IntraSearch.cpp

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

#if JVET_M0102_INTRA_SUBPARTITIONS
  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!" );
#endif

  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 JVET_M0102_INTRA_SUBPARTITIONS
      if( !CS::isDualITree( cs ) && cu.ispMode )
      {
        saveCS.clearCUs();
        saveCS.clearPUs();
      }
#endif

      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 JVET_M0102_INTRA_SUBPARTITIONS
      if( lumaUsesISP )
      {
        CodingUnit& auxCU = saveCS.addCU( cu, partitioner.chType );
        auxCU.ispMode = cu.ispMode;
        saveCS.sps = cu.cs->sps;
        saveCS.addPU( *cu.firstPU, partitioner.chType );
      }
#endif


      // 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 JVET_M0102_INTRA_SUBPARTITIONS
        if( lumaUsesISP || pu.contains( *ptu, CHANNEL_TYPE_CHROMA ) )
#else
        if( pu.contains( *ptu, CHANNEL_TYPE_CHROMA ) )
#endif
        {
          saveCS.addTU( *ptu, partitioner.chType );
          orgTUs.push_back( ptu );
        }
      }
#if JVET_M0102_INTRA_SUBPARTITIONS
      if( lumaUsesISP )
      {
        saveCS.clearCUs();
      }
#endif
      // 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 = true;
      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
        {
          predIntraAng(COMPONENT_Cb, predCb, pu, false);
        }

        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
        {
          predIntraAng(COMPONENT_Cr, predCr, pu, false);
        }
        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;

#if JVET_M0102_INTRA_SUBPARTITIONS
        xRecurIntraChromaCodingQT( cs, partitioner, bestCostSoFar, ispType );
        if( lumaUsesISP && cs.dist == MAX_UINT )
        {
          continue;
        }
#else
        xRecurIntraChromaCodingQT( cs, partitioner );
#endif

        if (cs.pps->getUseTransformSkip())
        {
          m_CABACEstimator->getCtx() = ctxStart;
        }

#if JVET_M0102_INTRA_SUBPARTITIONS
        uint64_t fracBits   = xGetIntraFracBitsQT( cs, partitioner, false, true, -1, ispType );
#else
        uint64_t fracBits   = xGetIntraFracBitsQT( cs, partitioner, false, true );
#endif
        Distortion uiDist = cs.dist;
        double    dCost   = m_pcRdCost->calcRdCost( fracBits, uiDist - baseDist );

        //----- compare -----
        if( dCost < dBestCost )
        {
#if JVET_M0102_INTRA_SUBPARTITIONS
          if( lumaUsesISP && dCost < bestCostSoFar )
          {
            bestCostSoFar = dCost;
          }
#endif
          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
#if JVET_M0427_INLOOP_RESHAPER
            saveCS.getPredBuf     ( area ).copyFrom( cs.getPredBuf   (area ) );
            cs.picture->getPredBuf( area ).copyFrom( cs.getPredBuf   (area ) );
#endif
            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
#if JVET_M0427_INLOOP_RESHAPER
        cs.getPredBuf         ( area ).copyFrom( saveCS.getPredBuf( area ) );
        cs.picture->getPredBuf( area ).copyFrom( cs.getPredBuf    ( area ) );
#endif

        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 JVET_M0102_INTRA_SUBPARTITIONS
  if( lumaUsesISP && bestCostSoFar >= maxCostAllowed )
  {
    cu.ispMode = 0;
  }
#endif
}

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);
#if JVET_M0427_INLOOP_RESHAPER
  cs.picture->getPredBuf(cs.area).copyFrom(cs.getPredBuf());
#endif
}

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");

  for (uint32_t uiY = 0; uiY < pcmBuf.height; uiY++)
  {
    for (uint32_t uiX = 0; uiX < pcmBuf.width; uiX++)
    {
      // Encode
      pcmBuf.at(uiX, uiY) = orgBuf.at(uiX, uiY) >> pcmShiftRight;
      // Reconstruction
      recBuf.at(uiX, uiY) = pcmBuf.at(uiX, uiY) << pcmShiftRight;
    }
  }
}

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

#if JVET_M0102_INTRA_SUBPARTITIONS
void IntraSearch::xEncIntraHeader( CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma, const int subTuIdx )
#else
void IntraSearch::xEncIntraHeader(CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma)
#endif
{
  CodingUnit &cu = *cs.getCU( partitioner.chType );

  if (bLuma)
  {
#if JVET_M0102_INTRA_SUBPARTITIONS
    bool isFirst = cu.ispMode ? subTuIdx == 0 : partitioner.currArea().lumaPos() == cs.area.lumaPos();
#else
    bool isFirst = partitioner.currArea().lumaPos() == cs.area.lumaPos();
#endif

    // CU header
    if( isFirst )
    {
#if JVET_M0483_IBC
      if ((!cs.slice->isIntra() || cs.slice->getSPS()->getIBCFlag())
#else
      if( !cs.slice->isIntra()
#endif
        && 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->extend_ref_line(cu);
#if JVET_M0102_INTRA_SUBPARTITIONS
      m_CABACEstimator->isp_mode      ( cu );
#endif
      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 );
    }
  }
}

#if JVET_M0102_INTRA_SUBPARTITIONS
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;
#else
void IntraSearch::xEncSubdivCbfQT(CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma)
{
  const UnitArea &currArea = partitioner.currArea();
  TransformUnit &currTU    = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType );
#if !JVET_M0464_UNI_MTS
  CodingUnit &currCU       = *currTU.cu;
#endif
#endif
  uint32_t currDepth           = partitioner.currTrDepth;

  const bool subdiv        = currTU.depth > currDepth;
#if JVET_M0102_INTRA_SUBPARTITIONS
  ComponentID compID = partitioner.chType == CHANNEL_TYPE_LUMA ? COMPONENT_Y : COMPONENT_Cb;
  const bool chromaCbfISP = currArea.blocks[COMPONENT_Cb].valid() && currCU.ispMode && !subdiv;
#endif

  if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
  {
    CHECK( !subdiv, "TU split implied" );
  }
  else
  {
#if JVET_M0102_INTRA_SUBPARTITIONS
    CHECK( subdiv && !currCU.ispMode && isLuma( compID ), "No TU subdivision is allowed with QTBT" );
  }

  if( bChroma && ( !currCU.ispMode || chromaCbfISP ) )
#else
    CHECK( subdiv, "No TU subdivision is allowed with QTBT" );
  }

  if (bChroma)
#endif
  {
    const uint32_t numberValidComponents = getNumberValidComponents(currArea.chromaFormat);
#if JVET_M0102_INTRA_SUBPARTITIONS
    const uint32_t cbfDepth = ( chromaCbfISP ? currDepth - 1 : currDepth );
#endif

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

#if JVET_M0102_INTRA_SUBPARTITIONS
      if( currDepth == 0 || TU::getCbfAtDepth( currTU, compID, currDepth - 1 ) || chromaCbfISP )
#else
      if( currDepth == 0 || TU::getCbfAtDepth( currTU, compID, currDepth - 1 ) )
#endif
      {
        const bool prevCbf = ( compID == COMPONENT_Cr ? TU::getCbfAtDepth( currTU, COMPONENT_Cb, currDepth ) : false );
#if JVET_M0102_INTRA_SUBPARTITIONS
        m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, compID, currDepth ), currArea.blocks[compID], cbfDepth, prevCbf );
#else
        m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, compID, currDepth ), currArea.blocks[compID], currDepth, prevCbf );
#endif

      }
    }
  }

  if (subdiv)
  {
#if !JVET_M0464_UNI_MTS
#if JVET_M0102_INTRA_SUBPARTITIONS
    if (!currCU.ispMode && isLuma( compID ) && currDepth == 0 && bLuma) m_CABACEstimator->emt_cu_flag( currCU );
#else
    if( currDepth == 0 && bLuma ) m_CABACEstimator->emt_cu_flag( currCU );
#endif
#endif

    if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
    {
      partitioner.splitCurrArea( TU_MAX_TR_SPLIT, cs );
    }
#if JVET_M0102_INTRA_SUBPARTITIONS
    else if( currCU.ispMode && isLuma( compID ) )
    {
      partitioner.splitCurrArea( ispType, cs );
    }
#endif
    else
    THROW( "Cannot perform an implicit split!" );

    do
    {
#if JVET_M0102_INTRA_SUBPARTITIONS
      xEncSubdivCbfQT( cs, partitioner, bLuma, bChroma, subTuCounter, ispType );
      subTuCounter += subTuCounter != -1 ? 1 : 0;
#else
      xEncSubdivCbfQT( cs, partitioner, bLuma, bChroma );
#endif
    } while( partitioner.nextPart( cs ) );

    partitioner.exitCurrSplit();
  }
  else
  {
#if !JVET_M0464_UNI_MTS
#if JVET_M0102_INTRA_SUBPARTITIONS
    if (!currCU.ispMode && isLuma( compID ) && currDepth == 0 && bLuma && TU::getCbfAtDepth( currTU, COMPONENT_Y, 0) ) m_CABACEstimator->emt_cu_flag( currCU );
#else
    if( currDepth == 0 && bLuma && TU::getCbfAtDepth( currTU, COMPONENT_Y, 0 ) ) m_CABACEstimator->emt_cu_flag( currCU );
#endif
#endif
    //===== Cbfs =====
    if (bLuma)
    {
#if JVET_M0102_INTRA_SUBPARTITIONS
      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 );
      }
#else
      m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, COMPONENT_Y, currDepth ), currTU.Y(), currTU.depth );
#endif
    }
  }
}

#if JVET_M0102_INTRA_SUBPARTITIONS
void IntraSearch::xEncCoeffQT( CodingStructure &cs, Partitioner &partitioner, const ComponentID compID, const int subTuIdx, const PartSplit ispType )
#else
void IntraSearch::xEncCoeffQT(CodingStructure &cs, Partitioner &partitioner, const ComponentID &compID)
#endif
{
  const UnitArea &currArea  = partitioner.currArea();

#if JVET_M0102_INTRA_SUBPARTITIONS
       int subTuCounter     = subTuIdx;
  TransformUnit &currTU     = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType, subTuIdx );
#else
  TransformUnit &currTU     = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType );
#endif
  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);
    }
#if JVET_M0102_INTRA_SUBPARTITIONS
    else if( currTU.cu->ispMode )
    {
      partitioner.splitCurrArea( ispType, cs );
    }
#endif
    else
      THROW("Implicit TU split not available!");

    do
    {
#if JVET_M0102_INTRA_SUBPARTITIONS
      xEncCoeffQT( cs, partitioner, compID, subTuCounter, ispType );
      subTuCounter += subTuCounter != -1 ? 1 : 0;
#else
      xEncCoeffQT( cs, partitioner, compID );
#endif
    } while( partitioner.nextPart( cs ) );

    partitioner.exitCurrSplit();
  }
  else

  if( currArea.blocks[compID].valid() )
  {
    if( TU::hasCrossCompPredInfo( currTU, compID ) )
    {
      m_CABACEstimator->cross_comp_pred( currTU, compID );
    }
    if( TU::getCbf( currTU, compID ) )
    {
      m_CABACEstimator->residual_coding( currTU, compID );
    }
  }
}

#if JVET_M0102_INTRA_SUBPARTITIONS
uint64_t IntraSearch::xGetIntraFracBitsQT( CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma, const int subTuIdx, const PartSplit ispType )
#else
uint64_t IntraSearch::xGetIntraFracBitsQT( CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma )
#endif
{
  m_CABACEstimator->resetBits();

#if JVET_M0102_INTRA_SUBPARTITIONS
  xEncIntraHeader( cs, partitioner, bLuma, bChroma, subTuIdx );
  xEncSubdivCbfQT( cs, partitioner, bLuma, bChroma, subTuIdx, ispType );
#else
  xEncIntraHeader( cs, partitioner, bLuma, bChroma );
  xEncSubdivCbfQT( cs, partitioner, bLuma, bChroma );
#endif


  if( bLuma )
  {
#if JVET_M0102_INTRA_SUBPARTITIONS
    xEncCoeffQT( cs, partitioner, COMPONENT_Y, subTuIdx, ispType );
#else
    xEncCoeffQT( cs, partitioner, COMPONENT_Y );
#endif
  }
  if( bChroma )
  {
#if JVET_M0102_INTRA_SUBPARTITIONS
    xEncCoeffQT( cs, partitioner, COMPONENT_Cb, subTuIdx, ispType );
    xEncCoeffQT( cs, partitioner, COMPONENT_Cr, subTuIdx, ispType );
#else
    xEncCoeffQT( cs, partitioner, COMPONENT_Cb );
    xEncCoeffQT( cs, partitioner, COMPONENT_Cr );
#endif
  }

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

#if JVET_M0102_INTRA_SUBPARTITIONS
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
  m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, compID, partitioner.currTrDepth ), currArea.blocks[compID], partitioner.currTrDepth - 1 );
  //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;
}
#endif

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

  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;
}

#if JVET_M0464_UNI_MTS
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)
#else
void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &compID, const bool &checkCrossCPrediction, Distortion& ruiDist, const int &default0Save1Load2, uint32_t* numSig )
#endif
{
  if (!tu.blocks[compID].valid())
  {
    return;
  }

  CodingStructure &cs                       = *tu.cs;

  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();
#if JVET_M0102_INTRA_SUBPARTITIONS
  const bool           ispSplitIsAllowed    = CU::canUseISPSplit( *tu.cu, compID );
#endif


  //===== init availability pattern =====
  PelBuf sharedPredTS( m_pSharedPredTransformSkip[compID], area );
  if( default0Save1Load2 != 2 )
  {
    const bool bUseFilteredPredictions = IntraPrediction::useFilteredIntraRefSamples( compID, pu, true, tu );
    initIntraPatternChType( *tu.cu, area, bUseFilteredPredictions );

    //===== get prediction signal =====
    if( compID != COMPONENT_Y && PU::isLMCMode( uiChFinalMode ) )
    {
      {
        xGetLumaRecPixels( pu, area );
      }
      predIntraChromaLM( compID, piPred, pu, area, uiChFinalMode );
    }
    else
    {
      predIntraAng( compID, piPred, pu, bUseFilteredPredictions );
    }


    // 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 );

#if JVET_M0427_INLOOP_RESHAPER
  const Slice           &slice = *cs.slice;
#if JVET_M0483_IBC
  bool flag = slice.getReshapeInfo().getUseSliceReshaper() && (slice.isIntra() || (!slice.isIntra() && m_pcReshape->getCTUFlag()));
#else
  bool flag = slice.getReshapeInfo().getUseSliceReshaper() && (slice.isIntra() || (!slice.isIntra() && m_pcReshape->getCTUFlag()) || (slice.getSliceType() == P_SLICE && slice.getSPS()->getIBCMode()));
#endif
  if (flag && slice.getReshapeInfo().getSliceReshapeChromaAdj() && 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 );
    PelBuf piPredY;
    piPredY = cs.picture->getPredBuf(areaY);
    const Pel avgLuma = piPredY.computeAvg();
    int adj = m_pcReshape->calculateChromaAdj(avgLuma);
    tu.setChromaAdj(adj);
  }
#endif
  //===== get residual signal =====
  piResi.copyFrom( piOrg  );
#if JVET_M0427_INLOOP_RESHAPER
  if (slice.getReshapeInfo().getUseSliceReshaper() && 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
#endif
  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);
  }

  //===== 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

#if JVET_M0427_INLOOP_RESHAPER
  flag =flag && (tu.blocks[compID].width*tu.blocks[compID].height > 4);
  if (flag && isChroma(compID) && slice.getReshapeInfo().getSliceReshapeChromaAdj() )
  {
    int cResScaleInv = tu.getChromaAdj();
    double cResScale = round((double)(1 << CSCALE_FP_PREC) / (double)cResScaleInv);
    m_pcTrQuant->setLambda(m_pcTrQuant->getLambda() / (cResScale*cResScale));
    piResi.scaleSignal(cResScaleInv, 1, tu.cu->cs->slice->clpRng(compID));
  }
#endif

#if JVET_M0102_INTRA_SUBPARTITIONS
  double diagRatio = 0, horVerRatio = 0;
#endif

#if JVET_M0464_UNI_MTS
  if( trModes )
  {
#if JVET_M0102_INTRA_SUBPARTITIONS
    m_pcTrQuant->transformNxN( tu, compID, cQP, trModes, CU::isIntra( *tu.cu ) ? m_pcEncCfg->getIntraMTSMaxCand() : m_pcEncCfg->getInterMTSMaxCand(), ispSplitIsAllowed ? &diagRatio : nullptr, ispSplitIsAllowed ? &horVerRatio : nullptr );
#else
    m_pcTrQuant->transformNxN( tu, compID, cQP, trModes, CU::isIntra( *tu.cu ) ? m_pcEncCfg->getIntraMTSMaxCand() : m_pcEncCfg->getInterMTSMaxCand() );
#endif
    tu.mtsIdx = trModes->at(0).first;
  }
#if JVET_M0102_INTRA_SUBPARTITIONS
  m_pcTrQuant->transformNxN( tu, compID, cQP, uiAbsSum, m_CABACEstimator->getCtx(), loadTr, &diagRatio, &horVerRatio );
#else
  m_pcTrQuant->transformNxN(tu, compID, cQP, uiAbsSum, m_CABACEstimator->getCtx(), loadTr);
#endif
#else
#if JVET_M0102_INTRA_SUBPARTITIONS
  m_pcTrQuant->transformNxN(tu, compID, cQP, uiAbsSum, m_CABACEstimator->getCtx(), ispSplitIsAllowed ? &diagRatio : nullptr, ispSplitIsAllowed ? &horVerRatio : nullptr);
#else
  m_pcTrQuant->transformNxN(tu, compID, cQP, uiAbsSum, m_CABACEstimator->getCtx());
#endif
#endif
#if JVET_M0102_INTRA_SUBPARTITIONS
  if (!tu.cu->ispMode && isLuma(compID) && ispSplitIsAllowed &&
#if JVET_M0464_UNI_MTS
    tu.mtsIdx == 0
#else
    !tu.cu->emtFlag
#endif
    )
  {
    m_intraModeDiagRatio        .push_back(diagRatio);
    m_intraModeHorVerRatio      .push_back(horVerRatio);
    m_intraModeTestedNormalIntra.push_back((int)uiChFinalMode);
  }
#endif


  DTRACE( g_trace_ctx, D_TU_ABS_SUM, "%d: comp=%d, abssum=%d\n", DTRACE_GET_COUNTER( g_trace_ctx, D_TU_ABS_SUM ), compID, uiAbsSum );


  //--- inverse transform ---
  if (uiAbsSum > 0)
  {
    m_pcTrQuant->invTransformNxN(tu, compID, piResi, cQP);
  }
  else
  {
    piResi.fill(0);
  }

  //===== reconstruction =====
#if JVET_M0427_INLOOP_RESHAPER
  if (flag && uiAbsSum > 0 && isChroma(compID) && slice.getReshapeInfo().getSliceReshapeChromaAdj() )
  {
    piResi.scaleSignal(tu.getChromaAdj(), 0, tu.cu->cs->slice->clpRng(compID));
  }
#endif
  if (bUseCrossCPrediction)
  {
    CrossComponentPrediction::crossComponentPrediction(tu, compID, cs.getResiBuf(tu.Y()), piResi, piResi, true);
  }

#if JVET_M0427_INLOOP_RESHAPER
  if (slice.getReshapeInfo().getUseSliceReshaper() && 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);
    piReco.reconstruct(tmpPred, piResi, cs.slice->clpRng(compID));
  }
  else
#endif
  piReco.reconstruct(piPred, piResi, cs.slice->clpRng( compID ));

  //===== update distortion =====
#if WCG_EXT
#if JVET_M0427_INLOOP_RESHAPER
  if (m_pcEncCfg->getLumaLevelToDeltaQPMapping().isEnabled() || (m_pcEncCfg->getReshaper()
    && slice.getReshapeInfo().getUseSliceReshaper() && (m_pcReshape->getCTUFlag() || (isChroma(compID) && m_pcEncCfg->getReshapeIntraCMD()))))
#else
  if( m_pcEncCfg->getLumaLevelToDeltaQPMapping().isEnabled() )
#endif
  {
    const CPelBuf orgLuma = cs.getOrgBuf( cs.area.blocks[COMPONENT_Y] );
#if JVET_M0427_INLOOP_RESHAPER
    if (compID == COMPONENT_Y)
    {
      CompArea      tmpArea1(COMPONENT_Y, area.chromaFormat, Position(0, 0), area.size());
      PelBuf tmpRecLuma = m_tmpStorageLCU.getBuf(tmpArea1);
      tmpRecLuma.copyFrom(piReco);
      tmpRecLuma.rspSignal(m_pcReshape->getInvLUT());
      ruiDist += m_pcRdCost->getDistPart(piOrg, tmpRecLuma, sps.getBitDepth(toChannelType(compID)), compID, DF_SSE_WTD, &orgLuma);
    }
    else
#endif
      ruiDist += m_pcRdCost->getDistPart(piOrg, piReco, bitDepth, compID, DF_SSE_WTD, &orgLuma);
  }
  else
#endif
  {
    ruiDist += m_pcRdCost->getDistPart( piOrg, piReco, bitDepth, compID, DF_SSE );
  }
}

#if JVET_M0102_INTRA_SUBPARTITIONS
void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &partitioner, const double bestCostSoFar, const int subTuIdx, const PartSplit ispType )
{
        int   subTuCounter = subTuIdx;
  const UnitArea &currArea = partitioner.currArea();
  const CodingUnit     &cu = *cs.getCU( currArea.lumaPos(), partitioner.chType );
        bool  earlySkipISP = false;
#else
void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &partitioner )
{
  const UnitArea &currArea = partitioner.currArea();