IntraSearch.cpp

        }
        tmpValidReturn = xRecurIntraCodingLumaQT( *csTemp, partitioner, uiBestPUMode.ispMod ? bestCurrentCost : MAX_DOUBLE, -1, TU_NO_ISP, uiBestPUMode.ispMod,
                                                  mtsCheckRangeFlag, mtsFirstCheckId, mtsLastCheckId, moreProbMTSIdxFirst );
      }

      if( cu.ispMode && !csTemp->cus[0]->firstTU->cbf[COMPONENT_Y] )
      {
        if( !sps.getUseLFNST() )
        {
          if ( cu.ispMode == HOR_INTRA_SUBPARTITIONS )
          {
            ispHorAllZeroCbfs |= ( m_pcEncCfg->getUseFastISP() && csTemp->tus[0]->lheight() > 2 && csTemp->cost >= bestCurrentCost );
          }
          else
          {
            ispVerAllZeroCbfs |= ( m_pcEncCfg->getUseFastISP() && csTemp->tus[0]->lwidth() > 2 && csTemp->cost >= bestCurrentCost );
          }
        }
        csTemp->cost = MAX_DOUBLE;
        csTemp->costDbOffset = 0;
        tmpValidReturn = false;
      }
      validReturn |= tmpValidReturn;

      if( sps.getUseLFNST() && mtsUsageFlag == 1 && !cu.ispMode && mode >= 0 )
      {
        m_modeCostStore[ lfnstIdx ][ testMip ? rdModeIdxList[ mode ] : mode ] = tmpValidReturn ? csTemp->cost : ( MAX_DOUBLE / 2.0 ); //(MAX_DOUBLE / 2.0) ??
      }


      DTRACE( g_trace_ctx, D_INTRA_COST, "IntraCost T %f (%d) \n", csTemp->cost, uiOrgMode.modeId );


      if( tmpValidReturn )
      {
        // check r-d cost
        if( csTemp->cost < csBest->cost )
        {
          std::swap( csTemp, csBest );

          uiBestPUMode  = uiOrgMode;
          bestBDPCMMode = cu.bdpcmMode;
          if( sps.getUseLFNST() && mtsUsageFlag == 1 && !cu.ispMode )
          {
            m_bestModeCostStore[ lfnstIdx ] = csBest->cost; //cs.cost;
          }
          if( csBest->cost < bestCurrentCost )
          {
            bestCurrentCost = csBest->cost;
          }
          if( !cu.ispMode && !cu.mtsFlag )
          {
            m_modeCtrl->setMtsFirstPassNoIspCost( csBest->cost );
          }
        }
        if( !cu.ispMode && !cu.bdpcmMode && csBest->cost < bestCostNonBDPCM )
        {
          bestCostNonBDPCM = csBest->cost;
          bestNormalIntraModeIndex = mode;
        }
      }

      csTemp->releaseIntermediateData();
    } // Mode loop
    cu.ispMode = uiBestPUMode.ispMod;

    if( validReturn )
    {
      cs.useSubStructure( *csBest, partitioner.chType, pu.singleChan( CHANNEL_TYPE_LUMA ), true, true, keepResi, keepResi );
    }
    csBest->releaseIntermediateData();
    if( validReturn )
    {
      //=== update PU data ====
      cu.mipFlag = uiBestPUMode.mipFlg;
      pu.multiRefIdx = uiBestPUMode.mRefId;
      pu.intraDir[ CHANNEL_TYPE_LUMA ] = uiBestPUMode.modeId;
      cu.bdpcmMode = bestBDPCMMode;
    }
  }

  //===== reset context models =====
  m_CABACEstimator->getCtx() = ctxStart;

  return validReturn;
}

void IntraSearch::estIntraPredChromaQT( CodingUnit &cu, Partitioner &partitioner, const double maxCostAllowed )
{
  const ChromaFormat format   = cu.chromaFormat;
  const uint32_t    numberValidComponents = getNumberValidComponents(format);
  CodingStructure &cs = *cu.cs;
  const TempCtx ctxStart  ( m_CtxCache, m_CABACEstimator->getCtx() );

  cs.setDecomp( cs.area.Cb(), false );

  double    bestCostSoFar = maxCostAllowed;
  bool      lumaUsesISP   = !CS::isDualITree( *cu.cs ) && cu.ispMode;
  PartSplit ispType       = lumaUsesISP ? CU::getISPType( cu, COMPONENT_Y ) : TU_NO_ISP;
  CHECK( cu.ispMode && bestCostSoFar < 0, "bestCostSoFar must be positive!" );

  auto &pu = *cu.firstPU;

  {
    uint32_t       uiBestMode = 0;
    Distortion uiBestDist = 0;
    double     dBestCost = MAX_DOUBLE;

    //----- init mode list ----
    {
      uint32_t  uiMinMode = 0;
      uint32_t  uiMaxMode = NUM_CHROMA_MODE;

      //----- check chroma modes -----
      uint32_t chromaCandModes[ NUM_CHROMA_MODE ];
      PU::getIntraChromaCandModes( pu, chromaCandModes );

      // create a temporary CS
      CodingStructure &saveCS = *m_pSaveCS[0];
      saveCS.pcv      = cs.pcv;
      saveCS.picture  = cs.picture;
      saveCS.area.repositionTo( cs.area );
      saveCS.clearTUs();

      if( !CS::isDualITree( cs ) && cu.ispMode )
      {
        saveCS.clearCUs();
        saveCS.clearPUs();
      }

      if( CS::isDualITree( cs ) )
      {
        if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
        {
          partitioner.splitCurrArea( TU_MAX_TR_SPLIT, cs );

          do
          {
            cs.addTU( CS::getArea( cs, partitioner.currArea(), partitioner.chType ), partitioner.chType ).depth = partitioner.currTrDepth;
          } while( partitioner.nextPart( cs ) );

          partitioner.exitCurrSplit();
        }
        else
        cs.addTU( CS::getArea( cs, partitioner.currArea(), partitioner.chType ), partitioner.chType );
      }

      std::vector<TransformUnit*> orgTUs;

      if( lumaUsesISP )
      {
        CodingUnit& auxCU = saveCS.addCU( cu, partitioner.chType );
        auxCU.ispMode = cu.ispMode;
        saveCS.sps = cu.cs->sps;
        saveCS.addPU( *cu.firstPU, partitioner.chType );
      }


      // create a store for the TUs
      for( const auto &ptu : cs.tus )
      {
        // for split TUs in HEVC, add the TUs without Chroma parts for correct setting of Cbfs
        if( lumaUsesISP || pu.contains( *ptu, CHANNEL_TYPE_CHROMA ) )
        {
          saveCS.addTU( *ptu, partitioner.chType );
          orgTUs.push_back( ptu );
        }
      }
      if( lumaUsesISP )
      {
        saveCS.clearCUs();
      }
      // SATD pre-selecting.
      int satdModeList[NUM_CHROMA_MODE];
      int64_t satdSortedCost[NUM_CHROMA_MODE];
      for (int i = 0; i < NUM_CHROMA_MODE; i++)
      {
        satdSortedCost[i] = 0; // for the mode not pre-select by SATD, do RDO by default, so set the initial value 0.
        satdModeList[i] = 0;
      }
      bool modeIsEnable[NUM_INTRA_MODE + 1]; // use intra mode idx to check whether enable
      for (int i = 0; i < NUM_INTRA_MODE + 1; i++)
      {
        modeIsEnable[i] = 1;
      }

      DistParam distParam;
      const bool useHadamard = !cu.transQuantBypass;
      pu.intraDir[1] = MDLM_L_IDX; // temporary assigned, just to indicate this is a MDLM mode. for luma down-sampling operation.

      initIntraPatternChType(cu, pu.Cb());
      initIntraPatternChType(cu, pu.Cr());
      xGetLumaRecPixels(pu, pu.Cb());

      for (int idx = uiMinMode; idx <= uiMaxMode - 1; idx++)
      {
        int mode = chromaCandModes[idx];
        satdModeList[idx] = mode;
        if (PU::isLMCMode(mode) && !PU::isLMCModeEnabled(pu, mode))
        {
          continue;
        }
        if ((mode == LM_CHROMA_IDX) || (mode == PLANAR_IDX) || (mode == DM_CHROMA_IDX)) // only pre-check regular modes and MDLM modes, not including DM ,Planar, and LM
        {
          continue;
        }
        pu.intraDir[1] = mode; // temporary assigned, for SATD checking.

        int64_t sad = 0;
        CodingStructure& cs = *(pu.cs);

        CompArea areaCb = pu.Cb();
        PelBuf orgCb = cs.getOrgBuf(areaCb);
        PelBuf predCb = cs.getPredBuf(areaCb);

        m_pcRdCost->setDistParam(distParam, orgCb, predCb, pu.cs->sps->getBitDepth(CHANNEL_TYPE_CHROMA), COMPONENT_Cb, useHadamard);
        distParam.applyWeight = false;

        if (PU::isLMCMode(mode))
        {
          predIntraChromaLM(COMPONENT_Cb, predCb, pu, areaCb, mode);
        }
        else
        {
          initPredIntraParams(pu, pu.Cb(), *pu.cs->sps);
          predIntraAng(COMPONENT_Cb, predCb, pu);
        }

        sad += distParam.distFunc(distParam);

        CompArea areaCr = pu.Cr();
        PelBuf orgCr = cs.getOrgBuf(areaCr);
        PelBuf predCr = cs.getPredBuf(areaCr);

        m_pcRdCost->setDistParam(distParam, orgCr, predCr, pu.cs->sps->getBitDepth(CHANNEL_TYPE_CHROMA), COMPONENT_Cr, useHadamard);
        distParam.applyWeight = false;

        if (PU::isLMCMode(mode))
        {
          predIntraChromaLM(COMPONENT_Cr, predCr, pu, areaCr, mode);
        }
        else
        {
          initPredIntraParams(pu, pu.Cr(), *pu.cs->sps);
          predIntraAng(COMPONENT_Cr, predCr, pu);
        }
        sad += distParam.distFunc(distParam);
        satdSortedCost[idx] = sad;
      }
      // sort the mode based on the cost from small to large.
      int tempIdx = 0;
      int64_t tempCost = 0;
      for (int i = uiMinMode; i <= uiMaxMode - 1; i++)
      {
        for (int j = i + 1; j <= uiMaxMode - 1; j++)
        {
          if (satdSortedCost[j] < satdSortedCost[i])
          {
            tempIdx = satdModeList[i];
            satdModeList[i] = satdModeList[j];
            satdModeList[j] = tempIdx;

            tempCost = satdSortedCost[i];
            satdSortedCost[i] = satdSortedCost[j];
            satdSortedCost[j] = tempCost;

          }
        }
      }
      int reducedModeNumber = 2; // reduce the number of chroma modes
      for (int i = 0; i < reducedModeNumber; i++)
      {
        modeIsEnable[satdModeList[uiMaxMode - 1 - i]] = 0; // disable the last reducedModeNumber modes
      }

      // save the dist
      Distortion baseDist = cs.dist;

      for (uint32_t uiMode = uiMinMode; uiMode < uiMaxMode; uiMode++)
      {
        const int chromaIntraMode = chromaCandModes[uiMode];
        if( PU::isLMCMode( chromaIntraMode ) && ! PU::isLMCModeEnabled( pu, chromaIntraMode ) )
        {
          continue;
        }
        if (!modeIsEnable[chromaIntraMode] && PU::isLMCModeEnabled(pu, chromaIntraMode)) // when CCLM is disable, then MDLM is disable. not use satd checking
        {
          continue;
        }
        cs.setDecomp( pu.Cb(), false );
        cs.dist = baseDist;
        //----- restore context models -----
        m_CABACEstimator->getCtx() = ctxStart;

        //----- chroma coding -----
        pu.intraDir[1] = chromaIntraMode;

        xRecurIntraChromaCodingQT( cs, partitioner, bestCostSoFar, ispType );
        if( lumaUsesISP && cs.dist == MAX_UINT )
        {
          continue;
        }

#if JVET_O1136_TS_BDPCM_SIGNALLING
        if (cs.sps->getTransformSkipEnabledFlag())
#else
        if (cs.pps->getUseTransformSkip())
#endif
        {
          m_CABACEstimator->getCtx() = ctxStart;
        }

        uint64_t fracBits   = xGetIntraFracBitsQT( cs, partitioner, false, true, -1, ispType );
        Distortion uiDist = cs.dist;
        double    dCost   = m_pcRdCost->calcRdCost( fracBits, uiDist - baseDist );

        //----- compare -----
        if( dCost < dBestCost )
        {
          if( lumaUsesISP && dCost < bestCostSoFar )
          {
            bestCostSoFar = dCost;
          }
          for( uint32_t i = getFirstComponentOfChannel( CHANNEL_TYPE_CHROMA ); i < numberValidComponents; i++ )
          {
            const CompArea &area = pu.blocks[i];

            saveCS.getRecoBuf     ( area ).copyFrom( cs.getRecoBuf   ( area ) );
#if KEEP_PRED_AND_RESI_SIGNALS
            saveCS.getPredBuf     ( area ).copyFrom( cs.getPredBuf   ( area ) );
            saveCS.getResiBuf     ( area ).copyFrom( cs.getResiBuf   ( area ) );
#endif
            saveCS.getPredBuf     ( area ).copyFrom( cs.getPredBuf   (area ) );
            cs.picture->getPredBuf( area ).copyFrom( cs.getPredBuf   (area ) );
            cs.picture->getRecoBuf( area ).copyFrom( cs.getRecoBuf( area ) );

            for( uint32_t j = 0; j < saveCS.tus.size(); j++ )
            {
              saveCS.tus[j]->copyComponentFrom( *orgTUs[j], area.compID );
            }
          }

          dBestCost  = dCost;
          uiBestDist = uiDist;
          uiBestMode = chromaIntraMode;
        }
      }

      for( uint32_t i = getFirstComponentOfChannel( CHANNEL_TYPE_CHROMA ); i < numberValidComponents; i++ )
      {
        const CompArea &area = pu.blocks[i];

        cs.getRecoBuf         ( area ).copyFrom( saveCS.getRecoBuf( area ) );
#if KEEP_PRED_AND_RESI_SIGNALS
        cs.getPredBuf         ( area ).copyFrom( saveCS.getPredBuf( area ) );
        cs.getResiBuf         ( area ).copyFrom( saveCS.getResiBuf( area ) );
#endif
        cs.getPredBuf         ( area ).copyFrom( saveCS.getPredBuf( area ) );
        cs.picture->getPredBuf( area ).copyFrom( cs.getPredBuf    ( area ) );

        cs.picture->getRecoBuf( area ).copyFrom( cs.    getRecoBuf( area ) );

        for( uint32_t j = 0; j < saveCS.tus.size(); j++ )
        {
          orgTUs[ j ]->copyComponentFrom( *saveCS.tus[ j ], area.compID );
        }
      }
    }

    pu.intraDir[1] = uiBestMode;
    cs.dist        = uiBestDist;
  }

  //----- restore context models -----
  m_CABACEstimator->getCtx() = ctxStart;
  if( lumaUsesISP && bestCostSoFar >= maxCostAllowed )
  {
    cu.ispMode = 0;
  }
}

void IntraSearch::IPCMSearch(CodingStructure &cs, Partitioner& partitioner)
{
  ComponentID compStr = (CS::isDualITree(cs) && !isLuma(partitioner.chType)) ? COMPONENT_Cb: COMPONENT_Y;
  ComponentID compEnd = (CS::isDualITree(cs) && isLuma(partitioner.chType)) ? COMPONENT_Y : COMPONENT_Cr;
  for( ComponentID compID = compStr; compID <= compEnd; compID = ComponentID(compID+1) )
  {

    xEncPCM(cs, partitioner, compID);
  }

  cs.getPredBuf().fill(0);
  cs.getResiBuf().fill(0);
  cs.getOrgResiBuf().fill(0);

  cs.dist     = 0;
  cs.fracBits = 0;
  cs.cost     = 0;

  cs.setDecomp(cs.area);
  cs.picture->getPredBuf(cs.area).copyFrom(cs.getPredBuf());
}

void IntraSearch::xEncPCM(CodingStructure &cs, Partitioner& partitioner, const ComponentID &compID)
{
  TransformUnit &tu = *cs.getTU( partitioner.chType );

  const int  channelBitDepth = cs.sps->getBitDepth(toChannelType(compID));
  const uint32_t uiPCMBitDepth = cs.sps->getPCMBitDepth(toChannelType(compID));

  const int pcmShiftRight = (channelBitDepth - int(uiPCMBitDepth));

  CompArea  area    = tu.blocks[compID];
  PelBuf    pcmBuf  = tu.getPcmbuf  (compID);
  PelBuf    recBuf  = cs.getRecoBuf ( area );
  CPelBuf   orgBuf  = cs.getOrgBuf  ( area );

  CHECK(pcmShiftRight < 0, "Negative shift");
  CompArea      tmpArea(COMPONENT_Y, area.chromaFormat, Position(0, 0), area.size());
  PelBuf tempOrgBuf = m_tmpStorageLCU.getBuf(tmpArea);
  tempOrgBuf.copyFrom(orgBuf);
  if (cs.slice->getLmcsEnabledFlag() && m_pcReshape->getCTUFlag() && compID == COMPONENT_Y)
  {
    tempOrgBuf.rspSignal(m_pcReshape->getFwdLUT());
  }
  for (uint32_t uiY = 0; uiY < pcmBuf.height; uiY++)
  {
    for (uint32_t uiX = 0; uiX < pcmBuf.width; uiX++)
    {
      // Encode
      pcmBuf.at(uiX, uiY) = tempOrgBuf.at(uiX, uiY) >> pcmShiftRight;
      // Reconstruction
      recBuf.at(uiX, uiY) = pcmBuf.at(uiX, uiY) << pcmShiftRight;
    }
  }
}

// -------------------------------------------------------------------------------------------------------------------
// Intra search
// -------------------------------------------------------------------------------------------------------------------

void IntraSearch::xEncIntraHeader( CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma, const int subTuIdx )
{
  CodingUnit &cu = *cs.getCU( partitioner.chType );

  if (bLuma)
  {
    bool isFirst = cu.ispMode ? subTuIdx == 0 : partitioner.currArea().lumaPos() == cs.area.lumaPos();

    // CU header
    if( isFirst )
    {
      if ((!cs.slice->isIntra() || cs.slice->getSPS()->getIBCFlag())
        && cu.Y().valid()
        )
      {
        if( cs.pps->getTransquantBypassEnabledFlag() )
        {
          m_CABACEstimator->cu_transquant_bypass_flag( cu );
        }
        m_CABACEstimator->cu_skip_flag( cu );
        m_CABACEstimator->pred_mode   ( cu );
      }
      m_CABACEstimator->bdpcm_mode  ( cu, ComponentID(partitioner.chType) );
      if( CU::isIntra(cu) )
      {
        m_CABACEstimator->pcm_data( cu, partitioner );
        if( cu.ipcm )
        {
          return;
        }
      }
    }

    PredictionUnit &pu = *cs.getPU(partitioner.currArea().lumaPos(), partitioner.chType);

    // luma prediction mode
    if (isFirst)
    {
      if ( !cu.Y().valid())
        m_CABACEstimator->pred_mode( cu );
      m_CABACEstimator->intra_luma_pred_mode( pu );
    }
  }

  if (bChroma)
  {
    bool isFirst = partitioner.currArea().Cb().valid() && partitioner.currArea().chromaPos() == cs.area.chromaPos();

    PredictionUnit &pu = *cs.getPU( partitioner.currArea().chromaPos(), CHANNEL_TYPE_CHROMA );

    if( isFirst )
    {
      m_CABACEstimator->intra_chroma_pred_mode( pu );
    }
  }
}

void IntraSearch::xEncSubdivCbfQT( CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma, const int subTuIdx, const PartSplit ispType )
{
  const UnitArea &currArea = partitioner.currArea();
          int subTuCounter = subTuIdx;
  TransformUnit &currTU = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType, subTuCounter );
  CodingUnit    &currCU = *currTU.cu;
  uint32_t currDepth           = partitioner.currTrDepth;

  const bool subdiv        = currTU.depth > currDepth;
  ComponentID compID = partitioner.chType == CHANNEL_TYPE_LUMA ? COMPONENT_Y : COMPONENT_Cb;
  const bool chromaCbfISP = currArea.blocks[COMPONENT_Cb].valid() && currCU.ispMode && !subdiv;

  if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
  {
    CHECK( !subdiv, "TU split implied" );
  }
  else
  {
    CHECK( subdiv && !currCU.ispMode && isLuma( compID ), "No TU subdivision is allowed with QTBT" );
  }

  if( bChroma && ( !currCU.ispMode || chromaCbfISP ) )
  {
    const uint32_t numberValidComponents = getNumberValidComponents(currArea.chromaFormat);
    const uint32_t cbfDepth = ( chromaCbfISP ? currDepth - 1 : currDepth );

    for (uint32_t ch = COMPONENT_Cb; ch < numberValidComponents; ch++)
    {
      const ComponentID compID = ComponentID(ch);

      if( currDepth == 0 || TU::getCbfAtDepth( currTU, compID, currDepth - 1 ) || chromaCbfISP )
      {
        const bool prevCbf = ( compID == COMPONENT_Cr ? TU::getCbfAtDepth( currTU, COMPONENT_Cb, currDepth ) : false );
        m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, compID, currDepth ), currArea.blocks[compID], cbfDepth, prevCbf );

      }
    }
  }

  if (subdiv)
  {

    if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
    {
      partitioner.splitCurrArea( TU_MAX_TR_SPLIT, cs );
    }
    else if( currCU.ispMode && isLuma( compID ) )
    {
      partitioner.splitCurrArea( ispType, cs );
    }
    else
    THROW( "Cannot perform an implicit split!" );

    do
    {
      xEncSubdivCbfQT( cs, partitioner, bLuma, bChroma, subTuCounter, ispType );
      subTuCounter += subTuCounter != -1 ? 1 : 0;
    } while( partitioner.nextPart( cs ) );

    partitioner.exitCurrSplit();
  }
  else
  {
    //===== Cbfs =====
    if (bLuma)
    {
      bool previousCbf       = false;
      bool lastCbfIsInferred = false;
      if( ispType != TU_NO_ISP )
      {
        bool rootCbfSoFar = false;
        uint32_t nTus = currCU.ispMode == HOR_INTRA_SUBPARTITIONS ? currCU.lheight() >> g_aucLog2[currTU.lheight()] : currCU.lwidth() >> g_aucLog2[currTU.lwidth()];
        if( subTuCounter == nTus - 1 )
        {
          TransformUnit* tuPointer = currCU.firstTU;
          for( int tuIdx = 0; tuIdx < nTus - 1; tuIdx++ )
          {
            rootCbfSoFar |= TU::getCbfAtDepth( *tuPointer, COMPONENT_Y, currDepth );
            tuPointer = tuPointer->next;
          }
          if( !rootCbfSoFar )
          {
            lastCbfIsInferred = true;
          }
        }
        if( !lastCbfIsInferred )
        {
          previousCbf = TU::getPrevTuCbfAtDepth( currTU, COMPONENT_Y, partitioner.currTrDepth );
        }
      }
      if( !lastCbfIsInferred )
      {
        m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, COMPONENT_Y, currDepth ), currTU.Y(), currTU.depth, previousCbf, currCU.ispMode );
      }
    }
  }
}

void IntraSearch::xEncCoeffQT( CodingStructure &cs, Partitioner &partitioner, const ComponentID compID, const int subTuIdx, const PartSplit ispType )
{
  const UnitArea &currArea  = partitioner.currArea();

       int subTuCounter     = subTuIdx;
  TransformUnit &currTU     = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType, subTuIdx );
  uint32_t      currDepth       = partitioner.currTrDepth;
  const bool subdiv         = currTU.depth > currDepth;

  if (subdiv)
  {
    if (partitioner.canSplit(TU_MAX_TR_SPLIT, cs))
    {
      partitioner.splitCurrArea(TU_MAX_TR_SPLIT, cs);
    }
    else if( currTU.cu->ispMode )
    {
      partitioner.splitCurrArea( ispType, cs );
    }
    else
      THROW("Implicit TU split not available!");

    do
    {
      xEncCoeffQT( cs, partitioner, compID, subTuCounter, ispType );
      subTuCounter += subTuCounter != -1 ? 1 : 0;
    } while( partitioner.nextPart( cs ) );

    partitioner.exitCurrSplit();
  }
  else

  if( currArea.blocks[compID].valid() )
  {
#if JVET_O0105_ICT
    if( compID == COMPONENT_Cr )
    {
      const int cbfMask = ( TU::getCbf( currTU, COMPONENT_Cb ) ? 2 : 0 ) + ( TU::getCbf( currTU, COMPONENT_Cr ) ? 1 : 0 );
      m_CABACEstimator->joint_cb_cr( currTU, cbfMask );
    }
#endif
    if( TU::hasCrossCompPredInfo( currTU, compID ) )
    {
      m_CABACEstimator->cross_comp_pred( currTU, compID );
    }
    if( TU::getCbf( currTU, compID ) )
    {
      m_CABACEstimator->residual_coding( currTU, compID );
    }
  }
}

uint64_t IntraSearch::xGetIntraFracBitsQT( CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma, const int subTuIdx, const PartSplit ispType )
{
  m_CABACEstimator->resetBits();

  xEncIntraHeader( cs, partitioner, bLuma, bChroma, subTuIdx );
  xEncSubdivCbfQT( cs, partitioner, bLuma, bChroma, subTuIdx, ispType );


  if( bLuma )
  {
    xEncCoeffQT( cs, partitioner, COMPONENT_Y, subTuIdx, ispType );
  }
  if( bChroma )
  {
    xEncCoeffQT( cs, partitioner, COMPONENT_Cb, subTuIdx, ispType );
    xEncCoeffQT( cs, partitioner, COMPONENT_Cr, subTuIdx, ispType );
  }

  uint64_t fracBits = m_CABACEstimator->getEstFracBits();
  return fracBits;
}

uint64_t IntraSearch::xGetIntraFracBitsQTSingleChromaComponent( CodingStructure &cs, Partitioner &partitioner, const ComponentID compID )
{
  m_CABACEstimator->resetBits();

  if( compID == COMPONENT_Cb )
  {
    //intra mode coding
    PredictionUnit &pu = *cs.getPU( partitioner.currArea().lumaPos(), partitioner.chType );
    m_CABACEstimator->intra_chroma_pred_mode( pu );
    //xEncIntraHeader(cs, partitioner, false, true);
  }
  CHECK( partitioner.currTrDepth != 1, "error in the depth!" );
  const UnitArea &currArea = partitioner.currArea();

  TransformUnit &currTU = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType );

  //cbf coding
#if JVET_O0105_ICT
  const bool prevCbf = ( compID == COMPONENT_Cr ? TU::getCbfAtDepth( currTU, COMPONENT_Cb, partitioner.currTrDepth ) : false );
  m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, compID, partitioner.currTrDepth ), currArea.blocks[compID], partitioner.currTrDepth - 1, prevCbf );
#else
  m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, compID, partitioner.currTrDepth ), currArea.blocks[compID], partitioner.currTrDepth - 1 );
#endif
  //coeffs coding and cross comp coding
  if( TU::hasCrossCompPredInfo( currTU, compID ) )
  {
    m_CABACEstimator->cross_comp_pred( currTU, compID );
  }
  if( TU::getCbf( currTU, compID ) )
  {
    m_CABACEstimator->residual_coding( currTU, compID );
  }

  uint64_t fracBits = m_CABACEstimator->getEstFracBits();
  return fracBits;
}

uint64_t IntraSearch::xGetIntraFracBitsQTChroma(TransformUnit& currTU, const ComponentID &compID)
{
  m_CABACEstimator->resetBits();

  if( TU::hasCrossCompPredInfo( currTU, compID ) )
  {
    m_CABACEstimator->cross_comp_pred( currTU, compID );
  }

  // Include Cbf and jointCbCr flags here as we make decisions across components
  CodingStructure &cs = *currTU.cs;

  if ( currTU.jointCbCr )
  {
#if JVET_O0105_ICT
    const int cbfMask = ( TU::getCbf( currTU, COMPONENT_Cb ) ? 2 : 0 ) + ( TU::getCbf( currTU, COMPONENT_Cr ) ? 1 : 0 );
    m_CABACEstimator->cbf_comp( cs, cbfMask>>1, currTU.blocks[ COMPONENT_Cb ], currTU.depth, false );
    m_CABACEstimator->cbf_comp( cs, cbfMask &1, currTU.blocks[ COMPONENT_Cr ], currTU.depth, cbfMask>>1 );
    if( cbfMask )
      m_CABACEstimator->joint_cb_cr( currTU, cbfMask );
    if( cbfMask >> 1 )
      m_CABACEstimator->residual_coding( currTU, COMPONENT_Cb );
    if( cbfMask & 1 )
      m_CABACEstimator->residual_coding( currTU, COMPONENT_Cr );
#else
    if ( TU::getCbf( currTU, COMPONENT_Cb ) )
    {
      m_CABACEstimator->cbf_comp( cs, true, currTU.blocks[ COMPONENT_Cb ], currTU.depth, false );
      m_CABACEstimator->cbf_comp( cs, true, currTU.blocks[ COMPONENT_Cr ], currTU.depth, true );
      m_CABACEstimator->joint_cb_cr( currTU );
    }
    else
    {
      m_CABACEstimator->cbf_comp( cs, false, currTU.blocks[ COMPONENT_Cb ], currTU.depth, false );
      m_CABACEstimator->cbf_comp( cs, false, currTU.blocks[ COMPONENT_Cr ], currTU.depth, false );
    }
#endif
  }
  else
  {
    if ( compID == COMPONENT_Cb )
      m_CABACEstimator->cbf_comp( cs, TU::getCbf( currTU, compID ), currTU.blocks[ compID ], currTU.depth, false );
    else
#if JVET_O0105_ICT
    {
      const bool cbCbf    = TU::getCbf( currTU, COMPONENT_Cb );
      const bool crCbf    = TU::getCbf( currTU, compID );
      const int  cbfMask  = ( cbCbf ? 2 : 0 ) + ( crCbf ? 1 : 0 );
      m_CABACEstimator->cbf_comp( cs, crCbf, currTU.blocks[ compID ], currTU.depth, cbCbf );
      m_CABACEstimator->joint_cb_cr( currTU, cbfMask );
    }
#else
      m_CABACEstimator->cbf_comp( cs, TU::getCbf( currTU, compID ), currTU.blocks[ compID ], currTU.depth, TU::getCbf( currTU, COMPONENT_Cb ) );
#endif
  }

#if JVET_O0105_ICT
  if( !currTU.jointCbCr && TU::getCbf( currTU, compID ) )
#else
  if( TU::getCbf( currTU, compID ) )
#endif
  {
    m_CABACEstimator->residual_coding( currTU, compID );
  }

  uint64_t fracBits = m_CABACEstimator->getEstFracBits();
  return fracBits;
}

void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &compID, const bool &checkCrossCPrediction, Distortion& ruiDist, const int &default0Save1Load2, uint32_t* numSig, std::vector<TrMode>* trModes, const bool loadTr)
{
  if (!tu.blocks[compID].valid())
  {
    return;
  }

  CodingStructure &cs                       = *tu.cs;
  m_pcRdCost->setChromaFormat(cs.sps->getChromaFormatIdc());

  const CompArea      &area                 = tu.blocks[compID];
  const SPS           &sps                  = *cs.sps;
  const PPS           &pps                  = *cs.pps;

  const ChannelType    chType               = toChannelType(compID);
  const int            bitDepth             = sps.getBitDepth(chType);

  PelBuf         piOrg                      = cs.getOrgBuf    (area);
  PelBuf         piPred                     = cs.getPredBuf   (area);
  PelBuf         piResi                     = cs.getResiBuf   (area);
  PelBuf         piOrgResi                  = cs.getOrgResiBuf(area);
  PelBuf         piReco                     = cs.getRecoBuf   (area);

  const PredictionUnit &pu                  = *cs.getPU(area.pos(), chType);
  const uint32_t           uiChFinalMode        = PU::getFinalIntraMode(pu, chType);

  const bool           bUseCrossCPrediction = pps.getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() && isChroma( compID ) && PU::isChromaIntraModeCrossCheckMode( pu ) && checkCrossCPrediction;
  const bool           ccUseRecoResi        = m_pcEncCfg->getUseReconBasedCrossCPredictionEstimate();
  const bool           ispSplitIsAllowed    = sps.getUseISP() && CU::canUseISP( *tu.cu, compID );


  //===== init availability pattern =====
#if JVET_O0105_ICT
  CHECK( tu.jointCbCr && compID == COMPONENT_Cr, "wrong combination of compID and jointCbCr" );
#endif
  bool jointCbCr = tu.jointCbCr && compID == COMPONENT_Cb;

  if ( compID == COMPONENT_Y )
  {
  PelBuf sharedPredTS( m_pSharedPredTransformSkip[compID], area );
  if( default0Save1Load2 != 2 )
  {
#if JVET_O0106_ISP_4xN_PREDREG_FOR_1xN_2xN
    bool predRegDiffFromTB = CU::isPredRegDiffFromTB(*tu.cu, compID);
    bool firstTBInPredReg = CU::isFirstTBInPredReg(*tu.cu, compID, area);
    CompArea areaPredReg(COMPONENT_Y, tu.chromaFormat, area);
    if (predRegDiffFromTB)
    {
      if (firstTBInPredReg)
      {
        CU::adjustPredArea(areaPredReg);
        initIntraPatternChType(*tu.cu, areaPredReg);
      }
    }
    else
#endif
      initIntraPatternChType(*tu.cu, area);

    //===== get prediction signal =====
    if( compID != COMPONENT_Y && PU::isLMCMode( uiChFinalMode ) )
    {
      {
        xGetLumaRecPixels( pu, area );
      }
      predIntraChromaLM( compID, piPred, pu, area, uiChFinalMode );
    }
    else
    {
      if( PU::isMIP( pu, chType ) )
      {
        predIntraMip( compID, piPred, pu );
      }
      else
      {
#if JVET_O0106_ISP_4xN_PREDREG_FOR_1xN_2xN
        if (predRegDiffFromTB)
        {
          if (firstTBInPredReg)
          {
            PelBuf piPredReg = cs.getPredBuf(areaPredReg);
            predIntraAng(compID, piPredReg, pu);
          }
        }
        else
#endif
          predIntraAng(compID, piPred, pu);
      }
    }


    // save prediction
    if( default0Save1Load2 == 1 )
    {
      sharedPredTS.copyFrom( piPred );
    }
  }
  else
  {
    // load prediction
    piPred.copyFrom( sharedPredTS );
  }
  }


  DTRACE( g_trace_ctx, D_PRED, "@(%4d,%4d) [%2dx%2d] IMode=%d\n", tu.lx(), tu.ly(), tu.lwidth(), tu.lheight(), uiChFinalMode );
  //DTRACE_PEL_BUF( D_PRED, piPred, tu, tu.cu->predMode, COMPONENT_Y );

  const Slice           &slice = *cs.slice;
  bool flag = slice.getLmcsEnabledFlag() && (slice.isIntra() || (!slice.isIntra() && m_pcReshape->getCTUFlag()));
#if JVET_O0105_ICT
  if (isLuma(compID))
  {
#endif
  if (flag && slice.getLmcsChromaResidualScaleFlag() && isChroma(compID))
  {
    const Area area = tu.Y().valid() ? tu.Y() : Area(recalcPosition(tu.chromaFormat, tu.chType, CHANNEL_TYPE_LUMA, tu.blocks[tu.chType].pos()), recalcSize(tu.chromaFormat, tu.chType, CHANNEL_TYPE_LUMA, tu.blocks[tu.chType].size()));
    const CompArea &areaY = CompArea(COMPONENT_Y, tu.chromaFormat, area );
#if JVET_O1109_UNFIY_CRS
    int adj = m_pcReshape->calculateChromaAdjVpduNei(tu, areaY);
#else
    PelBuf piPredY;
    piPredY = cs.picture->getPredBuf(areaY);
    const Pel avgLuma = piPredY.computeAvg();
    int adj = m_pcReshape->calculateChromaAdj(avgLuma);
#endif
    tu.setChromaAdj(adj);
  }
  //===== get residual signal =====
  piResi.copyFrom( piOrg  );
  if (slice.getLmcsEnabledFlag() && m_pcReshape->getCTUFlag() && compID == COMPONENT_Y)
  {
    CompArea      tmpArea(COMPONENT_Y, area.chromaFormat, Position(0, 0), area.size());
    PelBuf tmpPred = m_tmpStorageLCU.getBuf(tmpArea);
    tmpPred.copyFrom(piPred);
    piResi.rspSignal(m_pcReshape->getFwdLUT());
    piResi.subtract(tmpPred);
  }
  else
  piResi.subtract( piPred );

  if (pps.getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() && isLuma(compID))
  {
    piOrgResi.copyFrom (piResi);
  }

  if (bUseCrossCPrediction)
  {
    if (xCalcCrossComponentPredictionAlpha(tu, compID, ccUseRecoResi) == 0)
    {
      return;
    }
    CrossComponentPrediction::crossComponentPrediction(tu, compID, cs.getResiBuf(tu.Y()), piResi, piResi, false);
  }
#if JVET_O0105_ICT
  }
#endif

  //===== transform and quantization =====
  //--- init rate estimation arrays for RDOQ ---
  //--- transform and quantization           ---
  TCoeff uiAbsSum = 0;

  const QpParam cQP(tu, compID);

#if RDOQ_CHROMA_LAMBDA
  m_pcTrQuant->selectLambda(compID);
#endif

  flag =flag && (tu.blocks[compID].width*tu.blocks[compID].height > 4);
  if (flag && isChroma(compID) && slice.getLmcsChromaResidualScaleFlag() )
  {
    int cResScaleInv = tu.getChromaAdj();
#if JVET_O0429_CRS_LAMBDA_FIX
    double cResScale = (double)(1 << CSCALE_FP_PREC) / (double)cResScaleInv;
#else
    double cResScale = round((double)(1 << CSCALE_FP_PREC) / (double)cResScaleInv);
#endif
    m_pcTrQuant->setLambda(m_pcTrQuant->getLambda() / (cResScale*cResScale));
#if !JVET_O0105_ICT
    if ( !jointCbCr ) // Joint CbCr signal is to be scaled in the case of joint chroma
    piResi.scaleSignal(cResScaleInv, 1, tu.cu->cs->slice->clpRng(compID));
#endif
  }

  const CompArea &crArea = tu.blocks     [ COMPONENT_Cr ];
  PelBuf          crOrg  = cs.getOrgBuf  ( crArea );
  PelBuf          crPred = cs.getPredBuf ( crArea );
  PelBuf          crResi = cs.getResiBuf ( crArea );
  PelBuf          crReco = cs.getRecoBuf ( crArea );

  if ( jointCbCr )
  {
#if !JVET_O0105_ICT
    // Get Cr prediction and residual
    crResi.copyFrom( crOrg  );
    crResi.subtract( crPred );

    // Create joint residual and store it for Cb component: jointResi = (cbResi - crResi)/2
    piResi.subtractAndHalve( crResi );

    // Scale the joint signal
    if ( flag && slice.getLmcsChromaResidualScaleFlag() )
      piResi.scaleSignal(tu.getChromaAdj(), 1, tu.cu->cs->slice->clpRng(compID));
#endif
    // Lambda is loosened for the joint mode with respect to single modes as the same residual is used for both chroma blocks
#if JVET_O0105_ICT
    const int    absIct = abs( TU::getICTMode(tu) );
    const double lfact  = ( absIct == 1 || absIct == 3 ? 0.8 : 0.5 );
    m_pcTrQuant->setLambda( lfact * m_pcTrQuant->getLambda() );
#else
    m_pcTrQuant->setLambda( 0.60 * m_pcTrQuant->getLambda() );
#endif
  }
#if JVET_O0105_ICT
  if( isChroma(compID) && tu.cu->cs->slice->getSliceQp() > 18 )
  {
    m_pcTrQuant->setLambda( 1.3 * m_pcTrQuant->getLambda() );
  }
#else
  else if ( isChroma(compID) && tu.cu->cs->slice->getSliceQp() > 18 )
    m_pcTrQuant->setLambda( 1.10 * m_pcTrQuant->getLambda() );
#endif

#if JVET_O0105_ICT