InterPrediction.cpp

  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;
  const int   limit = (1<<(std::max<int>(5, bitDepth - 7)));

  int*     dotProductTemp1 = m_dotProduct1;
  int*     dotProductTemp2 = m_dotProduct2;
  int*     dotProductTemp3 = m_dotProduct3;
  int*     dotProductTemp5 = m_dotProduct5;
  int*     dotProductTemp6 = m_dotProduct6;

  xCalcBIOPar(srcY0Temp, srcY1Temp, gradX0, gradX1, gradY0, gradY1, dotProductTemp1, dotProductTemp2, dotProductTemp3, dotProductTemp5, dotProductTemp6, src0Stride, src1Stride, widthG, widthG, heightG, bitDepth);

  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++)
    {
#if !JVET_O0055_INT_DMVR_DIS_BDOF 
      if (m_bioPredSubBlkDist[yu*xUnit + xu] < m_bioSubBlkDistThres)
      {
        srcY0Temp = srcY0 + (stridePredMC + 1) + ((yu*src0Stride + xu) << 2);
        srcY1Temp = srcY1 + (stridePredMC + 1) + ((yu*src1Stride + xu) << 2);
        dstY0 = dstY + ((yu*dstStride + xu) << 2);
        PelBuf dstPelBuf(dstY0, dstStride, Size(4, 4));
        dstPelBuf.addAvg(CPelBuf(srcY0Temp, src0Stride, Size(4, 4)), CPelBuf(srcY1Temp, src1Stride, Size(4, 4)), clpRng);
        continue;
      }
#endif

      int     sGxdI = 0, sGydI = 0, sGxGy = 0, sGx2 = 0, sGy2 = 0;
      int     tmpx = 0, tmpy = 0;

      dotProductTemp1 = m_dotProduct1 + offsetPos + ((yu*widthG + xu) << 2);
      dotProductTemp2 = m_dotProduct2 + offsetPos + ((yu*widthG + xu) << 2);
      dotProductTemp3 = m_dotProduct3 + offsetPos + ((yu*widthG + xu) << 2);
      dotProductTemp5 = m_dotProduct5 + offsetPos + ((yu*widthG + xu) << 2);
      dotProductTemp6 = m_dotProduct6 + offsetPos + ((yu*widthG + xu) << 2);

      xCalcBlkGradient(xu << 2, yu << 2, dotProductTemp1, dotProductTemp2, dotProductTemp3, dotProductTemp5, dotProductTemp6, sGx2, sGy2, sGxGy, sGxdI, sGydI, widthG, heightG, (1 << 2));

      if (sGx2 > 0)
      {
        tmpx = rightShiftMSB(sGxdI << 3, sGx2);
        tmpx = Clip3(-limit, limit, tmpx);
      }
      if (sGy2 > 0)
      {
        int     mainsGxGy = sGxGy >> 12;
        int     secsGxGy = sGxGy & ((1 << 12) - 1);
        int     tmpData = tmpx * mainsGxGy;
        tmpData = ((tmpData << 12) + tmpx*secsGxGy) >> 1;
        tmpy = rightShiftMSB(((sGydI << 3) - tmpData), sGy2);
        tmpy = Clip3(-limit, limit, tmpy);
      }

      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
}


bool InterPrediction::xCalcBiPredSubBlkDist(const PredictionUnit &pu, const Pel* pYuvSrc0, const int src0Stride, const Pel* pYuvSrc1, const int src1Stride, const BitDepths &clipBitDepths)
{
  const int     width = pu.lwidth();
  const int     height = pu.lheight();
  const int     clipbd = clipBitDepths.recon[toChannelType(COMPONENT_Y)];
  const uint32_t distortionShift = DISTORTION_PRECISION_ADJUSTMENT(clipbd);
  const int     shift = std::max<int>(2, (IF_INTERNAL_PREC - clipbd));
  const int     xUnit = (width >> 2);
  const int     yUnit = (height >> 2);

  m_bioDistThres = (shift <= 5) ? (((32 << (clipbd - 8))*width*height) >> (5 - shift)) : (((32 << (clipbd - 8))*width*height) << (shift - 5));
  m_bioSubBlkDistThres = (shift <= 5) ? (((64 << (clipbd - 8)) << 4) >> (5 - shift)) : (((64 << (clipbd - 8)) << 4) << (shift - 5)); 

  m_bioDistThres >>= distortionShift;
  m_bioSubBlkDistThres >>= distortionShift;

  DistParam cDistParam;
  Distortion dist = 0;
  for (int yu = 0, blkIdx = 0; yu < yUnit; yu++)
  {
    for (int xu = 0; xu < xUnit; xu++, blkIdx++)
    {
      const Pel* pPred0 = pYuvSrc0 + ((yu*src0Stride + xu) << 2);
      const Pel* pPred1 = pYuvSrc1 + ((yu*src1Stride + xu) << 2);

      m_pcRdCost->setDistParam(cDistParam, pPred0, pPred1, src0Stride, src1Stride, clipbd, COMPONENT_Y, (1 << 2), (1 << 2), 0, 1, false, true);
      m_bioPredSubBlkDist[blkIdx] = cDistParam.distFunc(cDistParam);
      dist += m_bioPredSubBlkDist[blkIdx];
    }
  }

  return (dist >= m_bioDistThres);
}

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);
}

#if JVET_O0108_DIS_DMVR_BDOF_CIIP
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*/)
#else
void InterPrediction::xWeightedAverage(const PredictionUnit& pu, const CPelUnitBuf& pcYuvSrc0, const CPelUnitBuf& pcYuvSrc1, PelUnitBuf& pcYuvDst, const BitDepths& clipBitDepths, const ClpRngs& clpRngs, const bool& bioApplied )
#endif
{
  const int iRefIdx0 = pu.refIdx[0];
  const int iRefIdx1 = pu.refIdx[1];

  if( iRefIdx0 >= 0 && iRefIdx1 >= 0 )
  {
    if( pu.cu->GBiIdx != GBI_DEFAULT )
    {
      CHECK(bioApplied, "GBi is disallowed with BIO");
      pcYuvDst.addWeightedAvg(pcYuvSrc0, pcYuvSrc1, clpRngs, pu.cu->GBiIdx);
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
      if (yuvDstTmp)
        yuvDstTmp->copyFrom(pcYuvDst);
#endif
      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;

#if JVET_O0055_INT_DMVR_DIS_BDOF 
      bool bioEnabled = true;
#else
      bool bioEnabled = xCalcBiPredSubBlkDist(pu, pSrcY0, src0Stride, pSrcY1, src1Stride, clipBitDepths);
#endif
      if (bioEnabled)
      {
        applyBiOptFlow(pu, pcYuvSrc0, pcYuvSrc1, iRefIdx0, iRefIdx1, pcYuvDst, clipBitDepths);
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
        if (yuvDstTmp)
          yuvDstTmp->bufs[0].addAvg(CPelBuf(pSrcY0, src0Stride, pu.lumaSize()), CPelBuf(pSrcY1, src1Stride, pu.lumaSize()), clpRngs.comp[0]);
#endif
      }
      else
      {
        pcYuvDst.bufs[0].addAvg(CPelBuf(pSrcY0, src0Stride, pu.lumaSize()), CPelBuf(pSrcY1, src1Stride, pu.lumaSize()), clpRngs.comp[0]);
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
        if (yuvDstTmp)
          yuvDstTmp->bufs[0].copyFrom(pcYuvDst.bufs[0]);
#endif
      }
    }
    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 JVET_O0108_DIS_DMVR_BDOF_CIIP
    if (yuvDstTmp)
    {
      if (bioApplied)
      {
        yuvDstTmp->bufs[1].copyFrom(pcYuvDst.bufs[1]);
        yuvDstTmp->bufs[2].copyFrom(pcYuvDst.bufs[2]);
      }
      else
        yuvDstTmp->copyFrom(pcYuvDst);
    }
#endif
  }
  else if( iRefIdx0 >= 0 && iRefIdx1 < 0 )
  {
    if( pu.cu->triangle )
    {
      pcYuvDst.copyFrom( pcYuvSrc0 );
    }
    else
    pcYuvDst.copyClip( pcYuvSrc0, clpRngs );
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
    if (yuvDstTmp)
      yuvDstTmp->copyFrom(pcYuvDst);
#endif
  }
  else if( iRefIdx0 < 0 && iRefIdx1 >= 0 )
  {
    if( pu.cu->triangle )
    {
      pcYuvDst.copyFrom( pcYuvSrc1 );
    }
    else
    pcYuvDst.copyClip( pcYuvSrc1, clpRngs );
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
    if (yuvDstTmp)
      yuvDstTmp->copyFrom(pcYuvDst);
#endif
  }
}

void InterPrediction::motionCompensation( PredictionUnit &pu, PelUnitBuf &predBuf, const RefPicList &eRefPicList
  , const bool luma, const bool chroma
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
  , PelUnitBuf* predBufWOBIO /*= NULL*/
#endif
)
{
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
  CHECK(predBufWOBIO && pu.mhIntraFlag, "the case should not happen!");
#endif

  // dual tree handling for IBC as the only ref
  if ((!luma || !chroma) && eRefPicList == REF_PIC_LIST_0)
  {
    if (!luma && chroma)
    {
      xChromaMC(pu, predBuf);
      return;
    }
    else // (luma && !chroma)
    {
      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 )
  {
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
    CHECK(predBufWOBIO != NULL, "the case should not happen!");
#endif
    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())
    {

      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
          && PU::isBiPredFromDifferentDir(pu)
          && pu.Y().height != 4
          )
        {
          bioApplied = true;
        }
      }

#if JVET_O0108_DIS_DMVR_BDOF_CIIP
      if (bioApplied && pu.mhIntraFlag)
      {
        bioApplied = false;
      }
#endif

      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;
      }
    }
    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))
    {
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
      xSubPuBio(pu, predBuf, eRefPicList, predBufWOBIO);
#else
      xSubPuBio(pu, predBuf, eRefPicList);
#endif
    }
    else
    if (pu.mergeType != MRG_TYPE_DEFAULT_N && pu.mergeType != MRG_TYPE_IBC)
    {
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
      CHECK(predBufWOBIO != NULL, "the case should not happen!");
#endif 
      xSubPuMC( pu, predBuf, eRefPicList );
    }
    else if( xCheckIdenticalMotion( pu ) )
    {
      xPredInterUni( pu, REF_PIC_LIST_0, predBuf, false
        , false
        , true, true
      );
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
      if (predBufWOBIO)
        predBufWOBIO->copyFrom(predBuf);
#endif
    }
    else
    {
#if JVET_O0108_DIS_DMVR_BDOF_CIIP
      xPredInterBi(pu, predBuf, predBufWOBIO);
#else
      xPredInterBi( pu, predBuf );
#endif
    }
  }
  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)
{
  int     d;
  int msbIdx = 0;
  for (msbIdx = 0; msbIdx<32; msbIdx++)
  {
    if (denom < ((int)1 << msbIdx))
    {
      break;
    }
  }

  int shiftIdx = msbIdx - 1;
  d = (numer >> shiftIdx);

  return d;
}

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 )
  {
#if JVET_O0280_SIMD_TRIANGLE_WEIGHTING
    m_if.weightedTriangleBlk( pu, pu.lumaSize().width, pu.lumaSize().height, COMPONENT_Y, splitDir, predDst, predSrc0, predSrc1 );
#else
    xWeightedTriangleBlk( pu, pu.lumaSize().width, pu.lumaSize().height, COMPONENT_Y, splitDir, predDst, predSrc0, predSrc1 );
#endif
  }
  else if( channel == CHANNEL_TYPE_CHROMA )
  {
#if JVET_O0280_SIMD_TRIANGLE_WEIGHTING
    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
    xWeightedTriangleBlk( pu, pu.chromaSize().width, pu.chromaSize().height, COMPONENT_Cb, splitDir, predDst, predSrc0, predSrc1 );
    xWeightedTriangleBlk( pu, pu.chromaSize().width, pu.chromaSize().height, COMPONENT_Cr, splitDir, predDst, predSrc0, predSrc1 );
#endif
  }
  else
  {
#if JVET_O0280_SIMD_TRIANGLE_WEIGHTING
    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 );
#else
    xWeightedTriangleBlk( pu, pu.lumaSize().width,   pu.lumaSize().height,   COMPONENT_Y,  splitDir, predDst, predSrc0, predSrc1 );
    xWeightedTriangleBlk( pu, pu.chromaSize().width, pu.chromaSize().height, COMPONENT_Cb, splitDir, predDst, predSrc0, predSrc1 );
    xWeightedTriangleBlk( pu, pu.chromaSize().width, pu.chromaSize().height, COMPONENT_Cr, splitDir, predDst, predSrc0, predSrc1 );
#endif
  }
}

#if !JVET_O0280_SIMD_TRIANGLE_WEIGHTING
void InterPrediction::xWeightedTriangleBlk( const PredictionUnit &pu, const uint32_t width, const uint32_t height, const ComponentID compIdx, const bool splitDir, PelUnitBuf& predDst, PelUnitBuf& predSrc0, PelUnitBuf& predSrc1 )
{
  Pel*    dst        = predDst .get(compIdx).buf;
  Pel*    src0       = predSrc0.get(compIdx).buf;
  Pel*    src1       = predSrc1.get(compIdx).buf;
  int32_t strideDst  = predDst .get(compIdx).stride  - width;
  int32_t strideSrc0 = predSrc0.get(compIdx).stride  - width;
  int32_t strideSrc1 = predSrc1.get(compIdx).stride  - width;

  const char    log2WeightBase    = 3;
  const ClpRng  clipRng           = pu.cu->slice->clpRngs().comp[compIdx];
  const int32_t clipbd            = clipRng.bd;
  const int32_t shiftDefault      = std::max<int>(2, (IF_INTERNAL_PREC - clipbd));
  const int32_t offsetDefault     = (1<<(shiftDefault-1)) + IF_INTERNAL_OFFS;
  const int32_t shiftWeighted     = std::max<int>(2, (IF_INTERNAL_PREC - clipbd)) + log2WeightBase;
  const int32_t offsetWeighted    = (1 << (shiftWeighted - 1)) + (IF_INTERNAL_OFFS << log2WeightBase);

  const int32_t ratioWH           = (width > height) ? (width / height) : 1;
  const int32_t ratioHW           = (width > height) ? 1 : (height / width);

  const bool    longWeight        = (compIdx == COMPONENT_Y);
  const int32_t weightedLength    = longWeight ? 7 : 3;
        int32_t weightedStartPos  = ( splitDir == 0 ) ? ( 0 - (weightedLength >> 1) * ratioWH ) : ( width - ((weightedLength + 1) >> 1) * ratioWH );
        int32_t weightedEndPos    = weightedStartPos + weightedLength * ratioWH - 1;
        int32_t weightedPosoffset =( splitDir == 0 ) ? ratioWH : -ratioWH;

        Pel     tmpPelWeighted;
        int32_t weightIdx;
        int32_t x, y, tmpX, tmpY, tmpWeightedStart, tmpWeightedEnd;

  for( y = 0; y < height; y+= ratioHW )
  {
    for( tmpY = ratioHW; tmpY > 0; tmpY-- )
    {
      for( x = 0; x < weightedStartPos; x++ )
      {
        *dst++ = ClipPel( rightShift( (splitDir == 0 ? *src1 : *src0) + offsetDefault, shiftDefault), clipRng );
        src0++;
        src1++;
      }

      tmpWeightedStart = std::max((int32_t)0, weightedStartPos);
      tmpWeightedEnd   = std::min(weightedEndPos, (int32_t)(width - 1));
      weightIdx        = 1;
      if( weightedStartPos < 0 )
      {
        weightIdx     += abs(weightedStartPos) / ratioWH;
      }
      for( x = tmpWeightedStart; x <= tmpWeightedEnd; x+= ratioWH )
      {
        for( tmpX = ratioWH; tmpX > 0; tmpX-- )
        {
          tmpPelWeighted = Clip3( 1, 7, longWeight ? weightIdx : (weightIdx * 2));
          tmpPelWeighted = splitDir ? ( 8 - tmpPelWeighted ) : tmpPelWeighted;
          *dst++         = ClipPel( rightShift( (tmpPelWeighted*(*src0++) + ((8 - tmpPelWeighted) * (*src1++)) + offsetWeighted), shiftWeighted ), clipRng );
        }
        weightIdx ++;
      }

      for( x = weightedEndPos + 1; x < width; x++ )
      {
        *dst++ = ClipPel( rightShift( (splitDir == 0 ? *src0 : *src1) + offsetDefault, shiftDefault ), clipRng );
        src0++;
        src1++;
      }

      dst  += strideDst;
      src0 += strideSrc0;
      src1 += strideSrc1;
    }
    weightedStartPos += weightedPosoffset;
    weightedEndPos   += weightedPosoffset;
  }
}
#endif

void InterPrediction::xPrefetchPad(PredictionUnit& pu, PelUnitBuf &pcPad, RefPicList refId)
{
  int offset, width, height;
  int padsize;
  Mv cMv;
  const Picture* refPic = pu.cu->slice->getRefPic(refId, pu.refIdx[refId]);
  int mvShift = (MV_FRACTIONAL_BITS_INTERNAL);
  for (int compID = 0; compID < MAX_NUM_COMPONENT; compID++)
  {
    cMv = Mv(pu.mv[refId].getHor(), pu.mv[refId].getVer());
    pcPad.bufs[compID].stride = (MAX_CU_SIZE + (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);
    padsize = (DMVR_NUM_ITERATION) >> getComponentScaleX((ComponentID)compID, pu.chromaFormat);
    int mvshiftTemp = mvShift + getComponentScaleX((ComponentID)compID, pu.chromaFormat);
    width += (filtersize - 1);
    height += (filtersize - 1);
    cMv += Mv(-(((filtersize >> 1) - 1) << mvshiftTemp),
      -(((filtersize >> 1) - 1) << mvshiftTemp));
    bool wrapRef = false;
    if( pu.cs->sps->getWrapAroundEnabledFlag() ) 
    {
      wrapRef = wrapClipMv( cMv, pu.blocks[0].pos(), pu.blocks[0].size(), pu.cs->sps);   
    }
    else {
      clipMv(cMv, pu.lumaPos(), pu.lumaSize(),*pu.cs->sps);
    }
    /* Pre-fetch similar to HEVC*/
    {
      CPelBuf refBuf;
      Position Rec_offset = pu.blocks[compID].pos().offset(cMv.getHor() >> mvshiftTemp, cMv.getVer() >> mvshiftTemp);
      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);
#if JVET_J0090_MEMORY_BANDWITH_MEASURE
      JVET_J0090_SET_REF_PICTURE( refPic, (ComponentID)compID );
      for ( int row = 0 ; row < height ; row++ )
      {
        for ( int col = 0 ; col < width ; col++ )
        {
          JVET_J0090_CACHE_ACCESS( ((Pel *)refBuf.buf) + row * refBuf.stride + col, __FILE__, __LINE__ );
        }
      }
#endif
    }
    /*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]
)
{
  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]);
    Mv cMvClipped = cMv;
    clipMv(cMvClipped, pu.lumaPos(), pu.lumaSize(), *pu.cs->sps);

    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++)
    {
      int mvshiftTemp = mvShift + getComponentScaleX((ComponentID)compID, pu.chromaFormat);
      int leftPixelExtra;
      if (compID == COMPONENT_Y)
      {
        leftPixelExtra = (NTAPS_LUMA >> 1) - 1;
      }
      else
      {
        leftPixelExtra = (NTAPS_CHROMA >> 1) - 1;
      }

      deltaIntMvX = (cMv.getHor() >> mvshiftTemp) -
        (startMv.getHor() >> mvshiftTemp);
      deltaIntMvY = (cMv.getVer() >> mvshiftTemp) -
        (startMv.getVer() >> mvshiftTemp);

      CHECK((abs(deltaIntMvX) > DMVR_NUM_ITERATION) || (abs(deltaIntMvY) > DMVR_NUM_ITERATION), "not expected DMVR movement");

      offset = (DMVR_NUM_ITERATION + leftPixelExtra) * (pcPadTemp.bufs[compID].stride + 1);
      offset += (deltaIntMvY)* pcPadTemp.bufs[compID].stride;
      offset += (deltaIntMvX);
      PelBuf &srcBuf = pcPadTemp.bufs[compID];
      xPredInterBlk((ComponentID)compID, pu, refPic, cMvClipped, pcYUVTemp, true, pu.cs->slice->getClpRngs().comp[compID],
        bioApplied, false, 0, 0, 0, (srcBuf.buf + offset), pcPadTemp.bufs[compID].stride);
    }
    pcYUVTemp = pcYuvSrc1;
    pcPadTemp = pcPad1;
  }
}

uint64_t InterPrediction::xDMVRCost(int bitDepth, Pel* pOrg, uint32_t refStride, const Pel* pRef, uint32_t orgStride, int width, int height)
{
  DistParam cDistParam;
  cDistParam.applyWeight = false;
  cDistParam.useMR = false;
  m_pcRdCost->setDistParam(cDistParam, pOrg, pRef, orgStride, refStride, bitDepth, COMPONENT_Y, width, height, 1);
  uint64_t uiCost = cDistParam.distFunc(cDistParam);
  return uiCost;
}

void xDMVRSubPixelErrorSurface(bool notZeroCost, int16_t *totalDeltaMV, int16_t *deltaMV, uint64_t *pSADsArray)
{

  int sadStride = (((2 * DMVR_NUM_ITERATION) + 1));
  uint64_t sadbuffer[5];
  if (notZeroCost && (abs(totalDeltaMV[0]) != (2 << MV_FRACTIONAL_BITS_INTERNAL))
    && (abs(totalDeltaMV[1]) != (2 << MV_FRACTIONAL_BITS_INTERNAL)))
  {
    int32_t tempDeltaMv[2] = { 0,0 };
    sadbuffer[0] = pSADsArray[0];
    sadbuffer[1] = pSADsArray[-1];
    sadbuffer[2] = pSADsArray[-sadStride];
    sadbuffer[3] = pSADsArray[1];
    sadbuffer[4] = pSADsArray[sadStride];
    xSubPelErrorSrfc(sadbuffer, tempDeltaMv);
    totalDeltaMV[0] += tempDeltaMv[0];
    totalDeltaMV[1] += tempDeltaMv[1];
  }
}

void InterPrediction::xinitMC(PredictionUnit& pu, const ClpRngs &clpRngs)
{
  const int refIdx0 = pu.refIdx[0];
  const int refIdx1 = pu.refIdx[1];
  /*use merge MV as starting MV*/
  Mv mergeMVL0(pu.mv[REF_PIC_LIST_0]);
  Mv mergeMVL1(pu.mv[REF_PIC_LIST_1]);

  /*Clip the starting MVs*/
  clipMv(mergeMVL0, pu.lumaPos(), pu.lumaSize(), *pu.cs->sps);
  clipMv(mergeMVL1, pu.lumaPos(), pu.lumaSize(), *pu.cs->sps);

  /*L0 MC for refinement*/
  {
    int offset;
    int leftPixelExtra = (NTAPS_LUMA >> 1) - 1;
    offset = (DMVR_NUM_ITERATION + leftPixelExtra) * (m_cYuvRefBuffDMVRL0.bufs[COMPONENT_Y].stride + 1);
    offset += (-(int)DMVR_NUM_ITERATION)* (int)m_cYuvRefBuffDMVRL0.bufs[COMPONENT_Y].stride;
    offset += (-(int)DMVR_NUM_ITERATION);
    PelBuf srcBuf = m_cYuvRefBuffDMVRL0.bufs[COMPONENT_Y];
    PelUnitBuf yuvPredTempL0 = PelUnitBuf(pu.chromaFormat, PelBuf(m_cYuvPredTempDMVRL0,
      (MAX_CU_SIZE + (2 * DMVR_NUM_ITERATION)), pu.lwidth() + (2 * DMVR_NUM_ITERATION), pu.lheight() + (2 * DMVR_NUM_ITERATION)));

    xPredInterBlk(COMPONENT_Y, pu, pu.cu->slice->getRefPic(REF_PIC_LIST_0, refIdx0), mergeMVL0, yuvPredTempL0, true, clpRngs.comp[COMPONENT_Y],
      false, false, pu.lwidth() + (2 * DMVR_NUM_ITERATION), pu.lheight() + (2 * DMVR_NUM_ITERATION), true, ((Pel *)srcBuf.buf) + offset, srcBuf.stride
    );
  }

  /*L1 MC for refinement*/
  {
    int offset;
    int leftPixelExtra = (NTAPS_LUMA >> 1) - 1;
    offset = (DMVR_NUM_ITERATION + leftPixelExtra) * (m_cYuvRefBuffDMVRL1.bufs[COMPONENT_Y].stride + 1);
    offset += (-(int)DMVR_NUM_ITERATION)* (int)m_cYuvRefBuffDMVRL1.bufs[COMPONENT_Y].stride;
    offset += (-(int)DMVR_NUM_ITERATION);
    PelBuf srcBuf = m_cYuvRefBuffDMVRL1.bufs[COMPONENT_Y];
    PelUnitBuf yuvPredTempL1 = PelUnitBuf(pu.chromaFormat, PelBuf(m_cYuvPredTempDMVRL1,
      (MAX_CU_SIZE + (2 * DMVR_NUM_ITERATION)), pu.lwidth() + (2 * DMVR_NUM_ITERATION), pu.lheight() + (2 * DMVR_NUM_ITERATION)));

    xPredInterBlk(COMPONENT_Y, pu, pu.cu->slice->getRefPic(REF_PIC_LIST_1, refIdx1), mergeMVL1, yuvPredTempL1, true, clpRngs.comp[COMPONENT_Y],
      false, false, pu.lwidth() + (2 * DMVR_NUM_ITERATION), pu.lheight() + (2 * DMVR_NUM_ITERATION), true, ((Pel *)srcBuf.buf) + offset, srcBuf.stride
    );
  }
}

void InterPrediction::xProcessDMVR(PredictionUnit& pu, PelUnitBuf &pcYuvDst, const ClpRngs &clpRngs, const bool bioApplied)
{
  int iterationCount = 1;
  /*Always High Precision*/
  int mvShift = MV_FRACTIONAL_BITS_INTERNAL;

  /*use merge MV as starting MV*/
  Mv mergeMv[] = { pu.mv[REF_PIC_LIST_0] , pu.mv[REF_PIC_LIST_1] };

  m_biLinearBufStride = (MAX_CU_SIZE + (2 * DMVR_NUM_ITERATION));

  int dy = std::min<int>(pu.lumaSize().height, DMVR_SUBCU_HEIGHT);
  int dx = std::min<int>(pu.lumaSize().width,  DMVR_SUBCU_WIDTH);
  /*L0 Padding*/
  m_cYuvRefBuffDMVRL0 = (pu.chromaFormat == CHROMA_400 ?
    PelUnitBuf(pu.chromaFormat, PelBuf(m_cRefSamplesDMVRL0[0], pcYuvDst.Y())) :
    PelUnitBuf(pu.chromaFormat, PelBuf(m_cRefSamplesDMVRL0[0], pcYuvDst.Y()),
      PelBuf(m_cRefSamplesDMVRL0[1], pcYuvDst.Cb()), PelBuf(m_cRefSamplesDMVRL0[2], pcYuvDst.Cr())));

  xPrefetchPad(pu, m_cYuvRefBuffDMVRL0, REF_PIC_LIST_0);

  /*L1 Padding*/
  m_cYuvRefBuffDMVRL1 = (pu.chromaFormat == CHROMA_400 ?
    PelUnitBuf(pu.chromaFormat, PelBuf(m_cRefSamplesDMVRL1[0], pcYuvDst.Y())) :
    PelUnitBuf(pu.chromaFormat, PelBuf(m_cRefSamplesDMVRL1[0], pcYuvDst.Y()), PelBuf(m_cRefSamplesDMVRL1[1], pcYuvDst.Cb()),
      PelBuf(m_cRefSamplesDMVRL1[2], pcYuvDst.Cr())));

  xPrefetchPad(pu, m_cYuvRefBuffDMVRL1, REF_PIC_LIST_1);

  JVET_J0090_SET_CACHE_ENABLE( false );
  xinitMC(pu, clpRngs);

  // point mc buffer to cetre point to avoid multiplication to reach each iteration to the begining
  Pel *biLinearPredL0 = m_cYuvPredTempDMVRL0 + (DMVR_NUM_ITERATION * m_biLinearBufStride) + DMVR_NUM_ITERATION;
  Pel *biLinearPredL1 = m_cYuvPredTempDMVRL1 + (DMVR_NUM_ITERATION * m_biLinearBufStride) + DMVR_NUM_ITERATION;

  Position puPos = pu.lumaPos();

  int bd = pu.cs->slice->getClpRngs().comp[COMPONENT_Y].bd;

#if JVET_O0055_INT_DMVR_DIS_BDOF
  int            bioEnabledThres = 8 * (dy >> 1) * dx;
  bool           bioAppliedType[MAX_NUM_SUBCU_DMVR];
#endif
  {
    int num = 0;

    int yStart = 0;
    for (int y = puPos.y; y < (puPos.y + pu.lumaSize().height); y = y + dy, yStart = yStart + dy)
    {
      for (int x = puPos.x, xStart = 0; x < (puPos.x + pu.lumaSize().width); x = x + dx, xStart = xStart + dx)
      {
        uint64_t minCost = MAX_UINT64;
        bool notZeroCost = true;
        int16_t totalDeltaMV[2] = { 0,0 };
        int16_t deltaMV[2] = { 0, 0 };
        uint64_t  *pSADsArray;
        for (int i = 0; i < (((2 * DMVR_NUM_ITERATION) + 1) * ((2 * DMVR_NUM_ITERATION) + 1)); i++)
        {
          m_SADsArray[i] = MAX_UINT64;
        }
        pSADsArray = &m_SADsArray[(((2 * DMVR_NUM_ITERATION) + 1) * ((2 * DMVR_NUM_ITERATION) + 1)) >> 1];

        Pel *addrL0Centre = biLinearPredL0 + yStart * m_biLinearBufStride + xStart;
        Pel *addrL1Centre = biLinearPredL1 + yStart * m_biLinearBufStride + xStart;
        for (int i = 0; i < iterationCount; i++)
        {
          deltaMV[0] = 0;
          deltaMV[1] = 0;
          Pel *addrL0 = addrL0Centre + totalDeltaMV[0] + (totalDeltaMV[1] * m_biLinearBufStride);
          Pel *addrL1 = addrL1Centre - totalDeltaMV[0] - (totalDeltaMV[1] * m_biLinearBufStride);
          if (i == 0)
          {
            minCost = xDMVRCost(clpRngs.comp[COMPONENT_Y].bd, addrL0, m_biLinearBufStride, addrL1, m_biLinearBufStride, dx, dy);
#if JVET_O0590_REDUCE_DMVR_ORIG_MV_COST
            minCost -= (minCost >>2);            
#endif
            if (minCost < ((4 * dx * (dy >> 1/*for alternate line*/))))
            {
              notZeroCost = false;
              break;
            }
            pSADsArray[0] = minCost;
          }
          if (!minCost)