Memory.h

Go to the documentation of this file.

/*
 * Copyright (c) 2020, NVIDIA CORPORATION.  All rights reserved.
 *
 * NVIDIA CORPORATION and its licensors retain all intellectual property
 * and proprietary rights in and to this software, related documentation
 * and any modifications thereto.  Any use, reproduction, disclosure or
 * distribution of this software and related documentation without an express
 * license agreement from NVIDIA CORPORATION is strictly prohibited.
 */
 
#ifndef CVCORE_MEMORY_H
#define CVCORE_MEMORY_H
 
#include <cuda_runtime_api.h>
 
#include "Tensor.h"
 
namespace cvcore {
 
void TensorBaseCopy(TensorBase &dst, const TensorBase &src, cudaStream_t stream = 0);
 
void TensorBaseCopy2D(TensorBase &dst, const TensorBase &src, int dstPitch, int srcPitch, int widthInBytes, int height,
                      cudaStream_t stream = 0);
 
template<TensorLayout TL, ChannelCount CC, ChannelType CT,
         typename std::enable_if<TL != HWC && TL != CHW && TL != NHWC && TL != NCHW>::type * = nullptr>
void Copy(Tensor<TL, CC, CT> &dst, const Tensor<TL, CC, CT> &src, cudaStream_t stream = 0)
{
    TensorBaseCopy(dst, src, stream);
}
 
template<TensorLayout TL, ChannelCount CC, ChannelType CT, typename std::enable_if<TL == HWC>::type * = nullptr>
void Copy(Tensor<TL, CC, CT> &dst, const Tensor<TL, CC, CT> &src, cudaStream_t stream = 0)
{
    TensorBaseCopy2D(dst, src, dst.getStride(TensorDimension::HEIGHT) * GetChannelSize(CT),
                     src.getStride(TensorDimension::HEIGHT) * GetChannelSize(CT),
                     dst.getWidth() * dst.getChannelCount() * GetChannelSize(CT), src.getHeight(), stream);
}
 
template<TensorLayout TL, ChannelCount CC, ChannelType CT, typename std::enable_if<TL == NHWC>::type * = nullptr>
void Copy(Tensor<TL, CC, CT> &dst, const Tensor<TL, CC, CT> &src, cudaStream_t stream = 0)
{
    TensorBaseCopy2D(dst, src, dst.getStride(TensorDimension::HEIGHT) * GetChannelSize(CT),
                     src.getStride(TensorDimension::HEIGHT) * GetChannelSize(CT),
                     dst.getWidth() * dst.getChannelCount() * GetChannelSize(CT), src.getDepth() * src.getHeight(),
                     stream);
}
 
template<TensorLayout TL, ChannelCount CC, ChannelType CT, typename std::enable_if<TL == CHW>::type * = nullptr>
void Copy(Tensor<TL, CC, CT> &dst, const Tensor<TL, CC, CT> &src, cudaStream_t stream = 0)
{
    TensorBaseCopy2D(dst, src, dst.getStride(TensorDimension::HEIGHT) * GetChannelSize(CT),
                     src.getStride(TensorDimension::HEIGHT) * GetChannelSize(CT), dst.getWidth() * GetChannelSize(CT),
                     src.getChannelCount() * src.getHeight(), stream);
}
 
template<TensorLayout TL, ChannelCount CC, ChannelType CT, typename std::enable_if<TL == NCHW>::type * = nullptr>
void Copy(Tensor<TL, CC, CT> &dst, const Tensor<TL, CC, CT> &src, cudaStream_t stream = 0)
{
    TensorBaseCopy2D(dst, src, dst.getStride(TensorDimension::HEIGHT) * GetChannelSize(CT),
                     src.getStride(TensorDimension::HEIGHT) * GetChannelSize(CT), dst.getWidth() * GetChannelSize(CT),
                     src.getDepth() * src.getChannelCount() * src.getHeight(), stream);
}
 
} // namespace cvcore
 
#endif // CVCORE_MEMORY_H