2010-07-14 23:55:16 +08:00
|
|
|
/*M///////////////////////////////////////////////////////////////////////////////////////
|
|
|
|
//
|
|
|
|
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
|
|
|
|
//
|
|
|
|
// By downloading, copying, installing or using the software you agree to this license.
|
|
|
|
// If you do not agree to this license, do not download, install,
|
|
|
|
// copy or use the software.
|
|
|
|
//
|
|
|
|
//
|
|
|
|
// License Agreement
|
|
|
|
// For Open Source Computer Vision Library
|
|
|
|
//
|
|
|
|
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
|
|
|
|
// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
|
|
|
|
// Third party copyrights are property of their respective owners.
|
|
|
|
//
|
|
|
|
// Redistribution and use in source and binary forms, with or without modification,
|
|
|
|
// are permitted provided that the following conditions are met:
|
|
|
|
//
|
|
|
|
// * Redistribution's of source code must retain the above copyright notice,
|
|
|
|
// this list of conditions and the following disclaimer.
|
|
|
|
//
|
|
|
|
// * Redistribution's in binary form must reproduce the above copyright notice,
|
|
|
|
// this list of conditions and the following disclaimer in the documentation
|
|
|
|
// and/or other materials provided with the distribution.
|
|
|
|
//
|
|
|
|
// * The name of the copyright holders may not be used to endorse or promote products
|
|
|
|
// derived from this software without specific prior written permission.
|
|
|
|
//
|
|
|
|
// This software is provided by the copyright holders and contributors "as is" and
|
|
|
|
// any express or implied warranties, including, but not limited to, the implied
|
|
|
|
// warranties of merchantability and fitness for a particular purpose are disclaimed.
|
|
|
|
// In no event shall the Intel Corporation or contributors be liable for any direct,
|
|
|
|
// indirect, incidental, special, exemplary, or consequential damages
|
|
|
|
// (including, but not limited to, procurement of substitute goods or services;
|
|
|
|
// loss of use, data, or profits; or business interruption) however caused
|
|
|
|
// and on any theory of liability, whether in contract, strict liability,
|
|
|
|
// or tort (including negligence or otherwise) arising in any way out of
|
|
|
|
// the use of this software, even if advised of the possibility of such damage.
|
|
|
|
//
|
|
|
|
//M*/
|
|
|
|
|
|
|
|
#include "precomp.hpp"
|
|
|
|
|
|
|
|
using namespace cv;
|
|
|
|
using namespace cv::gpu;
|
|
|
|
|
2011-11-09 21:13:52 +08:00
|
|
|
cv::gpu::CudaMem::CudaMem()
|
|
|
|
: flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(0)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
|
|
|
|
cv::gpu::CudaMem::CudaMem(int _rows, int _cols, int _type, int _alloc_type)
|
|
|
|
: flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(0)
|
|
|
|
{
|
|
|
|
if( _rows > 0 && _cols > 0 )
|
|
|
|
create( _rows, _cols, _type, _alloc_type);
|
|
|
|
}
|
|
|
|
|
|
|
|
cv::gpu::CudaMem::CudaMem(Size _size, int _type, int _alloc_type)
|
|
|
|
: flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(0)
|
|
|
|
{
|
|
|
|
if( _size.height > 0 && _size.width > 0 )
|
|
|
|
create( _size.height, _size.width, _type, _alloc_type);
|
|
|
|
}
|
|
|
|
|
|
|
|
cv::gpu::CudaMem::CudaMem(const CudaMem& m)
|
|
|
|
: flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend), alloc_type(m.alloc_type)
|
|
|
|
{
|
|
|
|
if( refcount )
|
|
|
|
CV_XADD(refcount, 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
cv::gpu::CudaMem::CudaMem(const Mat& m, int _alloc_type)
|
|
|
|
: flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(0)
|
|
|
|
{
|
|
|
|
if( m.rows > 0 && m.cols > 0 )
|
|
|
|
create( m.size(), m.type(), _alloc_type);
|
|
|
|
|
|
|
|
Mat tmp = createMatHeader();
|
|
|
|
m.copyTo(tmp);
|
|
|
|
}
|
|
|
|
|
|
|
|
cv::gpu::CudaMem::~CudaMem()
|
|
|
|
{
|
|
|
|
release();
|
|
|
|
}
|
|
|
|
|
|
|
|
CudaMem& cv::gpu::CudaMem::operator = (const CudaMem& m)
|
|
|
|
{
|
|
|
|
if( this != &m )
|
|
|
|
{
|
|
|
|
if( m.refcount )
|
|
|
|
CV_XADD(m.refcount, 1);
|
|
|
|
release();
|
|
|
|
flags = m.flags;
|
|
|
|
rows = m.rows; cols = m.cols;
|
|
|
|
step = m.step; data = m.data;
|
|
|
|
datastart = m.datastart;
|
|
|
|
dataend = m.dataend;
|
|
|
|
refcount = m.refcount;
|
|
|
|
alloc_type = m.alloc_type;
|
|
|
|
}
|
|
|
|
return *this;
|
|
|
|
}
|
|
|
|
|
|
|
|
CudaMem cv::gpu::CudaMem::clone() const
|
|
|
|
{
|
|
|
|
CudaMem m(size(), type(), alloc_type);
|
|
|
|
Mat to = m;
|
|
|
|
Mat from = *this;
|
|
|
|
from.copyTo(to);
|
|
|
|
return m;
|
|
|
|
}
|
|
|
|
|
|
|
|
void cv::gpu::CudaMem::create(Size _size, int _type, int _alloc_type)
|
|
|
|
{
|
|
|
|
create(_size.height, _size.width, _type, _alloc_type);
|
|
|
|
}
|
|
|
|
|
|
|
|
Mat cv::gpu::CudaMem::createMatHeader() const
|
|
|
|
{
|
|
|
|
return Mat(size(), type(), data, step);
|
|
|
|
}
|
|
|
|
|
|
|
|
cv::gpu::CudaMem::operator Mat() const
|
|
|
|
{
|
|
|
|
return createMatHeader();
|
|
|
|
}
|
|
|
|
|
|
|
|
cv::gpu::CudaMem::operator GpuMat() const
|
|
|
|
{
|
|
|
|
return createGpuMatHeader();
|
|
|
|
}
|
|
|
|
|
|
|
|
bool cv::gpu::CudaMem::isContinuous() const
|
|
|
|
{
|
|
|
|
return (flags & Mat::CONTINUOUS_FLAG) != 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
size_t cv::gpu::CudaMem::elemSize() const
|
|
|
|
{
|
|
|
|
return CV_ELEM_SIZE(flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
size_t cv::gpu::CudaMem::elemSize1() const
|
|
|
|
{
|
|
|
|
return CV_ELEM_SIZE1(flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
int cv::gpu::CudaMem::type() const
|
|
|
|
{
|
|
|
|
return CV_MAT_TYPE(flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
int cv::gpu::CudaMem::depth() const
|
|
|
|
{
|
|
|
|
return CV_MAT_DEPTH(flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
int cv::gpu::CudaMem::channels() const
|
|
|
|
{
|
|
|
|
return CV_MAT_CN(flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
size_t cv::gpu::CudaMem::step1() const
|
|
|
|
{
|
|
|
|
return step/elemSize1();
|
|
|
|
}
|
|
|
|
|
|
|
|
Size cv::gpu::CudaMem::size() const
|
|
|
|
{
|
|
|
|
return Size(cols, rows);
|
|
|
|
}
|
|
|
|
|
|
|
|
bool cv::gpu::CudaMem::empty() const
|
|
|
|
{
|
|
|
|
return data == 0;
|
|
|
|
}
|
|
|
|
|
2012-10-02 02:37:20 +08:00
|
|
|
#if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)
|
2010-07-19 17:31:12 +08:00
|
|
|
|
2011-08-10 19:32:48 +08:00
|
|
|
void cv::gpu::registerPageLocked(Mat&) { throw_nogpu(); }
|
|
|
|
void cv::gpu::unregisterPageLocked(Mat&) { throw_nogpu(); }
|
2011-08-09 15:51:48 +08:00
|
|
|
void cv::gpu::CudaMem::create(int /*_rows*/, int /*_cols*/, int /*_type*/, int /*type_alloc*/) { throw_nogpu(); }
|
|
|
|
bool cv::gpu::CudaMem::canMapHostMemory() { throw_nogpu(); return false; }
|
|
|
|
void cv::gpu::CudaMem::release() { throw_nogpu(); }
|
|
|
|
GpuMat cv::gpu::CudaMem::createGpuMatHeader () const { throw_nogpu(); return GpuMat(); }
|
2010-07-19 17:31:12 +08:00
|
|
|
|
|
|
|
#else /* !defined (HAVE_CUDA) */
|
|
|
|
|
2012-06-30 17:29:33 +08:00
|
|
|
void cv::gpu::registerPageLocked(Mat& m)
|
2011-08-10 19:32:48 +08:00
|
|
|
{
|
|
|
|
cudaSafeCall( cudaHostRegister(m.ptr(), m.step * m.rows, cudaHostRegisterPortable) );
|
|
|
|
}
|
|
|
|
|
2012-06-30 17:29:33 +08:00
|
|
|
void cv::gpu::unregisterPageLocked(Mat& m)
|
2011-08-10 19:32:48 +08:00
|
|
|
{
|
|
|
|
cudaSafeCall( cudaHostUnregister(m.ptr()) );
|
|
|
|
}
|
2010-07-19 17:31:12 +08:00
|
|
|
|
2010-11-15 16:42:10 +08:00
|
|
|
bool cv::gpu::CudaMem::canMapHostMemory()
|
2010-07-19 17:31:12 +08:00
|
|
|
{
|
2010-08-17 18:39:18 +08:00
|
|
|
cudaDeviceProp prop;
|
2011-05-10 14:11:03 +08:00
|
|
|
cudaSafeCall( cudaGetDeviceProperties(&prop, getDevice()) );
|
2010-08-17 18:39:18 +08:00
|
|
|
return (prop.canMapHostMemory != 0) ? true : false;
|
|
|
|
}
|
|
|
|
|
2011-05-03 17:09:05 +08:00
|
|
|
namespace
|
|
|
|
{
|
2011-05-10 20:39:12 +08:00
|
|
|
size_t alignUpStep(size_t what, size_t alignment)
|
2011-05-03 17:09:05 +08:00
|
|
|
{
|
2011-05-10 20:39:12 +08:00
|
|
|
size_t alignMask = alignment-1;
|
|
|
|
size_t inverseAlignMask = ~alignMask;
|
|
|
|
size_t res = (what + alignMask) & inverseAlignMask;
|
2011-05-03 17:09:05 +08:00
|
|
|
return res;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-08-17 18:39:18 +08:00
|
|
|
void cv::gpu::CudaMem::create(int _rows, int _cols, int _type, int _alloc_type)
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-11-15 16:42:10 +08:00
|
|
|
if (_alloc_type == ALLOC_ZEROCOPY && !canMapHostMemory())
|
2010-08-17 18:39:18 +08:00
|
|
|
cv::gpu::error("ZeroCopy is not supported by current device", __FILE__, __LINE__);
|
|
|
|
|
2010-07-19 17:31:12 +08:00
|
|
|
_type &= TYPE_MASK;
|
|
|
|
if( rows == _rows && cols == _cols && type() == _type && data )
|
|
|
|
return;
|
|
|
|
if( data )
|
|
|
|
release();
|
|
|
|
CV_DbgAssert( _rows >= 0 && _cols >= 0 );
|
|
|
|
if( _rows > 0 && _cols > 0 )
|
2010-11-08 17:55:10 +08:00
|
|
|
{
|
2010-07-19 17:31:12 +08:00
|
|
|
flags = Mat::MAGIC_VAL + Mat::CONTINUOUS_FLAG + _type;
|
|
|
|
rows = _rows;
|
|
|
|
cols = _cols;
|
|
|
|
step = elemSize()*cols;
|
2011-05-03 17:09:05 +08:00
|
|
|
if (_alloc_type == ALLOC_ZEROCOPY)
|
|
|
|
{
|
|
|
|
cudaDeviceProp prop;
|
2011-05-10 14:11:03 +08:00
|
|
|
cudaSafeCall( cudaGetDeviceProperties(&prop, getDevice()) );
|
2011-05-10 20:39:12 +08:00
|
|
|
step = alignUpStep(step, prop.textureAlignment);
|
2011-05-03 17:09:05 +08:00
|
|
|
}
|
2010-07-19 17:31:12 +08:00
|
|
|
int64 _nettosize = (int64)step*rows;
|
|
|
|
size_t nettosize = (size_t)_nettosize;
|
|
|
|
if( _nettosize != (int64)nettosize )
|
|
|
|
CV_Error(CV_StsNoMem, "Too big buffer is allocated");
|
|
|
|
size_t datasize = alignSize(nettosize, (int)sizeof(*refcount));
|
|
|
|
|
2010-07-19 18:49:35 +08:00
|
|
|
//datastart = data = (uchar*)fastMalloc(datasize + sizeof(*refcount));
|
2010-08-17 18:39:18 +08:00
|
|
|
alloc_type = _alloc_type;
|
2010-07-19 17:31:12 +08:00
|
|
|
void *ptr;
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-08-17 18:39:18 +08:00
|
|
|
switch (alloc_type)
|
2010-08-13 22:52:50 +08:00
|
|
|
{
|
2010-08-17 18:39:18 +08:00
|
|
|
case ALLOC_PAGE_LOCKED: cudaSafeCall( cudaHostAlloc( &ptr, datasize, cudaHostAllocDefault) ); break;
|
|
|
|
case ALLOC_ZEROCOPY: cudaSafeCall( cudaHostAlloc( &ptr, datasize, cudaHostAllocMapped) ); break;
|
2010-08-13 22:52:50 +08:00
|
|
|
case ALLOC_WRITE_COMBINED: cudaSafeCall( cudaHostAlloc( &ptr, datasize, cudaHostAllocWriteCombined) ); break;
|
2010-08-17 18:39:18 +08:00
|
|
|
default: cv::gpu::error("Invalid alloc type", __FILE__, __LINE__);
|
2010-08-13 22:52:50 +08:00
|
|
|
}
|
2010-07-14 23:55:16 +08:00
|
|
|
|
2010-07-19 18:49:35 +08:00
|
|
|
datastart = data = (uchar*)ptr;
|
|
|
|
dataend = data + nettosize;
|
2010-07-14 23:55:16 +08:00
|
|
|
|
2010-07-19 17:31:12 +08:00
|
|
|
refcount = (int*)cv::fastMalloc(sizeof(*refcount));
|
|
|
|
*refcount = 1;
|
|
|
|
}
|
|
|
|
}
|
2010-07-14 23:55:16 +08:00
|
|
|
|
2010-08-19 01:13:01 +08:00
|
|
|
GpuMat cv::gpu::CudaMem::createGpuMatHeader () const
|
2010-08-13 22:52:50 +08:00
|
|
|
{
|
2010-08-17 18:39:18 +08:00
|
|
|
GpuMat res;
|
2010-08-13 22:52:50 +08:00
|
|
|
if (alloc_type == ALLOC_ZEROCOPY)
|
|
|
|
{
|
2010-08-17 18:39:18 +08:00
|
|
|
void *pdev;
|
|
|
|
cudaSafeCall( cudaHostGetDevicePointer( &pdev, data, 0 ) );
|
|
|
|
res = GpuMat(rows, cols, type(), pdev, step);
|
2010-08-13 22:52:50 +08:00
|
|
|
}
|
|
|
|
else
|
2010-08-17 18:39:18 +08:00
|
|
|
cv::gpu::error("Zero-copy is not supported or memory was allocated without zero-copy flag", __FILE__, __LINE__);
|
2010-11-08 17:55:10 +08:00
|
|
|
|
2010-08-17 18:39:18 +08:00
|
|
|
return res;
|
2010-08-13 22:52:50 +08:00
|
|
|
}
|
|
|
|
|
2010-08-17 18:39:18 +08:00
|
|
|
void cv::gpu::CudaMem::release()
|
2010-07-19 17:31:12 +08:00
|
|
|
{
|
|
|
|
if( refcount && CV_XADD(refcount, -1) == 1 )
|
|
|
|
{
|
|
|
|
cudaSafeCall( cudaFreeHost(datastart ) );
|
|
|
|
fastFree(refcount);
|
|
|
|
}
|
|
|
|
data = datastart = dataend = 0;
|
|
|
|
step = rows = cols = 0;
|
|
|
|
refcount = 0;
|
|
|
|
}
|
2010-07-14 23:55:16 +08:00
|
|
|
|
2010-07-19 18:49:35 +08:00
|
|
|
#endif /* !defined (HAVE_CUDA) */
|