wehub-resource-sync
121 lines
3.2 KiB
C++
121 lines
3.2 KiB
C++
/*
|
|
* Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
|
|
*
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*/
|
|
|
|
#ifndef TRT_SCOPED_CUDA_STREAM_H
|
|
#define TRT_SCOPED_CUDA_STREAM_H
|
|
|
|
#include <cstdint>
|
|
#include <cuda_runtime_api.h>
|
|
#include <memory>
|
|
#include <stdexcept>
|
|
|
|
namespace nvinfer1
|
|
{
|
|
namespace pluginInternal
|
|
{
|
|
|
|
// RAII wrapper for CUDA stream
|
|
// Automatically manages CUDA stream lifecycle - creation in constructor, destruction in destructor
|
|
// Provides safe stream management and prevents memory leaks
|
|
class ScopedCudaStream
|
|
{
|
|
public:
|
|
// Constructor that creates a new CUDA stream with default flags
|
|
ScopedCudaStream()
|
|
: mStream(nullptr)
|
|
{
|
|
cudaError_t result = cudaStreamCreate(&mStream);
|
|
if (result != cudaSuccess)
|
|
{
|
|
throw std::runtime_error("Failed to create CUDA stream: " + std::string(cudaGetErrorString(result)));
|
|
}
|
|
}
|
|
|
|
// Constructor that creates a new CUDA stream with custom flags
|
|
explicit ScopedCudaStream(uint32_t const flags)
|
|
: mStream(nullptr)
|
|
{
|
|
cudaError_t result = cudaStreamCreateWithFlags(&mStream, flags);
|
|
if (result != cudaSuccess)
|
|
{
|
|
mStream = nullptr;
|
|
}
|
|
}
|
|
|
|
// Destructor - automatically destroys the stream
|
|
~ScopedCudaStream()
|
|
{
|
|
if (mStream != nullptr)
|
|
{
|
|
cudaStreamDestroy(mStream);
|
|
}
|
|
}
|
|
|
|
// Delete copy constructor and assignment
|
|
ScopedCudaStream(ScopedCudaStream const&) = delete;
|
|
ScopedCudaStream& operator=(ScopedCudaStream const&) = delete;
|
|
|
|
// Move constructor
|
|
ScopedCudaStream(ScopedCudaStream&& other) noexcept
|
|
: mStream(std::exchange(other.mStream, nullptr))
|
|
{
|
|
}
|
|
|
|
// Move assignment
|
|
ScopedCudaStream& operator=(ScopedCudaStream&& other) noexcept
|
|
{
|
|
ScopedCudaStream tmp{std::move(other)};
|
|
std::swap(mStream, tmp.mStream);
|
|
return *this;
|
|
}
|
|
|
|
// Get the underlying CUDA stream handle
|
|
cudaStream_t get() const noexcept
|
|
{
|
|
return mStream;
|
|
}
|
|
|
|
// Implicit conversion to cudaStream_t
|
|
operator cudaStream_t() const noexcept
|
|
{
|
|
return mStream;
|
|
}
|
|
|
|
// Check if the stream is valid
|
|
bool isValid() const noexcept
|
|
{
|
|
return mStream != nullptr;
|
|
}
|
|
|
|
private:
|
|
cudaStream_t mStream;
|
|
};
|
|
|
|
// Helper function to create a unique_ptr to ScopedCudaStream
|
|
inline std::unique_ptr<ScopedCudaStream> makeScopedCudaStream(uint32_t const flags = cudaStreamDefault)
|
|
{
|
|
if (flags == cudaStreamDefault)
|
|
{
|
|
return std::make_unique<ScopedCudaStream>();
|
|
}
|
|
return std::make_unique<ScopedCudaStream>(flags);
|
|
}
|
|
|
|
} // namespace pluginInternal
|
|
} // namespace nvinfer1
|
|
|
|
#endif // TRT_SCOPED_CUDA_STREAM_H
|