EncCu.cpp

      if( !jobUsed[jId] || jId == bestJId ) continue;

      auto *jobBlkCache = dynamic_cast<BestEncInfoCache*>( m_pcEncLib->getCuEncoder( picture->scheduler.getSplitDataId( jId ) )->m_modeCtrl );
      CHECK( !jobBlkCache, "If own mode controller has blk info cache capability so should all other mode controllers!" );
      blkCache->BestEncInfoCache::copyState( *jobBlkCache, partitioner.currArea() );
    }

    blkCache->tick();
  }
#endif
}

void EncCu::copyState( EncCu* other, Partitioner& partitioner, const UnitArea& currArea, const bool isDist )
{
  const unsigned wIdx = gp_sizeIdxInfo->idxFrom( partitioner.currArea().lwidth () );
  const unsigned hIdx = gp_sizeIdxInfo->idxFrom( partitioner.currArea().lheight() );

  if( isDist )
  {
    other->m_pBestCS[wIdx][hIdx]->initSubStructure( *m_pBestCS[wIdx][hIdx], partitioner.chType, partitioner.currArea(), false );
    other->m_pTempCS[wIdx][hIdx]->initSubStructure( *m_pTempCS[wIdx][hIdx], partitioner.chType, partitioner.currArea(), false );
  }
  else
  {
          CodingStructure* dst =        m_pBestCS[wIdx][hIdx];
    const CodingStructure* src = other->m_pBestCS[wIdx][hIdx];
    bool keepResi = KEEP_PRED_AND_RESI_SIGNALS;
    bool keepPred = true;

    dst->useSubStructure( *src, partitioner.chType, currArea, keepPred, true, keepResi, keepResi );

    dst->cost           =  src->cost;
    dst->dist           =  src->dist;
    dst->fracBits       =  src->fracBits;
    dst->features       =  src->features;
  }

  if( isDist )
  {
    m_CurrCtx = m_CtxBuffer.data();
  }

  m_pcInterSearch->copyState( *other->m_pcInterSearch );
  m_modeCtrl     ->copyState( *other->m_modeCtrl, partitioner.currArea() );
  m_pcRdCost     ->copyState( *other->m_pcRdCost );
  m_pcTrQuant    ->copyState( *other->m_pcTrQuant );
  if( m_pcEncCfg->getReshaper() )
  {
    EncReshape *encReshapeThis  = dynamic_cast<EncReshape*>(       m_pcReshape);
    EncReshape *encReshapeOther = dynamic_cast<EncReshape*>(other->m_pcReshape);
    encReshapeThis->copyState( *encReshapeOther );
  }
  m_shareState    = other->m_shareState;
  m_shareBndPosX  = other->m_shareBndPosX;
  m_shareBndPosY  = other->m_shareBndPosY;
  m_shareBndSizeW = other->m_shareBndSizeW;
  m_shareBndSizeH = other->m_shareBndSizeH;
  setShareStateDec( other->getShareStateDec() );
  m_pcInterSearch->setShareState( other->m_pcInterSearch->getShareState() );

  m_CABACEstimator->getCtx() = other->m_CABACEstimator->getCtx();
}
#endif

void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
  const int qp                = encTestMode.qp;
  const Slice &slice          = *tempCS->slice;
  const bool bIsLosslessMode  = false; // False at this level. Next level down may set it to true.
  const int oldPrevQp         = tempCS->prevQP[partitioner.chType];
  const auto oldMotionLut     = tempCS->motionLut;
#if ENABLE_QPA_SUB_CTU
  const PPS &pps              = *tempCS->pps;
  const uint32_t currDepth    = partitioner.currDepth;
#endif

  const PartSplit split = getPartSplit( encTestMode );

  CHECK( split == CU_DONT_SPLIT, "No proper split provided!" );

  tempCS->initStructData( qp, bIsLosslessMode );

  m_CABACEstimator->getCtx() = m_CurrCtx->start;

  const TempCtx ctxStartSP( m_CtxCache, SubCtx( Ctx::SplitFlag,   m_CABACEstimator->getCtx() ) );
  const TempCtx ctxStartQt( m_CtxCache, SubCtx( Ctx::SplitQtFlag, m_CABACEstimator->getCtx() ) );
  const TempCtx ctxStartHv( m_CtxCache, SubCtx( Ctx::SplitHvFlag, m_CABACEstimator->getCtx() ) );
  const TempCtx ctxStart12( m_CtxCache, SubCtx( Ctx::Split12Flag, m_CABACEstimator->getCtx() ) );

  m_CABACEstimator->resetBits();

  m_CABACEstimator->split_cu_mode( split, *tempCS, partitioner );

  const double factor = ( tempCS->currQP[partitioner.chType] > 30 ? 1.1 : 1.075 );
  tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();
  if (!tempCS->useDbCost)
    CHECK(bestCS->costDbOffset != 0, "error");
  const double cost   = m_pcRdCost->calcRdCost( uint64_t( m_CABACEstimator->getEstFracBits() + ( ( bestCS->fracBits ) / factor ) ), Distortion( bestCS->dist / factor ) ) + bestCS->costDbOffset / factor;

  m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitFlag,   ctxStartSP );
  m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitQtFlag, ctxStartQt );
  m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitHvFlag, ctxStartHv );
  m_CABACEstimator->getCtx() = SubCtx( Ctx::Split12Flag, ctxStart12 );

  if (cost > bestCS->cost + bestCS->costDbOffset
#if ENABLE_QPA_SUB_CTU
    || (m_pcEncCfg->getUsePerceptQPA() && !m_pcEncCfg->getUseRateCtrl() && pps.getUseDQP() && (pps.getCuQpDeltaSubdiv() > 0) && (split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT) &&
        (currDepth == 0)) // force quad-split or no split at CTU level
#endif
    )
  {
    xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
    return;
  }

  int startShareThisLevel = 0;
  const uint32_t uiLPelX = tempCS->area.Y().lumaPos().x;
  const uint32_t uiTPelY = tempCS->area.Y().lumaPos().y;

  int splitRatio = 1;
  CHECK(!(split == CU_QUAD_SPLIT || split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT
    || split == CU_TRIH_SPLIT || split == CU_TRIV_SPLIT), "invalid split type");
  splitRatio = (split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT) ? 1 : 2;

  bool isOneChildSmall = ((tempCS->area.lwidth())*(tempCS->area.lheight()) >> splitRatio) < MRG_SHARELIST_SHARSIZE;

  if ((((tempCS->area.lwidth())*(tempCS->area.lheight())) > (MRG_SHARELIST_SHARSIZE * 1)))
  {
    m_shareState = NO_SHARE;
  }

  if (m_shareState == NO_SHARE)//init state
  {
    if (isOneChildSmall)
    {
      m_shareState = GEN_ON_SHARED_BOUND;//share start state
      startShareThisLevel = 1;
    }
  }
#if JVET_N0266_SMALL_BLOCKS
  if ( m_shareState == GEN_ON_SHARED_BOUND && slice.getSPS()->getIBCFlag() )
#else
  if ((m_shareState == GEN_ON_SHARED_BOUND) && (!slice.isIntra() || slice.getSPS()->getIBCFlag()))
#endif
  {
#if !JVET_N0266_SMALL_BLOCKS
    tempCS->motionLut.lutShare = tempCS->motionLut.lut;
#endif
    tempCS->motionLut.lutShareIbc = tempCS->motionLut.lutIbc;
    m_shareBndPosX = uiLPelX;
    m_shareBndPosY = uiTPelY;
    m_shareBndSizeW = tempCS->area.lwidth();
    m_shareBndSizeH = tempCS->area.lheight();
    m_shareState = SHARING;
  }


  m_pcInterSearch->setShareState(m_shareState);
  setShareStateDec(m_shareState);

  partitioner.splitCurrArea( split, *tempCS );
  bool qgEnableChildren = partitioner.currQgEnable(); // QG possible at children level

  m_CurrCtx++;

  tempCS->getRecoBuf().fill( 0 );

  tempCS->getPredBuf().fill(0);
  AffineMVInfo tmpMVInfo;
  bool isAffMVInfoSaved;
  m_pcInterSearch->savePrevAffMVInfo(0, tmpMVInfo, isAffMVInfoSaved);

  do
  {
    const auto &subCUArea  = partitioner.currArea();

    if( tempCS->picture->Y().contains( subCUArea.lumaPos() ) )
    {
      const unsigned wIdx    = gp_sizeIdxInfo->idxFrom( subCUArea.lwidth () );
      const unsigned hIdx    = gp_sizeIdxInfo->idxFrom( subCUArea.lheight() );

      CodingStructure *tempSubCS = m_pTempCS[wIdx][hIdx];
      CodingStructure *bestSubCS = m_pBestCS[wIdx][hIdx];

      tempCS->initSubStructure( *tempSubCS, partitioner.chType, subCUArea, false );
      tempCS->initSubStructure( *bestSubCS, partitioner.chType, subCUArea, false );
      tempSubCS->sharedBndPos.x = (m_shareState == SHARING) ? m_shareBndPosX : tempSubCS->area.Y().lumaPos().x;
      tempSubCS->sharedBndPos.y = (m_shareState == SHARING) ? m_shareBndPosY : tempSubCS->area.Y().lumaPos().y;
      tempSubCS->sharedBndSize.width = (m_shareState == SHARING) ? m_shareBndSizeW : tempSubCS->area.lwidth();
      tempSubCS->sharedBndSize.height = (m_shareState == SHARING) ? m_shareBndSizeH : tempSubCS->area.lheight();
      bestSubCS->sharedBndPos.x = (m_shareState == SHARING) ? m_shareBndPosX : tempSubCS->area.Y().lumaPos().x;
      bestSubCS->sharedBndPos.y = (m_shareState == SHARING) ? m_shareBndPosY : tempSubCS->area.Y().lumaPos().y;
      bestSubCS->sharedBndSize.width = (m_shareState == SHARING) ? m_shareBndSizeW : tempSubCS->area.lwidth();
      bestSubCS->sharedBndSize.height = (m_shareState == SHARING) ? m_shareBndSizeH : tempSubCS->area.lheight();
      xCompressCU( tempSubCS, bestSubCS, partitioner );

      if( bestSubCS->cost == MAX_DOUBLE )
      {
        CHECK( split == CU_QUAD_SPLIT, "Split decision reusing cannot skip quad split" );
        tempCS->cost = MAX_DOUBLE;
        tempCS->costDbOffset = 0;
        tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();
        m_CurrCtx--;
        partitioner.exitCurrSplit();
        xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
        return;
      }

      bool keepResi = KEEP_PRED_AND_RESI_SIGNALS;
      tempCS->useSubStructure( *bestSubCS, partitioner.chType, CS::getArea( *tempCS, subCUArea, partitioner.chType ), KEEP_PRED_AND_RESI_SIGNALS, true, keepResi, keepResi );

      if( partitioner.currQgEnable() )
      {
        tempCS->prevQP[partitioner.chType] = bestSubCS->prevQP[partitioner.chType];
      }

      tempSubCS->releaseIntermediateData();
      bestSubCS->releaseIntermediateData();
    }
  } while( partitioner.nextPart( *tempCS ) );

  partitioner.exitCurrSplit();

  if (startShareThisLevel == 1)
  {
    m_shareState = NO_SHARE;
    m_pcInterSearch->setShareState(m_shareState);
    setShareStateDec(m_shareState);
  }

  m_CurrCtx--;

  // Finally, generate split-signaling bits for RD-cost check
  const PartSplit implicitSplit = partitioner.getImplicitSplit( *tempCS );

  {
    bool enforceQT = implicitSplit == CU_QUAD_SPLIT;
#if HM_QTBT_REPRODUCE_FAST_LCTU_BUG

    // LARGE CTU bug
    if( m_pcEncCfg->getUseFastLCTU() )
    {
      unsigned minDepth = 0;
      unsigned maxDepth = g_aucLog2[tempCS->sps->getCTUSize()] - g_aucLog2[tempCS->sps->getMinQTSize(slice.getSliceType(), partitioner.chType)];

      if( auto ad = dynamic_cast<AdaptiveDepthPartitioner*>( &partitioner ) )
      {
        ad->setMaxMinDepth( minDepth, maxDepth, *tempCS );
      }

      if( minDepth > partitioner.currQtDepth )
      {
        // enforce QT
        enforceQT = true;
      }
    }
#endif

    if( !enforceQT )
    {
      m_CABACEstimator->resetBits();

      m_CABACEstimator->split_cu_mode( split, *tempCS, partitioner );

      tempCS->fracBits += m_CABACEstimator->getEstFracBits(); // split bits
    }
  }

  tempCS->cost = m_pcRdCost->calcRdCost( tempCS->fracBits, tempCS->dist );

  // Check Delta QP bits for splitted structure
  if( !qgEnableChildren ) // check at deepest QG level only
  xCheckDQP( *tempCS, partitioner, true );

  // If the configuration being tested exceeds the maximum number of bytes for a slice / slice-segment, then
  // a proper RD evaluation cannot be performed. Therefore, termination of the
  // slice/slice-segment must be made prior to this CTU.
  // This can be achieved by forcing the decision to be that of the rpcTempCU.
  // The exception is each slice / slice-segment must have at least one CTU.
  if (bestCS->cost != MAX_DOUBLE)
  {
    const TileMap& tileMap = *tempCS->picture->tileMap;
    const uint32_t CtuAddr             = CU::getCtuAddr( *bestCS->getCU( partitioner.chType ) );
    const bool isEndOfSlice        =    slice.getSliceMode() == FIXED_NUMBER_OF_BYTES
                                      && ((slice.getSliceBits() + CS::getEstBits(*bestCS)) > slice.getSliceArgument() << 3)
                                      && CtuAddr != tileMap.getCtuTsToRsAddrMap(slice.getSliceCurStartCtuTsAddr())
#if HEVC_DEPENDENT_SLICES
                                      && CtuAddr != tileMap.getCtuTsToRsAddrMap(slice.getSliceSegmentCurStartCtuTsAddr());
#else
                                      ;
#endif

#if HEVC_DEPENDENT_SLICES
    const bool isEndOfSliceSegment =    slice.getSliceSegmentMode() == FIXED_NUMBER_OF_BYTES
                                      && ((slice.getSliceSegmentBits() + CS::getEstBits(*bestCS)) > slice.getSliceSegmentArgument() << 3)
                                      && CtuAddr != tileMap.getCtuTsToRsAddrMap(slice.getSliceSegmentCurStartCtuTsAddr());
                                          // Do not need to check slice condition for slice-segment since a slice-segment is a subset of a slice.
    if (isEndOfSlice || isEndOfSliceSegment)
#else
    if(isEndOfSlice)
#endif
    {
      bestCS->cost = MAX_DOUBLE;
      bestCS->costDbOffset = 0;
    }
  }
  else
  {
    bestCS->costDbOffset = 0;
  }
  tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();

  // RD check for sub partitioned coding structure.
  xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );

  if (isAffMVInfoSaved)
    m_pcInterSearch->addAffMVInfo(tmpMVInfo);

  tempCS->motionLut = oldMotionLut;

  tempCS->releaseIntermediateData();

  tempCS->prevQP[partitioner.chType] = oldPrevQp;
}


void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
  const PPS &pps      = *tempCS->pps;

  bool   useIntraSubPartitions   = false;
  double maxCostAllowedForChroma = MAX_DOUBLE;
  const  CodingUnit *bestCU      = bestCS->getCU( partitioner.chType );
  Distortion interHad = m_modeCtrl->getInterHad();


  {

    tempCS->initStructData( encTestMode.qp, encTestMode.lossless );

    CodingUnit &cu      = tempCS->addCU( CS::getArea( *tempCS, tempCS->area, partitioner.chType ), partitioner.chType );

    partitioner.setCUData( cu );
    cu.slice            = tempCS->slice;
    cu.tileIdx          = tempCS->picture->tileMap->getTileIdxMap( tempCS->area.lumaPos() );
    cu.skip             = false;
    cu.mmvdSkip = false;
    cu.predMode         = MODE_INTRA;
    cu.transQuantBypass = encTestMode.lossless;
    cu.chromaQpAdj      = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
    cu.qp               = encTestMode.qp;
  //cu.ipcm             = false;
    cu.ispMode          = NOT_INTRA_SUBPARTITIONS;

    CU::addPUs( cu );

    tempCS->interHad    = interHad;

    m_bestModeUpdated = tempCS->useDbCost = bestCS->useDbCost = false;

    if( isLuma( partitioner.chType ) )
    {
      //the Intra SubPartitions mode uses the value of the best cost so far (luma if it is the fast version) to avoid test non-necessary lines
      const double bestCostSoFar = CS::isDualITree( *tempCS ) ? m_modeCtrl->getBestCostWithoutSplitFlags() : bestCU && bestCU->predMode == MODE_INTRA ? bestCS->lumaCost : bestCS->cost;
      m_pcIntraSearch->estIntraPredLumaQT( cu, partitioner, bestCostSoFar );

      useIntraSubPartitions = cu.ispMode != NOT_INTRA_SUBPARTITIONS;
      if( !CS::isDualITree( *tempCS ) )
      {
        tempCS->lumaCost = m_pcRdCost->calcRdCost( tempCS->fracBits, tempCS->dist );
        if( useIntraSubPartitions )
        {
          //the difference between the best cost so far and the current luma cost is stored to avoid testing the Cr component if the cost of luma + Cb is larger than the best cost
          maxCostAllowedForChroma = bestCS->cost < MAX_DOUBLE ? bestCS->cost - tempCS->lumaCost : MAX_DOUBLE;
        }
      }

      if (m_pcEncCfg->getUsePbIntraFast() && tempCS->dist == std::numeric_limits<Distortion>::max()
          && tempCS->interHad == 0)
      {
        interHad = 0;
        // JEM assumes only perfect reconstructions can from now on beat the inter mode
        m_modeCtrl->enforceInterHad( 0 );
        return;
      }

      if( !CS::isDualITree( *tempCS ) )
      {
        cu.cs->picture->getRecoBuf( cu.Y() ).copyFrom( cu.cs->getRecoBuf( COMPONENT_Y ) );
        cu.cs->picture->getPredBuf(cu.Y()).copyFrom(cu.cs->getPredBuf(COMPONENT_Y));
      }
    }

    if( tempCS->area.chromaFormat != CHROMA_400 && ( partitioner.chType == CHANNEL_TYPE_CHROMA || !CS::isDualITree( *tempCS ) ) )
    {
      TUIntraSubPartitioner subTuPartitioner( partitioner );
      m_pcIntraSearch->estIntraPredChromaQT( cu, ( !useIntraSubPartitions || ( CS::isDualITree( *cu.cs ) && !isLuma( CHANNEL_TYPE_CHROMA ) ) ) ? partitioner : subTuPartitioner, maxCostAllowedForChroma );
      if( useIntraSubPartitions && !cu.ispMode )
      {
        //At this point the temp cost is larger than the best cost. Therefore, we can already skip the remaining calculations
        return;
      }
    }

    cu.rootCbf = false;

    for( uint32_t t = 0; t < getNumberValidTBlocks( *cu.cs->pcv ); t++ )
    {
      cu.rootCbf |= cu.firstTU->cbf[t] != 0;
    }

    // Get total bits for current mode: encode CU
    m_CABACEstimator->resetBits();

    if( pps.getTransquantBypassEnabledFlag() )
    {
      m_CABACEstimator->cu_transquant_bypass_flag( cu );
    }

    if ((!cu.cs->slice->isIntra() || cu.cs->slice->getSPS()->getIBCFlag())
      && cu.Y().valid()
      )
    {
      m_CABACEstimator->cu_skip_flag ( cu );
    }
    m_CABACEstimator->pred_mode      ( cu );
    m_CABACEstimator->pcm_data       ( cu, partitioner );
    m_CABACEstimator->extend_ref_line( cu );
    m_CABACEstimator->isp_mode       ( cu );
    m_CABACEstimator->cu_pred_data   ( cu );

    // Encode Coefficients
    CUCtx cuCtx;
    cuCtx.isDQPCoded = true;
    cuCtx.isChromaQpAdjCoded = true;
    m_CABACEstimator->cu_residual( cu, partitioner, cuCtx );

    tempCS->fracBits = m_CABACEstimator->getEstFracBits();
    tempCS->cost     = m_pcRdCost->calcRdCost(tempCS->fracBits, tempCS->dist);

    const double tmpCostWithoutSplitFlags = tempCS->cost;
    xEncodeDontSplit( *tempCS, partitioner );

    xCheckDQP( *tempCS, partitioner );

    if( tempCS->cost < bestCS->cost )
    {
      m_modeCtrl->setBestCostWithoutSplitFlags( tmpCostWithoutSplitFlags );
    }

    xCalDebCost( *tempCS, partitioner );
    tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();


#if WCG_EXT
    DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda( true ) );
#else
    DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda() );
#endif
    xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );

  } //for emtCuFlag
}

void EncCu::xCheckIntraPCM(CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
  tempCS->initStructData( encTestMode.qp, encTestMode.lossless );

  CodingUnit &cu      = tempCS->addCU( CS::getArea( *tempCS, tempCS->area, partitioner.chType ), partitioner.chType );

  partitioner.setCUData( cu );
  cu.slice            = tempCS->slice;
  cu.tileIdx          = tempCS->picture->tileMap->getTileIdxMap( tempCS->area.lumaPos() );
  cu.skip             = false;
  cu.mmvdSkip = false;
  cu.predMode         = MODE_INTRA;
  cu.transQuantBypass = encTestMode.lossless;
  cu.chromaQpAdj      = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
  cu.qp               = encTestMode.qp;
  cu.ipcm             = true;

  tempCS->addPU( CS::getArea( *tempCS, tempCS->area, partitioner.chType ), partitioner.chType );

  tempCS->addTU( CS::getArea( *tempCS, tempCS->area, partitioner.chType ), partitioner.chType );

  m_pcIntraSearch->IPCMSearch(*tempCS, partitioner);

  m_CABACEstimator->getCtx() = m_CurrCtx->start;

  m_CABACEstimator->resetBits();

  if( tempCS->pps->getTransquantBypassEnabledFlag() )
  {
    m_CABACEstimator->cu_transquant_bypass_flag( cu );
  }

  if ((!cu.cs->slice->isIntra() || cu.cs->slice->getSPS()->getIBCFlag())
    && cu.Y().valid()
    )
  {
    m_CABACEstimator->cu_skip_flag ( cu );
  }
  m_CABACEstimator->pred_mode      ( cu );
  m_CABACEstimator->pcm_data       ( cu, partitioner );


  tempCS->fracBits = m_CABACEstimator->getEstFracBits();
  tempCS->cost     = m_pcRdCost->calcRdCost(tempCS->fracBits, tempCS->dist);

  xEncodeDontSplit( *tempCS, partitioner );

  xCheckDQP( *tempCS, partitioner );

  xCalDebCost( *tempCS, partitioner );
  tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();

#if WCG_EXT
  DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda( true ) );
#else
  DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda() );
#endif
  xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
}

void EncCu::xCheckDQP( CodingStructure& cs, Partitioner& partitioner, bool bKeepCtx )
{
  CHECK( bKeepCtx && cs.cus.size() <= 1 && partitioner.getImplicitSplit( cs ) == CU_DONT_SPLIT, "bKeepCtx should only be set in split case" );
  CHECK( !bKeepCtx && cs.cus.size() > 1, "bKeepCtx should never be set for non-split case" );

  if( !cs.pps->getUseDQP() )
  {
    return;
  }

  if (CS::isDualITree(cs) && isChroma(partitioner.chType))
  {
    return;
  }

  if( !partitioner.currQgEnable() ) // do not consider split or leaf/not leaf QG condition (checked by caller)
  {
    return;
  }


  CodingUnit* cuFirst = cs.getCU( partitioner.chType );

  CHECK( !cuFirst, "No CU available" );

  bool hasResidual = false;
  for( const auto &cu : cs.cus )
  {
    if( cu->rootCbf )
    {
      hasResidual = true;
      break;
    }
  }

  int predQP = CU::predictQP( *cuFirst, cs.prevQP[partitioner.chType] );

  if( hasResidual )
  {
    TempCtx ctxTemp( m_CtxCache );
    if( !bKeepCtx ) ctxTemp = SubCtx( Ctx::DeltaQP, m_CABACEstimator->getCtx() );

    m_CABACEstimator->resetBits();
    m_CABACEstimator->cu_qp_delta( *cuFirst, predQP, cuFirst->qp );

    cs.fracBits += m_CABACEstimator->getEstFracBits(); // dQP bits
    cs.cost      = m_pcRdCost->calcRdCost(cs.fracBits, cs.dist);


    if( !bKeepCtx ) m_CABACEstimator->getCtx() = SubCtx( Ctx::DeltaQP, ctxTemp );

    // NOTE: reset QPs for CUs without residuals up to first coded CU
    for( const auto &cu : cs.cus )
    {
      if( cu->rootCbf )
      {
        break;
      }
      cu->qp = predQP;
    }
  }
  else
  {
    // No residuals: reset CU QP to predicted value
    for( const auto &cu : cs.cus )
    {
      cu->qp = predQP;
    }
  }
}

void EncCu::xFillPCMBuffer( CodingUnit &cu )
{
  const ChromaFormat format        = cu.chromaFormat;
  const uint32_t numberValidComponents = getNumberValidComponents(format);

  for( auto &tu : CU::traverseTUs( cu ) )
  {
    for( uint32_t ch = 0; ch < numberValidComponents; ch++ )
    {
      const ComponentID compID = ComponentID( ch );

      const CompArea &compArea = tu.blocks[ compID ];

      const CPelBuf source      = tu.cs->getOrgBuf( compArea );
             PelBuf destination = tu.getPcmbuf( compID );

      if (tu.cs->slice->getReshapeInfo().getUseSliceReshaper() && m_pcReshape->getCTUFlag() && compID == COMPONENT_Y)
      {
        CompArea    tmpArea(COMPONENT_Y, compArea.chromaFormat, Position(0, 0), compArea.size());
        PelBuf tempOrgBuf = m_tmpStorageLCU->getBuf(tmpArea);
        tempOrgBuf.copyFrom(source);
        tempOrgBuf.rspSignal(m_pcReshape->getFwdLUT());
        destination.copyFrom(tempOrgBuf);
      }
      else
        destination.copyFrom( source );
    }
  }
}
void EncCu::xCheckRDCostHashInter( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
  bool isPerfectMatch = false;

  tempCS->initStructData(encTestMode.qp, encTestMode.lossless);
  m_pcInterSearch->resetBufferedUniMotions();
  m_pcInterSearch->setAffineModeSelected(false);
  CodingUnit &cu = tempCS->addCU(tempCS->area, partitioner.chType);

  partitioner.setCUData(cu);
  cu.slice = tempCS->slice;
  cu.skip = false;
  cu.predMode = MODE_INTER;
  cu.transQuantBypass = encTestMode.lossless;
  cu.chromaQpAdj = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
  cu.qp = encTestMode.qp;
  CU::addPUs(cu);
  cu.mmvdSkip = false;
  cu.firstPU->mmvdMergeFlag = false;

  if (m_pcInterSearch->predInterHashSearch(cu, partitioner, isPerfectMatch))
  {
    double equGBiCost = MAX_DOUBLE;

    m_bestModeUpdated = tempCS->useDbCost = bestCS->useDbCost = false;

    xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 0
      , 0
      , &equGBiCost
    );

    if ( m_bestModeUpdated && bestCS->cost != MAX_DOUBLE )
    {
      xCalDebCost( *bestCS, partitioner );
    }
  }
  tempCS->initStructData(encTestMode.qp, encTestMode.lossless);

  if (cu.lwidth() != 64)
  {
    isPerfectMatch = false;
  }
  m_modeCtrl->setIsHashPerfectMatch(isPerfectMatch);
}

void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
  const Slice &slice = *tempCS->slice;

  CHECK( slice.getSliceType() == I_SLICE, "Merge modes not available for I-slices" );

  tempCS->initStructData( encTestMode.qp, encTestMode.lossless );

  MergeCtx mergeCtx;
  const SPS &sps = *tempCS->sps;

  if( sps.getSBTMVPEnabledFlag() )
  {
    Size bufSize = g_miScaling.scale( tempCS->area.lumaSize() );
    mergeCtx.subPuMvpMiBuf    = MotionBuf( m_SubPuMiBuf,    bufSize );
  }

  Mv   refinedMvdL0[MAX_NUM_PARTS_IN_CTU][MRG_MAX_NUM_CANDS];
  setMergeBestSATDCost( MAX_DOUBLE );

  {
    // first get merge candidates
    CodingUnit cu( tempCS->area );
    cu.cs       = tempCS;
    cu.predMode = MODE_INTER;
    cu.slice    = tempCS->slice;
    cu.tileIdx  = tempCS->picture->tileMap->getTileIdxMap(tempCS->area.lumaPos());

    PredictionUnit pu( tempCS->area );
    pu.cu = &cu;
    pu.cs = tempCS;
    pu.shareParentPos = tempCS->sharedBndPos;
    pu.shareParentSize = tempCS->sharedBndSize;
    PU::getInterMergeCandidates(pu, mergeCtx
      , 0
    );
    PU::getInterMMVDMergeCandidates(pu, mergeCtx);
#if JVET_N0324_REGULAR_MRG_FLAG
    pu.regularMergeFlag = true;
#endif
  }
  bool candHasNoResidual[MRG_MAX_NUM_CANDS + MMVD_ADD_NUM];
  for (uint32_t ui = 0; ui < MRG_MAX_NUM_CANDS + MMVD_ADD_NUM; ui++)
  {
    candHasNoResidual[ui] = false;
  }

  bool                                        bestIsSkip = false;
  bool                                        bestIsMMVDSkip = true;
  PelUnitBuf                                  acMergeBuffer[MRG_MAX_NUM_CANDS];
  PelUnitBuf                                  acMergeRealBuffer[MMVD_MRG_MAX_RD_BUF_NUM];
  PelUnitBuf *                                acMergeTempBuffer[MMVD_MRG_MAX_RD_NUM];
  PelUnitBuf *                                singleMergeTempBuffer;
  int                                         insertPos;
  unsigned                                    uiNumMrgSATDCand = mergeCtx.numValidMergeCand + MMVD_ADD_NUM;

  static_vector<unsigned, MRG_MAX_NUM_CANDS + MMVD_ADD_NUM>  RdModeList;
  bool                                        mrgTempBufSet = false;

  for (unsigned i = 0; i < MRG_MAX_NUM_CANDS + MMVD_ADD_NUM; i++)
  {
    RdModeList.push_back(i);
  }

  const UnitArea localUnitArea(tempCS->area.chromaFormat, Area(0, 0, tempCS->area.Y().width, tempCS->area.Y().height));
  for (unsigned i = 0; i < MMVD_MRG_MAX_RD_BUF_NUM; i++)
  {
    acMergeRealBuffer[i] = m_acMergeBuffer[i].getBuf(localUnitArea);
    if (i < MMVD_MRG_MAX_RD_NUM)
    {
      acMergeTempBuffer[i] = acMergeRealBuffer + i;
    }
    else
    {
      singleMergeTempBuffer = acMergeRealBuffer + i;
    }
  }

  static_vector<unsigned, MRG_MAX_NUM_CANDS + MMVD_ADD_NUM>  RdModeList2; // store the Intra mode for Intrainter
  RdModeList2.clear();
  bool isIntrainterEnabled = sps.getUseMHIntra();
  if (bestCS->area.lwidth() * bestCS->area.lheight() < 64 || bestCS->area.lwidth() >= MAX_CU_SIZE || bestCS->area.lheight() >= MAX_CU_SIZE)
  {
    isIntrainterEnabled = false;
  }
  bool isTestSkipMerge[MRG_MAX_NUM_CANDS]; // record if the merge candidate has tried skip mode
  for (uint32_t idx = 0; idx < MRG_MAX_NUM_CANDS; idx++)
  {
    isTestSkipMerge[idx] = false;
  }
  if( m_pcEncCfg->getUseFastMerge() || isIntrainterEnabled)
  {
    uiNumMrgSATDCand = NUM_MRG_SATD_CAND;
    if (isIntrainterEnabled)
    {
      uiNumMrgSATDCand += 1;
    }
    bestIsSkip       = false;

    if( auto blkCache = dynamic_cast< CacheBlkInfoCtrl* >( m_modeCtrl ) )
    {
      if (slice.getSPS()->getIBCFlag())
      {
        ComprCUCtx cuECtx = m_modeCtrl->getComprCUCtx();
        bestIsSkip = blkCache->isSkip(tempCS->area) && cuECtx.bestCU;
      }
      else
      bestIsSkip = blkCache->isSkip( tempCS->area );
      bestIsMMVDSkip = blkCache->isMMVDSkip(tempCS->area);
    }

    if (isIntrainterEnabled) // always perform low complexity check
    {
      bestIsSkip = false;
    }

    static_vector<double, MRG_MAX_NUM_CANDS + MMVD_ADD_NUM> candCostList;

    // 1. Pass: get SATD-cost for selected candidates and reduce their count
    if( !bestIsSkip )
    {
      RdModeList.clear();
      mrgTempBufSet       = true;
      const double sqrtLambdaForFirstPass = m_pcRdCost->getMotionLambda( encTestMode.lossless );

      CodingUnit &cu      = tempCS->addCU( tempCS->area, partitioner.chType );
      const double sqrtLambdaForFirstPassIntra = m_pcRdCost->getMotionLambda(cu.transQuantBypass) / double(1 << SCALE_BITS);

      partitioner.setCUData( cu );
      cu.slice            = tempCS->slice;
      cu.tileIdx          = tempCS->picture->tileMap->getTileIdxMap( tempCS->area.lumaPos() );
      cu.skip             = false;
      cu.mmvdSkip = false;
      cu.triangle         = false;
    //cu.affine
      cu.predMode         = MODE_INTER;
    //cu.LICFlag
      cu.transQuantBypass = encTestMode.lossless;
      cu.chromaQpAdj      = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
      cu.qp               = encTestMode.qp;
    //cu.emtFlag  is set below

      PredictionUnit &pu  = tempCS->addPU( cu, partitioner.chType );

      DistParam distParam;
#if JVET_N0329_IBC_SEARCH_IMP
      const bool bUseHadamard = !encTestMode.lossless && !tempCS->slice->getDisableSATDForRD();
#else
      const bool bUseHadamard= !encTestMode.lossless;
#endif
      m_pcRdCost->setDistParam (distParam, tempCS->getOrgBuf().Y(), m_acMergeBuffer[0].Y(), sps.getBitDepth (CHANNEL_TYPE_LUMA), COMPONENT_Y, bUseHadamard);

      const UnitArea localUnitArea( tempCS->area.chromaFormat, Area( 0, 0, tempCS->area.Y().width, tempCS->area.Y().height) );
      for( uint32_t uiMergeCand = 0; uiMergeCand < mergeCtx.numValidMergeCand; uiMergeCand++ )
      {
        mergeCtx.setMergeInfo( pu, uiMergeCand );

        PU::spanMotionInfo( pu, mergeCtx );
        pu.mvRefine = true;
        distParam.cur = singleMergeTempBuffer->Y();
        m_pcInterSearch->motionCompensation(pu, *singleMergeTempBuffer);
        acMergeBuffer[uiMergeCand] = m_acRealMergeBuffer[uiMergeCand].getBuf(localUnitArea);
        acMergeBuffer[uiMergeCand].copyFrom(*singleMergeTempBuffer);
        pu.mvRefine = false;
        if( mergeCtx.interDirNeighbours[uiMergeCand] == 3 && mergeCtx.mrgTypeNeighbours[uiMergeCand] == MRG_TYPE_DEFAULT_N )
        {
          mergeCtx.mvFieldNeighbours[2*uiMergeCand].mv   = pu.mv[0];
          mergeCtx.mvFieldNeighbours[2*uiMergeCand+1].mv = pu.mv[1];
          {
            int dx, dy, i, j, num = 0;
            dy = std::min<int>(pu.lumaSize().height, DMVR_SUBCU_HEIGHT);
            dx = std::min<int>(pu.lumaSize().width, DMVR_SUBCU_WIDTH);
            if (PU::checkDMVRCondition(pu))
            {
              for (i = 0; i < (pu.lumaSize().height); i += dy)
              {
                for (j = 0; j < (pu.lumaSize().width); j += dx)
                {
                  refinedMvdL0[num][uiMergeCand] = pu.mvdL0SubPu[num];
                  num++;
                }
              }
            }
          }
        }

        Distortion uiSad = distParam.distFunc(distParam);
        uint32_t uiBitsCand = uiMergeCand + 1;
        if( uiMergeCand == tempCS->slice->getMaxNumMergeCand() - 1 )
        {
          uiBitsCand--;
        }
#if !JVET_MMVD_OFF_MACRO
#if JVET_N0127_MMVD_SPS_FLAG 
        if ( pu.cs->sps->getUseMMVD() )
          uiBitsCand++; // for mmvd_flag
#else
        uiBitsCand++; // for mmvd_flag
#endif
#endif
        double cost     = (double)uiSad + (double)uiBitsCand * sqrtLambdaForFirstPass;
        insertPos = -1;
        updateDoubleCandList(uiMergeCand, cost, RdModeList, candCostList, RdModeList2, (uint32_t)NUM_LUMA_MODE, uiNumMrgSATDCand, &insertPos);
        if (insertPos != -1)
        {
          if (insertPos == RdModeList.size() - 1)
          {
            swap(singleMergeTempBuffer, acMergeTempBuffer[insertPos]);
          }
          else
          {
            for (uint32_t i = uint32_t(RdModeList.size()) - 1; i > insertPos; i--)
            {
              swap(acMergeTempBuffer[i - 1], acMergeTempBuffer[i]);
            }
            swap(singleMergeTempBuffer, acMergeTempBuffer[insertPos]);
          }
        }
        CHECK(std::min(uiMergeCand + 1, uiNumMrgSATDCand) != RdModeList.size(), "");
      }

      if (isIntrainterEnabled)
      {
        int numTestIntraMode = 4;
        // prepare for Intra bits calculation
        const TempCtx ctxStart(m_CtxCache, m_CABACEstimator->getCtx());
        const TempCtx ctxStartIntraMode(m_CtxCache, SubCtx(Ctx::MHIntraPredMode, m_CABACEstimator->getCtx()));

        // for Intrainter fast, recored the best intra mode during the first round for mrege 0
        int bestMHIntraMode = -1;
        double bestMHIntraCost = MAX_DOUBLE;

        pu.mhIntraFlag = true;

        // save the to-be-tested merge candidates
        uint32_t MHIntraMergeCand[NUM_MRG_SATD_CAND];
        for (uint32_t mergeCnt = 0; mergeCnt < std::min(NUM_MRG_SATD_CAND, (const int)mergeCtx.numValidMergeCand); mergeCnt++)
        {
          MHIntraMergeCand[mergeCnt] = RdModeList[mergeCnt];
        }
        for (uint32_t mergeCnt = 0; mergeCnt < std::min(std::min(NUM_MRG_SATD_CAND, (const int)mergeCtx.numValidMergeCand), 4); mergeCnt++)
        {
          uint32_t mergeCand = MHIntraMergeCand[mergeCnt];
          acMergeBuffer[mergeCand] = m_acRealMergeBuffer[mergeCand].getBuf(localUnitArea);

          // estimate merge bits
          uint32_t bitsCand = mergeCand + 1;
          if (mergeCand == pu.cs->slice->getMaxNumMergeCand() - 1)
          {
            bitsCand--;
          }

          // first round
          for (uint32_t intraCnt = 0; intraCnt < numTestIntraMode; intraCnt++)
          {
            pu.intraDir[0] = (intraCnt < 2) ? intraCnt : ((intraCnt == 2) ? HOR_IDX : VER_IDX);

            // fast 2
            if (mergeCnt > 0 && bestMHIntraMode != pu.intraDir[0])
            {
              continue;
            }
            int narrowCase = PU::getNarrowShape(pu.lwidth(), pu.lheight());
            if (narrowCase == 1 && pu.intraDir[0] == HOR_IDX)
            {
              continue;
            }
            if (narrowCase == 2 && pu.intraDir[0] == VER_IDX)
            {
              continue;
            }
            // generate intrainter Y prediction
            if (mergeCnt == 0)
            {
              m_pcIntraSearch->initIntraPatternChType( *pu.cu, pu.Y());
              m_pcIntraSearch->predIntraAng(COMPONENT_Y, pu.cs->getPredBuf(pu).Y(), pu);
              m_pcIntraSearch->switchBuffer(pu, COMPONENT_Y, pu.cs->getPredBuf(pu).Y(), m_pcIntraSearch->getPredictorPtr2(COMPONENT_Y, intraCnt));
            }
            pu.cs->getPredBuf(pu).copyFrom(acMergeBuffer[mergeCand]);
            if (pu.cs->slice->getReshapeInfo().getUseSliceReshaper() && m_pcReshape->getCTUFlag())
            {
              pu.cs->getPredBuf(pu).Y().rspSignal(m_pcReshape->getFwdLUT());
            }
            m_pcIntraSearch->geneWeightedPred(COMPONENT_Y, pu.cs->getPredBuf(pu).Y(), pu, m_pcIntraSearch->getPredictorPtr2(COMPONENT_Y, intraCnt));

            // calculate cost
            if (pu.cs->slice->getReshapeInfo().getUseSliceReshaper() && m_pcReshape->getCTUFlag())
            {
               pu.cs->getPredBuf(pu).Y().rspSignal(m_pcReshape->getInvLUT());
            }
            distParam.cur = pu.cs->getPredBuf(pu).Y();
            Distortion sadValue = distParam.distFunc(distParam);
            if (pu.cs->slice->getReshapeInfo().getUseSliceReshaper() && m_pcReshape->getCTUFlag())
            {
              pu.cs->getPredBuf(pu).Y().rspSignal(m_pcReshape->getFwdLUT());
            }
            m_CABACEstimator->getCtx() = SubCtx(Ctx::MHIntraPredMode, ctxStartIntraMode);
            uint64_t fracModeBits = m_pcIntraSearch->xFracModeBitsIntra(pu, pu.intraDir[0], CHANNEL_TYPE_LUMA);
#if JVET_N0324_REGULAR_MRG_FLAG
            double cost = (double)sadValue + (double)(bitsCand + 9) * sqrtLambdaForFirstPass + (double)fracModeBits * sqrtLambdaForFirstPassIntra;
#else
            double cost = (double)sadValue + (double)(bitsCand + 1) * sqrtLambdaForFirstPass + (double)fracModeBits * sqrtLambdaForFirstPassIntra;
#endif
            insertPos = -1;
            updateDoubleCandList(mergeCand + MRG_MAX_NUM_CANDS + MMVD_ADD_NUM, cost, RdModeList, candCostList, RdModeList2, pu.intraDir[0], uiNumMrgSATDCand, &insertPos);
            if (insertPos != -1)
            {
              for (int i = int(RdModeList.size()) - 1; i > insertPos; i--)
              {
                swap(acMergeTempBuffer[i - 1], acMergeTempBuffer[i]);
              }
              swap(singleMergeTempBuffer, acMergeTempBuffer[insertPos]);
            }
            // fast 2
            if (mergeCnt == 0 && cost < bestMHIntraCost)
            {
              bestMHIntraMode = pu.intraDir[0];
              bestMHIntraCost = cost;
            }
          }
        }
        pu.mhIntraFlag = false;
        m_CABACEstimator->getCtx() = ctxStart;
      }
#if !JVET_MMVD_OFF_MACRO
#if JVET_N0127_MMVD_SPS_FLAG 
      if ( pu.cs->sps->getUseMMVD() )
      {
#endif 
        cu.mmvdSkip = true;
#if JVET_N0448_N0380