRNMode_t lrnMode, |
| const void *alpha, |
| const cudnnTensorDescriptor_t yDesc, |
| const void *y, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t dxDesc, |
| void *dx) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnLRNDescriptor_t, cudnnLRNMode_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnLRNCrossChannelBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, normDesc, lrnMode, alpha, yDesc, y, dyDesc, dy, xDesc, x, beta, dxDesc, dx); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDivisiveNormalizationForward(cudnnHandle_t handle, |
| cudnnLRNDescriptor_t normDesc, |
| cudnnDivNormMode_t mode, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, /* same desc for means, temp, temp2 */ |
| const void *x, |
| const void *means, /* if NULL, means are assumed to be zero */ |
| void *temp, |
| void *temp2, |
| const void *beta, |
| const cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnLRNDescriptor_t, cudnnDivNormMode_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, void *, void *, const void *, const cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDivisiveNormalizationForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, normDesc, mode, alpha, xDesc, x, means, temp, temp2, beta, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDivisiveNormalizationBackward(cudnnHandle_t handle, |
| cudnnLRNDescriptor_t normDesc, |
| cudnnDivNormMode_t mode, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, /* same desc for x, means, dy, temp, temp2 */ |
| const void *x, |
| const void *means, /* if NULL, means are assumed to be zero */ |
| const void *dy, |
| void *temp, |
| void *temp2, |
| const void *beta, |
| const cudnnTensorDescriptor_t dXdMeansDesc, /* same desc for dx, dMeans */ |
| void *dx, /* output x differential */ |
| void *dMeans) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnLRNDescriptor_t, cudnnDivNormMode_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const void *, void *, void *, const void *, const cudnnTensorDescriptor_t, void *, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDivisiveNormalizationBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, normDesc, mode, alpha, xDesc, x, means, dy, temp, temp2, beta, dXdMeansDesc, dx, dMeans); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDeriveBNTensorDescriptor(cudnnTensorDescriptor_t derivedBnDesc, |
| const cudnnTensorDescriptor_t xDesc, |
| cudnnBatchNormMode_t mode) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnTensorDescriptor_t, const cudnnTensorDescriptor_t, cudnnBatchNormMode_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDeriveBNTensorDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(derivedBnDesc, xDesc, mode); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnBatchNormalizationForwardTraining( |
| cudnnHandle_t handle, |
| cudnnBatchNormMode_t mode, |
| |
| const void *alpha, /* alpha[0] = result blend factor */ |
| const void *beta, /* beta[0] = dest layer blend factor */ |
| |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, /* NxCxHxW */ |
| const cudnnTensorDescriptor_t yDesc, |
| void *y, /* NxCxHxW */ |
| |
| /* Shared desc for the next 6 tensors in the argument list. |
| Data type to be set as follows: |
| type = (typeOf(x) == double) ? double : float |
| Dimensions for this descriptor depend on normalization mode |
| - Spatial Normalization : tensors are expected to have dims 1xCx1x1 |
| (normalization is performed across NxHxW) |
| - Per-Activation Normalization : tensors are expected to have dims of 1xCxHxW |
| (normalization is performed across N) */ |
| const cudnnTensorDescriptor_t bnScaleBiasMeanVarDesc, |
| |
| /* 'Gamma' and 'Beta' respectively in Ioffe and Szegedy's paper's notation */ |
| const void *bnScale, |
| const void *bnBias, |
| |
| /* MUST use factor=1 in the very first call of a complete training cycle. |
| Use a factor=1/(1+n) at N-th call to the function to get |
| Cumulative Moving Average (CMA) behavior |
| CMA[n] = (x[1]+...+x[n])/n |
| Since CMA[n+1] = (n*CMA[n]+x[n+1])/(n+1) = |
| ((n+1)*CMA[n]-CMA[n])/(n+1) + x[n+1]/(n+1) = |
| CMA[n]*(1-1/(n+1)) + x[n+1]*1/(n+1) */ |
| double exponentialAverageFactor, |
| |
| /* Used in Training phase only. |
| runningMean = newMean*factor + runningMean*(1-factor) */ |
| void *resultRunningMean, |
| /* Output in training mode, input in inference. Is the moving average |
| of variance[x] (factor is applied in the same way as for runningMean) */ |
| void *resultRunningVariance, |
| |
| /* Has to be >= CUDNN_BN_MIN_EPSILON. Should be the same in forward and backward functions. */ |
| double epsilon, |
| |
| /* Optionally save intermediate results from the forward pass here |
| - can be reused to speed up backward pass. NULL if unused */ |
| void *resultSaveMean, |
| void *resultSaveInvVariance) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnBatchNormMode_t, const void *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, const void *, const void *, double, void *, void *, double, void *, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnBatchNormalizationForwardTraining"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, mode, alpha, beta, xDesc, x, yDesc, y, bnScaleBiasMeanVarDesc, bnScale, bnBias, exponentialAverageFactor, resultRunningMean, resultRunningVariance, epsilon, resultSaveMean, resultSaveInvVariance); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnBatchNormalizationForwardInference(cudnnHandle_t handle, |
| cudnnBatchNormMode_t mode, |
| const void *alpha, /* alpha[0] = result blend factor */ |
| const void *beta, /* beta[0] = dest layer blend factor */ |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, /* NxCxHxW */ |
| const cudnnTensorDescriptor_t yDesc, |
| void *y, /* NxCxHxW */ |
| const cudnnTensorDescriptor_t bnScaleBiasMeanVarDesc, |
| const void *bnScale, |
| const void *bnBias, |
| const void *estimatedMean, |
| const void *estimatedVariance, |
| double epsilon) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnBatchNormMode_t, const void *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, const void *, const void *, const void *, const void *, double); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnBatchNormalizationForwardInference"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, mode, alpha, beta, xDesc, x, yDesc, y, bnScaleBiasMeanVarDesc, bnScale, bnBias, estimatedMean, estimatedVariance, epsilon); |
| } |
| |
| 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, /* same desc for x, dx, dy */ |
| const void *x, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const cudnnTensorDescriptor_t dxDesc, |
| void *dx, |
| /* Shared tensor desc for the 4 tensors below */ |
| const cudnnTensorDescriptor_t dBnScaleBiasDesc, |
| const void *bnScale, /* bnBias doesn't affect backpropagation */ |
| /* scale and bias diff are not backpropagated below this layer */ |
| void *dBnScaleResult, |
| void *dBnBiasResult, |
| /* Same epsilon as forward pass */ |
| double epsilon, |
| |
| /* Optionally cached intermediate results from |
| forward pass */ |
| const void *savedMean, |
| const void *savedInvVariance) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnBatchNormMode_t, const void *, const void *, const void *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, const void *, void *, void *, double, const void *, const void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnBatchNormalizationBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, mode, alphaDataDiff, betaDataDiff, alphaParamDiff, betaParamDiff, xDesc, x, dyDesc, dy, dxDesc, dx, dBnScaleBiasDesc, bnScale, dBnScaleResult, dBnBiasResult, epsilon, savedMean, savedInvVariance); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateSpatialTransformerDescriptor(cudnnSpatialTransformerDescriptor_t *stDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnSpatialTransformerDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateSpatialTransformerDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(stDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetSpatialTransformerNdDescriptor(cudnnSpatialTransformerDescriptor_t stDesc, |
| cudnnSamplerType_t samplerType, |
| cudnnDataType_t dataType, |
| const int nbDims, |
| const int dimA[]) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnSpatialTransformerDescriptor_t, cudnnSamplerType_t, cudnnDataType_t, const int, const int []); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetSpatialTransformerNdDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(stDesc, samplerType, dataType, nbDims, dimA); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroySpatialTransformerDescriptor(cudnnSpatialTransformerDescriptor_t stDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnSpatialTransformerDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroySpatialTransformerDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(stDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSpatialTfGridGeneratorForward(cudnnHandle_t handle, |
| const cudnnSpatialTransformerDescriptor_t stDesc, |
| const void *theta, |
| void *grid) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnSpatialTransformerDescriptor_t, const void *, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSpatialTfGridGeneratorForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, stDesc, theta, grid); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSpatialTfGridGeneratorBackward(cudnnHandle_t handle, |
| const cudnnSpatialTransformerDescriptor_t stDesc, |
| const void *dgrid, |
| void *dtheta) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnSpatialTransformerDescriptor_t, const void *, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSpatialTfGridGeneratorBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, stDesc, dgrid, dtheta); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSpatialTfSamplerForward(cudnnHandle_t handle, |
| cudnnSpatialTransformerDescriptor_t stDesc, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *grid, |
| const void *beta, |
| cudnnTensorDescriptor_t yDesc, |
| void *y) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnSpatialTransformerDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const void *, cudnnTensorDescriptor_t, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSpatialTfSamplerForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, stDesc, alpha, xDesc, x, grid, beta, yDesc, y); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSpatialTfSamplerBackward(cudnnHandle_t handle, |
| cudnnSpatialTransformerDescriptor_t stDesc, |
| const void *alpha, |
| const cudnnTensorDescriptor_t xDesc, |
| const void *x, |
| const void *beta, |
| const cudnnTensorDescriptor_t dxDesc, |
| void *dx, |
| const void *alphaDgrid, |
| const cudnnTensorDescriptor_t dyDesc, |
| const void *dy, |
| const void *grid, |
| const void *betaDgrid, |
| void *dgrid) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnSpatialTransformerDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const cudnnTensorDescriptor_t, void *, const void *, const cudnnTensorDescriptor_t, const void *, const void *, const void *, void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSpatialTfSamplerBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, stDesc, alpha, xDesc, x, beta, dxDesc, dx, alphaDgrid, dyDesc, dy, grid, betaDgrid, dgrid); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateDropoutDescriptor(cudnnDropoutDescriptor_t *dropoutDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnDropoutDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateDropoutDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(dropoutDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyDropoutDescriptor(cudnnDropoutDescriptor_t dropoutDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnDropoutDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyDropoutDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(dropoutDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDropoutGetStatesSize(cudnnHandle_t handle, size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDropoutGetStatesSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDropoutGetReserveSpaceSize(cudnnTensorDescriptor_t xdesc, size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnTensorDescriptor_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDropoutGetReserveSpaceSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(xdesc, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetDropoutDescriptor(cudnnDropoutDescriptor_t dropoutDesc, |
| cudnnHandle_t handle, |
| float dropout, |
| void *states, |
| size_t stateSizeInBytes, |
| unsigned long long seed) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnDropoutDescriptor_t, cudnnHandle_t, float, void *, size_t, unsigned long long); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetDropoutDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(dropoutDesc, handle, dropout, states, stateSizeInBytes, seed); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRestoreDropoutDescriptor(cudnnDropoutDescriptor_t dropoutDesc, |
| cudnnHandle_t handle, |
| float dropout, |
| void *states, |
| size_t stateSizeInBytes, |
| unsigned long long seed) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnDropoutDescriptor_t, cudnnHandle_t, float, void *, size_t, unsigned long long); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRestoreDropoutDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(dropoutDesc, handle, dropout, states, stateSizeInBytes, seed); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetDropoutDescriptor(cudnnDropoutDescriptor_t dropoutDesc, |
| cudnnHandle_t handle, |
| float *dropout, |
| void **states, |
| unsigned long long *seed) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnDropoutDescriptor_t, cudnnHandle_t, float *, void **, unsigned long long *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetDropoutDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(dropoutDesc, handle, dropout, states, seed); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDropoutForward(cudnnHandle_t handle, |
| const cudnnDropoutDescriptor_t dropoutDesc, |
| const cudnnTensorDescriptor_t xdesc, |
| const void *x, |
| const cudnnTensorDescriptor_t ydesc, |
| void *y, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnDropoutDescriptor_t, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, void *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDropoutForward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, dropoutDesc, xdesc, x, ydesc, y, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDropoutBackward(cudnnHandle_t handle, |
| const cudnnDropoutDescriptor_t dropoutDesc, |
| const cudnnTensorDescriptor_t dydesc, |
| const void *dy, |
| const cudnnTensorDescriptor_t dxdesc, |
| void *dx, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnDropoutDescriptor_t, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, void *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDropoutBackward"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, dropoutDesc, dydesc, dy, dxdesc, dx, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateRNNDescriptor(cudnnRNNDescriptor_t *rnnDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateRNNDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyRNNDescriptor(cudnnRNNDescriptor_t rnnDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyRNNDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNForwardInferenceAlgorithmMaxCount(cudnnHandle_t handle, const cudnnRNNDescriptor_t rnnDesc, int *count) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNForwardInferenceAlgorithmMaxCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, count); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindRNNForwardInferenceAlgorithmEx(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t *yDesc, |
| void *y, |
| const cudnnTensorDescriptor_t hyDesc, |
| void *hy, |
| const cudnnTensorDescriptor_t cyDesc, |
| void *cy, |
| const float findIntensity, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnAlgorithmPerformance_t *perfResults, |
| void *workspace, |
| size_t workSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t *, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const float, const int, int *, cudnnAlgorithmPerformance_t *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindRNNForwardInferenceAlgorithmEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, xDesc, x, hxDesc, hx, cxDesc, cx, wDesc, w, yDesc, y, hyDesc, hy, cyDesc, cy, findIntensity, requestedAlgoCount, returnedAlgoCount, perfResults, workspace, workSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNForwardTrainingAlgorithmMaxCount(cudnnHandle_t handle, const cudnnRNNDescriptor_t rnnDesc, int *count) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNForwardTrainingAlgorithmMaxCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, count); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindRNNForwardTrainingAlgorithmEx(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t *yDesc, |
| void *y, |
| const cudnnTensorDescriptor_t hyDesc, |
| void *hy, |
| const cudnnTensorDescriptor_t cyDesc, |
| void *cy, |
| const float findIntensity, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnAlgorithmPerformance_t *perfResults, |
| void *workspace, |
| size_t workSpaceSizeInBytes, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t *, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const float, const int, int *, cudnnAlgorithmPerformance_t *, void *, size_t, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindRNNForwardTrainingAlgorithmEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, xDesc, x, hxDesc, hx, cxDesc, cx, wDesc, w, yDesc, y, hyDesc, hy, cyDesc, cy, findIntensity, requestedAlgoCount, returnedAlgoCount, perfResults, workspace, workSpaceSizeInBytes, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNBackwardDataAlgorithmMaxCount(cudnnHandle_t handle, const cudnnRNNDescriptor_t rnnDesc, int *count) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNBackwardDataAlgorithmMaxCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, count); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindRNNBackwardDataAlgorithmEx(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *yDesc, |
| const void *y, |
| const cudnnTensorDescriptor_t *dyDesc, |
| const void *dy, |
| const cudnnTensorDescriptor_t dhyDesc, |
| const void *dhy, |
| const cudnnTensorDescriptor_t dcyDesc, |
| const void *dcy, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnTensorDescriptor_t *dxDesc, |
| void *dx, |
| const cudnnTensorDescriptor_t dhxDesc, |
| void *dhx, |
| const cudnnTensorDescriptor_t dcxDesc, |
| void *dcx, |
| const float findIntensity, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnAlgorithmPerformance_t *perfResults, |
| void *workspace, |
| size_t workSpaceSizeInBytes, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t *, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const float, const int, int *, cudnnAlgorithmPerformance_t *, void *, size_t, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindRNNBackwardDataAlgorithmEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, yDesc, y, dyDesc, dy, dhyDesc, dhy, dcyDesc, dcy, wDesc, w, hxDesc, hx, cxDesc, cx, dxDesc, dx, dhxDesc, dhx, dcxDesc, dcx, findIntensity, requestedAlgoCount, returnedAlgoCount, perfResults, workspace, workSpaceSizeInBytes, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNBackwardWeightsAlgorithmMaxCount(cudnnHandle_t handle, const cudnnRNNDescriptor_t rnnDesc, int *count) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNBackwardWeightsAlgorithmMaxCount"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, count); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnFindRNNBackwardWeightsAlgorithmEx(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t *yDesc, |
| const void *y, |
| const float findIntensity, |
| const int requestedAlgoCount, |
| int *returnedAlgoCount, |
| cudnnAlgorithmPerformance_t *perfResults, |
| const void *workspace, |
| size_t workSpaceSizeInBytes, |
| const cudnnFilterDescriptor_t dwDesc, |
| void *dw, |
| const void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t *, const void *, const float, const int, int *, cudnnAlgorithmPerformance_t *, const void *, size_t, const cudnnFilterDescriptor_t, void *, const void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnFindRNNBackwardWeightsAlgorithmEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, xDesc, x, hxDesc, hx, yDesc, y, findIntensity, requestedAlgoCount, returnedAlgoCount, perfResults, workspace, workSpaceSizeInBytes, dwDesc, dw, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreatePersistentRNNPlan(cudnnRNNDescriptor_t rnnDesc, |
| const int minibatch, |
| const cudnnDataType_t dataType, |
| cudnnPersistentRNNPlan_t *plan) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t, const int, const cudnnDataType_t, cudnnPersistentRNNPlan_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreatePersistentRNNPlan"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc, minibatch, dataType, plan); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetPersistentRNNPlan(cudnnRNNDescriptor_t rnnDesc, cudnnPersistentRNNPlan_t plan) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t, cudnnPersistentRNNPlan_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetPersistentRNNPlan"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc, plan); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyPersistentRNNPlan(cudnnPersistentRNNPlan_t plan) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnPersistentRNNPlan_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyPersistentRNNPlan"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(plan); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetRNNDescriptor(cudnnHandle_t handle, |
| cudnnRNNDescriptor_t rnnDesc, |
| const int hiddenSize, |
| const int numLayers, |
| cudnnDropoutDescriptor_t dropoutDesc, /* Between layers, not between recurrent steps. */ |
| cudnnRNNInputMode_t inputMode, |
| cudnnDirectionMode_t direction, |
| cudnnRNNMode_t mode, |
| cudnnRNNAlgo_t algo, |
| cudnnDataType_t dataType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnRNNDescriptor_t, const int, const int, cudnnDropoutDescriptor_t, cudnnRNNInputMode_t, cudnnDirectionMode_t, cudnnRNNMode_t, cudnnRNNAlgo_t, cudnnDataType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetRNNDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, hiddenSize, numLayers, dropoutDesc, inputMode, direction, mode, algo, dataType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetRNNProjectionLayers(cudnnHandle_t handle, |
| cudnnRNNDescriptor_t rnnDesc, |
| const int recProjSize, |
| const int outProjSize) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnRNNDescriptor_t, const int, const int); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetRNNProjectionLayers"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, recProjSize, outProjSize); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNProjectionLayers(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| int *recProjSize, |
| int *outProjSize) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, int *, int *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNProjectionLayers"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, recProjSize, outProjSize); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetRNNAlgorithmDescriptor(cudnnHandle_t handle, cudnnRNNDescriptor_t rnnDesc, cudnnAlgorithmDescriptor_t algoDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnRNNDescriptor_t, cudnnAlgorithmDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetRNNAlgorithmDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, algoDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNDescriptor(cudnnHandle_t handle, |
| cudnnRNNDescriptor_t rnnDesc, |
| int *hiddenSize, |
| int *numLayers, |
| cudnnDropoutDescriptor_t *dropoutDesc, |
| cudnnRNNInputMode_t *inputMode, |
| cudnnDirectionMode_t *direction, |
| cudnnRNNMode_t *mode, |
| cudnnRNNAlgo_t *algo, |
| cudnnDataType_t *dataType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnRNNDescriptor_t, int *, int *, cudnnDropoutDescriptor_t *, cudnnRNNInputMode_t *, cudnnDirectionMode_t *, cudnnRNNMode_t *, cudnnRNNAlgo_t *, cudnnDataType_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, hiddenSize, numLayers, dropoutDesc, inputMode, direction, mode, algo, dataType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetRNNMatrixMathType(cudnnRNNDescriptor_t rnnDesc, cudnnMathType_t mType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t, cudnnMathType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetRNNMatrixMathType"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc, mType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNMatrixMathType(cudnnRNNDescriptor_t rnnDesc, cudnnMathType_t *mType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t, cudnnMathType_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNMatrixMathType"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc, mType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNWorkspaceSize(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *xDesc, |
| size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNWorkspaceSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, xDesc, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNTrainingReserveSize(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *xDesc, |
| size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNTrainingReserveSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, xDesc, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNParamsSize(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const cudnnTensorDescriptor_t xDesc, |
| size_t *sizeInBytes, |
| cudnnDataType_t dataType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const cudnnTensorDescriptor_t, size_t *, cudnnDataType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNParamsSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, xDesc, sizeInBytes, dataType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNLinLayerMatrixParams(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int pseudoLayer, |
| const cudnnTensorDescriptor_t xDesc, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const int linLayerID, |
| cudnnFilterDescriptor_t linLayerMatDesc, |
| void **linLayerMat) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t, const cudnnFilterDescriptor_t, const void *, const int, cudnnFilterDescriptor_t, void **); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNLinLayerMatrixParams"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, pseudoLayer, xDesc, wDesc, w, linLayerID, linLayerMatDesc, linLayerMat); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNLinLayerBiasParams(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int pseudoLayer, |
| const cudnnTensorDescriptor_t xDesc, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const int linLayerID, |
| cudnnFilterDescriptor_t linLayerBiasDesc, |
| void **linLayerBias) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t, const cudnnFilterDescriptor_t, const void *, const int, cudnnFilterDescriptor_t, void **); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNLinLayerBiasParams"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, pseudoLayer, xDesc, wDesc, w, linLayerID, linLayerBiasDesc, linLayerBias); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNForwardInference(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t *yDesc, |
| void *y, |
| const cudnnTensorDescriptor_t hyDesc, |
| void *hy, |
| const cudnnTensorDescriptor_t cyDesc, |
| void *cy, |
| void *workspace, |
| size_t workSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t *, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNForwardInference"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, xDesc, x, hxDesc, hx, cxDesc, cx, wDesc, w, yDesc, y, hyDesc, hy, cyDesc, cy, workspace, workSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNForwardTraining(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t *yDesc, |
| void *y, |
| const cudnnTensorDescriptor_t hyDesc, |
| void *hy, |
| const cudnnTensorDescriptor_t cyDesc, |
| void *cy, |
| void *workspace, |
| size_t workSpaceSizeInBytes, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t *, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, void *, size_t, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNForwardTraining"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, xDesc, x, hxDesc, hx, cxDesc, cx, wDesc, w, yDesc, y, hyDesc, hy, cyDesc, cy, workspace, workSpaceSizeInBytes, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNBackwardData(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *yDesc, |
| const void *y, |
| const cudnnTensorDescriptor_t *dyDesc, |
| const void *dy, |
| const cudnnTensorDescriptor_t dhyDesc, |
| const void *dhy, |
| const cudnnTensorDescriptor_t dcyDesc, |
| const void *dcy, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnTensorDescriptor_t *dxDesc, |
| void *dx, |
| const cudnnTensorDescriptor_t dhxDesc, |
| void *dhx, |
| const cudnnTensorDescriptor_t dcxDesc, |
| void *dcx, |
| void *workspace, |
| size_t workSpaceSizeInBytes, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t *, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, void *, size_t, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNBackwardData"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, yDesc, y, dyDesc, dy, dhyDesc, dhy, dcyDesc, dcy, wDesc, w, hxDesc, hx, cxDesc, cx, dxDesc, dx, dhxDesc, dhx, dcxDesc, dcx, workspace, workSpaceSizeInBytes, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNBackwardWeights(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const int seqLength, |
| const cudnnTensorDescriptor_t *xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t *yDesc, |
| const void *y, |
| const void *workspace, |
| size_t workSpaceSizeInBytes, |
| const cudnnFilterDescriptor_t dwDesc, |
| void *dw, |
| const void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const int, const cudnnTensorDescriptor_t *, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t *, const void *, const void *, size_t, const cudnnFilterDescriptor_t, void *, const void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNBackwardWeights"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, seqLength, xDesc, x, hxDesc, hx, yDesc, y, workspace, workSpaceSizeInBytes, dwDesc, dw, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateCTCLossDescriptor(cudnnCTCLossDescriptor_t *ctcLossDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnCTCLossDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateCTCLossDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(ctcLossDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetCTCLossDescriptor(cudnnCTCLossDescriptor_t ctcLossDesc, cudnnDataType_t compType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnCTCLossDescriptor_t, cudnnDataType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetCTCLossDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(ctcLossDesc, compType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetCTCLossDescriptor(cudnnCTCLossDescriptor_t ctcLossDesc, cudnnDataType_t *compType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnCTCLossDescriptor_t, cudnnDataType_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetCTCLossDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(ctcLossDesc, compType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyCTCLossDescriptor(cudnnCTCLossDescriptor_t ctcLossDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnCTCLossDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyCTCLossDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(ctcLossDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCTCLoss( |
| cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t |
| probsDesc, /* Tensor descriptor for probabilities, the dimensions are T,N,A (T is the timing steps, N is the |
| mini batch size, A is the alphabet size) */ |
| const void *probs, /* probabilities after softmax, in GPU memory */ |
| const int *labels, /* labels, in CPU memory */ |
| const int *labelLengths, /* the length of each label, in CPU memory */ |
| const int *inputLengths, /* the lengths of timing steps in each batch, in CPU memory */ |
| void *costs, /* the returned costs of CTC, in GPU memory */ |
| const cudnnTensorDescriptor_t gradientsDesc, /* Tensor descriptor for gradients, the dimensions are T,N,A */ |
| const void *gradients, /* the returned CTC gradients, in GPU memory, to compute costs only, set it to NULL */ |
| cudnnCTCLossAlgo_t algo, /* algorithm selected, supported now 0 and 1 */ |
| cudnnCTCLossDescriptor_t ctcLossDesc, |
| void *workspace, /* pointer to the workspace, in GPU memory */ |
| size_t workSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const void *, const int *, const int *, const int *, void *, const cudnnTensorDescriptor_t, const void *, cudnnCTCLossAlgo_t, cudnnCTCLossDescriptor_t, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCTCLoss"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, probsDesc, probs, labels, labelLengths, inputLengths, costs, gradientsDesc, gradients, algo, ctcLossDesc, workspace, workSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetCTCLossWorkspaceSize( |
| cudnnHandle_t handle, |
| const cudnnTensorDescriptor_t probsDesc, /* Tensor descriptor for probabilities, the dimensions are T,N,A (T is the |
| timing steps, N is the mini batch size, A is the alphabet size) */ |
| const cudnnTensorDescriptor_t gradientsDesc, /* Tensor descriptor for gradients, the |
| dimensions are T,N,A. To compute costs |
| only, set it to NULL */ |
| const int *labels, /* labels, in CPU memory */ |
| const int *labelLengths, /* the length of each label, in CPU memory */ |
| const int *inputLengths, /* the lengths of timing steps in each batch, in CPU memory */ |
| cudnnCTCLossAlgo_t algo, /* algorithm selected, supported now 0 and 1 */ |
| cudnnCTCLossDescriptor_t ctcLossDesc, |
| size_t *sizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnTensorDescriptor_t, const cudnnTensorDescriptor_t, const int *, const int *, const int *, cudnnCTCLossAlgo_t, cudnnCTCLossDescriptor_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetCTCLossWorkspaceSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, probsDesc, gradientsDesc, labels, labelLengths, inputLengths, algo, ctcLossDesc, sizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateAlgorithmDescriptor(cudnnAlgorithmDescriptor_t *algoDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnAlgorithmDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateAlgorithmDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(algoDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetAlgorithmDescriptor(cudnnAlgorithmDescriptor_t algoDesc, cudnnAlgorithm_t algorithm) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnAlgorithmDescriptor_t, cudnnAlgorithm_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetAlgorithmDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(algoDesc, algorithm); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetAlgorithmDescriptor(const cudnnAlgorithmDescriptor_t algoDesc, cudnnAlgorithm_t *algorithm) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnAlgorithmDescriptor_t, cudnnAlgorithm_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetAlgorithmDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(algoDesc, algorithm); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCopyAlgorithmDescriptor(const cudnnAlgorithmDescriptor_t src, cudnnAlgorithmDescriptor_t dest) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnAlgorithmDescriptor_t, cudnnAlgorithmDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCopyAlgorithmDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(src, dest); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyAlgorithmDescriptor(cudnnAlgorithmDescriptor_t algoDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnAlgorithmDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyAlgorithmDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(algoDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateAlgorithmPerformance(cudnnAlgorithmPerformance_t *algoPerf, int numberToCreate) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnAlgorithmPerformance_t *, int); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateAlgorithmPerformance"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(algoPerf, numberToCreate); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetAlgorithmPerformance(cudnnAlgorithmPerformance_t algoPerf, |
| cudnnAlgorithmDescriptor_t algoDesc, |
| cudnnStatus_t status, |
| float time, |
| size_t memory) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnAlgorithmPerformance_t, cudnnAlgorithmDescriptor_t, cudnnStatus_t, float, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetAlgorithmPerformance"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(algoPerf, algoDesc, status, time, memory); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetAlgorithmPerformance(const cudnnAlgorithmPerformance_t algoPerf, |
| cudnnAlgorithmDescriptor_t *algoDesc, |
| cudnnStatus_t *status, |
| float *time, |
| size_t *memory) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(const cudnnAlgorithmPerformance_t, cudnnAlgorithmDescriptor_t *, cudnnStatus_t *, float *, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetAlgorithmPerformance"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(algoPerf, algoDesc, status, time, memory); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyAlgorithmPerformance(cudnnAlgorithmPerformance_t *algoPerf, int numberToDestroy) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnAlgorithmPerformance_t *, int); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyAlgorithmPerformance"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(algoPerf, numberToDestroy); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetAlgorithmSpaceSize(cudnnHandle_t handle, cudnnAlgorithmDescriptor_t algoDesc, size_t *algoSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnAlgorithmDescriptor_t, size_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetAlgorithmSpaceSize"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, algoDesc, algoSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSaveAlgorithm(cudnnHandle_t handle, |
| cudnnAlgorithmDescriptor_t algoDesc, |
| void *algoSpace, |
| size_t algoSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnAlgorithmDescriptor_t, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSaveAlgorithm"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, algoDesc, algoSpace, algoSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRestoreAlgorithm(cudnnHandle_t handle, |
| void *algoSpace, |
| size_t algoSpaceSizeInBytes, |
| cudnnAlgorithmDescriptor_t algoDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, void *, size_t, cudnnAlgorithmDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRestoreAlgorithm"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, algoSpace, algoSpaceSizeInBytes, algoDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNSetClip(cudnnHandle_t handle, |
| cudnnRNNDescriptor_t rnnDesc, |
| cudnnRNNClipMode_t clipMode, |
| cudnnNanPropagation_t clipNanOpt, |
| double lclip, |
| double rclip) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnRNNDescriptor_t, cudnnRNNClipMode_t, cudnnNanPropagation_t, double, double); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNSetClip"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, clipMode, clipNanOpt, lclip, rclip); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNGetClip(cudnnHandle_t handle, |
| cudnnRNNDescriptor_t rnnDesc, |
| cudnnRNNClipMode_t *clipMode, |
| cudnnNanPropagation_t *clipNanOpt, |
| double *lclip, |
| double *rclip) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnRNNDescriptor_t, cudnnRNNClipMode_t *, cudnnNanPropagation_t *, double *, double *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNGetClip"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, clipMode, clipNanOpt, lclip, rclip); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetCallback(unsigned mask, void *udata, cudnnCallback_t fptr) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(unsigned int, void *, cudnnCallback_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetCallback"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(mask, udata, fptr); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetCallback(unsigned *mask, void **udata, cudnnCallback_t *fptr) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(unsigned int *, void **, cudnnCallback_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetCallback"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(mask, udata, fptr); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetRNNPaddingMode(cudnnRNNDescriptor_t rnnDesc, cudnnRNNPaddingMode_t paddingMode) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t, cudnnRNNPaddingMode_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetRNNPaddingMode"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc, paddingMode); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNPaddingMode(cudnnRNNDescriptor_t rnnDesc, cudnnRNNPaddingMode_t *paddingMode) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t, cudnnRNNPaddingMode_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNPaddingMode"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc, paddingMode); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnCreateRNNDataDescriptor(cudnnRNNDataDescriptor_t *RNNDataDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDataDescriptor_t *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnCreateRNNDataDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(RNNDataDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnDestroyRNNDataDescriptor(cudnnRNNDataDescriptor_t RNNDataDesc) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDataDescriptor_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnDestroyRNNDataDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(RNNDataDesc); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetRNNDataDescriptor(cudnnRNNDataDescriptor_t RNNDataDesc, |
| cudnnDataType_t dataType, |
| cudnnRNNDataLayout_t layout, |
| int maxSeqLength, |
| int batchSize, |
| int vectorSize, |
| const int seqLengthArray[], /* length of each sequence in the batch */ |
| void *paddingFill) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDataDescriptor_t, cudnnDataType_t, cudnnRNNDataLayout_t, int, int, int, const int [], void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetRNNDataDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(RNNDataDesc, dataType, layout, maxSeqLength, batchSize, vectorSize, seqLengthArray, paddingFill); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnGetRNNDataDescriptor(cudnnRNNDataDescriptor_t RNNDataDesc, |
| cudnnDataType_t *dataType, |
| cudnnRNNDataLayout_t *layout, |
| int *maxSeqLength, |
| int *batchSize, |
| int *vectorSize, |
| int arrayLengthRequested, |
| int seqLengthArray[], |
| void *paddingFill) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDataDescriptor_t, cudnnDataType_t *, cudnnRNNDataLayout_t *, int *, int *, int *, int, int [], void *); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnGetRNNDataDescriptor"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(RNNDataDesc, dataType, layout, maxSeqLength, batchSize, vectorSize, arrayLengthRequested, seqLengthArray, paddingFill); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNForwardTrainingEx(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const cudnnRNNDataDescriptor_t xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnRNNDataDescriptor_t yDesc, |
| void *y, |
| const cudnnTensorDescriptor_t hyDesc, |
| void *hy, |
| const cudnnTensorDescriptor_t cyDesc, |
| void *cy, |
| const cudnnRNNDataDescriptor_t kDesc, /* reserved, should pass NULL */ |
| const void *keys, /* reserved, should pass NULL */ |
| const cudnnRNNDataDescriptor_t cDesc, /* reserved, should pass NULL */ |
| void *cAttn, /* reserved, should pass NULL */ |
| const cudnnRNNDataDescriptor_t iDesc, /* reserved, should pass NULL */ |
| void *iAttn, /* reserved, should pass NULL */ |
| const cudnnRNNDataDescriptor_t qDesc, /* reserved, should pass NULL */ |
| void *queries, /* reserved, should pass NULL */ |
| void *workSpace, |
| size_t workSpaceSizeInBytes, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const cudnnRNNDataDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnRNNDataDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const cudnnRNNDataDescriptor_t, const void *, const cudnnRNNDataDescriptor_t, void *, const cudnnRNNDataDescriptor_t, void *, const cudnnRNNDataDescriptor_t, void *, void *, size_t, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNForwardTrainingEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, xDesc, x, hxDesc, hx, cxDesc, cx, wDesc, w, yDesc, y, hyDesc, hy, cyDesc, cy, kDesc, keys, cDesc, cAttn, iDesc, iAttn, qDesc, queries, workSpace, workSpaceSizeInBytes, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNForwardInferenceEx(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const cudnnRNNDataDescriptor_t xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnRNNDataDescriptor_t yDesc, |
| void *y, |
| const cudnnTensorDescriptor_t hyDesc, |
| void *hy, |
| const cudnnTensorDescriptor_t cyDesc, |
| void *cy, |
| const cudnnRNNDataDescriptor_t kDesc, /* reserved, should pass NULL */ |
| const void *keys, /* reserved, should pass NULL */ |
| const cudnnRNNDataDescriptor_t cDesc, /* reserved, should pass NULL */ |
| void *cAttn, /* reserved, should pass NULL */ |
| const cudnnRNNDataDescriptor_t iDesc, /* reserved, should pass NULL */ |
| void *iAttn, /* reserved, should pass NULL */ |
| const cudnnRNNDataDescriptor_t qDesc, /* reserved, should pass NULL */ |
| void *queries, /* reserved, should pass NULL */ |
| void *workSpace, |
| size_t workSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const cudnnRNNDataDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnRNNDataDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const cudnnRNNDataDescriptor_t, const void *, const cudnnRNNDataDescriptor_t, void *, const cudnnRNNDataDescriptor_t, void *, const cudnnRNNDataDescriptor_t, void *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNForwardInferenceEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, xDesc, x, hxDesc, hx, cxDesc, cx, wDesc, w, yDesc, y, hyDesc, hy, cyDesc, cy, kDesc, keys, cDesc, cAttn, iDesc, iAttn, qDesc, queries, workSpace, workSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNBackwardDataEx(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const cudnnRNNDataDescriptor_t yDesc, |
| const void *y, |
| const cudnnRNNDataDescriptor_t dyDesc, |
| const void *dy, |
| const cudnnRNNDataDescriptor_t dcDesc, /* reserved, should pass NULL */ |
| const void *dcAttn, /* reserved, should pass NULL */ |
| const cudnnTensorDescriptor_t dhyDesc, |
| const void *dhy, |
| const cudnnTensorDescriptor_t dcyDesc, |
| const void *dcy, |
| const cudnnFilterDescriptor_t wDesc, |
| const void *w, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnTensorDescriptor_t cxDesc, |
| const void *cx, |
| const cudnnRNNDataDescriptor_t dxDesc, |
| void *dx, |
| const cudnnTensorDescriptor_t dhxDesc, |
| void *dhx, |
| const cudnnTensorDescriptor_t dcxDesc, |
| void *dcx, |
| const cudnnRNNDataDescriptor_t dkDesc, /* reserved, should pass NULL */ |
| void *dkeys, /* reserved, should pass NULL */ |
| void *workSpace, |
| size_t workSpaceSizeInBytes, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const cudnnRNNDataDescriptor_t, const void *, const cudnnRNNDataDescriptor_t, const void *, const cudnnRNNDataDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnFilterDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnRNNDataDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const cudnnTensorDescriptor_t, void *, const cudnnRNNDataDescriptor_t, void *, void *, size_t, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNBackwardDataEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, yDesc, y, dyDesc, dy, dcDesc, dcAttn, dhyDesc, dhy, dcyDesc, dcy, wDesc, w, hxDesc, hx, cxDesc, cx, dxDesc, dx, dhxDesc, dhx, dcxDesc, dcx, dkDesc, dkeys, workSpace, workSpaceSizeInBytes, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnRNNBackwardWeightsEx(cudnnHandle_t handle, |
| const cudnnRNNDescriptor_t rnnDesc, |
| const cudnnRNNDataDescriptor_t xDesc, |
| const void *x, |
| const cudnnTensorDescriptor_t hxDesc, |
| const void *hx, |
| const cudnnRNNDataDescriptor_t yDesc, |
| const void *y, |
| void *workSpace, |
| size_t workSpaceSizeInBytes, |
| const cudnnFilterDescriptor_t dwDesc, |
| void *dw, |
| void *reserveSpace, |
| size_t reserveSpaceSizeInBytes) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, const cudnnRNNDescriptor_t, const cudnnRNNDataDescriptor_t, const void *, const cudnnTensorDescriptor_t, const void *, const cudnnRNNDataDescriptor_t, const void *, void *, size_t, const cudnnFilterDescriptor_t, void *, void *, size_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnRNNBackwardWeightsEx"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, xDesc, x, hxDesc, hx, yDesc, y, workSpace, workSpaceSizeInBytes, dwDesc, dw, reserveSpace, reserveSpaceSizeInBytes); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetRNNDescriptor_v6(cudnnHandle_t handle, |
| cudnnRNNDescriptor_t rnnDesc, |
| const int hiddenSize, |
| const int numLayers, |
| cudnnDropoutDescriptor_t dropoutDesc, |
| cudnnRNNInputMode_t inputMode, |
| cudnnDirectionMode_t direction, |
| cudnnRNNMode_t mode, |
| cudnnRNNAlgo_t algo, |
| cudnnDataType_t dataType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnHandle_t, cudnnRNNDescriptor_t, const int, const int, cudnnDropoutDescriptor_t, cudnnRNNInputMode_t, cudnnDirectionMode_t, cudnnRNNMode_t, cudnnRNNAlgo_t, cudnnDataType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetRNNDescriptor_v6"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(handle, rnnDesc, hiddenSize, numLayers, dropoutDesc, inputMode, direction, mode, algo, dataType); |
| } |
| |
| cudnnStatus_t CUDNNWINAPI |
| cudnnSetRNNDescriptor_v5(cudnnRNNDescriptor_t rnnDesc, |
| int hiddenSize, |
| int numLayers, |
| cudnnDropoutDescriptor_t dropoutDesc, |
| cudnnRNNInputMode_t inputMode, |
| cudnnDirectionMode_t direction, |
| cudnnRNNMode_t mode, |
| cudnnDataType_t dataType) { |
| using FuncPtr = cudnnStatus_t (CUDNNWINAPI *)(cudnnRNNDescriptor_t, int, int, cudnnDropoutDescriptor_t, cudnnRNNInputMode_t, cudnnDirectionMode_t, cudnnRNNMode_t, cudnnDataType_t); |
| static auto func_ptr = LoadSymbol<FuncPtr>("cudnnSetRNNDescriptor_v5"); |
| if (!func_ptr) return GetSymbolNotFoundError(); |
| return func_ptr(rnnDesc, hiddenSize, numLayers, dropoutDesc, inputMode, direction, mode, dataType); |
| } |
| |
| } // extern "C" |
