CudaTypes.h

Go to the documentation of this file.

/*
* SPDX-FileCopyrightText: Copyright (c) 2020-2021 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS IN THE SOFTWARE.
*/
 
 
#ifndef NVNEURAL_CUDATYPES_H
#define NVNEURAL_CUDATYPES_H
 
#include <nvneural/CoreTypes.h>
#include <cuda.h>
 
namespace nvneural {
 
class INetworkBackendCuda;
#define NVNEURAL_INETWORKBACKENDCUDA_OBJECTCLASS "com.nvidia.backendcuda"
#define NVNEURAL_ICUDAMEMORYALLOCATOR_OBJECTCLASS "com.nvidia.memoryallocatorcuda"
 
class ICudaMemoryAllocator : public IRefObject
{
public:
    static const IRefObject::TypeId typeID = 0x121e6098096e5c97ul;
 
    virtual NeuralResult allocateMemoryBlock(MemoryHandle* pHandleOut, std::size_t byteCount, MemorySemantic semantic) noexcept = 0;
 
    virtual NeuralResult freeMemoryBlock(MemoryHandle handle) noexcept = 0;
 
    virtual void* getAddressForMemoryBlock(MemoryHandle handle) const noexcept = 0;
 
    virtual std::size_t getSizeForMemoryBlock(MemoryHandle handle) const noexcept = 0;
 
    virtual NeuralResult lockMemoryBlock(MemoryHandle handle) noexcept = 0;
 
    virtual NeuralResult unlockMemoryBlock(MemoryHandle handle) noexcept = 0;
 
    virtual NeuralResult compactMemory() noexcept = 0;
};
 
class ICudaCompiledFunction : public IRefObject
{
public:
    static const IRefObject::TypeId typeID = 0x467d6d0e91bcc332ul;
 
    virtual NeuralResult launch(
        INetworkBackendCuda* pBackend,
        std::size_t gridSizeX,
        std::size_t gridSizeY,
        std::size_t gridSizeZ,
        std::size_t blockSizeX,
        std::size_t blockSizeY,
        std::size_t blockSizeZ,
        void** ppArguments,
        std::uint32_t smem) const noexcept = 0;
 
    virtual CUmodule module() const noexcept = 0;
 
    virtual CUfunction function() const noexcept = 0;
 
    virtual std::size_t compiledBinarySize() const noexcept = 0;
 
    virtual const void* compiledBinary() const noexcept = 0;
};
 
class ICudaRuntimeCompiler : public IRefObject
{
public:
    static const IRefObject::TypeId typeID = 0xe39c2816f916d342ul;
 
    struct CompilationDetails
    {
        const char* pModuleName = nullptr;
        const char* pModuleSource = nullptr;
        const char* pModuleEntryPoint = nullptr;
        IStringList* pHeaderNames = nullptr;
        IStringList* pHeaderContents = nullptr;
        IStringList* pAdditionalCompilerOptions = nullptr;
    };
 
    virtual const char* targetArchitecture() const noexcept = 0;
 
    virtual NeuralResult setTargetArchitecture(const char* pTargetArch) noexcept = 0;
 
    virtual NeuralResult compile(ICudaCompiledFunction** ppCompiledFunctionOut, const CompilationDetails& compilationDetails) noexcept = 0;
 
    virtual NeuralResult loadCubin(ICudaCompiledFunction** ppCompiledFunctionOut, std::uint8_t* pCode, std::size_t codeSize, const char* pEntryPoint) noexcept = 0;
};
 
class INetworkBackendCuda : public IRefObject
{
public:
    static const IRefObject::TypeId typeID = 0x61f19c57a3032f9ul;
 
    virtual ICudaMemoryAllocator* getAllocator() const noexcept = 0;
 
    virtual NeuralResult setAllocator(ICudaMemoryAllocator* pAllocator) noexcept = 0;
 
    virtual CUcontext getCudaContext() const noexcept = 0;
 
    virtual CUdevice getCudaDevice() const noexcept = 0;
 
    virtual CUstream getCudaStream() const noexcept = 0;
 
    virtual ICudaRuntimeCompiler* runtimeCompiler() const noexcept = 0;
};
 
} // namespace nvneural
 
#endif // NVNEURAL_CUDATYPES_H