opencv/modules/core/src/gpumat.cpp

/*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"
#include "opencv2/core/gpumat.hpp"

using namespace std;
using namespace cv;
using namespace cv::gpu;

cv::gpu::GpuMat::GpuMat(const GpuMat& 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)
{
    if (refcount)
        CV_XADD(refcount, 1);
}

cv::gpu::GpuMat::GpuMat(int rows_, int cols_, int type_, void* data_, size_t step_) : 
    flags(Mat::MAGIC_VAL + (type_ & TYPE_MASK)), rows(rows_), cols(cols_), 
    step(step_), data((uchar*)data_), refcount(0),
    datastart((uchar*)data_), dataend((uchar*)data_)
{
    size_t minstep = cols * elemSize();

    if (step == Mat::AUTO_STEP)
    {
        step = minstep;
        flags |= Mat::CONTINUOUS_FLAG;
    }
    else
    {
        if (rows == 1) 
            step = minstep;

        CV_DbgAssert(step >= minstep);

        flags |= step == minstep ? Mat::CONTINUOUS_FLAG : 0;
    }
    dataend += step * (rows - 1) + minstep;
}

cv::gpu::GpuMat::GpuMat(Size size_, int type_, void* data_, size_t step_) : 
    flags(Mat::MAGIC_VAL + (type_ & TYPE_MASK)), rows(size_.height), cols(size_.width),
    step(step_), data((uchar*)data_), refcount(0),
    datastart((uchar*)data_), dataend((uchar*)data_)
{
    size_t minstep = cols * elemSize();

    if (step == Mat::AUTO_STEP)
    {
        step = minstep;
        flags |= Mat::CONTINUOUS_FLAG;
    }
    else
    {
        if (rows == 1) 
            step = minstep;

        CV_DbgAssert(step >= minstep);

        flags |= step == minstep ? Mat::CONTINUOUS_FLAG : 0;
    }
    dataend += step * (rows - 1) + minstep;
}

cv::gpu::GpuMat::GpuMat(const GpuMat& m, Range rowRange, Range colRange)
{
    flags = m.flags;
    step = m.step; refcount = m.refcount;
    data = m.data; datastart = m.datastart; dataend = m.dataend;

    if (rowRange == Range::all())
        rows = m.rows;
    else
    {
        CV_Assert(0 <= rowRange.start && rowRange.start <= rowRange.end && rowRange.end <= m.rows);

        rows = rowRange.size();
        data += step*rowRange.start;
    }

    if (colRange == Range::all())
        cols = m.cols;
    else
    {
        CV_Assert(0 <= colRange.start && colRange.start <= colRange.end && colRange.end <= m.cols);

        cols = colRange.size();
        data += colRange.start*elemSize();
        flags &= cols < m.cols ? ~Mat::CONTINUOUS_FLAG : -1;
    }

    if (rows == 1)
        flags |= Mat::CONTINUOUS_FLAG;

    if (refcount)
        CV_XADD(refcount, 1);

    if (rows <= 0 || cols <= 0)
        rows = cols = 0;
}

cv::gpu::GpuMat::GpuMat(const GpuMat& m, Rect roi) : 
    flags(m.flags), rows(roi.height), cols(roi.width),
    step(m.step), data(m.data + roi.y*step), refcount(m.refcount),
    datastart(m.datastart), dataend(m.dataend)
{
    flags &= roi.width < m.cols ? ~Mat::CONTINUOUS_FLAG : -1;
    data += roi.x * elemSize();

    CV_Assert(0 <= roi.x && 0 <= roi.width && roi.x + roi.width <= m.cols && 0 <= roi.y && 0 <= roi.height && roi.y + roi.height <= m.rows);

    if (refcount)
        CV_XADD(refcount, 1);

    if (rows <= 0 || cols <= 0)
        rows = cols = 0;
}

cv::gpu::GpuMat::GpuMat(const Mat& m) : 
    flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0) 
{ 
    upload(m); 
}

GpuMat& cv::gpu::GpuMat::operator = (const GpuMat& m)
{
    if (this != &m)
    {
        GpuMat temp(m);
        swap(temp);
    }

    return *this;
}

void cv::gpu::GpuMat::swap(GpuMat& b)
{
    std::swap(flags, b.flags);
    std::swap(rows, b.rows); 
    std::swap(cols, b.cols);
    std::swap(step, b.step); 
    std::swap(data, b.data);
    std::swap(datastart, b.datastart);
    std::swap(dataend, b.dataend);
    std::swap(refcount, b.refcount);
}

void cv::gpu::GpuMat::locateROI(Size& wholeSize, Point& ofs) const
{
    size_t esz = elemSize();
    ptrdiff_t delta1 = data - datastart;
    ptrdiff_t delta2 = dataend - datastart;

    CV_DbgAssert(step > 0);

    if (delta1 == 0)
        ofs.x = ofs.y = 0;
    else
    {
        ofs.y = static_cast<int>(delta1 / step);
        ofs.x = static_cast<int>((delta1 - step * ofs.y) / esz);

        CV_DbgAssert(data == datastart + ofs.y * step + ofs.x * esz);
    }

    size_t minstep = (ofs.x + cols) * esz;

    wholeSize.height = std::max(static_cast<int>((delta2 - minstep) / step + 1), ofs.y + rows);
    wholeSize.width = std::max(static_cast<int>((delta2 - step * (wholeSize.height - 1)) / esz), ofs.x + cols);
}

GpuMat& cv::gpu::GpuMat::adjustROI(int dtop, int dbottom, int dleft, int dright)
{
    Size wholeSize; 
    Point ofs;
    locateROI(wholeSize, ofs);

    size_t esz = elemSize();

    int row1 = std::max(ofs.y - dtop, 0); 
    int row2 = std::min(ofs.y + rows + dbottom, wholeSize.height);

    int col1 = std::max(ofs.x - dleft, 0);
    int col2 = std::min(ofs.x + cols + dright, wholeSize.width);

    data += (row1 - ofs.y) * step + (col1 - ofs.x) * esz;
    rows = row2 - row1; 
    cols = col2 - col1;

    if (esz * cols == step || rows == 1)
        flags |= Mat::CONTINUOUS_FLAG;
    else
        flags &= ~Mat::CONTINUOUS_FLAG;

    return *this;
}

GpuMat cv::gpu::GpuMat::reshape(int new_cn, int new_rows) const
{
    GpuMat hdr = *this;

    int cn = channels();
    if (new_cn == 0)
        new_cn = cn;

    int total_width = cols * cn;

    if ((new_cn > total_width || total_width % new_cn != 0) && new_rows == 0)
        new_rows = rows * total_width / new_cn;

    if (new_rows != 0 && new_rows != rows)
    {
        int total_size = total_width * rows;

        if (!isContinuous())
            CV_Error(CV_BadStep, "The matrix is not continuous, thus its number of rows can not be changed");

        if ((unsigned)new_rows > (unsigned)total_size)
            CV_Error(CV_StsOutOfRange, "Bad new number of rows");

        total_width = total_size / new_rows;

        if (total_width * new_rows != total_size)
            CV_Error(CV_StsBadArg, "The total number of matrix elements is not divisible by the new number of rows");

        hdr.rows = new_rows;
        hdr.step = total_width * elemSize1();
    }

    int new_width = total_width / new_cn;

    if (new_width * new_cn != total_width)
        CV_Error(CV_BadNumChannels, "The total width is not divisible by the new number of channels");

    hdr.cols = new_width;
    hdr.flags = (hdr.flags & ~CV_MAT_CN_MASK) | ((new_cn - 1) << CV_CN_SHIFT);

    return hdr;
}

cv::Mat::Mat(const GpuMat& m) : flags(0), dims(0), rows(0), cols(0), data(0), refcount(0), datastart(0), dataend(0), datalimit(0), allocator(0), size(&rows)
{
    m.download(*this);
}

namespace
{
    void throw_nogpu() 
    { 
        CV_Error(CV_GpuNotSupported, "The library is compiled without GPU support"); 
    }

    class EmptyFuncTable : public GpuFuncTable
    {
    public:
        void copy(const Mat&, GpuMat&) const { throw_nogpu(); }
        void copy(const GpuMat&, Mat&) const { throw_nogpu(); }
        void copy(const GpuMat&, GpuMat&) const { throw_nogpu(); }

        void copyWithMask(const GpuMat&, GpuMat&, const GpuMat&) const { throw_nogpu(); }

        void convert(const GpuMat&, GpuMat&) const { throw_nogpu(); }
        void convert(const GpuMat&, GpuMat&, double, double) const { throw_nogpu(); }

        void setTo(GpuMat&, Scalar, const GpuMat&) const { throw_nogpu(); }

        void mallocPitch(void**, size_t*, size_t, size_t) const { throw_nogpu(); }
        void free(void*) const {}
    };

    const GpuFuncTable* g_funcTbl = 0;

    const GpuFuncTable* gpuFuncTable()
    {
        static EmptyFuncTable empty;
        return g_funcTbl ? g_funcTbl : &empty;
    }
}

void cv::gpu::setGpuFuncTable(const GpuFuncTable* funcTbl)
{
    g_funcTbl = funcTbl;
}

void cv::gpu::GpuMat::upload(const Mat& m)
{
    CV_DbgAssert(!m.empty());

    create(m.size(), m.type());

    gpuFuncTable()->copy(m, *this);
}

void cv::gpu::GpuMat::download(Mat& m) const
{
    CV_DbgAssert(!empty());

    m.create(size(), type());

    gpuFuncTable()->copy(*this, m);
}

void cv::gpu::GpuMat::copyTo(GpuMat& m) const
{
    CV_DbgAssert(!empty());

    m.create(size(), type());

    gpuFuncTable()->copy(*this, m);
}

void cv::gpu::GpuMat::copyTo(GpuMat& mat, const GpuMat& mask) const
{
    if (mask.empty())
        copyTo(mat);
    else
    {
        mat.create(size(), type());

        gpuFuncTable()->copyWithMask(*this, mat, mask);
    }
}

void cv::gpu::GpuMat::convertTo(GpuMat& dst, int rtype, double alpha, double beta) const
{
    bool noScale = fabs(alpha - 1) < numeric_limits<double>::epsilon() && fabs(beta) < numeric_limits<double>::epsilon();

    if (rtype < 0)
        rtype = type();
    else
        rtype = CV_MAKETYPE(CV_MAT_DEPTH(rtype), channels());

    int sdepth = depth();
    int ddepth = CV_MAT_DEPTH(rtype);
    if (sdepth == ddepth && noScale)
    {
        copyTo(dst);
        return;
    }

    GpuMat temp;
    const GpuMat* psrc = this;
    if (sdepth != ddepth && psrc == &dst)
    {
        temp = *this;
        psrc = &temp;
    }

    dst.create(size(), rtype);

    if (noScale)
        gpuFuncTable()->convert(*psrc, dst);
    else
        gpuFuncTable()->convert(*psrc, dst, alpha, beta);
}

GpuMat& cv::gpu::GpuMat::setTo(Scalar s, const GpuMat& mask)
{
    CV_Assert(mask.empty() || mask.type() == CV_8UC1);
    CV_DbgAssert(!empty());

    gpuFuncTable()->setTo(*this, s, mask);    

    return *this;
}

void cv::gpu::GpuMat::create(int _rows, int _cols, int _type)
{
    _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)
    {
        flags = Mat::MAGIC_VAL + _type;
        rows = _rows;
        cols = _cols;

        size_t esz = elemSize();

        void* devPtr;
        gpuFuncTable()->mallocPitch(&devPtr, &step, esz * cols, rows);

        // Single row must be continuous
        if (rows == 1)
            step = esz * cols;

        if (esz * cols == step)
            flags |= Mat::CONTINUOUS_FLAG;

        int64 _nettosize = static_cast<int64>(step) * rows;
        size_t nettosize = static_cast<size_t>(_nettosize);

        datastart = data = static_cast<uchar*>(devPtr);
        dataend = data + nettosize;

        refcount = static_cast<int*>(fastMalloc(sizeof(*refcount)));
        *refcount = 1;
    }
}

void cv::gpu::GpuMat::release()
{
    if (refcount && CV_XADD(refcount, -1) == 1)
    {
        fastFree(refcount);

        gpuFuncTable()->free(datastart);
    }

    data = datastart = dataend = 0;
    step = rows = cols = 0;
    refcount = 0;
}