NvInferRuntimeCommon.h

Go to the documentation of this file.

/*
 * Copyright 1993-2020 NVIDIA Corporation.  All rights reserved.
 *
 * NOTICE TO LICENSEE:
 *
 * This source code and/or documentation ("Licensed Deliverables") are
 * subject to NVIDIA intellectual property rights under U.S. and
 * international Copyright laws.
 *
 * These Licensed Deliverables contained herein is PROPRIETARY and
 * CONFIDENTIAL to NVIDIA and is being provided under the terms and
 * conditions of a form of NVIDIA software license agreement by and
 * between NVIDIA and Licensee ("License Agreement") or electronically
 * accepted by Licensee.  Notwithstanding any terms or conditions to
 * the contrary in the License Agreement, reproduction or disclosure
 * of the Licensed Deliverables to any third party without the express
 * written consent of NVIDIA is prohibited.
 *
 * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
 * LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
 * SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE.  IT IS
 * PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
 * NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
 * DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
 * NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
 * NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
 * LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
 * SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
 * DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
 * WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
 * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
 * OF THESE LICENSED DELIVERABLES.
 *
 * U.S. Government End Users.  These Licensed Deliverables are a
 * "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
 * 1995), consisting of "commercial computer software" and "commercial
 * computer software documentation" as such terms are used in 48
 * C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
 * only as a commercial end item.  Consistent with 48 C.F.R.12.212 and
 * 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
 * U.S. Government End Users acquire the Licensed Deliverables with
 * only those rights set forth herein.
 *
 * Any use of the Licensed Deliverables in individual and commercial
 * software must include, in the user documentation and internal
 * comments to the code, the above Disclaimer and U.S. Government End
 * Users Notice.
 */
 
#ifndef NV_INFER_RUNTIME_COMMON_H
#define NV_INFER_RUNTIME_COMMON_H
 
#include <cstddef>
#include <cstdint>
#include "NvInferVersion.h"
 
#if __cplusplus >= 201103L
#define _TENSORRT_FINAL final
#define _TENSORRT_OVERRIDE override
#else
#define _TENSORRT_FINAL
#define _TENSORRT_OVERRIDE
#endif
 
#if __cplusplus >= 201402L
#define TRT_DEPRECATED [[deprecated]]
#if __GNUC__ < 6
#define TRT_DEPRECATED_ENUM
#else
#define TRT_DEPRECATED_ENUM TRT_DEPRECATED
#endif
#ifdef _MSC_VER
#define TRT_DEPRECATED_API __declspec(dllexport)
#else
#define TRT_DEPRECATED_API [[deprecated]] __attribute__((visibility("default")))
#endif
#else
#ifdef _MSC_VER
#define TRT_DEPRECATED
#define TRT_DEPRECATED_ENUM
#define TRT_DEPRECATED_API __declspec(dllexport)
#else
#define TRT_DEPRECATED __attribute__((deprecated))
#define TRT_DEPRECATED_ENUM
#define TRT_DEPRECATED_API __attribute__((deprecated, visibility("default")))
#endif
#endif
 
#ifdef TENSORRT_BUILD_LIB
#ifdef _MSC_VER
#define TENSORRTAPI __declspec(dllexport)
#else
#define TENSORRTAPI __attribute__((visibility("default")))
#endif
#else
#define TENSORRTAPI
#endif
 
#define TRTNOEXCEPT
 
// forward declare some CUDA types to avoid an include dependency
 
struct cublasContext;
struct cudnnContext;
 
typedef struct CUstream_st* cudaStream_t; 
typedef struct CUevent_st* cudaEvent_t;   
 
static const int32_t NV_TENSORRT_VERSION
    = (NV_TENSORRT_MAJOR * 1000) + (NV_TENSORRT_MINOR * 100) + NV_TENSORRT_PATCH; // major, minor, patch
 
namespace nvinfer1
{
 
class IErrorRecorder; 
class IGpuAllocator;  
 
template <typename T>
constexpr inline int32_t EnumMax();
 
enum class ActivationType : int32_t
{
    kRELU = 0,             
    kSIGMOID = 1,          
    kTANH = 2,             
    kLEAKY_RELU = 3,       
    kELU = 4,              
    kSELU = 5,             
    kSOFTSIGN = 6,         
    kSOFTPLUS = 7,         
    kCLIP = 8,             
    kHARD_SIGMOID = 9,     
    kSCALED_TANH = 10,     
    kTHRESHOLDED_RELU = 11 
};
 
template <>
constexpr inline int32_t EnumMax<ActivationType>()
{
    return 12;
}
 
enum class DataType : int32_t
{
    kFLOAT = 0,
 
    kHALF = 1,
 
    kINT8 = 2,
 
    kINT32 = 3,
 
    kBOOL = 4
};
 
template <>
constexpr inline int32_t EnumMax<DataType>()
{
    return 5;
}
 
enum class DimensionType : int32_t
{
    kSPATIAL = 0, 
    kCHANNEL = 1, 
    kINDEX = 2,   
    kSEQUENCE = 3 
};
 
template <>
constexpr inline int32_t EnumMax<DimensionType>()
{
    return 4;
}
 
class Dims
{
public:
    static const int32_t MAX_DIMS = 8;           
    int32_t nbDims;                              
    int32_t d[MAX_DIMS];                         
    TRT_DEPRECATED DimensionType type[MAX_DIMS]; 
};
 
typedef uint32_t TensorFormats;
 
enum class TensorFormat : int32_t
{
    kLINEAR = 0,
    kNCHW TRT_DEPRECATED_ENUM = kLINEAR, 
 
    kCHW2 = 1,
    kNC2HW2 TRT_DEPRECATED_ENUM = kCHW2, 
 
    kHWC8 = 2,
    kNHWC8 TRT_DEPRECATED_ENUM = kHWC8, 
 
    kCHW4 = 3,
 
    kCHW16 = 4,
 
    kCHW32 = 5,
 
    kDHWC8 = 6,
 
    kCDHW32 = 7,
 
    kHWC = 8
};
 
using PluginFormat = TensorFormat;
 
template <>
constexpr inline int32_t EnumMax<TensorFormat>()
{
    return 9;
}
 
struct PluginTensorDesc
{
    Dims dims;
    DataType type; 
    TensorFormat format;
    float scale;
};
 
enum class PluginVersion : uint8_t
{
    kV2 = 0,            
    kV2_EXT = 1,        
    kV2_IOEXT = 2,      
    kV2_DYNAMICEXT = 3, 
};
 
class IPluginV2
{
public:
    virtual int32_t getTensorRTVersion() const TRTNOEXCEPT
    {
        return NV_TENSORRT_VERSION;
    }
 
    virtual const char* getPluginType() const TRTNOEXCEPT = 0;
 
    virtual const char* getPluginVersion() const TRTNOEXCEPT = 0;
 
    virtual int32_t getNbOutputs() const TRTNOEXCEPT = 0;
 
    virtual Dims getOutputDimensions(int32_t index, const Dims* inputs, int32_t nbInputDims) TRTNOEXCEPT = 0;
 
    virtual bool supportsFormat(DataType type, PluginFormat format) const TRTNOEXCEPT = 0;
 
    virtual void configureWithFormat(const Dims* inputDims, int32_t nbInputs, const Dims* outputDims, int32_t nbOutputs,
        DataType type, PluginFormat format, int32_t maxBatchSize) TRTNOEXCEPT = 0;
 
    virtual int32_t initialize() TRTNOEXCEPT = 0;
 
    virtual void terminate() TRTNOEXCEPT = 0;
 
    virtual size_t getWorkspaceSize(int32_t maxBatchSize) const TRTNOEXCEPT = 0;
 
    virtual int32_t enqueue(int32_t batchSize, const void* const* inputs, void** outputs, void* workspace,
        cudaStream_t stream) TRTNOEXCEPT = 0;
 
    virtual size_t getSerializationSize() const TRTNOEXCEPT = 0;
 
    virtual void serialize(void* buffer) const TRTNOEXCEPT = 0;
 
    virtual void destroy() TRTNOEXCEPT = 0;
 
    virtual IPluginV2* clone() const TRTNOEXCEPT = 0;
 
    virtual void setPluginNamespace(const char* pluginNamespace) TRTNOEXCEPT = 0;
 
    virtual const char* getPluginNamespace() const TRTNOEXCEPT = 0;
 
protected:
    virtual ~IPluginV2() {}
};
 
class IPluginV2Ext : public IPluginV2
{
public:
    virtual nvinfer1::DataType getOutputDataType(
        int32_t index, const nvinfer1::DataType* inputTypes, int32_t nbInputs) const TRTNOEXCEPT = 0;
 
    virtual bool isOutputBroadcastAcrossBatch(
        int32_t outputIndex, const bool* inputIsBroadcasted, int32_t nbInputs) const TRTNOEXCEPT = 0;
 
    virtual bool canBroadcastInputAcrossBatch(int32_t inputIndex) const TRTNOEXCEPT = 0;
 
 
    virtual void configurePlugin(const Dims* inputDims, int32_t nbInputs, const Dims* outputDims, int32_t nbOutputs,
        const DataType* inputTypes, const DataType* outputTypes, const bool* inputIsBroadcast,
        const bool* outputIsBroadcast, PluginFormat floatFormat, int32_t maxBatchSize) TRTNOEXCEPT = 0;
 
    virtual ~IPluginV2Ext() {}
 
    virtual void attachToContext(cudnnContext* /*cudnn*/, cublasContext* /*cublas*/, IGpuAllocator* /*allocator*/) TRTNOEXCEPT {}
 
    virtual void detachFromContext() TRTNOEXCEPT {}
 
    virtual IPluginV2Ext* clone() const _TENSORRT_OVERRIDE TRTNOEXCEPT = 0;
 
protected:
    int32_t getTensorRTVersion() const _TENSORRT_OVERRIDE TRTNOEXCEPT
    {
        return (static_cast<int32_t>(PluginVersion::kV2_EXT) << 24 | (NV_TENSORRT_VERSION & 0xFFFFFF));
    }
 
    void configureWithFormat(const Dims* /*inputDims*/, int32_t /*nbInputs*/, const Dims* /*outputDims*/,
        int32_t /*nbOutputs*/, DataType /*type*/, PluginFormat /*format*/,
        int32_t /*maxBatchSize*/) _TENSORRT_OVERRIDE TRTNOEXCEPT
    {
    }
};
 
class IPluginV2IOExt : public IPluginV2Ext
{
public:
    virtual void configurePlugin(
        const PluginTensorDesc* in, int32_t nbInput, const PluginTensorDesc* out, int32_t nbOutput) TRTNOEXCEPT = 0;
 
    virtual bool supportsFormatCombination(
        int32_t pos, const PluginTensorDesc* inOut, int32_t nbInputs, int32_t nbOutputs) const TRTNOEXCEPT = 0;
 
protected:
    TRT_DEPRECATED
    int32_t getTensorRTVersion() const _TENSORRT_OVERRIDE
    {
        return (static_cast<int32_t>(PluginVersion::kV2_IOEXT) << 24 | (NV_TENSORRT_VERSION & 0xFFFFFF));
    }
 
    TRT_DEPRECATED
    void configureWithFormat(
        const Dims*, int32_t, const Dims*, int32_t, DataType, PluginFormat, int32_t) _TENSORRT_OVERRIDE _TENSORRT_FINAL
    {
    }
 
    TRT_DEPRECATED
    void configurePlugin(const Dims*, int32_t, const Dims*, int32_t, const DataType*, const DataType*, const bool*,
        const bool*, PluginFormat, int32_t) _TENSORRT_OVERRIDE _TENSORRT_FINAL
    {
    }
 
    TRT_DEPRECATED
    bool supportsFormat(DataType, PluginFormat) const _TENSORRT_OVERRIDE _TENSORRT_FINAL
    {
        return false;
    }
};
 
 
enum class PluginFieldType : int32_t
{
    kFLOAT16 = 0, 
    kFLOAT32 = 1, 
    kFLOAT64 = 2, 
    kINT8 = 3,    
    kINT16 = 4,   
    kINT32 = 5,   
    kCHAR = 6,    
    kDIMS = 7,    
    kUNKNOWN = 8
};
 
class PluginField
{
public:
    const char* name{nullptr};
    const void* data{nullptr};
    PluginFieldType type{PluginFieldType::kUNKNOWN};
    int32_t length{0};
 
    PluginField(const char* name_ = nullptr, const void* data_ = nullptr, const PluginFieldType type_ = PluginFieldType::kUNKNOWN, int32_t length_ = 0)
        : name(name_)
        , data(data_)
        , type(type_)
        , length(length_)
    {
    }
};
 
struct PluginFieldCollection
{
    int32_t nbFields;          
    const PluginField* fields; 
};
 
 
class IPluginCreator
{
public:
    virtual int32_t getTensorRTVersion() const TRTNOEXCEPT
    {
        return NV_TENSORRT_VERSION;
    }
 
    virtual const char* getPluginName() const TRTNOEXCEPT = 0;
 
    virtual const char* getPluginVersion() const TRTNOEXCEPT = 0;
 
    virtual const PluginFieldCollection* getFieldNames() TRTNOEXCEPT = 0;
 
    virtual IPluginV2* createPlugin(const char* name, const PluginFieldCollection* fc) TRTNOEXCEPT = 0;
 
    virtual IPluginV2* deserializePlugin(const char* name, const void* serialData, size_t serialLength) TRTNOEXCEPT = 0;
 
    virtual void setPluginNamespace(const char* pluginNamespace) TRTNOEXCEPT = 0;
 
    virtual const char* getPluginNamespace() const TRTNOEXCEPT = 0;
 
    virtual ~IPluginCreator() {}
};
 
 
class IPluginRegistry
{
public:
    virtual bool registerCreator(IPluginCreator& creator, const char* pluginNamespace) noexcept = 0;
 
    virtual IPluginCreator* const* getPluginCreatorList(int32_t* numCreators) const noexcept = 0;
 
    virtual IPluginCreator* getPluginCreator(const char* pluginType, const char* pluginVersion, const char* pluginNamespace = "") noexcept = 0;
 
protected:
    virtual ~IPluginRegistry() noexcept {}
 
public:
    //
    virtual void setErrorRecorder(IErrorRecorder* recorder) noexcept = 0;
 
    virtual IErrorRecorder* getErrorRecorder() const noexcept = 0;
};
 
enum class TensorLocation : int32_t
{
    kDEVICE = 0, 
    kHOST = 1,   
};
 
template <>
constexpr inline int32_t EnumMax<TensorLocation>()
{
    return 2;
}
 
class IGpuAllocator
{
public:
    virtual void* allocate(uint64_t size, uint64_t alignment, uint32_t flags) TRTNOEXCEPT = 0;
 
    virtual void free(void* memory) TRTNOEXCEPT = 0;
 
    virtual ~IGpuAllocator() {}
};
 
class ILogger
{
public:
    enum class Severity : int32_t
    {
        kINTERNAL_ERROR = 0, 
        kERROR = 1,          
        kWARNING = 2, 
        kINFO = 3,    
        kVERBOSE = 4, 
    };
 
    virtual void log(Severity severity, const char* msg) TRTNOEXCEPT = 0;
 
    virtual ~ILogger() {}
};
 
template <>
constexpr inline int32_t EnumMax<ILogger::Severity>()
{
    return 5;
}
 
enum class ErrorCode : int32_t
{
    kSUCCESS = 0,
 
    kUNSPECIFIED_ERROR = 1,
 
    kINTERNAL_ERROR = 2,
 
    kINVALID_ARGUMENT = 3,
 
    kINVALID_CONFIG = 4,
 
    kFAILED_ALLOCATION = 5,
 
    kFAILED_INITIALIZATION = 6,
 
    kFAILED_EXECUTION = 7,
 
    kFAILED_COMPUTATION = 8,
 
    kINVALID_STATE = 9,
 
    kUNSUPPORTED_STATE = 10,
 
};
 
template <>
constexpr inline int32_t EnumMax<ErrorCode>()
{
    return 11;
}
 
class IErrorRecorder
{
public:
    using ErrorDesc = const char*;
 
    using RefCount = int32_t;
 
    virtual ~IErrorRecorder() noexcept {};
 
    // Public API’s used to retrieve information from the error recorder.
 
    virtual int32_t getNbErrors() const noexcept = 0;
 
    virtual ErrorCode getErrorCode(int32_t errorIdx) const noexcept = 0;
 
    virtual ErrorDesc getErrorDesc(int32_t errorIdx) const noexcept = 0;
 
    virtual bool hasOverflowed() const noexcept = 0;
 
    virtual void clear() noexcept = 0;
 
    // API’s used by TensorRT to report Error information to the application.
 
    virtual bool reportError(ErrorCode val, ErrorDesc desc) noexcept = 0;
 
    virtual RefCount incRefCount() noexcept = 0;
 
    virtual RefCount decRefCount() noexcept = 0;
 
}; // class IErrorRecorder
 
} // namespace nvinfer1
 
extern "C" TENSORRTAPI void* createSafeInferRuntime_INTERNAL(void* logger, int32_t version);
 
extern "C" TENSORRTAPI nvinfer1::ILogger* getLogger();
 
extern "C" TENSORRTAPI int32_t getInferLibVersion();
 
extern "C" TENSORRTAPI nvinfer1::IPluginRegistry* getPluginRegistry();
 
namespace nvinfer1
{
 
template <typename T>
class PluginRegistrar
{
public:
    PluginRegistrar() { getPluginRegistry()->registerCreator(instance, ""); }
private:
    T instance{};
};
 
#define REGISTER_TENSORRT_PLUGIN(name) \
    static nvinfer1::PluginRegistrar<name> pluginRegistrar##name {}
 
} // namespace nvinfer1
 
#endif // NV_INFER_RUNTIME_COMMON_H