Picture.cpp

#if JVET_AB0082
  int filterLengthsLuma[3] = { 8, 12, 12 };
  int filterLengthsChroma[3] = { 4, 6, 6 };
  const int filterLength = downsampling ? 12 : rescaleForDisplay ? (useLumaFilter ? filterLengthsLuma[upscaleFilterForDisplay] : filterLengthsChroma[upscaleFilterForDisplay]) : (useLumaFilter ? NTAPS_LUMA(0) : NTAPS_CHROMA);
#else
  const int filterLength = downsampling ? 12 : (useLumaFilter ? NTAPS_LUMA(1) : NTAPS_CHROMA);
#endif
#else
  const int filterLength = downsampling ? 12 : (useLumaFilter ? NTAPS_LUMA : NTAPS_CHROMA);
#endif
#if RPR_ENABLE && IF_12TAP
  const int log2Norm = downsampling ? 14 : 16;
#else
  const int log2Norm = downsampling ? 14 : 12;
#endif

  int *buf = new int[orgHeight * scaledWidth];
  int maxVal = ( 1 << bitDepth ) - 1;

  CHECK( bitDepth > 17, "Overflow may happen!" );

  for( int i = 0; i < scaledWidth; i++ )
  {
    const Pel* org = orgSrc;
    int refPos = ( ( ( i << compScale.first ) - afterScaleLeftOffset ) * scalingRatio.first + addX ) >> posShiftX;
    int integer = refPos >> numFracShift;
    int frac = refPos & numFracPositions;
    int* tmp = buf + i;

    for( int j = 0; j < orgHeight; j++ )
    {
      int sum = 0;
      const TFilterCoeff* f = filterHor + frac * filterLength;

      for( int k = 0; k < filterLength; k++ )
      {
        int xInt = std::min<int>( std::max( 0, integer + k - filterLength / 2 + 1 ), orgWidth - 1 );
        sum += f[k] * org[xInt]; // postpone horizontal filtering gain removal after vertical filtering
      }

      *tmp = sum;

      tmp += scaledWidth;
      org += orgStride;
    }
  }

  Pel* dst = scaledSrc;

  for( int j = 0; j < scaledHeight; j++ )
  {
    int refPos = ( ( ( j << compScale.second ) - afterScaleTopOffset ) * scalingRatio.second + addY ) >> posShiftY;
    int integer = refPos >> numFracShift;
    int frac = refPos & numFracPositions;

    for( int i = 0; i < scaledWidth; i++ )
    {
#if RPR_ENABLE && IF_12TAP
      uint64_t sum = 0;
#else
      int sum = 0;
#endif
      int* tmp = buf + i;
      const TFilterCoeff* f = filterVer + frac * filterLength;

      for( int k = 0; k < filterLength; k++ )
      {
        int yInt = std::min<int>( std::max( 0, integer + k - filterLength / 2 + 1 ), orgHeight - 1 );
        sum += f[k] * tmp[yInt*scaledWidth];
      }

#if RPR_ENABLE && IF_12TAP
      const uint64_t one = 1;
      int sumS = (int)((sum + (one << (log2Norm - 1))) >> log2Norm);
      dst[i] = std::min<int>(std::max(0, sumS), maxVal);
#else
      dst[i] = std::min<int>( std::max( 0, ( sum + ( 1 << ( log2Norm - 1 ) ) ) >> log2Norm ), maxVal );
#endif
    }

    dst += scaledStride;
  }

  delete[] buf;
}

void Picture::rescalePicture( const std::pair<int, int> scalingRatio,
                              const CPelUnitBuf& beforeScaling, const Window& scalingWindowBefore,
                              const PelUnitBuf& afterScaling, const Window& scalingWindowAfter,
                              const ChromaFormat chromaFormatIDC, const BitDepths& bitDepths, const bool useLumaFilter, const bool downsampling,
                              const bool horCollocatedChromaFlag, const bool verCollocatedChromaFlag
#if JVET_AB0082
                            , bool rescaleForDisplay, int upscaleFilterForDisplay
#endif
)
{
  for( int comp = 0; comp < ::getNumberValidComponents( chromaFormatIDC ); comp++ )
  {
    ComponentID compID = ComponentID( comp );
    const CPelBuf& beforeScale = beforeScaling.get( compID );
    const PelBuf& afterScale = afterScaling.get( compID );

    sampleRateConv( scalingRatio, std::pair<int, int>( ::getComponentScaleX( compID, chromaFormatIDC ), ::getComponentScaleY( compID, chromaFormatIDC ) ),
                    beforeScale, scalingWindowBefore.getWindowLeftOffset() * SPS::getWinUnitX( chromaFormatIDC ), scalingWindowBefore.getWindowTopOffset() * SPS::getWinUnitY( chromaFormatIDC ),
                    afterScale, scalingWindowAfter.getWindowLeftOffset() * SPS::getWinUnitX( chromaFormatIDC ), scalingWindowAfter.getWindowTopOffset() * SPS::getWinUnitY( chromaFormatIDC ),
                    bitDepths.recon[toChannelType(compID)], downsampling || useLumaFilter ? true : isLuma( compID ), downsampling,
                    isLuma( compID ) ? 1 : horCollocatedChromaFlag, isLuma( compID ) ? 1 : verCollocatedChromaFlag
#if JVET_AB0082
                  , rescaleForDisplay, upscaleFilterForDisplay
#endif
    );
  }
}

void Picture::saveSubPicBorder(int POC, int subPicX0, int subPicY0, int subPicWidth, int subPicHeight)
{
  // 1.1 set up margin for back up memory allocation
  int xMargin = margin >> getComponentScaleX(COMPONENT_Y, cs->area.chromaFormat);
  int yMargin = margin >> getComponentScaleY(COMPONENT_Y, cs->area.chromaFormat);

  // 1.2 measure the size of back up memory
  Area areaAboveBelow(0, 0, subPicWidth + 2 * xMargin, yMargin);
  Area areaLeftRight(0, 0, xMargin, subPicHeight);
  UnitArea unitAreaAboveBelow(cs->area.chromaFormat, areaAboveBelow);
  UnitArea unitAreaLeftRight(cs->area.chromaFormat, areaLeftRight);

  // 1.3 create back up memory
  m_bufSubPicAbove.create(unitAreaAboveBelow);
  m_bufSubPicBelow.create(unitAreaAboveBelow);
  m_bufSubPicLeft.create(unitAreaLeftRight);
  m_bufSubPicRight.create(unitAreaLeftRight);
  m_bufWrapSubPicAbove.create(unitAreaAboveBelow);
  m_bufWrapSubPicBelow.create(unitAreaAboveBelow);

  for (int comp = 0; comp < getNumberValidComponents(cs->area.chromaFormat); comp++)
  {
    ComponentID compID = ComponentID(comp);

    // 2.1 measure the margin for each component
    int xmargin = margin >> getComponentScaleX(compID, cs->area.chromaFormat);
    int ymargin = margin >> getComponentScaleY(compID, cs->area.chromaFormat);

    // 2.2 calculate the origin of the subpicture
    int left = subPicX0 >> getComponentScaleX(compID, cs->area.chromaFormat);
    int top = subPicY0 >> getComponentScaleY(compID, cs->area.chromaFormat);

    // 2.3 calculate the width/height of the subPic
    int width = subPicWidth >> getComponentScaleX(compID, cs->area.chromaFormat);
    int height = subPicHeight >> getComponentScaleY(compID, cs->area.chromaFormat);


    // 3.1.1 set reconstructed picture
    PelBuf s = M_BUFS(0, PIC_RECONSTRUCTION).get(compID);
    Pel *src = s.bufAt(left, top);

    // 3.2.1 set back up buffer for left
    PelBuf dBufLeft   = m_bufSubPicLeft.getBuf(compID);
    Pel    *dstLeft   = dBufLeft.bufAt(0, 0);


    // 3.2.2 set back up buffer for right
    PelBuf dBufRight  = m_bufSubPicRight.getBuf(compID);
    Pel    *dstRight  = dBufRight.bufAt(0, 0);

    // 3.2.3 copy to recon picture to back up buffer
    Pel *srcLeft  = src - xmargin;
    Pel *srcRight = src + width;
    for (int y = 0; y < height; y++)
    {
      ::memcpy(dstLeft  + y *  dBufLeft.stride, srcLeft  + y * s.stride, sizeof(Pel) * xmargin);
      ::memcpy(dstRight + y * dBufRight.stride, srcRight + y * s.stride, sizeof(Pel) * xmargin);
    }

    // 3.3.1 set back up buffer for above
    PelBuf dBufTop = m_bufSubPicAbove.getBuf(compID);
    Pel    *dstTop = dBufTop.bufAt(0, 0);

    // 3.3.2 set back up buffer for below
    PelBuf dBufBottom = m_bufSubPicBelow.getBuf(compID);
    Pel    *dstBottom = dBufBottom.bufAt(0, 0);

    // 3.3.3 copy to recon picture to back up buffer
    Pel *srcTop    = src - xmargin - ymargin * s.stride;
    Pel *srcBottom = src - xmargin +  height * s.stride;
    for (int y = 0; y < ymargin; y++)
    {
      ::memcpy(dstTop    + y *    dBufTop.stride, srcTop    + y * s.stride, sizeof(Pel) * (2 * xmargin + width));
      ::memcpy(dstBottom + y * dBufBottom.stride, srcBottom + y * s.stride, sizeof(Pel) * (2 * xmargin + width));
    }

    // back up recon wrap buffer
    if (cs->sps->getWrapAroundEnabledFlag())
    {
      PelBuf sWrap = M_BUFS(0, PIC_RECON_WRAP).get(compID);
      Pel *srcWrap = sWrap.bufAt(left, top);

      // 3.4.1 set back up buffer for above
      PelBuf dBufTopWrap = m_bufWrapSubPicAbove.getBuf(compID);
      Pel    *dstTopWrap = dBufTopWrap.bufAt(0, 0);

      // 3.4.2 set back up buffer for below
      PelBuf dBufBottomWrap = m_bufWrapSubPicBelow.getBuf(compID);
      Pel    *dstBottomWrap = dBufBottomWrap.bufAt(0, 0);

      // 3.4.3 copy recon wrap picture to back up buffer
      Pel *srcTopWrap    = srcWrap - xmargin - ymargin * sWrap.stride;
      Pel *srcBottomWrap = srcWrap - xmargin +  height * sWrap.stride;
      for (int y = 0; y < ymargin; y++)
      {
        ::memcpy(dstTopWrap    + y *    dBufTopWrap.stride, srcTopWrap    + y * sWrap.stride, sizeof(Pel) * (2 * xmargin + width));
        ::memcpy(dstBottomWrap + y * dBufBottomWrap.stride, srcBottomWrap + y * sWrap.stride, sizeof(Pel) * (2 * xmargin + width));
      }
    }
  }
}

void Picture::extendSubPicBorder(int POC, int subPicX0, int subPicY0, int subPicWidth, int subPicHeight)
{
  for (int comp = 0; comp < getNumberValidComponents(cs->area.chromaFormat); comp++)
  {
    ComponentID compID = ComponentID(comp);

    // 2.1 measure the margin for each component
    int xmargin = margin >> getComponentScaleX(compID, cs->area.chromaFormat);
    int ymargin = margin >> getComponentScaleY(compID, cs->area.chromaFormat);

    // 2.2 calculate the origin of the Subpicture
    int left = subPicX0 >> getComponentScaleX(compID, cs->area.chromaFormat);
    int top = subPicY0 >> getComponentScaleY(compID, cs->area.chromaFormat);

    // 2.3 calculate the width/height of the Subpicture
    int width = subPicWidth >> getComponentScaleX(compID, cs->area.chromaFormat);
    int height = subPicHeight >> getComponentScaleY(compID, cs->area.chromaFormat);

#if JVET_Z0118_GDR
    int numPt = (cs->isGdrEnabled()) ? 2 : 1;    
    for (int i = 0; i < numPt; i++)
    {
      PelBuf s = M_BUFS(0, PIC_RECONSTRUCTION+i).get(compID);
      Pel *src = s.bufAt(left, top);
#else
      // 3.1 set reconstructed picture
      PelBuf s = M_BUFS(0, PIC_RECONSTRUCTION).get(compID);
      Pel *src = s.bufAt(left, top);
#endif

      // 4.1 apply padding for left and right    
      Pel *dstLeft  = src - xmargin;
      Pel *dstRight = src + width;
      Pel *srcLeft  = src + 0;
      Pel *srcRight = src + width - 1;

      for (int y = 0; y < height; y++)
      {
        for (int x = 0; x < xmargin; x++)
        {
          dstLeft[x]  = *srcLeft;
          dstRight[x] = *srcRight;
        }
        dstLeft += s.stride;
        dstRight += s.stride;
        srcLeft += s.stride;
        srcRight += s.stride;
      }    

      // 4.2 apply padding on bottom
      Pel *srcBottom = src + s.stride * (height - 1) - xmargin;
      Pel *dstBottom = srcBottom + s.stride;
      for (int y = 0; y < ymargin; y++)
      {
        ::memcpy(dstBottom, srcBottom, sizeof(Pel)*(2 * xmargin + width));
        dstBottom += s.stride;
      }

      // 4.3 apply padding for top
      // si is still (-marginX, SubpictureHeight-1)
      Pel *srcTop = src - xmargin;
      Pel *dstTop = srcTop - s.stride;
      // si is now (-marginX, 0)
      for (int y = 0; y < ymargin; y++)
      {
        ::memcpy(dstTop, srcTop, sizeof(Pel)*(2 * xmargin + width));
        dstTop -= s.stride;
      }
#if JVET_Z0118_GDR
    } // for loop
#endif 

    // Appy padding for recon wrap buffer
    if (cs->sps->getWrapAroundEnabledFlag())
    {
      // set recon wrap picture
      PelBuf sWrap = M_BUFS(0, PIC_RECON_WRAP).get(compID);
      Pel *srcWrap = sWrap.bufAt(left, top);

      // apply padding on bottom
      Pel *srcBottomWrap = srcWrap + sWrap.stride * (height - 1) - xmargin;
      Pel *dstBottomWrap = srcBottomWrap + sWrap.stride;
      for (int y = 0; y < ymargin; y++)
      {
        ::memcpy(dstBottomWrap, srcBottomWrap, sizeof(Pel)*(2 * xmargin + width));
        dstBottomWrap += sWrap.stride;
      }

      // apply padding for top
      // si is still (-marginX, SubpictureHeight-1)
      Pel *srcTopWrap = srcWrap - xmargin;
      Pel *dstTopWrap = srcTopWrap - sWrap.stride;
      // si is now (-marginX, 0)
      for (int y = 0; y < ymargin; y++)
      {
        ::memcpy(dstTopWrap, srcTopWrap, sizeof(Pel)*(2 * xmargin + width));
        dstTopWrap -= sWrap.stride;
      }
    }
  } // end of for
}

void Picture::restoreSubPicBorder(int POC, int subPicX0, int subPicY0, int subPicWidth, int subPicHeight)
{
  for (int comp = 0; comp < getNumberValidComponents(cs->area.chromaFormat); comp++)
  {
    ComponentID compID = ComponentID(comp);

    // 2.1 measure the margin for each component
    int xmargin = margin >> getComponentScaleX(compID, cs->area.chromaFormat);
    int ymargin = margin >> getComponentScaleY(compID, cs->area.chromaFormat);

    // 2.2 calculate the origin of the subpicture
    int left = subPicX0 >> getComponentScaleX(compID, cs->area.chromaFormat);
    int top = subPicY0 >> getComponentScaleY(compID, cs->area.chromaFormat);

    // 2.3 calculate the width/height of the subpicture
    int width = subPicWidth >> getComponentScaleX(compID, cs->area.chromaFormat);
    int height = subPicHeight >> getComponentScaleY(compID, cs->area.chromaFormat);

    // 3.1 set reconstructed picture
    PelBuf s = M_BUFS(0, PIC_RECONSTRUCTION).get(compID);
    Pel *src = s.bufAt(left, top);

    // 4.2.1 copy from back up buffer to recon picture
    PelBuf dBufLeft = m_bufSubPicLeft.getBuf(compID);
    Pel    *dstLeft = dBufLeft.bufAt(0, 0);

    // 4.2.2 set back up buffer for right
    PelBuf dBufRight = m_bufSubPicRight.getBuf(compID);
    Pel    *dstRight = dBufRight.bufAt(0, 0);

    // 4.2.3 copy to recon picture to back up buffer
    Pel *srcLeft  = src - xmargin;
    Pel *srcRight = src + width;

    for (int y = 0; y < height; y++)
    {
      // the destination and source position is reversed on purpose
      ::memcpy(srcLeft  + y * s.stride,  dstLeft + y *  dBufLeft.stride, sizeof(Pel) * xmargin);
      ::memcpy(srcRight + y * s.stride, dstRight + y * dBufRight.stride, sizeof(Pel) * xmargin);
    }


    // 4.3.1 set back up buffer for above
    PelBuf dBufTop = m_bufSubPicAbove.getBuf(compID);
    Pel    *dstTop = dBufTop.bufAt(0, 0);

    // 4.3.2 set back up buffer for below
    PelBuf dBufBottom = m_bufSubPicBelow.getBuf(compID);
    Pel    *dstBottom = dBufBottom.bufAt(0, 0);

    // 4.3.3 copy to recon picture to back up buffer
    Pel *srcTop = src - xmargin - ymargin * s.stride;
    Pel *srcBottom = src - xmargin + height * s.stride;

    for (int y = 0; y < ymargin; y++)
    {
      ::memcpy(srcTop    + y * s.stride, dstTop    + y *    dBufTop.stride, sizeof(Pel) * (2 * xmargin + width));
      ::memcpy(srcBottom + y * s.stride, dstBottom + y * dBufBottom.stride, sizeof(Pel) * (2 * xmargin + width));
    }

    // restore recon wrap buffer
    if (cs->sps->getWrapAroundEnabledFlag())
    {
      // set recon wrap picture
      PelBuf sWrap = M_BUFS(0, PIC_RECON_WRAP).get(compID);
      Pel *srcWrap = sWrap.bufAt(left, top);

      // set back up buffer for above
      PelBuf dBufTopWrap = m_bufWrapSubPicAbove.getBuf(compID);
      Pel    *dstTopWrap = dBufTopWrap.bufAt(0, 0);

      // set back up buffer for below
      PelBuf dBufBottomWrap = m_bufWrapSubPicBelow.getBuf(compID);
      Pel    *dstBottomWrap = dBufBottomWrap.bufAt(0, 0);

      // copy to recon wrap picture from back up buffer
      Pel *srcTopWrap = srcWrap - xmargin - ymargin * sWrap.stride;
      Pel *srcBottomWrap = srcWrap - xmargin + height * sWrap.stride;

      for (int y = 0; y < ymargin; y++)
      {
        ::memcpy(srcTopWrap    + y * sWrap.stride, dstTopWrap    + y *    dBufTopWrap.stride, sizeof(Pel) * (2 * xmargin + width));
        ::memcpy(srcBottomWrap + y * sWrap.stride, dstBottomWrap + y * dBufBottomWrap.stride, sizeof(Pel) * (2 * xmargin + width));
      }
    }
  }

  // 5.0 destroy the back up memory
  m_bufSubPicAbove.destroy();
  m_bufSubPicBelow.destroy();
  m_bufSubPicLeft.destroy();
  m_bufSubPicRight.destroy();
  m_bufWrapSubPicAbove.destroy();
  m_bufWrapSubPicBelow.destroy();
}

void Picture::extendPicBorder( const PPS *pps )
{
  if ( m_bIsBorderExtended )
  {
    if( isWrapAroundEnabled( pps ) && ( !m_wrapAroundValid || m_wrapAroundOffset != pps->getWrapAroundOffset() ) )
    {
      extendWrapBorder( pps );
    }
    return;
  }


#if JVET_AK0085_TM_BOUNDARY_PADDING
    int picWidth = cs->sps->getMaxPicWidthInLumaSamples();
    int picHeight = cs->sps->getMaxPicHeightInLumaSamples();

    PelUnitBuf s = M_BUFS(0, PIC_RECONSTRUCTION);

    BoundaryTop topB = 0;
    BoundaryBottom bottomB = 0;
    BoundaryLeft leftB = 0;
    BoundaryRight rightB = 0;

    Area picAreaTopBottom = Area(Position(0, 0), Size(picWidth, picHeight));
    TemplateMatchingPadding(s, topB, picAreaTopBottom);
    TemplateMatchingPadding(s, bottomB, picAreaTopBottom);

    Area picAreaLeftRight = Area(Position(0, -TMP_PADSIZE), Size(picWidth, picHeight+2*TMP_PADSIZE));
    TemplateMatchingPadding(s, leftB, picAreaLeftRight);
    TemplateMatchingPadding(s, rightB, picAreaLeftRight);
#endif


#if JVET_Z0118_GDR
  int numPt = (cs->isGdrEnabled()) ? PIC_RECONSTRUCTION_1 : PIC_RECONSTRUCTION_0;  
  for (int pt = (int) PIC_RECONSTRUCTION_0; pt <= (int) numPt; pt++)
  {
    for (int comp = 0; comp < getNumberValidComponents(cs->area.chromaFormat); comp++)
    {
      ComponentID compID = ComponentID(comp);
      PelBuf p = M_BUFS(0, (PictureType) pt).get(compID);
#if JVET_AK0085_TM_BOUNDARY_PADDING
      int width = p.width + ((2*TMP_PADSIZE) >> getComponentScaleX(compID, cs->area.chromaFormat));
      int height = p.height + ((2*TMP_PADSIZE) >> getComponentScaleY(compID, cs->area.chromaFormat));

      int tmpOffsetX = TMP_PADSIZE >> getComponentScaleX( compID, cs->area.chromaFormat );
      int tmpOffsetY = TMP_PADSIZE >> getComponentScaleY( compID, cs->area.chromaFormat );

      Pel *piTxt = p.bufAt(-tmpOffsetX,-tmpOffsetY);

      int xmargin = (margin - TMP_PADSIZE) >> getComponentScaleX( compID, cs->area.chromaFormat );
      int ymargin = (margin - TMP_PADSIZE) >> getComponentScaleY( compID, cs->area.chromaFormat );
#else
      int width = p.width;
      int height = p.height;

      Pel *piTxt = p.bufAt(0, 0);
      int xmargin = margin >> getComponentScaleX(compID, cs->area.chromaFormat);
      int ymargin = margin >> getComponentScaleY(compID, cs->area.chromaFormat);
#endif

      Pel*  pi = piTxt;
      // do left and right margins
      for (int y = 0; y < height; y++)
      {
        for (int x = 0; x < xmargin; x++)
        {
          pi[-xmargin + x] = pi[0];
          pi[width + x] = pi[width - 1];
        }
        pi += p.stride;
      }

      // pi is now the (0,height) (bottom left of image within bigger picture
      pi -= (p.stride + xmargin);
      // pi is now the (-marginX, height-1)
      for (int y = 0; y < ymargin; y++)
      {
        ::memcpy(pi + (y + 1)*p.stride, pi, sizeof(Pel)*(width + (xmargin << 1)));
      }

      // pi is still (-marginX, height-1)
      pi -= ((height - 1) * p.stride);
      // pi is now (-marginX, 0)
      for (int y = 0; y < ymargin; y++)
      {
        ::memcpy(pi - (y + 1)*p.stride, pi, sizeof(Pel)*(width + (xmargin << 1)));
      }

      // reference picture with horizontal wrapped boundary
      if (isWrapAroundEnabled(pps))
      {
        extendWrapBorder(pps);
      }
      else
      {
        m_wrapAroundValid = false;
        m_wrapAroundOffset = 0;
      }
    }
  }
#else
  for(int comp=0; comp<getNumberValidComponents( cs->area.chromaFormat ); comp++)
  {
    ComponentID compID = ComponentID( comp );
    PelBuf p = M_BUFS( 0, PIC_RECONSTRUCTION ).get( compID );

#if JVET_AK0085_TM_BOUNDARY_PADDING
    int width = p.width + ((2*TMP_PADSIZE) >> getComponentScaleX(compID, cs->area.chromaFormat));
    int height = p.height + ((2*TMP_PADSIZE) >> getComponentScaleY(compID, cs->area.chromaFormat));

    Pel *piTxt = p.bufAt(-TMP_PADSIZE,-TMP_PADSIZE);
    int xmargin = (margin - TMP_PADSIZE) >> getComponentScaleX( compID, cs->area.chromaFormat );
    int ymargin = (margin - TMP_PADSIZE) >> getComponentScaleY( compID, cs->area.chromaFormat );
#else
    int width = p.width;
    int height = p.height;

    Pel *piTxt = p.bufAt(0,0);
    int xmargin = margin >> getComponentScaleX( compID, cs->area.chromaFormat );
    int ymargin = margin >> getComponentScaleY( compID, cs->area.chromaFormat );
#endif

    Pel*  pi = piTxt;
    // do left and right margins
    for (int y = 0; y < height; y++)
    {
      for (int x = 0; x < xmargin; x++)
      {
        pi[-xmargin + x] = pi[0];
        pi[width + x]  = pi[width - 1];
      }
      pi += p.stride;
    }

    // pi is now the (0,height) (bottom left of image within bigger picture
    pi -= (p.stride + xmargin);
    // pi is now the (-marginX, height-1)
    for (int y = 0; y < ymargin; y++ )
    {
      ::memcpy( pi + (y+1)*p.stride, pi, sizeof(Pel)*(width + (xmargin << 1)));
    }

    // pi is still (-marginX, height-1)
    pi -= ((height-1) * p.stride);
    // pi is now (-marginX, 0)
    for (int y = 0; y < ymargin; y++ )
    {
      ::memcpy( pi - (y+1)*p.stride, pi, sizeof(Pel)*(width + (xmargin<<1)) );
    }

    // reference picture with horizontal wrapped boundary
    if ( isWrapAroundEnabled( pps ) )
    {
      extendWrapBorder( pps );
    }
    else
    {
      m_wrapAroundValid = false;
      m_wrapAroundOffset = 0;
    }
  }
#endif

  m_bIsBorderExtended = true;
}

void Picture::extendWrapBorder( const PPS *pps )
{
  for(int comp=0; comp<getNumberValidComponents( cs->area.chromaFormat ); comp++)
  {
    ComponentID compID = ComponentID( comp );
    PelBuf p = M_BUFS( 0, PIC_RECON_WRAP ).get( compID );
    p.copyFrom(M_BUFS( 0, PIC_RECONSTRUCTION ).get( compID ));
    Pel *piTxt = p.bufAt(0,0);
    int xmargin = margin >> getComponentScaleX( compID, cs->area.chromaFormat );
    int ymargin = margin >> getComponentScaleY( compID, cs->area.chromaFormat );
    Pel*  pi = piTxt;
    int xoffset = pps->getWrapAroundOffset() >> getComponentScaleX( compID, cs->area.chromaFormat );
    for (int y = 0; y < p.height; y++)
    {
      for (int x = 0; x < xmargin; x++ )
      {
        if( x < xoffset )
        {
          pi[ -x - 1 ] = pi[ -x - 1 + xoffset ];
          pi[  p.width + x ] = pi[ p.width + x - xoffset ];
        }
        else
        {
          pi[ -x - 1 ] = pi[ 0 ];
          pi[  p.width + x ] = pi[ p.width - 1 ];
        }
      }
      pi += p.stride;
    }
    pi -= (p.stride + xmargin);
    for (int y = 0; y < ymargin; y++ )
    {
      ::memcpy( pi + (y+1)*p.stride, pi, sizeof(Pel)*(p.width + (xmargin << 1)));
    }
    pi -= ((p.height-1) * p.stride);
    for (int y = 0; y < ymargin; y++ )
    {
      ::memcpy( pi - (y+1)*p.stride, pi, sizeof(Pel)*(p.width + (xmargin<<1)) );
    }
  }
  m_wrapAroundValid = true;
  m_wrapAroundOffset = pps->getWrapAroundOffset();
}

PelBuf Picture::getBuf( const ComponentID compID, const PictureType &type )
{
#if JVET_Z0118_GDR
  if (type == PIC_RECONSTRUCTION_0 || type == PIC_RECONSTRUCTION_1)
  {
    return M_BUFS(scheduler.getSplitPicId(), type).getBuf(compID);
  }
#endif

  return M_BUFS( ( type == PIC_ORIGINAL || type == PIC_TRUE_ORIGINAL || type == PIC_FILTERED_ORIGINAL || type == PIC_ORIGINAL_INPUT || type == PIC_TRUE_ORIGINAL_INPUT || type == PIC_FILTERED_ORIGINAL_INPUT ) ? 0 : scheduler.getSplitPicId(), type ).getBuf( compID );
}

const CPelBuf Picture::getBuf( const ComponentID compID, const PictureType &type ) const
{
#if JVET_Z0118_GDR
  if (type == PIC_RECONSTRUCTION_0 || type == PIC_RECONSTRUCTION_1)
  {
    return M_BUFS(scheduler.getSplitPicId(), type).getBuf(compID);
  }
#endif

  return M_BUFS( ( type == PIC_ORIGINAL || type == PIC_TRUE_ORIGINAL || type == PIC_FILTERED_ORIGINAL || type == PIC_ORIGINAL_INPUT || type == PIC_TRUE_ORIGINAL_INPUT || type == PIC_FILTERED_ORIGINAL_INPUT ) ? 0 : scheduler.getSplitPicId(), type ).getBuf( compID );
}

PelBuf Picture::getBuf( const CompArea &blk, const PictureType &type )
{
  if( !blk.valid() )
  {
    return PelBuf();
  }

#if ENABLE_SPLIT_PARALLELISM
  const int jId = ( type == PIC_ORIGINAL || type == PIC_TRUE_ORIGINAL || type == PIC_ORIGINAL_INPUT || type == PIC_TRUE_ORIGINAL_INPUT ) ? 0 : scheduler.getSplitPicId();
#endif
#if !KEEP_PRED_AND_RESI_SIGNALS
#if JVET_AC0162_ALF_RESIDUAL_SAMPLES_INPUT
  if (type == PIC_PREDICTION)
#else
  if( type == PIC_RESIDUAL || type == PIC_PREDICTION )
#endif
  {
    CompArea localBlk = blk;
    localBlk.x &= ( cs->pcv->maxCUWidthMask  >> getComponentScaleX( blk.compID, blk.chromaFormat ) );
    localBlk.y &= ( cs->pcv->maxCUHeightMask >> getComponentScaleY( blk.compID, blk.chromaFormat ) );

    return M_BUFS( jId, type ).getBuf( localBlk );
  }
#endif

  return M_BUFS( jId, type ).getBuf( blk );
}

const CPelBuf Picture::getBuf( const CompArea &blk, const PictureType &type ) const
{
  if( !blk.valid() )
  {
    return PelBuf();
  }

#if ENABLE_SPLIT_PARALLELISM
  const int jId = ( type == PIC_ORIGINAL || type == PIC_TRUE_ORIGINAL ) ? 0 : scheduler.getSplitPicId();

#endif
#if !KEEP_PRED_AND_RESI_SIGNALS
#if JVET_AC0162_ALF_RESIDUAL_SAMPLES_INPUT
  if (type == PIC_PREDICTION)
#else
  if( type == PIC_RESIDUAL || type == PIC_PREDICTION )
#endif
  {
    CompArea localBlk = blk;
    localBlk.x &= ( cs->pcv->maxCUWidthMask  >> getComponentScaleX( blk.compID, blk.chromaFormat ) );
    localBlk.y &= ( cs->pcv->maxCUHeightMask >> getComponentScaleY( blk.compID, blk.chromaFormat ) );

    return M_BUFS( jId, type ).getBuf( localBlk );
  }
#endif

  return M_BUFS( jId, type ).getBuf( blk );
}

PelUnitBuf Picture::getBuf( const UnitArea &unit, const PictureType &type )
{
  if( chromaFormat == CHROMA_400 )
  {
    return PelUnitBuf( chromaFormat, getBuf( unit.Y(), type ) );
  }
  else
  {
    return PelUnitBuf( chromaFormat, getBuf( unit.Y(), type ), getBuf( unit.Cb(), type ), getBuf( unit.Cr(), type ) );
  }
}

const CPelUnitBuf Picture::getBuf( const UnitArea &unit, const PictureType &type ) const
{
  if( chromaFormat == CHROMA_400 )
  {
    return CPelUnitBuf( chromaFormat, getBuf( unit.Y(), type ) );
  }
  else
  {
    return CPelUnitBuf( chromaFormat, getBuf( unit.Y(), type ), getBuf( unit.Cb(), type ), getBuf( unit.Cr(), type ) );
  }
}

Pel* Picture::getOrigin( const PictureType &type, const ComponentID compID ) const
{
#if ENABLE_SPLIT_PARALLELISM
  const int jId = ( type == PIC_ORIGINAL || type == PIC_TRUE_ORIGINAL ) ? 0 : scheduler.getSplitPicId();
#endif
  return M_BUFS( jId, type ).getOrigin( compID );
}

void Picture::createSpliceIdx(int nums)
{
  m_ctuNums = nums;
  m_spliceIdx = new int[m_ctuNums];
  memset(m_spliceIdx, 0, m_ctuNums * sizeof(int));
}

bool Picture::getSpliceFull()
{
  int count = 0;
  for (int i = 0; i < m_ctuNums; i++)
  {
    if (m_spliceIdx[i] != 0)
    {
      count++;
    }
  }
  if (count < m_ctuNums * 0.25)
  {
    return false;
  }
  return true;
}

void Picture::addPictureToHashMapForInter()
{
  int picWidth = slices[0]->getPPS()->getPicWidthInLumaSamples();
  int picHeight = slices[0]->getPPS()->getPicHeightInLumaSamples();
  uint32_t* blockHashValues[2][2];
  bool* bIsBlockSame[2][3];

  for (int i = 0; i < 2; i++)
  {
    for (int j = 0; j < 2; j++)
    {
      blockHashValues[i][j] = new uint32_t[picWidth*picHeight];
    }

    for (int j = 0; j < 3; j++)
    {
      bIsBlockSame[i][j] = new bool[picWidth*picHeight];
    }
  }
  m_hashMap.create(picWidth, picHeight);
  m_hashMap.generateBlock2x2HashValue(getOrigBuf(), picWidth, picHeight, slices[0]->getSPS()->getBitDepths(), blockHashValues[0], bIsBlockSame[0]);//2x2
  m_hashMap.generateBlockHashValue(picWidth, picHeight, 4, 4, blockHashValues[0], blockHashValues[1], bIsBlockSame[0], bIsBlockSame[1]);//4x4
  m_hashMap.addToHashMapByRowWithPrecalData(blockHashValues[1], bIsBlockSame[1][2], picWidth, picHeight, 4, 4);

  m_hashMap.generateBlockHashValue(picWidth, picHeight, 8, 8, blockHashValues[1], blockHashValues[0], bIsBlockSame[1], bIsBlockSame[0]);//8x8
  m_hashMap.addToHashMapByRowWithPrecalData(blockHashValues[0], bIsBlockSame[0][2], picWidth, picHeight, 8, 8);

  m_hashMap.generateBlockHashValue(picWidth, picHeight, 16, 16, blockHashValues[0], blockHashValues[1], bIsBlockSame[0], bIsBlockSame[1]);//16x16
  m_hashMap.addToHashMapByRowWithPrecalData(blockHashValues[1], bIsBlockSame[1][2], picWidth, picHeight, 16, 16);

  m_hashMap.generateBlockHashValue(picWidth, picHeight, 32, 32, blockHashValues[1], blockHashValues[0], bIsBlockSame[1], bIsBlockSame[0]);//32x32
  m_hashMap.addToHashMapByRowWithPrecalData(blockHashValues[0], bIsBlockSame[0][2], picWidth, picHeight, 32, 32);

  m_hashMap.generateBlockHashValue(picWidth, picHeight, 64, 64, blockHashValues[0], blockHashValues[1], bIsBlockSame[0], bIsBlockSame[1]);//64x64
  m_hashMap.addToHashMapByRowWithPrecalData(blockHashValues[1], bIsBlockSame[1][2], picWidth, picHeight, 64, 64);

  m_hashMap.setInitial();

  for (int i = 0; i < 2; i++)
  {
    for (int j = 0; j < 2; j++)
    {
      delete[] blockHashValues[i][j];
    }

    for (int j = 0; j < 3; j++)
    {
      delete[] bIsBlockSame[i][j];
    }
  }
}

#if JVET_Z0118_GDR
void Picture::initCleanCurPicture()
{   
  if (!cs->isGdrEnabled())
  {
    return;
  }

  const int picWidth = getPicWidthInLumaSamples();
  const int picHight = getPicHeightInLumaSamples();
  const int bitDepth = slices[0]->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA);
  const Pel dirtyPelVal = 1 << (bitDepth - 1);
      
  UnitArea wholePictureArea = UnitArea(chromaFormat, Area(Position(0, 0), Size(picWidth, picHight)));

  getBuf(wholePictureArea, PIC_RECONSTRUCTION_0).fill(dirtyPelVal);
  getBuf(wholePictureArea, PIC_RECONSTRUCTION_1).fill(dirtyPelVal);

  cs->getMotionBuf(wholePictureArea, PIC_RECONSTRUCTION_0).fill(0);
  cs->getMotionBuf(wholePictureArea, PIC_RECONSTRUCTION_1).fill(0);
   
#if JVET_W0123_TIMD_FUSION
  cs->getIpmBuf(wholePictureArea, PIC_RECONSTRUCTION_0).fill(0);
  cs->getIpmBuf(wholePictureArea, PIC_RECONSTRUCTION_1).fill(0);
#endif
}

void Picture::copyCleanCurPicture()
{
  if (!cs->isGdrEnabled())
  {
    return;
  }

  if (cs->isInGdrIntervalOrRecoveryPoc())
  {
    ChromaFormat chromaFormat = cs->sps->getChromaFormatIdc();
    int gdrEndX = cs->picHeader->getGdrEndX();
    int gdrEndY = cs->pps->getPicHeightInLumaSamples();

    UnitArea cleanArea = UnitArea(chromaFormat, Area(Position(0, 0), Size(gdrEndX, gdrEndY)));

    PelUnitBuf picBuf0 = getBuf(cleanArea, PIC_RECONSTRUCTION_0);
    PelUnitBuf picBuf1 = getBuf(cleanArea, PIC_RECONSTRUCTION_1);

    picBuf1.copyFrom(picBuf0);
  }
}
#endif

#if JVET_AG0145_ADAPTIVE_CLIPPING
void Picture::calcLumaClpParams()
{
  int pelMax = getLumaClpRng().max;
  int pelMin = getLumaClpRng().min;
#if JVET_AI0096_ADAPTIVE_CLIPPING_BIT_DEPTH_FIX
  int targetMin = 16 * (1 << (cs->sps->getBitDepth(toChannelType(COMPONENT_Y)) - 8));
  int targetMax = 235 * (1 << (cs->sps->getBitDepth(toChannelType(COMPONENT_Y)) - 8));
#else
  int targetMin = 64, targetMax = 940;
#endif
  if (cs->slice->getSliceType() != I_SLICE)
  {
    const Picture *const pColPic = cs->slice->getRefPic(RefPicList(1 - cs->slice->getColFromL0Flag()), cs->slice->getColRefIdx())->unscaledPic;
    ClpRng colLumaClpRng = pColPic->getLumaClpRng();
    targetMin            = colLumaClpRng.min;
    targetMax            = colLumaClpRng.max;
  }
  int clipDeltaShift = 0;
  if (cs->slice->getSliceType() != I_SLICE && cs->slice->getCheckLDC())
  {
    clipDeltaShift = ADAPTIVE_CLIP_SHIFT_DELTA_VALUE_1;
    cs->slice->setAdaptiveClipQuant(true);
  }
  else
  {
    clipDeltaShift = ADAPTIVE_CLIP_SHIFT_DELTA_VALUE_0;
    cs->slice->setAdaptiveClipQuant(false);
  }
#if JVET_AJ0237_INTERNAL_12BIT
  clipDeltaShift += std::max(0, cs->sps->getBitDepth(toChannelType(COMPONENT_Y)) - 10);
#endif
  int       pelMaxOF  = 0;
  int       pelMinOF  = (1 << cs->sps->getBitDepth(toChannelType(COMPONENT_Y))) - 1;
  const int orgPelMin = pelMin;
  {
    int deltaMinToSignal = (pelMin - targetMin);
    if (deltaMinToSignal < 0)
    {
      int absDelta = ((targetMin - pelMin) >> clipDeltaShift) << clipDeltaShift;
      pelMin       = targetMin - absDelta;
      while (pelMin > orgPelMin)
      {
        pelMin -= (1 << clipDeltaShift);
      }
      while (pelMin < 0)
      {
        pelMinOF = pelMin;
        pelMin   = 0;
      }
      CHECK(pelMin < 0, "this is not possible");
    }
    else if (deltaMinToSignal > 0)
    {
      int absDelta = (deltaMinToSignal >> clipDeltaShift) << clipDeltaShift;
      pelMin       = targetMin + absDelta;
      CHECK(pelMin > orgPelMin, "this is not possible");
      CHECK(pelMin < 0, "this is not possible");
    }
    else
    {
      CHECK(pelMin != targetMin, "this is not possible");
    }
  }

  const int orgPelMax = pelMax;
  {
    int deltaMaxToSignal = (pelMax - targetMax);
    if (deltaMaxToSignal < 0)
    {
      int absDelta = ((targetMax - pelMax) >> clipDeltaShift) << clipDeltaShift;
      pelMax       = targetMax - absDelta;
      CHECK(pelMax < orgPelMax, "this is not possible");
      CHECK(pelMax > (1 << cs->sps->getBitDepth(toChannelType(COMPONENT_Y))) - 1, "this is not possible");
    }
    else if (deltaMaxToSignal > 0)
    {
      int absDelta = (deltaMaxToSignal >> clipDeltaShift) << clipDeltaShift;
      pelMax       = targetMax + absDelta;
      while (pelMax < orgPelMax)
      {
        pelMax += (1 << clipDeltaShift);
      }
      while (pelMax >= (1 << cs->sps->getBitDepth(toChannelType(COMPONENT_Y))))
      {
        pelMaxOF = pelMax;
        pelMax   = (1 << cs->sps->getBitDepth(toChannelType(COMPONENT_Y))) - 1;
      }
      CHECK(pelMax > (1 << cs->sps->getBitDepth(toChannelType(COMPONENT_Y))) - 1, "this is not possible");
    }
    else
    {
      CHECK(pelMax != targetMax, "this is not possible");
    }
  }
  cs->slice->setLumaPelMax(pelMax);
  cs->slice->setLumaPelMin(pelMin);
  lumaClpRng.min         = pelMin;
  lumaClpRng.max         = pelMax;
  lumaClpRngforQuant.min = std::min(pelMin, pelMinOF);
  lumaClpRngforQuant.max = std::max(pelMax, pelMaxOF);
}
#endif