InterPrediction.cpp

        {
          iMvScaleTmpHor = iMvScaleHor + iDMvHorX * ( cxWidth >> 1 ) + iDMvVerX * ( cxHeight >> 1 );
          iMvScaleTmpVer = iMvScaleVer + iDMvHorY * ( cxWidth >> 1 ) + iDMvVerY * ( cxHeight >> 1 );
        }
        roundAffineMv(iMvScaleTmpHor, iMvScaleTmpVer, shift);
        Mv tmpMv(iMvScaleTmpHor, iMvScaleTmpVer);
        tmpMv.clipToStorageBitDepth();
        iMvScaleTmpHor = tmpMv.getHor();
        iMvScaleTmpVer = tmpMv.getVer();

        // clip and scale
        if (sps.getWrapAroundEnabledFlag())
        {
          m_storedMv[h / AFFINE_MIN_BLOCK_SIZE * MVBUFFER_SIZE + w / AFFINE_MIN_BLOCK_SIZE].set(iMvScaleTmpHor, iMvScaleTmpVer);
          Mv tmpMv(iMvScaleTmpHor, iMvScaleTmpVer);
          wrapRef = wrapClipMv( tmpMv, Position( pu.Y().x + w, pu.Y().y + h ), Size( blockWidth, blockHeight ), &sps, pu.cs->pps );
          iMvScaleTmpHor = tmpMv.getHor();
          iMvScaleTmpVer = tmpMv.getVer();
        }
        else
        {
          wrapRef = false;
          m_storedMv[h / AFFINE_MIN_BLOCK_SIZE * MVBUFFER_SIZE + w / AFFINE_MIN_BLOCK_SIZE].set(iMvScaleTmpHor, iMvScaleTmpVer);
          if( scalingRatio == SCALE_1X ) 
          {
            iMvScaleTmpHor = std::min<int>(iHorMax, std::max<int>(iHorMin, iMvScaleTmpHor));
            iMvScaleTmpVer = std::min<int>(iVerMax, std::max<int>(iVerMin, iMvScaleTmpVer));
          }
        }
      }
      else
      {
        Mv curMv = m_storedMv[((h << iScaleY) / AFFINE_MIN_BLOCK_SIZE) * MVBUFFER_SIZE + ((w << iScaleX) / AFFINE_MIN_BLOCK_SIZE)] +
          m_storedMv[((h << iScaleY) / AFFINE_MIN_BLOCK_SIZE + iScaleY)* MVBUFFER_SIZE + ((w << iScaleX) / AFFINE_MIN_BLOCK_SIZE + iScaleX)];
        roundAffineMv(curMv.hor, curMv.ver, 1);
        if (sps.getWrapAroundEnabledFlag())
        {
          wrapRef = wrapClipMv( curMv, Position( pu.Y().x + ( w << iScaleX ), pu.Y().y + ( h << iScaleY ) ), Size( blockWidth << iScaleX, blockHeight << iScaleY ), &sps, pu.cs->pps );
        }
        else
        {
          wrapRef = false;
          if( scalingRatio == SCALE_1X ) 
          {
            curMv.hor = std::min<int>(iHorMax, std::max<int>(iHorMin, curMv.hor));
            curMv.ver = std::min<int>(iVerMax, std::max<int>(iVerMin, curMv.ver));
          }
        }
        iMvScaleTmpHor = curMv.hor;
        iMvScaleTmpVer = curMv.ver;
      }

      if( xPredInterBlkRPR( scalingRatio, *pu.cs->pps, CompArea( compID, chFmt, pu.blocks[compID].offset( w, h ), Size( blockWidth, blockHeight ) ), refPic, Mv( iMvScaleTmpHor, iMvScaleTmpVer ), dstBuf.buf + w + h * dstBuf.stride, dstBuf.stride, bi, wrapRef, clpRng, 2 ) )
      {
        CHECK( enablePROF, "PROF should be disabled with RPR" );
      }
      else
      {
      // get the MV in high precision
      int xFrac, yFrac, xInt, yInt;

      if (!iScaleX)
      {
        xInt  = iMvScaleTmpHor >> 4;
        xFrac = iMvScaleTmpHor & 15;
      }
      else
      {
        xInt  = iMvScaleTmpHor >> 5;
        xFrac = iMvScaleTmpHor & 31;
      }
      if (!iScaleY)
      {
        yInt  = iMvScaleTmpVer >> 4;
        yFrac = iMvScaleTmpVer & 15;
      }
      else
      {
        yInt  = iMvScaleTmpVer >> 5;
        yFrac = iMvScaleTmpVer & 31;
      }

      const CPelBuf refBuf = refPic->getRecoBuf( CompArea( compID, chFmt, pu.blocks[compID].offset(xInt + w, yInt + h), pu.blocks[compID] ), wrapRef );

      Pel* ref = (Pel*) refBuf.buf;
      Pel* dst = dstBuf.buf + w + h * dstBuf.stride;

      int refStride = refBuf.stride;
      int dstStride = dstBuf.stride;

      int bw = blockWidth;
      int bh = blockHeight;

      if (enablePROF)
      {
        dst = dstExtBuf.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
        dstStride = dstExtBuf.stride;
      }

      if ( yFrac == 0 )
      {
        m_if.filterHor( compID, (Pel*) ref, refStride, dst, dstStride, bw, bh, xFrac, isLast, chFmt, clpRng);
      }
      else if ( xFrac == 0 )
      {
        m_if.filterVer( compID, (Pel*) ref, refStride, dst, dstStride, bw, bh, yFrac, true, isLast, chFmt, clpRng);
      }
      else
      {
        m_if.filterHor( compID, (Pel*)ref - ((vFilterSize>>1) -1)*refStride, refStride, tmpBuf.buf, tmpBuf.stride, bw, bh+vFilterSize-1, xFrac, false,      chFmt, clpRng);
        JVET_J0090_SET_CACHE_ENABLE( false );
        m_if.filterVer( compID, tmpBuf.buf + ((vFilterSize>>1) -1)*tmpBuf.stride, tmpBuf.stride, dst, dstStride, bw, bh, yFrac, false, isLast, chFmt, clpRng);
        JVET_J0090_SET_CACHE_ENABLE( true );
      }
      if (enablePROF)
      {
        const int shift = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
        const int xOffset = xFrac >> 3;
        const int yOffset = yFrac >> 3;

        const int refOffset = (blockHeight + 1) * refStride;
        const int dstOffset = (blockHeight + 1)* dstStride;

        const Pel* refPel = ref - (1 - yOffset) * refStride + xOffset - 1;
        Pel* dstPel = dst - dstStride - 1;
        for (int pw = 0; pw < blockWidth + 2; pw++)
        {
          dstPel[pw] = leftShift_round(refPel[pw], shift) - (Pel)IF_INTERNAL_OFFS;
          dstPel[pw+dstOffset] = leftShift_round(refPel[pw+refOffset], shift) - (Pel)IF_INTERNAL_OFFS;
        }

        refPel = ref + yOffset * refBuf.stride + xOffset;
        dstPel = dst;
        for (int ph = 0; ph < blockHeight; ph++, refPel += refStride, dstPel += dstStride)
        {
          dstPel[-1] = leftShift_round(refPel[-1], shift) - (Pel)IF_INTERNAL_OFFS;
          dstPel[blockWidth] = leftShift_round(refPel[blockWidth], shift) - (Pel)IF_INTERNAL_OFFS;
        }

#if JVET_P0154_PROF_SAMPLE_OFFSET_CLIPPING
        PelBuf gradXBuf = gradXExt.subBuf(0, 0, blockWidth + 2, blockHeight + 2);
        PelBuf gradYBuf = gradYExt.subBuf(0, 0, blockWidth + 2, blockHeight + 2);
#else
        PelBuf gradXBuf = gradXExt.subBuf(w, h, blockWidth + 2, blockHeight + 2);
        PelBuf gradYBuf = gradYExt.subBuf(w, h, blockWidth + 2, blockHeight + 2);
#endif
        g_pelBufOP.profGradFilter(dstExtBuf.buf, dstExtBuf.stride, blockWidth + 2, blockHeight + 2, gradXBuf.stride, gradXBuf.buf, gradYBuf.buf, clpRng.bd);

        const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
        const Pel offset = (1 << (shiftNum - 1)) + IF_INTERNAL_OFFS;
        Pel* src = dstExtBuf.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
        Pel* gX = gradXBuf.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
        Pel* gY = gradYBuf.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);

        Pel * dstY = dstBuf.bufAt(w, h);

#if JVET_P0154_PROF_SAMPLE_OFFSET_CLIPPING
        g_pelBufOP.applyPROF(dstY, dstBuf.stride, src, dstExtBuf.stride, blockWidth, blockHeight, gX, gY, gradXBuf.stride, dMvScaleHor, dMvScaleVer, blockWidth, bi, shiftNum, offset, clpRng);
#else
        if (!bi)
        {
          g_pelBufOP.applyPROF(dstY, dstBuf.stride, src, dstExtBuf.stride, blockWidth, blockHeight, gX, gY, gradXBuf.stride, dMvScaleHor, dMvScaleVer, blockWidth, shiftNum, offset, clpRng);
        }
        else
        {
          PelBuf srcExtBuf(src, dstExtBuf.stride, Size(blockWidth, blockHeight));
          PelBuf destBuf(dstY, dstBuf.stride, Size(blockWidth, blockHeight));
          destBuf.copyFrom(srcExtBuf);
        }
#endif
      }
      }
    }
  }
}

void InterPrediction::applyBiOptFlow(const PredictionUnit &pu, const CPelUnitBuf &yuvSrc0, const CPelUnitBuf &yuvSrc1, const int &refIdx0, const int &refIdx1, PelUnitBuf &yuvDst, const BitDepths &clipBitDepths)
{
  const int     height = yuvDst.Y().height;
  const int     width = yuvDst.Y().width;
  int           heightG = height + 2 * BIO_EXTEND_SIZE;
  int           widthG = width + 2 * BIO_EXTEND_SIZE;
  int           offsetPos = widthG*BIO_EXTEND_SIZE + BIO_EXTEND_SIZE;

  Pel*          gradX0 = m_gradX0;
  Pel*          gradX1 = m_gradX1;
  Pel*          gradY0 = m_gradY0;
  Pel*          gradY1 = m_gradY1;

  int           stridePredMC = widthG + 2;
  const Pel*    srcY0 = m_filteredBlockTmp[2][COMPONENT_Y] + stridePredMC + 1;
  const Pel*    srcY1 = m_filteredBlockTmp[3][COMPONENT_Y] + stridePredMC + 1;
  const int     src0Stride = stridePredMC;
  const int     src1Stride = stridePredMC;

  Pel*          dstY = yuvDst.Y().buf;
  const int     dstStride = yuvDst.Y().stride;
  const Pel*    srcY0Temp = srcY0;
  const Pel*    srcY1Temp = srcY1;

  for (int refList = 0; refList < NUM_REF_PIC_LIST_01; refList++)
  {
    Pel* dstTempPtr = m_filteredBlockTmp[2 + refList][COMPONENT_Y] + stridePredMC + 1;
    Pel* gradY = (refList == 0) ? m_gradY0 : m_gradY1;
    Pel* gradX = (refList == 0) ? m_gradX0 : m_gradX1;

    xBioGradFilter(dstTempPtr, stridePredMC, widthG, heightG, widthG, gradX, gradY, clipBitDepths.recon[toChannelType(COMPONENT_Y)]);
    Pel* padStr = m_filteredBlockTmp[2 + refList][COMPONENT_Y] + 2 * stridePredMC + 2;
    for (int y = 0; y< height; y++)
    {
      padStr[-1] = padStr[0];
      padStr[width] = padStr[width - 1];
      padStr += stridePredMC;
    }

    padStr = m_filteredBlockTmp[2 + refList][COMPONENT_Y] + 2 * stridePredMC + 1;
    ::memcpy(padStr - stridePredMC, padStr, sizeof(Pel)*(widthG));
    ::memcpy(padStr + height*stridePredMC, padStr + (height - 1)*stridePredMC, sizeof(Pel)*(widthG));
  }

  const ClpRng& clpRng = pu.cu->cs->slice->clpRng(COMPONENT_Y);
  const int   bitDepth = clipBitDepths.recon[toChannelType(COMPONENT_Y)];
  const int   shiftNum = IF_INTERNAL_PREC + 1 - bitDepth;
  const int   offset = (1 << (shiftNum - 1)) + 2 * IF_INTERNAL_OFFS;
#if JVET_P0653_BDOF_PROF_PARA_DEV
#if JVET_P0491_BDOFPROF_MVD_RANGE
#if JVET_P0091_REMOVE_BDOF_OFFSET_SHIFT
  const int   limit = ( 1 << 4 ) - 1;
#else
  const int   limit = ( 1 << 5 ) - 1;
#endif
#else
  const int   limit = (1 << 5);
#endif
#else
  const int   limit = (1<<(std::max<int>(5, bitDepth - 7)));
#endif

  int xUnit = (width >> 2);
  int yUnit = (height >> 2);

  Pel *dstY0 = dstY;
  gradX0 = m_gradX0; gradX1 = m_gradX1;
  gradY0 = m_gradY0; gradY1 = m_gradY1;

  for (int yu = 0; yu < yUnit; yu++)
  {
    for (int xu = 0; xu < xUnit; xu++)
    {
      int tmpx = 0, tmpy = 0;
      int sumAbsGX = 0, sumAbsGY = 0, sumDIX = 0, sumDIY = 0;
      int sumSignGY_GX = 0;

      Pel* pGradX0Tmp = m_gradX0 + (xu << 2) + (yu << 2) * widthG;
      Pel* pGradX1Tmp = m_gradX1 + (xu << 2) + (yu << 2) * widthG;
      Pel* pGradY0Tmp = m_gradY0 + (xu << 2) + (yu << 2) * widthG;
      Pel* pGradY1Tmp = m_gradY1 + (xu << 2) + (yu << 2) * widthG;
      const Pel* SrcY1Tmp = srcY1 + (xu << 2) + (yu << 2) * src1Stride;
      const Pel* SrcY0Tmp = srcY0 + (xu << 2) + (yu << 2) * src0Stride;

      g_pelBufOP.calcBIOSums(SrcY0Tmp, SrcY1Tmp, pGradX0Tmp, pGradX1Tmp, pGradY0Tmp, pGradY1Tmp, xu, yu, src0Stride, src1Stride, widthG, bitDepth, &sumAbsGX, &sumAbsGY, &sumDIX, &sumDIY, &sumSignGY_GX);
#if JVET_P0091_REMOVE_BDOF_OFFSET_SHIFT
      tmpx = (sumAbsGX == 0 ? 0 : rightShiftMSB(sumDIX << 2, sumAbsGX));
#else
      tmpx = (sumAbsGX == 0 ? 0 : rightShiftMSB(sumDIX << 3, sumAbsGX));
#endif
#if JVET_P0057_BDOF_PROF_HARMONIZATION && !JVET_P0491_BDOFPROF_MVD_RANGE
      tmpx = Clip3(-limit, limit - 1, tmpx);
#else
      tmpx = Clip3(-limit, limit, tmpx);
#endif

      int     mainsGxGy = sumSignGY_GX >> 12;
      int     secsGxGy = sumSignGY_GX & ((1 << 12) - 1);
      int     tmpData = tmpx * mainsGxGy;
      tmpData = ((tmpData << 12) + tmpx*secsGxGy) >> 1;
#if JVET_P0091_REMOVE_BDOF_OFFSET_SHIFT
      tmpy = (sumAbsGY == 0 ? 0 : rightShiftMSB(((sumDIY << 2) - tmpData), sumAbsGY));
#else
      tmpy = (sumAbsGY == 0 ? 0 : rightShiftMSB(((sumDIY << 3) - tmpData), sumAbsGY));
#endif
#if JVET_P0057_BDOF_PROF_HARMONIZATION && !JVET_P0491_BDOFPROF_MVD_RANGE
      tmpy = Clip3(-limit, limit - 1, tmpy);
#else
      tmpy = Clip3(-limit, limit, tmpy);
#endif
      srcY0Temp = srcY0 + (stridePredMC + 1) + ((yu*src0Stride + xu) << 2);
      srcY1Temp = srcY1 + (stridePredMC + 1) + ((yu*src0Stride + xu) << 2);
      gradX0 = m_gradX0 + offsetPos + ((yu*widthG + xu) << 2);
      gradX1 = m_gradX1 + offsetPos + ((yu*widthG + xu) << 2);
      gradY0 = m_gradY0 + offsetPos + ((yu*widthG + xu) << 2);
      gradY1 = m_gradY1 + offsetPos + ((yu*widthG + xu) << 2);

      dstY0 = dstY + ((yu*dstStride + xu) << 2);
      xAddBIOAvg4(srcY0Temp, src0Stride, srcY1Temp, src1Stride, dstY0, dstStride, gradX0, gradX1, gradY0, gradY1, widthG, (1 << 2), (1 << 2), (int)tmpx, (int)tmpy, shiftNum, offset, clpRng);
    }  // xu
  }  // yu
}



void InterPrediction::xAddBIOAvg4(const Pel* src0, int src0Stride, const Pel* src1, int src1Stride, Pel *dst, int dstStride, const Pel *gradX0, const Pel *gradX1, const Pel *gradY0, const Pel*gradY1, int gradStride, int width, int height, int tmpx, int tmpy, int shift, int offset, const ClpRng& clpRng)
{
  g_pelBufOP.addBIOAvg4(src0, src0Stride, src1, src1Stride, dst, dstStride, gradX0, gradX1, gradY0, gradY1, gradStride, width, height, tmpx, tmpy, shift, offset, clpRng);
}

void InterPrediction::xBioGradFilter(Pel* pSrc, int srcStride, int width, int height, int gradStride, Pel* gradX, Pel* gradY, int bitDepth)
{
  g_pelBufOP.bioGradFilter(pSrc, srcStride, width, height, gradStride, gradX, gradY, bitDepth);
}

void InterPrediction::xCalcBIOPar(const Pel* srcY0Temp, const Pel* srcY1Temp, const Pel* gradX0, const Pel* gradX1, const Pel* gradY0, const Pel* gradY1, int* dotProductTemp1, int* dotProductTemp2, int* dotProductTemp3, int* dotProductTemp5, int* dotProductTemp6, const int src0Stride, const int src1Stride, const int gradStride, const int widthG, const int heightG, int bitDepth)
{
  g_pelBufOP.calcBIOPar(srcY0Temp, srcY1Temp, gradX0, gradX1, gradY0, gradY1, dotProductTemp1, dotProductTemp2, dotProductTemp3, dotProductTemp5, dotProductTemp6, src0Stride, src1Stride, gradStride, widthG, heightG, bitDepth);
}

void InterPrediction::xCalcBlkGradient(int sx, int sy, int    *arraysGx2, int     *arraysGxGy, int     *arraysGxdI, int     *arraysGy2, int     *arraysGydI, int     &sGx2, int     &sGy2, int     &sGxGy, int     &sGxdI, int     &sGydI, int width, int height, int unitSize)
{
  g_pelBufOP.calcBlkGradient(sx, sy, arraysGx2, arraysGxGy, arraysGxdI, arraysGy2, arraysGydI, sGx2, sGy2, sGxGy, sGxdI, sGydI, width, height, unitSize);
}

void InterPrediction::xWeightedAverage(const PredictionUnit& pu, const CPelUnitBuf& pcYuvSrc0, const CPelUnitBuf& pcYuvSrc1, PelUnitBuf& pcYuvDst, const BitDepths& clipBitDepths, const ClpRngs& clpRngs, const bool& bioApplied, PelUnitBuf* yuvDstTmp /*= NULL*/)
{
  const int iRefIdx0 = pu.refIdx[0];
  const int iRefIdx1 = pu.refIdx[1];

  if( iRefIdx0 >= 0 && iRefIdx1 >= 0 )
  {
#if !JVET_P0154_PROF_SAMPLE_OFFSET_CLIPPING
    if (pu.cu->affine && (m_applyPROF[0] || m_applyPROF[1]))
    {
      xApplyBiPROF(pu, pcYuvSrc0.bufs[COMPONENT_Y], pcYuvSrc1.bufs[COMPONENT_Y], pcYuvDst.bufs[COMPONENT_Y], clpRngs.comp[COMPONENT_Y]);
      pcYuvDst.addWeightedAvg(pcYuvSrc0, pcYuvSrc1, clpRngs, pu.cu->GBiIdx, true);
      CHECK(yuvDstTmp, "yuvDstTmp is disallowed with PROF");
      return;
    }
#endif
    if( pu.cu->GBiIdx != GBI_DEFAULT && (yuvDstTmp || !pu.mhIntraFlag) )
    {
      CHECK(bioApplied, "GBi is disallowed with BIO");
      pcYuvDst.addWeightedAvg(pcYuvSrc0, pcYuvSrc1, clpRngs, pu.cu->GBiIdx);
      if (yuvDstTmp)
        yuvDstTmp->addAvg(pcYuvSrc0, pcYuvSrc1, clpRngs, false);
      return;
    }
    if (bioApplied)
    {
      const int  src0Stride = pu.lwidth() + 2 * BIO_EXTEND_SIZE + 2;
      const int  src1Stride = pu.lwidth() + 2 * BIO_EXTEND_SIZE + 2;
      const Pel* pSrcY0 = m_filteredBlockTmp[2][COMPONENT_Y] + 2 * src0Stride + 2;
      const Pel* pSrcY1 = m_filteredBlockTmp[3][COMPONENT_Y] + 2 * src1Stride + 2;

      bool bioEnabled = true;
      if (bioEnabled)
      {
        applyBiOptFlow(pu, pcYuvSrc0, pcYuvSrc1, iRefIdx0, iRefIdx1, pcYuvDst, clipBitDepths);
        if (yuvDstTmp)
          yuvDstTmp->bufs[0].addAvg(CPelBuf(pSrcY0, src0Stride, pu.lumaSize()), CPelBuf(pSrcY1, src1Stride, pu.lumaSize()), clpRngs.comp[0]);
      }
      else
      {
        pcYuvDst.bufs[0].addAvg(CPelBuf(pSrcY0, src0Stride, pu.lumaSize()), CPelBuf(pSrcY1, src1Stride, pu.lumaSize()), clpRngs.comp[0]);
        if (yuvDstTmp)
          yuvDstTmp->bufs[0].copyFrom(pcYuvDst.bufs[0]);
      }
    }
    if (pu.cs->pps->getWPBiPred())
    {
      const int iRefIdx0 = pu.refIdx[0];
      const int iRefIdx1 = pu.refIdx[1];
      WPScalingParam  *pwp0;
      WPScalingParam  *pwp1;
      getWpScaling(pu.cu->slice, iRefIdx0, iRefIdx1, pwp0, pwp1);
      if (!bioApplied)
      {
        addWeightBiComponent(pcYuvSrc0, pcYuvSrc1, pu.cu->slice->clpRngs(), pwp0, pwp1, pcYuvDst, true, COMPONENT_Y);
      }
      addWeightBiComponent(pcYuvSrc0, pcYuvSrc1, pu.cu->slice->clpRngs(), pwp0, pwp1, pcYuvDst, true, COMPONENT_Cb);
      addWeightBiComponent(pcYuvSrc0, pcYuvSrc1, pu.cu->slice->clpRngs(), pwp0, pwp1, pcYuvDst, true, COMPONENT_Cr);
    }
    else
    {
      pcYuvDst.addAvg(pcYuvSrc0, pcYuvSrc1, clpRngs, bioApplied);
    }
    if (yuvDstTmp)
    {
      if (bioApplied)
      {
        yuvDstTmp->bufs[1].copyFrom(pcYuvDst.bufs[1]);
        yuvDstTmp->bufs[2].copyFrom(pcYuvDst.bufs[2]);
      }
      else
        yuvDstTmp->copyFrom(pcYuvDst);
    }
  }
  else if( iRefIdx0 >= 0 && iRefIdx1 < 0 )
  {
    if( pu.cu->triangle )
    {
      pcYuvDst.copyFrom( pcYuvSrc0 );
    }
    else
    pcYuvDst.copyClip( pcYuvSrc0, clpRngs );
    if (yuvDstTmp)
      yuvDstTmp->copyFrom(pcYuvDst);
  }
  else if( iRefIdx0 < 0 && iRefIdx1 >= 0 )
  {
    if( pu.cu->triangle )
    {
      pcYuvDst.copyFrom( pcYuvSrc1 );
    }
    else
    pcYuvDst.copyClip( pcYuvSrc1, clpRngs );
    if (yuvDstTmp)
      yuvDstTmp->copyFrom(pcYuvDst);
  }
}

#if !JVET_P0154_PROF_SAMPLE_OFFSET_CLIPPING
void InterPrediction::xApplyBiPROF(const PredictionUnit &pu, const CPelBuf& pcYuvSrc0, const CPelBuf& pcYuvSrc1, PelBuf& pcYuvDst, const ClpRng& clpRng)
{
  int blockWidth = AFFINE_MIN_BLOCK_SIZE;
  int blockHeight = AFFINE_MIN_BLOCK_SIZE;

  CHECK(!m_applyPROF[0] && !m_applyPROF[1], "xApplyBiPROF() applies PROF for at least one list.");
  const int width = pu.Y().width;
  const int height = pu.Y().height;

  const int bit = MAX_CU_DEPTH;
#if JVET_P0653_BDOF_PROF_PARA_DEV
  const int mvShift  = 8;
#if JVET_P0491_BDOFPROF_MVD_RANGE
  const int dmvLimit = ( 1 << 5 ) - 1;
#else
  const int dmvLimit = (1 << 5);
#endif
#else
  const int shift = bit - 4 + MV_FRACTIONAL_BITS_INTERNAL;
#if JVET_P0057_BDOF_PROF_HARMONIZATION 
  const int mvShift = shift + MV_FRACTIONAL_BITS_INTERNAL + 2 - std::max<int>(5, clpRng.bd - 7);
  const int dmvLimit = (1 << (std::max<int>(5, clpRng.bd - 7)));
#else
  const int bdlimit = std::max<int>(6, clpRng.bd - 6);
  const int dmvLimit = 1 << bdlimit;
#endif
#endif

  for (int list = 0; list < 2; list++)
  {
    if (m_applyPROF[list])
    {
      Mv mvLT = pu.mvAffi[list][0];
      Mv mvRT = pu.mvAffi[list][1];
      Mv mvLB = pu.mvAffi[list][2];

      int dMvHorX, dMvHorY, dMvVerX, dMvVerY;
      dMvHorX = (mvRT - mvLT).getHor() << (bit - floorLog2(width));
      dMvHorY = (mvRT - mvLT).getVer() << (bit - floorLog2(width));
      if (pu.cu->affineType == AFFINEMODEL_6PARAM)
      {
        dMvVerX = (mvLB - mvLT).getHor() << (bit - floorLog2(height));
        dMvVerY = (mvLB - mvLT).getVer() << (bit - floorLog2(height));
      }
      else
      {
        dMvVerX = -dMvHorY;
        dMvVerY = dMvHorX;
      }

      int *dMvScaleHor = m_dMvBuf[list];
      int *dMvScaleVer = m_dMvBuf[list] + 16;

      int* dMvH = dMvScaleHor;
      int* dMvV = dMvScaleVer;
      int  quadHorX = dMvHorX << 2;
      int  quadHorY = dMvHorY << 2;
      int  quadVerX = dMvVerX << 2;
      int  quadVerY = dMvVerY << 2;

      dMvH[0] = ((dMvHorX + dMvVerX) << 1) - ((quadHorX + quadVerX) << 1);
      dMvV[0] = ((dMvHorY + dMvVerY) << 1) - ((quadHorY + quadVerY) << 1);

      for (int w = 1; w < blockWidth; w++)
      {
        dMvH[w] = dMvH[w - 1] + quadHorX;
        dMvV[w] = dMvV[w - 1] + quadHorY;
      }

      dMvH += blockWidth;
      dMvV += blockWidth;
      for (int h = 1; h < blockHeight; h++)
      {
        for (int w = 0; w < blockWidth; w++)
        {
          dMvH[w] = dMvH[w - blockWidth] + quadVerX;
          dMvV[w] = dMvV[w - blockWidth] + quadVerY;
        }
        dMvH += blockWidth;
        dMvV += blockWidth;
      }

      if (!g_pelBufOP.roundIntVector)
      {
        for (int idx = 0; idx < blockWidth * blockHeight; idx++)
        {
#if JVET_P0057_BDOF_PROF_HARMONIZATION 
          roundAffineMv(dMvScaleHor[idx], dMvScaleVer[idx], mvShift);
#else
          roundAffineMv(dMvScaleHor[idx], dMvScaleVer[idx], shift);
#endif
#if JVET_P0491_BDOFPROF_MVD_RANGE
          dMvScaleHor[idx] = Clip3( -dmvLimit, dmvLimit, dMvScaleHor[idx] );
          dMvScaleVer[idx] = Clip3( -dmvLimit, dmvLimit, dMvScaleVer[idx] );
#else
          dMvScaleHor[idx] = Clip3(-dmvLimit, dmvLimit - 1, dMvScaleHor[idx]);
          dMvScaleVer[idx] = Clip3(-dmvLimit, dmvLimit - 1, dMvScaleVer[idx]);
#endif
        }
      }
      else
      {
        int sz = blockWidth * blockHeight;
#if JVET_P0057_BDOF_PROF_HARMONIZATION 
        g_pelBufOP.roundIntVector(dMvScaleHor, sz, mvShift, dmvLimit);
        g_pelBufOP.roundIntVector(dMvScaleVer, sz, mvShift, dmvLimit);
#else
        g_pelBufOP.roundIntVector(dMvScaleHor, sz, shift, dmvLimit);
        g_pelBufOP.roundIntVector(dMvScaleVer, sz, shift, dmvLimit);
#endif
      }
    }
  }

  const int cuExtW = width + PROF_BORDER_EXT_W * 2;
  const int cuExtH = height + PROF_BORDER_EXT_H * 2;

  PelBuf gradXExt0 = PelBuf(m_gradBuf[REF_PIC_LIST_0][0], cuExtW, cuExtH);
  PelBuf gradYExt0 = PelBuf(m_gradBuf[REF_PIC_LIST_0][1], cuExtW, cuExtH);
  PelBuf gradXExt1 = PelBuf(m_gradBuf[REF_PIC_LIST_1][0], cuExtW, cuExtH);
  PelBuf gradYExt1 = PelBuf(m_gradBuf[REF_PIC_LIST_1][1], cuExtW, cuExtH);

  Pel* gX0 = gradXExt0.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
  Pel* gY0 = gradYExt0.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
  Pel* gX1 = gradXExt1.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
  Pel* gY1 = gradYExt1.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);

  int *dMvX0 = m_dMvBuf[REF_PIC_LIST_0];
  int *dMvY0 = m_dMvBuf[REF_PIC_LIST_0] + 16;
  int *dMvX1 = m_dMvBuf[REF_PIC_LIST_1];
  int *dMvY1 = m_dMvBuf[REF_PIC_LIST_1] + 16;

  const Pel* srcY0 = pcYuvSrc0.bufAt(0, 0);
  const Pel* srcY1 = pcYuvSrc1.bufAt(0, 0);
  Pel* dstY = pcYuvDst.bufAt(0, 0);

  if(m_applyPROF[0] && m_applyPROF[1])
    g_pelBufOP.applyBiPROF[1](dstY, pcYuvDst.stride, srcY0, srcY1, pcYuvSrc0.stride, width, height, gX0, gY0, gX1, gY1, gradXExt0.stride, dMvX0, dMvY0, dMvX1, dMvY1, blockWidth, getGbiWeight(pu.cu->GBiIdx, REF_PIC_LIST_0), clpRng);
  else if (m_applyPROF[0])
    g_pelBufOP.applyBiPROF[0](dstY, pcYuvDst.stride, srcY0, srcY1, pcYuvSrc0.stride, width, height, gX0, gY0, gX1, gY1, gradXExt0.stride, dMvX0, dMvY0, dMvX1, dMvY1, blockWidth, getGbiWeight(pu.cu->GBiIdx, REF_PIC_LIST_0), clpRng);
  else
    g_pelBufOP.applyBiPROF[0](dstY, pcYuvDst.stride, srcY1, srcY0, pcYuvSrc0.stride, width, height, gX1, gY1, gX0, gY0, gradXExt0.stride, dMvX1, dMvY1, dMvX0, dMvY0, blockWidth, getGbiWeight(pu.cu->GBiIdx, REF_PIC_LIST_1), clpRng);
}
#endif

void InterPrediction::motionCompensation( PredictionUnit &pu, PelUnitBuf &predBuf, const RefPicList &eRefPicList
  , const bool luma, const bool chroma
  , PelUnitBuf* predBufWOBIO /*= NULL*/
)
{
  CHECK(predBufWOBIO && pu.mhIntraFlag, "the case should not happen!");

  if (!pu.cs->pcv->isEncoder)
  {
    if (CU::isIBC(*pu.cu))
    {
      CHECK(!luma, "IBC only for Chroma is not allowed.");
      xIntraBlockCopy(pu, predBuf, COMPONENT_Y);
      if (chroma)
      {
        xIntraBlockCopy(pu, predBuf, COMPONENT_Cb);
        xIntraBlockCopy(pu, predBuf, COMPONENT_Cr);
      }
      return;
    }
  }
  // dual tree handling for IBC as the only ref
  if ((!luma || !chroma) && eRefPicList == REF_PIC_LIST_0)
  {
      xPredInterUni(pu, eRefPicList, predBuf, false
        , false
        , luma, chroma);
      return;
  }
  // else, go with regular MC below
        CodingStructure &cs = *pu.cs;
  const PPS &pps            = *cs.pps;
  const SliceType sliceType =  cs.slice->getSliceType();

  if( eRefPicList != REF_PIC_LIST_X )
  {
    CHECK(predBufWOBIO != NULL, "the case should not happen!");
    if( ( ( sliceType == P_SLICE && pps.getUseWP() ) || ( sliceType == B_SLICE && pps.getWPBiPred() ) ) )
    {
      xPredInterUni         ( pu,          eRefPicList, predBuf, true
        , false
        , true, true
      );
      xWeightedPredictionUni( pu, predBuf, eRefPicList, predBuf, -1, m_maxCompIDToPred );
    }
    else
    {
      xPredInterUni( pu, eRefPicList, predBuf, false
        , false
        , true, true
      );
    }
  }
  else
  {

    CHECK( !pu.cu->affine && pu.refIdx[0] >= 0 && pu.refIdx[1] >= 0 && ( pu.lwidth() + pu.lheight() == 12 ), "invalid 4x8/8x4 bi-predicted blocks" );
    WPScalingParam *wp0;
    WPScalingParam *wp1;
    int refIdx0 = pu.refIdx[REF_PIC_LIST_0];
    int refIdx1 = pu.refIdx[REF_PIC_LIST_1];
    pu.cs->slice->getWpScaling(REF_PIC_LIST_0, refIdx0, wp0);
    pu.cs->slice->getWpScaling(REF_PIC_LIST_1, refIdx1, wp1);
    bool bioApplied = false;
    const Slice &slice = *pu.cs->slice;
    if (pu.cs->sps->getBDOFEnabledFlag() && (!pu.cs->slice->getDisBdofDmvrFlag()))
    {

      if (pu.cu->affine || m_subPuMC)
      {
        bioApplied = false;
      }
      else
      {
        const bool biocheck0 = !((wp0[COMPONENT_Y].bPresentFlag || wp1[COMPONENT_Y].bPresentFlag) && slice.getSliceType() == B_SLICE);
        const bool biocheck1 = !(pps.getUseWP() && slice.getSliceType() == P_SLICE);
        if (biocheck0
          && biocheck1
#if JVET_P1023_DMVR_BDOF_RP_CONDITION
          && PU::isBiPredFromDifferentDirEqDistPoc(pu)
#else
          && PU::isBiPredFromDifferentDir(pu)
#endif
          && (pu.Y().height >= 8)
          && (pu.Y().width >= 8)
          && ((pu.Y().height * pu.Y().width) >= 128)
          )
        {
          bioApplied = true;
        }
      }

      if (bioApplied && pu.mhIntraFlag)
      {
        bioApplied = false;
      }

      if (bioApplied && pu.cu->smvdMode)
      {
        bioApplied = false;
      }
      if (pu.cu->cs->sps->getUseGBi() && bioApplied && pu.cu->GBiIdx != GBI_DEFAULT)
      {
        bioApplied = false;
      }
      if (pu.mmvdEncOptMode == 2 && pu.mmvdMergeFlag)
      {
        bioApplied = false;
      }
    }
    bioApplied = PU::isRefPicSameSize( pu ) ? bioApplied : false;
    bool dmvrApplied = false;
    dmvrApplied = (pu.mvRefine) && PU::checkDMVRCondition(pu);
    if ((pu.lumaSize().width > MAX_BDOF_APPLICATION_REGION || pu.lumaSize().height > MAX_BDOF_APPLICATION_REGION) && pu.mergeType != MRG_TYPE_SUBPU_ATMVP && (bioApplied && !dmvrApplied))
    {
      xSubPuBio(pu, predBuf, eRefPicList, predBufWOBIO);
    }
    else
    if (pu.mergeType != MRG_TYPE_DEFAULT_N && pu.mergeType != MRG_TYPE_IBC)
    {
      CHECK(predBufWOBIO != NULL, "the case should not happen!");
      xSubPuMC( pu, predBuf, eRefPicList );
    }
    else if( xCheckIdenticalMotion( pu ) )
    {
      xPredInterUni( pu, REF_PIC_LIST_0, predBuf, false
        , false
        , true, true
      );
      if (predBufWOBIO)
        predBufWOBIO->copyFrom(predBuf);
    }
    else
    {
      xPredInterBi(pu, predBuf, predBufWOBIO);
    }
  }
  return;
}

void InterPrediction::motionCompensation( CodingUnit &cu, const RefPicList &eRefPicList
  , const bool luma, const bool chroma
)
{
  for( auto &pu : CU::traversePUs( cu ) )
  {
    PelUnitBuf predBuf = cu.cs->getPredBuf( pu );
    pu.mvRefine = true;
    motionCompensation( pu, predBuf, eRefPicList
      , luma, chroma
    );
    pu.mvRefine = false;
  }
}

void InterPrediction::motionCompensation( PredictionUnit &pu, const RefPicList &eRefPicList /*= REF_PIC_LIST_X*/
  , const bool luma, const bool chroma
)
{
  PelUnitBuf predBuf = pu.cs->getPredBuf( pu );
  motionCompensation( pu, predBuf, eRefPicList
    , luma, chroma
  );
}

int InterPrediction::rightShiftMSB(int numer, int denom)
{
  return numer >> floorLog2(denom);
}

void InterPrediction::motionCompensation4Triangle( CodingUnit &cu, MergeCtx &triangleMrgCtx, const bool splitDir, const uint8_t candIdx0, const uint8_t candIdx1 )
{
  for( auto &pu : CU::traversePUs( cu ) )
  {
    const UnitArea localUnitArea( cu.cs->area.chromaFormat, Area( 0, 0, pu.lwidth(), pu.lheight() ) );
    PelUnitBuf tmpTriangleBuf = m_triangleBuf.getBuf( localUnitArea );
    PelUnitBuf predBuf        = cu.cs->getPredBuf( pu );

    triangleMrgCtx.setMergeInfo( pu, candIdx0 );
    PU::spanMotionInfo( pu );
    motionCompensation( pu, tmpTriangleBuf );

    {
      if( g_mctsDecCheckEnabled && !MCTSHelper::checkMvBufferForMCTSConstraint( pu, true ) )
      {
        printf( "DECODER_TRIANGLE_PU: pu motion vector across tile boundaries (%d,%d,%d,%d)\n", pu.lx(), pu.ly(), pu.lwidth(), pu.lheight() );
      }
    }

    triangleMrgCtx.setMergeInfo( pu, candIdx1 );
    PU::spanMotionInfo( pu );
    motionCompensation( pu, predBuf );

    {
      if( g_mctsDecCheckEnabled && !MCTSHelper::checkMvBufferForMCTSConstraint( pu, true ) )
      {
        printf( "DECODER_TRIANGLE_PU: pu motion vector across tile boundaries (%d,%d,%d,%d)\n", pu.lx(), pu.ly(), pu.lwidth(), pu.lheight() );
      }
    }
    weightedTriangleBlk( pu, splitDir, MAX_NUM_CHANNEL_TYPE, predBuf, tmpTriangleBuf, predBuf );
  }
}

void InterPrediction::weightedTriangleBlk( PredictionUnit &pu, const bool splitDir, int32_t channel, PelUnitBuf& predDst, PelUnitBuf& predSrc0, PelUnitBuf& predSrc1 )
{
  if( channel == CHANNEL_TYPE_LUMA )
  {
    m_if.weightedTriangleBlk( pu, pu.lumaSize().width, pu.lumaSize().height, COMPONENT_Y, splitDir, predDst, predSrc0, predSrc1 );
  }
  else if( channel == CHANNEL_TYPE_CHROMA )
  {
    m_if.weightedTriangleBlk( pu, pu.chromaSize().width, pu.chromaSize().height, COMPONENT_Cb, splitDir, predDst, predSrc0, predSrc1 );
    m_if.weightedTriangleBlk( pu, pu.chromaSize().width, pu.chromaSize().height, COMPONENT_Cr, splitDir, predDst, predSrc0, predSrc1 );
  }
  else
  {
    m_if.weightedTriangleBlk( pu, pu.lumaSize().width,   pu.lumaSize().height,   COMPONENT_Y,  splitDir, predDst, predSrc0, predSrc1 );
    m_if.weightedTriangleBlk( pu, pu.chromaSize().width, pu.chromaSize().height, COMPONENT_Cb, splitDir, predDst, predSrc0, predSrc1 );
    m_if.weightedTriangleBlk( pu, pu.chromaSize().width, pu.chromaSize().height, COMPONENT_Cr, splitDir, predDst, predSrc0, predSrc1 );
  }
}


void InterPrediction::xPrefetch(PredictionUnit& pu, PelUnitBuf &pcPad, RefPicList refId, bool forLuma)
{
  int offset, width, height;
  Mv cMv;
  const Picture* refPic = pu.cu->slice->getRefPic( refId, pu.refIdx[refId] )->unscaledPic;
  int mvShift = (MV_FRACTIONAL_BITS_INTERNAL);

  int start = 0;
  int end = MAX_NUM_COMPONENT;

  start = forLuma ? 0 : 1;
  end = forLuma ? 1 : MAX_NUM_COMPONENT;

  for (int compID = start; compID < end; compID++)
  {
    cMv = Mv(pu.mv[refId].getHor(), pu.mv[refId].getVer());
    pcPad.bufs[compID].stride = (pcPad.bufs[compID].width + (2 * DMVR_NUM_ITERATION) + NTAPS_LUMA);
    int filtersize = (compID == (COMPONENT_Y)) ? NTAPS_LUMA : NTAPS_CHROMA;
    width = pcPad.bufs[compID].width;
    height = pcPad.bufs[compID].height;
    offset = (DMVR_NUM_ITERATION) * (pcPad.bufs[compID].stride + 1);
    int mvshiftTempHor = mvShift + getComponentScaleX((ComponentID)compID, pu.chromaFormat);
    int mvshiftTempVer = mvShift + getComponentScaleY((ComponentID)compID, pu.chromaFormat);
    width += (filtersize - 1);
    height += (filtersize - 1);
    cMv += Mv(-(((filtersize >> 1) - 1) << mvshiftTempHor),
      -(((filtersize >> 1) - 1) << mvshiftTempVer));
    bool wrapRef = false;
    if( pu.cs->sps->getWrapAroundEnabledFlag() )
    {
      wrapRef = wrapClipMv( cMv, pu.blocks[0].pos(), pu.blocks[0].size(), pu.cs->sps, pu.cs->pps );
    }
    else
    {
      clipMv( cMv, pu.lumaPos(), pu.lumaSize(), *pu.cs->sps, *pu.cs->pps );
    }
    /* Pre-fetch similar to HEVC*/
    {
      CPelBuf refBuf;
      Position Rec_offset = pu.blocks[compID].pos().offset(cMv.getHor() >> mvshiftTempHor, cMv.getVer() >> mvshiftTempVer);
      refBuf = refPic->getRecoBuf(CompArea((ComponentID)compID, pu.chromaFormat, Rec_offset, pu.blocks[compID].size()), wrapRef);
      PelBuf &dstBuf = pcPad.bufs[compID];
      g_pelBufOP.copyBuffer((Pel *)refBuf.buf, refBuf.stride, ((Pel *)dstBuf.buf) + offset, dstBuf.stride, width, height);
    }
  }
}
void InterPrediction::xPad(PredictionUnit& pu, PelUnitBuf &pcPad, RefPicList refId)
{
  int offset = 0, width, height;
  int padsize;
  Mv cMv;
  for (int compID = 0; compID < MAX_NUM_COMPONENT; compID++)
  {
    int filtersize = (compID == (COMPONENT_Y)) ? NTAPS_LUMA : NTAPS_CHROMA;
    width = pcPad.bufs[compID].width;
    height = pcPad.bufs[compID].height;
    offset = (DMVR_NUM_ITERATION) * (pcPad.bufs[compID].stride + 1);
    /*using the larger padsize for 422*/
    padsize = (DMVR_NUM_ITERATION) >> getComponentScaleY((ComponentID)compID, pu.chromaFormat);
    width += (filtersize - 1);
    height += (filtersize - 1);
    /*padding on all side of size DMVR_PAD_LENGTH*/
    {
      g_pelBufOP.padding(pcPad.bufs[compID].buf + offset, pcPad.bufs[compID].stride, width, height, padsize);
    }
  }
}
inline int32_t div_for_maxq7(int64_t N, int64_t D)
{
  int32_t sign, q;
  sign = 0;
  if (N < 0)
  {
    sign = 1;
    N = -N;
  }

  q = 0;
  D = (D << 3);
  if (N >= D)
  {
    N -= D;
    q++;
  }
  q = (q << 1);

  D = (D >> 1);
  if (N >= D)
  {
    N -= D;
    q++;
  }
  q = (q << 1);

  if (N >= (D >> 1))
    q++;

  if (sign)
    return (-q);
  return(q);
}

void xSubPelErrorSrfc(uint64_t *sadBuffer, int32_t *deltaMv)
{
  int64_t numerator, denominator;
  int32_t mvDeltaSubPel;
  int32_t mvSubPelLvl = 4;/*1: half pel, 2: Qpel, 3:1/8, 4: 1/16*/
                                                        /*horizontal*/
    numerator = (int64_t)((sadBuffer[1] - sadBuffer[3]) << mvSubPelLvl);
    denominator = (int64_t)((sadBuffer[1] + sadBuffer[3] - (sadBuffer[0] << 1)));

    if (0 != denominator)
    {
      if ((sadBuffer[1] != sadBuffer[0]) && (sadBuffer[3] != sadBuffer[0]))
      {
        mvDeltaSubPel = div_for_maxq7(numerator, denominator);
        deltaMv[0] = (mvDeltaSubPel);
      }
      else
      {
        if (sadBuffer[1] == sadBuffer[0])
        {
          deltaMv[0] = -8;// half pel
        }
        else
        {
          deltaMv[0] = 8;// half pel
        }
      }
    }

    /*vertical*/
    numerator = (int64_t)((sadBuffer[2] - sadBuffer[4]) << mvSubPelLvl);
    denominator = (int64_t)((sadBuffer[2] + sadBuffer[4] - (sadBuffer[0] << 1)));
    if (0 != denominator)
    {
      if ((sadBuffer[2] != sadBuffer[0]) && (sadBuffer[4] != sadBuffer[0]))
      {
        mvDeltaSubPel = div_for_maxq7(numerator, denominator);
        deltaMv[1] = (mvDeltaSubPel);
      }
      else
      {
        if (sadBuffer[2] == sadBuffer[0])
        {
          deltaMv[1] = -8;// half pel
        }
        else
        {
          deltaMv[1] = 8;// half pel
        }
      }
    }
  return;
}

void InterPrediction::xBIPMVRefine(int bd, Pel *pRefL0, Pel *pRefL1, uint64_t& minCost, int16_t *deltaMV, uint64_t *pSADsArray, int width, int height)
{
  const int32_t refStrideL0 = m_biLinearBufStride;
  const int32_t refStrideL1 = m_biLinearBufStride;
  Pel *pRefL0Orig = pRefL0;
  Pel *pRefL1Orig = pRefL1;
  for (int nIdx = 0; (nIdx < 25); ++nIdx)
  {
    int32_t sadOffset = ((m_pSearchOffset[nIdx].getVer() * ((2 * DMVR_NUM_ITERATION) + 1)) + m_pSearchOffset[nIdx].getHor());
    pRefL0 = pRefL0Orig + m_pSearchOffset[nIdx].hor + (m_pSearchOffset[nIdx].ver * refStrideL0);
    pRefL1 = pRefL1Orig - m_pSearchOffset[nIdx].hor - (m_pSearchOffset[nIdx].ver * refStrideL1);
    if (*(pSADsArray + sadOffset) == MAX_UINT64)
    {
      const uint64_t cost = xDMVRCost(bd, pRefL0, refStrideL0, pRefL1, refStrideL1, width, height);
      *(pSADsArray + sadOffset) = cost;
    }
    if (*(pSADsArray + sadOffset) < minCost)
    {
      minCost = *(pSADsArray + sadOffset);
      deltaMV[0] = m_pSearchOffset[nIdx].getHor();
      deltaMV[1] = m_pSearchOffset[nIdx].getVer();
    }
  }
}

void InterPrediction::xFinalPaddedMCForDMVR(PredictionUnit& pu, PelUnitBuf &pcYuvSrc0, PelUnitBuf &pcYuvSrc1, PelUnitBuf &pcPad0, PelUnitBuf &pcPad1, const bool bioApplied
  , const Mv mergeMV[NUM_REF_PIC_LIST_01]
  , bool blockMoved
)
{
  int offset, deltaIntMvX, deltaIntMvY;

  PelUnitBuf pcYUVTemp = pcYuvSrc0;
  PelUnitBuf pcPadTemp = pcPad0;
  /*always high precision MVs are used*/
  int mvShift = MV_FRACTIONAL_BITS_INTERNAL;

  for (int k = 0; k < NUM_REF_PIC_LIST_01; k++)
  {
    RefPicList refId = (RefPicList)k;
    Mv cMv = pu.mv[refId];
    m_iRefListIdx = refId;
    const Picture* refPic = pu.cu->slice->getRefPic( refId, pu.refIdx[refId] )->unscaledPic;
    Mv cMvClipped = cMv;
    clipMv( cMvClipped, pu.lumaPos(), pu.lumaSize(), *pu.cs->sps, *pu.cs->pps );

    Mv startMv = mergeMV[refId];

    if( g_mctsDecCheckEnabled && !MCTSHelper::checkMvForMCTSConstraint( pu, startMv, MV_PRECISION_INTERNAL ) )
    {
      const Area& tileArea = pu.cs->picture->mctsInfo.getTileArea();
      printf( "Attempt an access over tile boundary at block %d,%d %d,%d with MV %d,%d (in Tile TL: %d,%d BR: %d,%d)\n",
        pu.lx(), pu.ly(), pu.lwidth(), pu.lheight(), startMv.getHor(), startMv.getVer(), tileArea.topLeft().x, tileArea.topLeft().y, tileArea.bottomRight().x, tileArea.bottomRight().y );
      THROW( "MCTS constraint failed!" );
    }
    for (int compID = 0; compID < MAX_NUM_COMPONENT; compID++)