UnitTools.cpp

    && puLeft->mergeType == MRG_TYPE_DEFAULT_N
    )
  {
    npu[num++] = puLeft;
    return num;
  }

  return num;
}

const int getAvailableAffineNeighboursForAbovePredictor( const PredictionUnit &pu, const PredictionUnit* npu[], int numAffNeighLeft )
{
  const Position posLT = pu.Y().topLeft();
  const Position posRT = pu.Y().topRight();
  int num = numAffNeighLeft;

  const PredictionUnit* puAboveRight = pu.cs->getPURestricted( posRT.offset( 1, -1 ), pu, pu.chType );
  if ( puAboveRight && puAboveRight->cu->affine
    && puAboveRight->mergeType == MRG_TYPE_DEFAULT_N
    )
  {
    npu[num++] = puAboveRight;
    return num;
  }

  const PredictionUnit* puAbove = pu.cs->getPURestricted( posRT.offset( 0, -1 ), pu, pu.chType );
  if ( puAbove && puAbove->cu->affine
    && puAbove->mergeType == MRG_TYPE_DEFAULT_N
    )
  {
    npu[num++] = puAbove;
    return num;
  }

  const PredictionUnit *puAboveLeft = pu.cs->getPURestricted( posLT.offset( -1, -1 ), pu, pu.chType );
  if ( puAboveLeft && puAboveLeft->cu->affine
    && puAboveLeft->mergeType == MRG_TYPE_DEFAULT_N
    )
  {
    npu[num++] = puAboveLeft;
    return num;
  }

  return num;
}

void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx, const int mrgCandIdx )
{
  const CodingStructure &cs = *pu.cs;
  const Slice &slice = *pu.cs->slice;
  const uint32_t maxNumAffineMergeCand = slice.getMaxNumAffineMergeCand();

  for ( int i = 0; i < maxNumAffineMergeCand; i++ )
  {
    for ( int mvNum = 0; mvNum < 3; mvNum++ )
    {
      affMrgCtx.mvFieldNeighbours[(i << 1) + 0][mvNum].setMvField( Mv(), -1 );
      affMrgCtx.mvFieldNeighbours[(i << 1) + 1][mvNum].setMvField( Mv(), -1 );
    }
    affMrgCtx.interDirNeighbours[i] = 0;
    affMrgCtx.affineType[i] = AFFINEMODEL_4PARAM;
    affMrgCtx.mergeType[i] = MRG_TYPE_DEFAULT_N;
    affMrgCtx.GBiIdx[i] = GBI_DEFAULT;
  }

  affMrgCtx.numValidMergeCand = 0;
  affMrgCtx.maxNumMergeCand = maxNumAffineMergeCand;

  bool enableSubPuMvp = slice.getSPS()->getSBTMVPEnabledFlag() && !(slice.getPOC() == slice.getRefPic(REF_PIC_LIST_0, 0)->getPOC() && slice.isIRAP());
  bool isAvailableSubPu = false;
  if ( enableSubPuMvp && slice.getEnableTMVPFlag() )
  {
    MergeCtx mrgCtx = *affMrgCtx.mrgCtx;
    bool tmpLICFlag = false;

    CHECK( mrgCtx.subPuMvpMiBuf.area() == 0 || !mrgCtx.subPuMvpMiBuf.buf, "Buffer not initialized" );
    mrgCtx.subPuMvpMiBuf.fill( MotionInfo() );

    int pos = 0;
    // Get spatial MV
    const Position posCurLB = pu.Y().bottomLeft();
    MotionInfo miLeft;

    //left
    const PredictionUnit* puLeft = cs.getPURestricted( posCurLB.offset( -1, 0 ), pu, pu.chType );
    const bool isAvailableA1 = puLeft && isDiffMER( pu, *puLeft ) && pu.cu != puLeft->cu && CU::isInter( *puLeft->cu );
    if ( isAvailableA1 )
    {
      miLeft = puLeft->getMotionInfo( posCurLB.offset( -1, 0 ) );
      // get Inter Dir
      mrgCtx.interDirNeighbours[pos] = miLeft.interDir;

      // get Mv from Left
      mrgCtx.mvFieldNeighbours[pos << 1].setMvField( miLeft.mv[0], miLeft.refIdx[0] );

      if ( slice.isInterB() )
      {
        mrgCtx.mvFieldNeighbours[(pos << 1) + 1].setMvField( miLeft.mv[1], miLeft.refIdx[1] );
      }
      pos++;
    }

    mrgCtx.numValidMergeCand = pos;

    isAvailableSubPu = getInterMergeSubPuMvpCand( pu, mrgCtx, tmpLICFlag, pos
      , 0
    );
    if ( isAvailableSubPu )
    {
      for ( int mvNum = 0; mvNum < 3; mvNum++ )
      {
        affMrgCtx.mvFieldNeighbours[(affMrgCtx.numValidMergeCand << 1) + 0][mvNum].setMvField( mrgCtx.mvFieldNeighbours[(pos << 1) + 0].mv, mrgCtx.mvFieldNeighbours[(pos << 1) + 0].refIdx );
        affMrgCtx.mvFieldNeighbours[(affMrgCtx.numValidMergeCand << 1) + 1][mvNum].setMvField( mrgCtx.mvFieldNeighbours[(pos << 1) + 1].mv, mrgCtx.mvFieldNeighbours[(pos << 1) + 1].refIdx );
      }
      affMrgCtx.interDirNeighbours[affMrgCtx.numValidMergeCand] = mrgCtx.interDirNeighbours[pos];

      affMrgCtx.affineType[affMrgCtx.numValidMergeCand] = AFFINE_MODEL_NUM;
      affMrgCtx.mergeType[affMrgCtx.numValidMergeCand] = MRG_TYPE_SUBPU_ATMVP;
      if ( affMrgCtx.numValidMergeCand == mrgCandIdx )
      {
        return;
      }

      affMrgCtx.numValidMergeCand++;

      // early termination
      if ( affMrgCtx.numValidMergeCand == maxNumAffineMergeCand )
      {
        return;
      }
    }
  }

  if ( slice.getSPS()->getUseAffine() )
  {
    ///> Start: inherited affine candidates
    const PredictionUnit* npu[5];
    int numAffNeighLeft = getAvailableAffineNeighboursForLeftPredictor( pu, npu );
    int numAffNeigh = getAvailableAffineNeighboursForAbovePredictor( pu, npu, numAffNeighLeft );
    for ( int idx = 0; idx < numAffNeigh; idx++ )
    {
      // derive Mv from Neigh affine PU
      Mv cMv[2][3];
      const PredictionUnit* puNeigh = npu[idx];
      pu.cu->affineType = puNeigh->cu->affineType;
      if ( puNeigh->interDir != 2 )
      {
        xInheritedAffineMv( pu, puNeigh, REF_PIC_LIST_0, cMv[0] );
      }
      if ( slice.isInterB() )
      {
        if ( puNeigh->interDir != 1 )
        {
          xInheritedAffineMv( pu, puNeigh, REF_PIC_LIST_1, cMv[1] );
        }
      }

      for ( int mvNum = 0; mvNum < 3; mvNum++ )
      {
        affMrgCtx.mvFieldNeighbours[(affMrgCtx.numValidMergeCand << 1) + 0][mvNum].setMvField( cMv[0][mvNum], puNeigh->refIdx[0] );
        affMrgCtx.mvFieldNeighbours[(affMrgCtx.numValidMergeCand << 1) + 1][mvNum].setMvField( cMv[1][mvNum], puNeigh->refIdx[1] );
      }
      affMrgCtx.interDirNeighbours[affMrgCtx.numValidMergeCand] = puNeigh->interDir;
      affMrgCtx.affineType[affMrgCtx.numValidMergeCand] = (EAffineModel)(puNeigh->cu->affineType);
      affMrgCtx.GBiIdx[affMrgCtx.numValidMergeCand] = puNeigh->cu->GBiIdx;

      if ( affMrgCtx.numValidMergeCand == mrgCandIdx )
      {
        return;
      }

      // early termination
      affMrgCtx.numValidMergeCand++;
      if ( affMrgCtx.numValidMergeCand == maxNumAffineMergeCand )
      {
        return;
      }
    }
    ///> End: inherited affine candidates

    ///> Start: Constructed affine candidates
    {
      MotionInfo mi[4];
      bool isAvailable[4] = { false };

#if JVET_O0366_AFFINE_BCW
      int8_t neighGbi[2] = { GBI_DEFAULT, GBI_DEFAULT };
#else
      int8_t neighGbi[4] = { GBI_DEFAULT, GBI_DEFAULT, GBI_DEFAULT, GBI_DEFAULT };
#endif
      // control point: LT B2->B3->A2
      const Position posLT[3] = { pu.Y().topLeft().offset( -1, -1 ), pu.Y().topLeft().offset( 0, -1 ), pu.Y().topLeft().offset( -1, 0 ) };
      for ( int i = 0; i < 3; i++ )
      {
        const Position pos = posLT[i];
        const PredictionUnit* puNeigh = cs.getPURestricted( pos, pu, pu.chType );

        if ( puNeigh && CU::isInter( *puNeigh->cu )
          )
        {
          isAvailable[0] = true;
          mi[0] = puNeigh->getMotionInfo( pos );
          neighGbi[0] = puNeigh->cu->GBiIdx;
          break;
        }
      }

      // control point: RT B1->B0
      const Position posRT[2] = { pu.Y().topRight().offset( 0, -1 ), pu.Y().topRight().offset( 1, -1 ) };
      for ( int i = 0; i < 2; i++ )
      {
        const Position pos = posRT[i];
        const PredictionUnit* puNeigh = cs.getPURestricted( pos, pu, pu.chType );


        if ( puNeigh && CU::isInter( *puNeigh->cu )
          )
        {
          isAvailable[1] = true;
          mi[1] = puNeigh->getMotionInfo( pos );
          neighGbi[1] = puNeigh->cu->GBiIdx;
          break;
        }
      }

      // control point: LB A1->A0
      const Position posLB[2] = { pu.Y().bottomLeft().offset( -1, 0 ), pu.Y().bottomLeft().offset( -1, 1 ) };
      for ( int i = 0; i < 2; i++ )
      {
        const Position pos = posLB[i];
        const PredictionUnit* puNeigh = cs.getPURestricted( pos, pu, pu.chType );


        if ( puNeigh && CU::isInter( *puNeigh->cu )
          )
        {
          isAvailable[2] = true;
          mi[2] = puNeigh->getMotionInfo( pos );
#if !JVET_O0366_AFFINE_BCW
          neighGbi[2] = puNeigh->cu->GBiIdx;
#endif
          break;
        }
      }

      // control point: RB
      if ( slice.getEnableTMVPFlag() )
      {
        //>> MTK colocated-RightBottom
        // offset the pos to be sure to "point" to the same position the uiAbsPartIdx would've pointed to
        Position posRB = pu.Y().bottomRight().offset( -3, -3 );

        const PreCalcValues& pcv = *cs.pcv;
        Position posC0;
        bool C0Avail = false;

        if ( ((posRB.x + pcv.minCUWidth) < pcv.lumaWidth) && ((posRB.y + pcv.minCUHeight) < pcv.lumaHeight) )
        {
          int posYInCtu = posRB.y & pcv.maxCUHeightMask;
          if (posYInCtu + 4 < pcv.maxCUHeight)
          {
            posC0 = posRB.offset(4, 4);
            C0Avail = true;
          }
        }

        Mv        cColMv;
        int       refIdx = 0;
        bool      bExistMV = C0Avail && getColocatedMVP( pu, REF_PIC_LIST_0, posC0, cColMv, refIdx );
        if ( bExistMV )
        {
          mi[3].mv[0] = cColMv;
          mi[3].refIdx[0] = refIdx;
          mi[3].interDir = 1;
          isAvailable[3] = true;
        }

        if ( slice.isInterB() )
        {
          bExistMV = C0Avail && getColocatedMVP( pu, REF_PIC_LIST_1, posC0, cColMv, refIdx );
          if ( bExistMV )
          {
            mi[3].mv[1] = cColMv;
            mi[3].refIdx[1] = refIdx;
            mi[3].interDir |= 2;
            isAvailable[3] = true;
          }
        }
      }

      //-------------------  insert model  -------------------//
      int order[6] = { 0, 1, 2, 3, 4, 5 };
      int modelNum = 6;
      int model[6][4] = {
        { 0, 1, 2 },          // 0:  LT, RT, LB
        { 0, 1, 3 },          // 1:  LT, RT, RB
        { 0, 2, 3 },          // 2:  LT, LB, RB
        { 1, 2, 3 },          // 3:  RT, LB, RB
        { 0, 1 },             // 4:  LT, RT
        { 0, 2 },             // 5:  LT, LB
      };

      int verNum[6] = { 3, 3, 3, 3, 2, 2 };
      int startIdx = pu.cs->sps->getUseAffineType() ? 0 : 4;
      for ( int idx = startIdx; idx < modelNum; idx++ )
      {
        int modelIdx = order[idx];
#if JVET_O0366_AFFINE_BCW
        getAffineControlPointCand(pu, mi, isAvailable, model[modelIdx], ((modelIdx == 3) ? neighGbi[1] : neighGbi[0]), modelIdx, verNum[modelIdx], affMrgCtx);
#else
        getAffineControlPointCand(pu, mi, neighGbi, isAvailable, model[modelIdx], modelIdx, verNum[modelIdx], affMrgCtx);
#endif
        if ( affMrgCtx.numValidMergeCand != 0 && affMrgCtx.numValidMergeCand - 1 == mrgCandIdx )
        {
          return;
        }

        // early termination
        if ( affMrgCtx.numValidMergeCand == maxNumAffineMergeCand )
        {
          return;
        }
      }
    }
    ///> End: Constructed affine candidates
  }

  ///> zero padding
  int cnt = affMrgCtx.numValidMergeCand;
  while ( cnt < maxNumAffineMergeCand )
  {
    for ( int mvNum = 0; mvNum < 3; mvNum++ )
    {
      affMrgCtx.mvFieldNeighbours[(cnt << 1) + 0][mvNum].setMvField( Mv( 0, 0 ), 0 );
    }
    affMrgCtx.interDirNeighbours[cnt] = 1;

    if ( slice.isInterB() )
    {
      for ( int mvNum = 0; mvNum < 3; mvNum++ )
      {
        affMrgCtx.mvFieldNeighbours[(cnt << 1) + 1][mvNum].setMvField( Mv( 0, 0 ), 0 );
      }
      affMrgCtx.interDirNeighbours[cnt] = 3;
    }
    affMrgCtx.affineType[cnt] = AFFINEMODEL_4PARAM;

    if ( cnt == mrgCandIdx )
    {
      return;
    }
    cnt++;
    affMrgCtx.numValidMergeCand++;
  }
}

void PU::setAllAffineMvField( PredictionUnit &pu, MvField *mvField, RefPicList eRefList )
{
  // Set Mv
  Mv mv[3];
  for ( int i = 0; i < 3; i++ )
  {
    mv[i] = mvField[i].mv;
  }
  setAllAffineMv( pu, mv[0], mv[1], mv[2], eRefList );

  // Set RefIdx
  CHECK( mvField[0].refIdx != mvField[1].refIdx || mvField[0].refIdx != mvField[2].refIdx, "Affine mv corners don't have the same refIdx." );
  pu.refIdx[eRefList] = mvField[0].refIdx;
}

void PU::setAllAffineMv(PredictionUnit& pu, Mv affLT, Mv affRT, Mv affLB, RefPicList eRefList, bool clipCPMVs)
{
  int width  = pu.Y().width;
  int shift = MAX_CU_DEPTH;
  if (clipCPMVs)
  {
    affLT.mvCliptoStorageBitDepth();
    affRT.mvCliptoStorageBitDepth();
    if (pu.cu->affineType == AFFINEMODEL_6PARAM)
    {
      affLB.mvCliptoStorageBitDepth();
    }
  }
  int deltaMvHorX, deltaMvHorY, deltaMvVerX, deltaMvVerY;
  deltaMvHorX = (affRT - affLT).getHor() << (shift - g_aucLog2[width]);
  deltaMvHorY = (affRT - affLT).getVer() << (shift - g_aucLog2[width]);
  int height = pu.Y().height;
  if ( pu.cu->affineType == AFFINEMODEL_6PARAM )
  {
    deltaMvVerX = (affLB - affLT).getHor() << (shift - g_aucLog2[height]);
    deltaMvVerY = (affLB - affLT).getVer() << (shift - g_aucLog2[height]);
  }
  else
  {
    deltaMvVerX = -deltaMvHorY;
    deltaMvVerY = deltaMvHorX;
  }

  int mvScaleHor = affLT.getHor() << shift;
  int mvScaleVer = affLT.getVer() << shift;

  int blockWidth = AFFINE_MIN_BLOCK_SIZE;
  int blockHeight = AFFINE_MIN_BLOCK_SIZE;
  const int halfBW = blockWidth >> 1;
  const int halfBH = blockHeight >> 1;

  MotionBuf mb = pu.getMotionBuf();
  int mvScaleTmpHor, mvScaleTmpVer;
  const bool subblkMVSpreadOverLimit = InterPrediction::isSubblockVectorSpreadOverLimit( deltaMvHorX, deltaMvHorY, deltaMvVerX, deltaMvVerY, pu.interDir );
  for ( int h = 0; h < pu.Y().height; h += blockHeight )
  {
    for ( int w = 0; w < pu.Y().width; w += blockWidth )
    {
      if ( !subblkMVSpreadOverLimit )
      {
        mvScaleTmpHor = mvScaleHor + deltaMvHorX * (halfBW + w) + deltaMvVerX * (halfBH + h);
        mvScaleTmpVer = mvScaleVer + deltaMvHorY * (halfBW + w) + deltaMvVerY * (halfBH + h);

      }
      else
      {
        mvScaleTmpHor = mvScaleHor + deltaMvHorX * ( pu.Y().width >> 1 ) + deltaMvVerX * ( pu.Y().height >> 1 );
        mvScaleTmpVer = mvScaleVer + deltaMvHorY * ( pu.Y().width >> 1 ) + deltaMvVerY * ( pu.Y().height >> 1 );
      }
      roundAffineMv( mvScaleTmpHor, mvScaleTmpVer, shift );
      Mv curMv(mvScaleTmpHor, mvScaleTmpVer);
      curMv.clipToStorageBitDepth();

      for ( int y = (h >> MIN_CU_LOG2); y < ((h + blockHeight) >> MIN_CU_LOG2); y++ )
      {
        for ( int x = (w >> MIN_CU_LOG2); x < ((w + blockWidth) >> MIN_CU_LOG2); x++ )
        {
          mb.at(x, y).mv[eRefList] = curMv;
        }
      }
    }
  }

  pu.mvAffi[eRefList][0] = affLT;
  pu.mvAffi[eRefList][1] = affRT;
  pu.mvAffi[eRefList][2] = affLB;
}

static bool deriveScaledMotionTemporal( const Slice&      slice,
                                        const Position&   colPos,
                                        const Picture*    pColPic,
                                        const RefPicList  eCurrRefPicList,
                                        Mv&         cColMv,
                                        const RefPicList  eFetchRefPicList)
{
  const MotionInfo &mi = pColPic->cs->getMotionInfo(colPos);
  const Slice *pColSlice = nullptr;

  for (const auto &pSlice : pColPic->slices)
  {
    if (pSlice->getIndependentSliceIdx() == mi.sliceIdx)
    {
      pColSlice = pSlice;
      break;
    }
  }

  CHECK(pColSlice == nullptr, "Couldn't find the colocated slice");

  int iColPOC, iColRefPOC, iCurrPOC, iCurrRefPOC, iScale;
  bool bAllowMirrorMV = true;
  RefPicList eColRefPicList = slice.getCheckLDC() ? eCurrRefPicList : RefPicList(1 - eFetchRefPicList);
  if (pColPic == slice.getRefPic(RefPicList(slice.isInterB() ? 1 - slice.getColFromL0Flag() : 0), slice.getColRefIdx()))
  {
    eColRefPicList = eCurrRefPicList;   //67 -> disable, 64 -> enable
    bAllowMirrorMV = false;
  }

  // Although it might make sense to keep the unavailable motion field per direction still be unavailable, I made the MV prediction the same way as in TMVP
  // So there is an interaction between MV0 and MV1 of the corresponding blocks identified by TV.

  // Grab motion and do necessary scaling.{{
  iCurrPOC = slice.getPOC();

  int iColRefIdx = mi.refIdx[eColRefPicList];

  if (iColRefIdx < 0 && (slice.getCheckLDC() || bAllowMirrorMV))
  {
    eColRefPicList = RefPicList(1 - eColRefPicList);
    iColRefIdx = mi.refIdx[eColRefPicList];

    if (iColRefIdx < 0)
    {
      return false;
    }
  }

  if (iColRefIdx >= 0 && slice.getNumRefIdx(eCurrRefPicList) > 0)
  {
    iColPOC = pColSlice->getPOC();
    iColRefPOC = pColSlice->getRefPOC(eColRefPicList, iColRefIdx);
    if (iColPOC == iColRefPOC)
      return false;
    ///////////////////////////////////////////////////////////////
    // Set the target reference index to 0, may be changed later //
    ///////////////////////////////////////////////////////////////
    iCurrRefPOC = slice.getRefPic(eCurrRefPicList, 0)->getPOC();
    // Scale the vector.
    cColMv = mi.mv[eColRefPicList];
    cColMv.setHor(roundMvComp(cColMv.getHor()));
    cColMv.setVer(roundMvComp(cColMv.getVer()));
    //pcMvFieldSP[2*iPartition + eCurrRefPicList].getMv();
    // Assume always short-term for now
    iScale = xGetDistScaleFactor(iCurrPOC, iCurrRefPOC, iColPOC, iColRefPOC);

    if (iScale != 4096)
    {

      cColMv = cColMv.scaleMv(iScale);
    }

    return true;
  }
  return false;
}

void clipColPos(int& posX, int& posY, const PredictionUnit& pu)
{
  Position puPos = pu.lumaPos();
  int log2CtuSize = g_aucLog2[pu.cs->sps->getCTUSize()];
  int ctuX = ((puPos.x >> log2CtuSize) << log2CtuSize);
  int ctuY = ((puPos.y >> log2CtuSize) << log2CtuSize);
  int horMax = std::min((int)pu.cs->sps->getPicWidthInLumaSamples() - 1, ctuX + (int)pu.cs->sps->getCTUSize() + 3);
  int horMin = std::max((int)0, ctuX);
  int verMax = std::min((int)pu.cs->sps->getPicHeightInLumaSamples() - 1, ctuY + (int)pu.cs->sps->getCTUSize() - 1);
  int verMin = std::max((int)0, ctuY);

  posX = std::min(horMax, std::max(horMin, posX));
  posY = std::min(verMax, std::max(verMin, posY));
}

bool PU::getInterMergeSubPuMvpCand(const PredictionUnit &pu, MergeCtx& mrgCtx, bool& LICFlag, const int count
  , int mmvdList
)
{
  const Slice   &slice = *pu.cs->slice;
  const unsigned scale = 4 * std::max<int>(1, 4 * AMVP_DECIMATION_FACTOR / 4);
  const unsigned mask = ~(scale - 1);

  const Picture *pColPic = slice.getRefPic(RefPicList(slice.isInterB() ? 1 - slice.getColFromL0Flag() : 0), slice.getColRefIdx());
  Mv cTMv;
  RefPicList fetchRefPicList = RefPicList(slice.isInterB() ? 1 - slice.getColFromL0Flag() : 0);

#if JVET_O0163_REMOVE_SWITCHING_TMV
  if ( count )
  {
    if ( (mrgCtx.interDirNeighbours[0] & (1 << REF_PIC_LIST_0)) && slice.getRefPic( REF_PIC_LIST_0, mrgCtx.mvFieldNeighbours[REF_PIC_LIST_0].refIdx ) == pColPic )
    {
      cTMv = mrgCtx.mvFieldNeighbours[REF_PIC_LIST_0].mv;
      fetchRefPicList = REF_PIC_LIST_0;
    }
    else if ( slice.isInterB() && (mrgCtx.interDirNeighbours[0] & (1 << REF_PIC_LIST_1)) && slice.getRefPic( REF_PIC_LIST_1, mrgCtx.mvFieldNeighbours[REF_PIC_LIST_1].refIdx ) == pColPic )
    {
      cTMv = mrgCtx.mvFieldNeighbours[REF_PIC_LIST_1].mv;
      fetchRefPicList = REF_PIC_LIST_1;
    }
  }
#else
  bool terminate = false;
  for (unsigned currRefListId = 0; currRefListId < (slice.getSliceType() == B_SLICE ? 2 : 1) && !terminate; currRefListId++)
  {
    if ( count )
    {
      RefPicList currRefPicList = RefPicList(slice.getCheckLDC() ? (slice.getColFromL0Flag() ? currRefListId : 1 - currRefListId) : currRefListId);

      if ((mrgCtx.interDirNeighbours[0] & (1 << currRefPicList)) && slice.getRefPic(currRefPicList, mrgCtx.mvFieldNeighbours[0 * 2 + currRefPicList].refIdx) == pColPic)
      {
        cTMv = mrgCtx.mvFieldNeighbours[0 * 2 + currRefPicList].mv;
        terminate = true;
        fetchRefPicList = currRefPicList;
        break;
      }
    }
  }
#endif

  ///////////////////////////////////////////////////////////////////////
  ////////          GET Initial Temporal Vector                  ////////
  ///////////////////////////////////////////////////////////////////////
  int mvPrec = MV_FRACTIONAL_BITS_INTERNAL;

  Mv cTempVector = cTMv;
  bool  tempLICFlag = false;

  // compute the location of the current PU
  Position puPos = pu.lumaPos();
  Size puSize = pu.lumaSize();
  int numPartLine = std::max(puSize.width >> ATMVP_SUB_BLOCK_SIZE, 1u);
  int numPartCol = std::max(puSize.height >> ATMVP_SUB_BLOCK_SIZE, 1u);
  int puHeight = numPartCol == 1 ? puSize.height : 1 << ATMVP_SUB_BLOCK_SIZE;
  int puWidth = numPartLine == 1 ? puSize.width : 1 << ATMVP_SUB_BLOCK_SIZE;

  Mv cColMv;
  // use coldir.
  bool     bBSlice = slice.isInterB();

  Position centerPos;

  bool found = false;
  cTempVector = cTMv;
  int tempX = cTempVector.getHor() >> mvPrec;
  int tempY = cTempVector.getVer() >> mvPrec;

  centerPos.x = puPos.x + (puSize.width >> 1) + tempX;
  centerPos.y = puPos.y + (puSize.height >> 1) + tempY;

  clipColPos(centerPos.x, centerPos.y, pu);

  centerPos = Position{ PosType(centerPos.x & mask), PosType(centerPos.y & mask) };

  // derivation of center motion parameters from the collocated CU
  const MotionInfo &mi = pColPic->cs->getMotionInfo(centerPos);

  if (mi.isInter && mi.isIBCmot == false)
  {
    mrgCtx.interDirNeighbours[count] = 0;

    for (unsigned currRefListId = 0; currRefListId < (bBSlice ? 2 : 1); currRefListId++)
    {
      RefPicList  currRefPicList = RefPicList(currRefListId);

      if (deriveScaledMotionTemporal(slice, centerPos, pColPic, currRefPicList, cColMv, fetchRefPicList))
      {
        // set as default, for further motion vector field spanning
        mrgCtx.mvFieldNeighbours[(count << 1) + currRefListId].setMvField(cColMv, 0);
        mrgCtx.interDirNeighbours[count] |= (1 << currRefListId);
        LICFlag = tempLICFlag;
        mrgCtx.GBiIdx[count] = GBI_DEFAULT;
        found = true;
      }
      else
      {
        mrgCtx.mvFieldNeighbours[(count << 1) + currRefListId].setMvField(Mv(), NOT_VALID);
        mrgCtx.interDirNeighbours[count] &= ~(1 << currRefListId);
      }
    }
  }

  if (!found)
  {
    return false;
  }
  if (mmvdList != 1)
  {
  int xOff = (puWidth >> 1) + tempX;
  int yOff = (puHeight >> 1) + tempY;

  MotionBuf& mb = mrgCtx.subPuMvpMiBuf;

  const bool isBiPred = isBipredRestriction(pu);

  for (int y = puPos.y; y < puPos.y + puSize.height; y += puHeight)
  {
    for (int x = puPos.x; x < puPos.x + puSize.width; x += puWidth)
    {
      Position colPos{ x + xOff, y + yOff };

      clipColPos(colPos.x, colPos.y, pu);

      colPos = Position{ PosType(colPos.x & mask), PosType(colPos.y & mask) };

      const MotionInfo &colMi = pColPic->cs->getMotionInfo(colPos);

      MotionInfo mi;

      found = false;
      mi.isInter = true;
      mi.sliceIdx = slice.getIndependentSliceIdx();
      mi.isIBCmot = false;
      if (colMi.isInter && colMi.isIBCmot == false)
      {
        for (unsigned currRefListId = 0; currRefListId < (bBSlice ? 2 : 1); currRefListId++)
        {
          RefPicList currRefPicList = RefPicList(currRefListId);
          if (deriveScaledMotionTemporal(slice, colPos, pColPic, currRefPicList, cColMv, fetchRefPicList))
          {
            mi.refIdx[currRefListId] = 0;
            mi.mv[currRefListId] = cColMv;
            found = true;
          }
        }
      }
      if (!found)
      {
        mi.mv[0] = mrgCtx.mvFieldNeighbours[(count << 1) + 0].mv;
        mi.mv[1] = mrgCtx.mvFieldNeighbours[(count << 1) + 1].mv;
        mi.refIdx[0] = mrgCtx.mvFieldNeighbours[(count << 1) + 0].refIdx;
        mi.refIdx[1] = mrgCtx.mvFieldNeighbours[(count << 1) + 1].refIdx;
      }

      mi.interDir = (mi.refIdx[0] != -1 ? 1 : 0) + (mi.refIdx[1] != -1 ? 2 : 0);

      if (isBiPred && mi.interDir == 3)
      {
        mi.interDir = 1;
        mi.mv[1] = Mv();
        mi.refIdx[1] = NOT_VALID;
      }

      mb.subBuf(g_miScaling.scale(Position{ x, y } -pu.lumaPos()), g_miScaling.scale(Size(puWidth, puHeight))).fill(mi);
      }
    }
  }
  return true;
  }

void PU::spanMotionInfo( PredictionUnit &pu, const MergeCtx &mrgCtx )
{
  MotionBuf mb = pu.getMotionBuf();

  if( !pu.mergeFlag || pu.mergeType == MRG_TYPE_DEFAULT_N
    || pu.mergeType == MRG_TYPE_IBC
    )
  {
    MotionInfo mi;

    mi.isInter = !CU::isIntra(*pu.cu);
    mi.isIBCmot = CU::isIBC(*pu.cu);
    mi.sliceIdx = pu.cu->slice->getIndependentSliceIdx();

    if( mi.isInter )
    {
      mi.interDir = pu.interDir;

      for( int i = 0; i < NUM_REF_PIC_LIST_01; i++ )
      {
        mi.mv[i]     = pu.mv[i];
        mi.refIdx[i] = pu.refIdx[i];
      }
      if (mi.isIBCmot)
      {
        mi.bv = pu.bv;
      }
    }

    if( pu.cu->affine )
    {
      for( int y = 0; y < mb.height; y++ )
      {
        for( int x = 0; x < mb.width; x++ )
        {
          MotionInfo &dest = mb.at( x, y );
          dest.isInter  = mi.isInter;
          dest.isIBCmot = false;
          dest.interDir = mi.interDir;
          dest.sliceIdx = mi.sliceIdx;
          for( int i = 0; i < NUM_REF_PIC_LIST_01; i++ )
          {
            if( mi.refIdx[i] == -1 )
            {
              dest.mv[i] = Mv();
            }
            dest.refIdx[i] = mi.refIdx[i];
          }
        }
      }
    }
    else
    {
      mb.fill( mi );
    }
  }
  else if (pu.mergeType == MRG_TYPE_SUBPU_ATMVP)
  {
    CHECK(mrgCtx.subPuMvpMiBuf.area() == 0 || !mrgCtx.subPuMvpMiBuf.buf, "Buffer not initialized");
    mb.copyFrom(mrgCtx.subPuMvpMiBuf);
  }
  else
  {

    if( isBipredRestriction( pu ) )
    {
      for( int y = 0; y < mb.height; y++ )
      {
        for( int x = 0; x < mb.width; x++ )
        {
          MotionInfo &mi = mb.at( x, y );
          if( mi.interDir == 3 )
          {
            mi.interDir  = 1;
            mi.mv    [1] = Mv();
            mi.refIdx[1] = NOT_VALID;
          }
        }
      }
    }
  }
}

void PU::applyImv( PredictionUnit& pu, MergeCtx &mrgCtx, InterPrediction *interPred )
{
  if( !pu.mergeFlag )
  {
    if( pu.interDir != 2 /* PRED_L1 */ )
    {
      pu.mvd[0].changeTransPrecAmvr2Internal(pu.cu->imv);
      unsigned mvp_idx = pu.mvpIdx[0];
      AMVPInfo amvpInfo;
      if (CU::isIBC(*pu.cu))
      {
        PU::fillIBCMvpCand(pu, amvpInfo);
      }
      else
      PU::fillMvpCand(pu, REF_PIC_LIST_0, pu.refIdx[0], amvpInfo);
      pu.mvpNum[0] = amvpInfo.numCand;
      pu.mvpIdx[0] = mvp_idx;
      pu.mv    [0] = amvpInfo.mvCand[mvp_idx] + pu.mvd[0];
      pu.mv[0].mvCliptoStorageBitDepth();
    }

    if (pu.interDir != 1 /* PRED_L0 */)
    {
      if( !( pu.cu->cs->slice->getMvdL1ZeroFlag() && pu.interDir == 3 ) && pu.cu->imv )/* PRED_BI */
      {
        pu.mvd[1].changeTransPrecAmvr2Internal(pu.cu->imv);
      }
      unsigned mvp_idx = pu.mvpIdx[1];
      AMVPInfo amvpInfo;
      PU::fillMvpCand(pu, REF_PIC_LIST_1, pu.refIdx[1], amvpInfo);
      pu.mvpNum[1] = amvpInfo.numCand;
      pu.mvpIdx[1] = mvp_idx;
      pu.mv    [1] = amvpInfo.mvCand[mvp_idx] + pu.mvd[1];
      pu.mv[1].mvCliptoStorageBitDepth();
    }
  }
  else
  {
    // this function is never called for merge
    THROW("unexpected");
    PU::getInterMergeCandidates ( pu, mrgCtx
      , 0
    );

    mrgCtx.setMergeInfo( pu, pu.mergeIdx );
  }

  PU::spanMotionInfo( pu, mrgCtx );
}

bool PU::isBiPredFromDifferentDir( const PredictionUnit& pu )
{
  if ( pu.refIdx[0] >= 0 && pu.refIdx[1] >= 0 )
  {
    const int iPOC0 = pu.cu->slice->getRefPOC( REF_PIC_LIST_0, pu.refIdx[0] );
    const int iPOC1 = pu.cu->slice->getRefPOC( REF_PIC_LIST_1, pu.refIdx[1] );
    const int iPOC  = pu.cu->slice->getPOC();
    if ( (iPOC - iPOC0)*(iPOC - iPOC1) < 0 )
    {
      return true;
    }
  }

  return false;
}
bool PU::isBiPredFromDifferentDirEqDistPoc(const PredictionUnit& pu)
{
  if (pu.refIdx[0] >= 0 && pu.refIdx[1] >= 0)
  {
    const int poc0 = pu.cu->slice->getRefPOC(REF_PIC_LIST_0, pu.refIdx[0]);
    const int poc1 = pu.cu->slice->getRefPOC(REF_PIC_LIST_1, pu.refIdx[1]);
    const int poc = pu.cu->slice->getPOC();
    if ((poc - poc0)*(poc - poc1) < 0)
    {
      if (abs(poc - poc0) == abs(poc - poc1))
      {
        return true;
      }
    }
  }
  return false;
}
void PU::restrictBiPredMergeCands( const PredictionUnit &pu, MergeCtx& mergeCtx )
{
  if( PU::isBipredRestriction( pu ) )
  {
    for( uint32_t mergeCand = 0; mergeCand < mergeCtx.numValidMergeCand; ++mergeCand )
    {
      if( mergeCtx.interDirNeighbours[ mergeCand ] == 3 )
      {
        mergeCtx.interDirNeighbours[ mergeCand ] = 1;
        mergeCtx.mvFieldNeighbours[( mergeCand << 1 ) + 1].setMvField( Mv( 0, 0 ), -1 );
        mergeCtx.GBiIdx[mergeCand] = GBI_DEFAULT;
      }
    }
  }
}

void PU::restrictBiPredMergeCandsOne(PredictionUnit &pu)
{
  if (PU::isBipredRestriction(pu))
  {
    if (pu.interDir == 3)
    {
      pu.interDir = 1;
      pu.refIdx[1] = -1;
      pu.mv[1] = Mv(0, 0);
      pu.cu->GBiIdx = GBI_DEFAULT;
    }
  }
}

void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangleMrgCtx )
{
  MergeCtx tmpMergeCtx;

  const Slice &slice = *pu.cs->slice;
  const uint32_t maxNumMergeCand = slice.getMaxNumMergeCand();

  triangleMrgCtx.numValidMergeCand = 0;

  for (int32_t i = 0; i < TRIANGLE_MAX_NUM_UNI_CANDS; i++)
  {
    triangleMrgCtx.GBiIdx[i] = GBI_DEFAULT;
    triangleMrgCtx.interDirNeighbours[i] = 0;
    triangleMrgCtx.mrgTypeNeighbours[i] = MRG_TYPE_DEFAULT_N;
    triangleMrgCtx.mvFieldNeighbours[(i << 1)].refIdx = NOT_VALID;
    triangleMrgCtx.mvFieldNeighbours[(i << 1) + 1].refIdx = NOT_VALID;
    triangleMrgCtx.mvFieldNeighbours[(i << 1)].mv = Mv();
    triangleMrgCtx.mvFieldNeighbours[(i << 1) + 1].mv = Mv();
  }

  PU::getInterMergeCandidates(pu, tmpMergeCtx, 0);

  for (int32_t i = 0; i < maxNumMergeCand; i++)
  {
    int parity = i & 1;
    if (tmpMergeCtx.interDirNeighbours[i] & (0x01 + parity))
    {
      triangleMrgCtx.interDirNeighbours[triangleMrgCtx.numValidMergeCand] = 1 + parity;
      triangleMrgCtx.mrgTypeNeighbours[triangleMrgCtx.numValidMergeCand] = MRG_TYPE_DEFAULT_N;
      triangleMrgCtx.mvFieldNeighbours[(triangleMrgCtx.numValidMergeCand << 1) + !parity].mv = Mv(0, 0);
      triangleMrgCtx.mvFieldNeighbours[(triangleMrgCtx.numValidMergeCand << 1) + parity].mv = tmpMergeCtx.mvFieldNeighbours[(i << 1) + parity].mv;
      triangleMrgCtx.mvFieldNeighbours[(triangleMrgCtx.numValidMergeCand << 1) + !parity].refIdx = -1;
      triangleMrgCtx.mvFieldNeighbours[(triangleMrgCtx.numValidMergeCand << 1) + parity].refIdx = tmpMergeCtx.mvFieldNeighbours[(i << 1) + parity].refIdx;
      triangleMrgCtx.numValidMergeCand++;
      if (triangleMrgCtx.numValidMergeCand == TRIANGLE_MAX_NUM_UNI_CANDS)
      {
        return;
      }
      continue;
    }

    if (tmpMergeCtx.interDirNeighbours[i] & (0x02 - parity))
    {
      triangleMrgCtx.interDirNeighbours[triangleMrgCtx.numValidMergeCand] = 2 - parity;
      triangleMrgCtx.mrgTypeNeighbours[triangleMrgCtx.numValidMergeCand] = MRG_TYPE_DEFAULT_N;
      triangleMrgCtx.mvFieldNeighbours[(triangleMrgCtx.numValidMergeCand << 1) + !parity].mv = tmpMergeCtx.mvFieldNeighbours[(i << 1) + !parity].mv;
      triangleMrgCtx.mvFieldNeighbours[(triangleMrgCtx.numValidMergeCand << 1) + parity].mv = Mv(0, 0);
      triangleMrgCtx.mvFieldNeighbours[(triangleMrgCtx.numValidMergeCand << 1) + !parity].refIdx = tmpMergeCtx.mvFieldNeighbours[(i << 1) + !parity].refIdx;
      triangleMrgCtx.mvFieldNeighbours[(triangleMrgCtx.numValidMergeCand << 1) + parity].refIdx = -1;
      triangleMrgCtx.numValidMergeCand++;
      if (triangleMrgCtx.numValidMergeCand == TRIANGLE_MAX_NUM_UNI_CANDS)
      {
        return;
      }
    }
  }
}

bool PU::isUniqueTriangleCandidates( const PredictionUnit &pu, MergeCtx& triangleMrgCtx )
{
  int newCand = triangleMrgCtx.numValidMergeCand;
  for( int32_t i = 0; i < newCand; i++ )
  {
    int32_t predFlagCur  = triangleMrgCtx.interDirNeighbours[i] == 1 ? 0 : 1;
    int32_t predFlagNew  = triangleMrgCtx.interDirNeighbours[newCand] == 1 ? 0 : 1;
    int32_t refPicPocCur = pu.cs->slice->getRefPOC( (RefPicList)predFlagCur, triangleMrgCtx.mvFieldNeighbours[(i << 1) + predFlagCur].refIdx );
    int32_t refPicPocNew = pu.cs->slice->getRefPOC( (RefPicList)predFlagNew, triangleMrgCtx.mvFieldNeighbours[(newCand << 1) + predFlagNew].refIdx);
    if( refPicPocCur == refPicPocNew && triangleMrgCtx.mvFieldNeighbours[(i << 1) + predFlagCur].mv == triangleMrgCtx.mvFieldNeighbours[(newCand << 1) + predFlagNew].mv )
    {
      return false;
    }
  }
  return true;
}


void PU::spanTriangleMotionInfo( PredictionUnit &pu, MergeCtx &triangleMrgCtx, const bool splitDir, const uint8_t candIdx0, const uint8_t candIdx1 )
{
  pu.triangleSplitDir = splitDir;
  pu.triangleMergeIdx0 = candIdx0;
  pu.triangleMergeIdx1 = candIdx1;
  MotionBuf mb = pu.getMotionBuf();

  MotionInfo biMv;
  biMv.isInter  = true;
  biMv.sliceIdx = pu.cs->slice->getIndependentSliceIdx();

  if( triangleMrgCtx.interDirNeighbours[candIdx0] == 1 && triangleMrgCtx.interDirNeighbours[candIdx1] == 2 )
  {
    biMv.interDir  = 3;
    biMv.mv[0]     = triangleMrgCtx.mvFieldNeighbours[ candIdx0 << 1     ].mv;
    biMv.mv[1]     = triangleMrgCtx.mvFieldNeighbours[(candIdx1 << 1) + 1].mv;
    biMv.refIdx[0] = triangleMrgCtx.mvFieldNeighbours[ candIdx0 << 1     ].refIdx;