CABACWriter.cpp

    if( imv == 2 )//IMV_4PEL
    {
      CHECK( (horMvd % 4) != 0 && (verMvd % 4) != 0, "IMV: MVD is not a multiple of 16" );
      horMvd >>= 2;
      verMvd >>= 2;
    }
#endif
  }
  unsigned  horAbs  = unsigned( horMvd < 0 ? -horMvd : horMvd );
  unsigned  verAbs  = unsigned( verMvd < 0 ? -verMvd : verMvd );


  // abs_mvd_greater0_flag[ 0 | 1 ]
  m_BinEncoder.encodeBin( (horAbs > 0), Ctx::Mvd() );
  m_BinEncoder.encodeBin( (verAbs > 0), Ctx::Mvd() );

  // abs_mvd_greater1_flag[ 0 | 1 ]
  if( horAbs > 0 )
  {
    m_BinEncoder.encodeBin( (horAbs > 1), Ctx::Mvd(1) );
  }
  if( verAbs > 0 )
  {
    m_BinEncoder.encodeBin( (verAbs > 1), Ctx::Mvd(1) );
  }

  // abs_mvd_minus2[ 0 | 1 ] and mvd_sign_flag[ 0 | 1 ]
  if( horAbs > 0 )
  {
    if( horAbs > 1 )
    {
      exp_golomb_eqprob( horAbs - 2, 1 );
    }
    m_BinEncoder.encodeBinEP( (horMvd < 0) );
  }
  if( verAbs > 0 )
  {
    if( verAbs > 1 )
    {
      exp_golomb_eqprob( verAbs - 2, 1 );
    }
    m_BinEncoder.encodeBinEP( (verMvd < 0) );
  }
}




//================================================================================
//  clause 7.3.8.10
//--------------------------------------------------------------------------------
//    void  transform_unit      ( tu, cuCtx, chromaCbfs )
//    void  cu_qp_delta         ( cu )
//    void  cu_chroma_qp_offset ( cu )
//================================================================================
#if JVET_O0596_CBF_SIG_ALIGN_TO_SPEC
void CABACWriter::transform_unit( const TransformUnit& tu, CUCtx& cuCtx, Partitioner& partitioner, const int subTuCounter)
#else
void CABACWriter::transform_unit( const TransformUnit& tu, CUCtx& cuCtx, ChromaCbfs& chromaCbfs )
#endif
{
#if JVET_O0596_CBF_SIG_ALIGN_TO_SPEC
  const CodingStructure&  cs = *tu.cs;
  const CodingUnit&       cu = *tu.cu;
  const UnitArea&         area = partitioner.currArea();
  const unsigned          trDepth = partitioner.currTrDepth;
  const bool              chromaCbfISP = area.blocks[COMPONENT_Cb].valid() && cu.ispMode;
  ChromaCbfs              chromaCbfs;
  CHECK(tu.depth != trDepth, " transform unit should be not be futher partitioned");

  // cbf_cb & cbf_cr
#if JVET_O0050_LOCAL_DUAL_TREE
  if (area.chromaFormat != CHROMA_400 && area.blocks[COMPONENT_Cb].valid() && (!cu.isSepTree() || partitioner.chType == CHANNEL_TYPE_CHROMA) && (!cu.ispMode || chromaCbfISP))
#else
  if (area.chromaFormat != CHROMA_400 && area.blocks[COMPONENT_Cb].valid() && (!CS::isDualITree(cs) || partitioner.chType == CHANNEL_TYPE_CHROMA) && (!cu.ispMode || chromaCbfISP))
#endif
  {
    {
      unsigned cbfDepth = chromaCbfISP ? trDepth - 1 : trDepth;
      {
        chromaCbfs.Cb = TU::getCbfAtDepth(tu, COMPONENT_Cb, trDepth);
        //if (!(cu.sbtInfo && trDepth == 1))
        if (!(cu.sbtInfo && tu.noResidual))
          cbf_comp(cs, chromaCbfs.Cb, area.blocks[COMPONENT_Cb], cbfDepth);
      }

      {
        chromaCbfs.Cr = TU::getCbfAtDepth(tu, COMPONENT_Cr, trDepth);
        //if (!(cu.sbtInfo && trDepth == 1))
        if (!(cu.sbtInfo && tu.noResidual))
          cbf_comp(cs, chromaCbfs.Cr, area.blocks[COMPONENT_Cr], cbfDepth, chromaCbfs.Cb);
      }
    }
  }
#if JVET_O0050_LOCAL_DUAL_TREE
  else if (cu.isSepTree())
#else
  else if (CS::isDualITree(cs))
#endif
  {
    chromaCbfs = ChromaCbfs(false);
  }

  if (!isChroma(partitioner.chType))
  {
    if (!CU::isIntra(cu) && trDepth == 0 && !chromaCbfs.sigChroma(area.chromaFormat))
    {
      CHECK(!TU::getCbfAtDepth(tu, COMPONENT_Y, trDepth), "Luma cbf must be true for inter units with no chroma coeffs");
    }
    else if (cu.sbtInfo && tu.noResidual)
    {
      CHECK(TU::getCbfAtDepth(tu, COMPONENT_Y, trDepth), "Luma cbf must be false for inter sbt no-residual tu");
    }
    else if (cu.sbtInfo && !chromaCbfs.sigChroma(area.chromaFormat))
    {
      assert(!tu.noResidual);
      CHECK(!TU::getCbfAtDepth(tu, COMPONENT_Y, trDepth), "Luma cbf must be true for inter sbt residual tu");
    }
    else
    {
      bool previousCbf = false;
      bool rootCbfSoFar = false;
      bool lastCbfIsInferred = false;
      if (cu.ispMode)
      {
        uint32_t nTus = cu.ispMode == HOR_INTRA_SUBPARTITIONS ? cu.lheight() >> g_aucLog2[tu.lheight()] : cu.lwidth() >> g_aucLog2[tu.lwidth()];
        if (subTuCounter == nTus - 1)
        {
          TransformUnit* tuPointer = cu.firstTU;
          for (int tuIdx = 0; tuIdx < subTuCounter; tuIdx++)
          {
            rootCbfSoFar |= TU::getCbfAtDepth(*tuPointer, COMPONENT_Y, trDepth);
            tuPointer = tuPointer->next;
          }
          if (!rootCbfSoFar)
          {
            lastCbfIsInferred = true;
          }
        }
        if (!lastCbfIsInferred)
        {
          previousCbf = TU::getPrevTuCbfAtDepth(tu, COMPONENT_Y, partitioner.currTrDepth);
        }
      }
      if (!lastCbfIsInferred)
      {
        cbf_comp(cs, TU::getCbfAtDepth(tu, COMPONENT_Y, trDepth), tu.Y(), trDepth, previousCbf, cu.ispMode);
      }
    }
  }
#else
  CodingUnit& cu        = *tu.cu;
#endif
  bool        lumaOnly  = ( cu.chromaFormat == CHROMA_400 || !tu.blocks[COMPONENT_Cb].valid() );
  bool        cbf[3]    = { TU::getCbf( tu, COMPONENT_Y ), chromaCbfs.Cb, chromaCbfs.Cr };
  bool        cbfLuma   = ( cbf[ COMPONENT_Y ] != 0 );
  bool        cbfChroma = false;

  if( !lumaOnly )
  {
    if( tu.blocks[COMPONENT_Cb].valid() )
    {
      cbf   [ COMPONENT_Cb  ] = TU::getCbf( tu, COMPONENT_Cb );
      cbf   [ COMPONENT_Cr  ] = TU::getCbf( tu, COMPONENT_Cr );
    }
    cbfChroma = ( cbf[ COMPONENT_Cb ] || cbf[ COMPONENT_Cr ] );
  }

#if JVET_O0046_DQ_SIGNALLING
  if( ( cu.lwidth() > 64 || cu.lheight() > 64 || cbfLuma || cbfChroma ) &&
#else
  if( ( cbfLuma || cbfChroma ) &&
#endif
#if JVET_O0050_LOCAL_DUAL_TREE
    (!tu.cu->isSepTree() || isLuma(tu.chType)) )
#else
    (!CS::isDualITree(*tu.cs) || isLuma(tu.chType)) )
#endif
  {
    if( cu.cs->pps->getUseDQP() && !cuCtx.isDQPCoded )
    {
      cu_qp_delta(cu, cuCtx.qp, cu.qp);
      cuCtx.qp = cu.qp;
      cuCtx.isDQPCoded = true;
    }
  }
#if JVET_O1168_CU_CHROMA_QP_OFFSET
    if (cu.cs->slice->getUseChromaQpAdj() && cbfChroma && !cuCtx.isChromaQpAdjCoded)
#else
    if( cu.cs->slice->getUseChromaQpAdj() && cbfChroma && !cu.transQuantBypass && !cuCtx.isChromaQpAdjCoded )
#endif
    {
      cu_chroma_qp_offset( cu );
      cuCtx.isChromaQpAdjCoded = true;
    }

#if JVET_O0105_ICT
  if( !lumaOnly )
  {
    joint_cb_cr( tu, ( cbf[COMPONENT_Cb] ? 2 : 0 ) + ( cbf[COMPONENT_Cr] ? 1 : 0 ) );
  }
#endif

    if( cbfLuma )
    {
#if JVET_O0094_LFNST_ZERO_PRIM_COEFFS || JVET_O0472_LFNST_SIGNALLING_LAST_SCAN_POS
      residual_coding( tu, COMPONENT_Y, &cuCtx );
#else
      residual_coding( tu, COMPONENT_Y );
#endif
    }
    if( !lumaOnly )
    {
      for( ComponentID compID = COMPONENT_Cb; compID <= COMPONENT_Cr; compID = ComponentID( compID + 1 ) )
      {
        if( TU::hasCrossCompPredInfo( tu, compID ) )
        {
          cross_comp_pred( tu, compID );
        }
        if( cbf[ compID ] )
        {
#if JVET_O0094_LFNST_ZERO_PRIM_COEFFS || JVET_O0472_LFNST_SIGNALLING_LAST_SCAN_POS
          residual_coding( tu, compID, &cuCtx );
#else
          residual_coding( tu, compID );
#endif
      }
    }
  }
}

void CABACWriter::cu_qp_delta( const CodingUnit& cu, int predQP, const int8_t qp )
{
  CHECK(!( predQP != std::numeric_limits<int>::max()), "Unspecified error");
  int       DQp         = qp - predQP;
  int       qpBdOffsetY = cu.cs->sps->getQpBDOffset( CHANNEL_TYPE_LUMA );
  DQp                   = ( DQp + (MAX_QP + 1) + (MAX_QP + 1) / 2 + qpBdOffsetY + (qpBdOffsetY / 2)) % ((MAX_QP + 1) + qpBdOffsetY) - (MAX_QP + 1) / 2 - (qpBdOffsetY / 2);
  unsigned  absDQP      = unsigned( DQp < 0 ? -DQp : DQp );
  unsigned  unaryDQP    = std::min<unsigned>( absDQP, CU_DQP_TU_CMAX );

  unary_max_symbol( unaryDQP, Ctx::DeltaQP(), Ctx::DeltaQP(1), CU_DQP_TU_CMAX );
  if( absDQP >= CU_DQP_TU_CMAX )
  {
    exp_golomb_eqprob( absDQP - CU_DQP_TU_CMAX, CU_DQP_EG_k );
  }
  if( absDQP > 0 )
  {
    m_BinEncoder.encodeBinEP( DQp < 0 );
  }

  DTRACE_COND( ( isEncoding() ), g_trace_ctx, D_DQP, "x=%d, y=%d, d=%d, pred_qp=%d, DQp=%d, qp=%d\n", cu.blocks[cu.chType].lumaPos().x, cu.blocks[cu.chType].lumaPos().y, cu.qtDepth, predQP, DQp, qp );
}


void CABACWriter::cu_chroma_qp_offset( const CodingUnit& cu )
{
  // cu_chroma_qp_offset_flag
  unsigned qpAdj = cu.chromaQpAdj;
  if( qpAdj == 0 )
  {
    m_BinEncoder.encodeBin( 0, Ctx::ChromaQpAdjFlag() );
  }
  else
  {
    m_BinEncoder.encodeBin( 1, Ctx::ChromaQpAdjFlag() );
    int length = cu.cs->pps->getPpsRangeExtension().getChromaQpOffsetListLen();
    if( length > 1 )
    {
      unary_max_symbol( qpAdj-1, Ctx::ChromaQpAdjIdc(), Ctx::ChromaQpAdjIdc(), length-1 );
    }
  }
}





//================================================================================
//  clause 7.3.8.11
//--------------------------------------------------------------------------------
//    void        residual_coding         ( tu, compID )
//    void        transform_skip_flag     ( tu, compID )
//    void        explicit_rdpcm_mode     ( tu, compID )
//    void        last_sig_coeff          ( coeffCtx )
//    void        residual_coding_subblock( coeffCtx )
//================================================================================

#if JVET_O0105_ICT
void CABACWriter::joint_cb_cr( const TransformUnit& tu, const int cbfMask )
{
#if JVET_O0376_SPS_JOINTCBCR_FLAG
  if ( !tu.cu->slice->getSPS()->getJointCbCrEnabledFlag() )
  {
    return;
  }
#endif

  CHECK( tu.jointCbCr && tu.jointCbCr != cbfMask, "wrong value of jointCbCr (" << (int)tu.jointCbCr << " vs " << (int)cbfMask << ")" );
#if JVET_O0543_ICT_ICU_ONLY
  if( ( CU::isIntra( *tu.cu ) && cbfMask ) || ( cbfMask == 3 ) )
#else
  if( cbfMask )
#endif
  {
    m_BinEncoder.encodeBin( tu.jointCbCr ? 1 : 0, Ctx::JointCbCrFlag( cbfMask - 1 ) );
  }
}
#else
void CABACWriter::joint_cb_cr( const TransformUnit& tu )
{
  m_BinEncoder.encodeBin( tu.jointCbCr ? 1 : 0, Ctx::JointCbCrFlag( 0 ) );
}
#endif

#if JVET_O0094_LFNST_ZERO_PRIM_COEFFS || JVET_O0472_LFNST_SIGNALLING_LAST_SCAN_POS
void CABACWriter::residual_coding( const TransformUnit& tu, ComponentID compID, CUCtx* cuCtx )
#else
void CABACWriter::residual_coding( const TransformUnit& tu, ComponentID compID)
#endif
{
  const CodingUnit& cu = *tu.cu;
  DTRACE( g_trace_ctx, D_SYNTAX, "residual_coding() etype=%d pos=(%d,%d) size=%dx%d predMode=%d\n", tu.blocks[compID].compID, tu.blocks[compID].x, tu.blocks[compID].y, tu.blocks[compID].width, tu.blocks[compID].height, cu.predMode );

#if JVET_O0105_ICT
  if( compID == COMPONENT_Cr && tu.jointCbCr == 3 )
    return;
#else
  // Joint Cb-Cr residual mode is signalled if both Cb and Cr cbfs are true
  if ( compID == COMPONENT_Cr && TU::getCbf( tu, COMPONENT_Cb ) )
  {
    joint_cb_cr( tu );

    // No Cr residual in bitstream in joint Cb-Cr residual mode
    if ( tu.jointCbCr )
      return;
  }
#endif

  // code transform skip and explicit rdpcm mode
  mts_coding         ( tu, compID );
  explicit_rdpcm_mode( tu, compID );

  if( isLuma( compID ) && ( tu.mtsIdx == MTS_SKIP || tu.cu->bdpcmMode ) )
  {
    residual_codingTS( tu, compID );
    return;
  }

  // determine sign hiding
  bool signHiding  = ( cu.cs->slice->getSignDataHidingEnabledFlag() && !cu.transQuantBypass && tu.rdpcm[compID] == RDPCM_OFF );
  if(  signHiding && CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && tu.mtsIdx==MTS_SKIP )
  {
    const ChannelType chType    = toChannelType( compID );
    const unsigned    intraMode = PU::getFinalIntraMode( *cu.cs->getPU( tu.blocks[compID].pos(), chType ), chType );
    if( intraMode == HOR_IDX || intraMode == VER_IDX )
    {
      signHiding = false;
    }
  }

  // init coeff coding context
  CoeffCodingContext  cctx    ( tu, compID, signHiding );
  const TCoeff*       coeff   = tu.getCoeffs( compID ).buf;

  // determine and set last coeff position and sig group flags
  int                      scanPosLast = -1;
  std::bitset<MLS_GRP_NUM> sigGroupFlags;
  for( int scanPos = 0; scanPos < cctx.maxNumCoeff(); scanPos++)
  {
    unsigned blkPos = cctx.blockPos( scanPos );
    if( coeff[blkPos] )
    {
      scanPosLast = scanPos;
      sigGroupFlags.set( scanPos >> cctx.log2CGSize() );
    }
  }
  CHECK( scanPosLast < 0, "Coefficient coding called for empty TU" );
  cctx.setScanPosLast(scanPosLast);

#if JVET_O0094_LFNST_ZERO_PRIM_COEFFS
  if( cuCtx && tu.mtsIdx != MTS_SKIP && tu.blocks[ compID ].height >= 4 && tu.blocks[ compID ].width >= 4 )
  {
    const int maxLfnstPos = ((tu.blocks[compID].height == 4 && tu.blocks[compID].width == 4) || (tu.blocks[compID].height == 8 && tu.blocks[compID].width == 8)) ? 7 : 15;
    cuCtx->violatesLfnstConstrained[ toChannelType(compID) ] |= cctx.scanPosLast() > maxLfnstPos;
  }
#endif
#if JVET_O0472_LFNST_SIGNALLING_LAST_SCAN_POS
  if( cuCtx && tu.mtsIdx != MTS_SKIP && tu.blocks[ compID ].height >= 4 && tu.blocks[ compID ].width >= 4 )
  {
    const int lfnstLastScanPosTh = isLuma( compID ) ? LFNST_LAST_SIG_LUMA : LFNST_LAST_SIG_CHROMA;
    cuCtx->lfnstLastScanPos |= cctx.scanPosLast() >= lfnstLastScanPosTh;
  }
#endif
  // code last coeff position
  last_sig_coeff( cctx, tu, compID );

  // code subblocks
  const int stateTab  = ( tu.cs->slice->getDepQuantEnabledFlag() ? 32040 : 0 );
  int       state     = 0;

#if JVET_O0052_TU_LEVEL_CTX_CODED_BIN_CONSTRAINT
  int ctxBinSampleRatio = (compID == COMPONENT_Y) ? MAX_TU_LEVEL_CTX_CODED_BIN_CONSTRAINT_LUMA : MAX_TU_LEVEL_CTX_CODED_BIN_CONSTRAINT_CHROMA;
  cctx.regBinLimit = (tu.getTbAreaAfterCoefZeroOut(compID) * ctxBinSampleRatio) >> 4;
#endif

  for( int subSetId = ( cctx.scanPosLast() >> cctx.log2CGSize() ); subSetId >= 0; subSetId--)
  {
    cctx.initSubblock       ( subSetId, sigGroupFlags[subSetId] );
    if( ( tu.mtsIdx > MTS_SKIP || ( tu.cu->sbtInfo != 0 && tu.blocks[ compID ].height <= 32 && tu.blocks[ compID ].width <= 32 ) ) && !tu.cu->transQuantBypass && compID == COMPONENT_Y )
    {
      if( ( tu.blocks[ compID ].height == 32 && cctx.cgPosY() >= ( 16 >> cctx.log2CGHeight() ) )
       || ( tu.blocks[ compID ].width  == 32 && cctx.cgPosX() >= ( 16 >> cctx.log2CGWidth()  ) ) )
      {
        continue;
      }
    }
    residual_coding_subblock( cctx, coeff, stateTab, state );

  }


}

void CABACWriter::mts_coding( const TransformUnit& tu, ComponentID compID )
{
#if !JVET_O0294_TRANSFORM_CLEANUP
  const CodingUnit  &cu = *tu.cu;
#endif
  const bool  tsAllowed = TU::isTSAllowed ( tu, compID );
  const bool mtsAllowed = TU::isMTSAllowed( tu, compID );

  if( !mtsAllowed && !tsAllowed ) return;

  int symbol  = 0;
  int ctxIdx  = 0;

  if( tsAllowed )
  {
#if JVET_O0294_TRANSFORM_CLEANUP
    symbol = (tu.mtsIdx == MTS_SKIP) ? 1 : 0;
#else
    symbol = (tu.mtsIdx == MTS_SKIP) ? 0 : 1;
#endif
    ctxIdx = 6;
    m_BinEncoder.encodeBin( symbol, Ctx::MTSIndex( ctxIdx ) );
  }

  if( tu.mtsIdx != MTS_SKIP )
  {
    if( mtsAllowed )
    {
      symbol = tu.mtsIdx != MTS_DCT2_DCT2 ? 1 : 0;
#if JVET_O0294_TRANSFORM_CLEANUP
      ctxIdx = 0;
#else
      ctxIdx = std::min( (int)cu.qtDepth, 5 );
#endif
      m_BinEncoder.encodeBin( symbol, Ctx::MTSIndex( ctxIdx ) );

      if( symbol )
      {
        ctxIdx = 7;
        for( int i = 0; i < 3; i++, ctxIdx++ )
        {
          symbol = tu.mtsIdx > i + MTS_DST7_DST7 ? 1 : 0;
          m_BinEncoder.encodeBin( symbol, Ctx::MTSIndex( ctxIdx ) );

          if( !symbol )
          {
            break;
          }
        }
      }
    }
  }
#if JVET_O0294_TRANSFORM_CLEANUP
  DTRACE( g_trace_ctx, D_SYNTAX, "mts_coding() etype=%d pos=(%d,%d) mtsIdx=%d\n", COMPONENT_Y, tu.cu->lx(), tu.cu->ly(), tu.mtsIdx);
#else
  DTRACE( g_trace_ctx, D_SYNTAX, "mts_coding() etype=%d pos=(%d,%d) mtsIdx=%d\n", COMPONENT_Y, cu.lx(), cu.ly(), tu.mtsIdx );
#endif
}

void CABACWriter::isp_mode( const CodingUnit& cu )
{
  if( !CU::isIntra( cu ) || !isLuma( cu.chType ) || cu.firstPU->multiRefIdx || cu.ipcm || !cu.cs->sps->getUseISP() || cu.bdpcmMode || !CU::canUseISP( cu, getFirstComponentOfChannel( cu.chType ) ) )
  {
    CHECK( cu.ispMode != NOT_INTRA_SUBPARTITIONS, "cu.ispMode != 0" );
    return;
  }
  if ( cu.ispMode == NOT_INTRA_SUBPARTITIONS )
  {
    m_BinEncoder.encodeBin( 0, Ctx::ISPMode( 0 ) );
  }
  else
  {
    m_BinEncoder.encodeBin( 1, Ctx::ISPMode( 0 ) );
    m_BinEncoder.encodeBin( cu.ispMode - 1, Ctx::ISPMode( 1 ) );
  }
  DTRACE( g_trace_ctx, D_SYNTAX, "intra_subPartitions() etype=%d pos=(%d,%d) ispIdx=%d\n", cu.chType, cu.blocks[cu.chType].x, cu.blocks[cu.chType].y, (int)cu.ispMode );
}

void CABACWriter::explicit_rdpcm_mode( const TransformUnit& tu, ComponentID compID )
{
  const CodingUnit& cu = *tu.cu;
  if( !CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && ( tu.mtsIdx==MTS_SKIP || cu.transQuantBypass ) )
  {
    ChannelType chType = toChannelType( compID );
    switch( tu.rdpcm[compID] )
    {
    case RDPCM_VER:
      m_BinEncoder.encodeBin( 1, Ctx::RdpcmFlag(chType) );
      m_BinEncoder.encodeBin( 1, Ctx::RdpcmDir (chType) );
      break;
    case RDPCM_HOR:
      m_BinEncoder.encodeBin( 1, Ctx::RdpcmFlag(chType) );
      m_BinEncoder.encodeBin( 0, Ctx::RdpcmDir (chType) );
      break;
    default: // RDPCM_OFF
      m_BinEncoder.encodeBin( 0, Ctx::RdpcmFlag(chType) );
    }
  }
}

void CABACWriter::residual_lfnst_mode( const CodingUnit& cu, CUCtx& cuCtx )
{
#if JVET_O0213_RESTRICT_LFNST_TO_MAX_TB_SIZE
  int chIdx = CS::isDualITree( *cu.cs ) && cu.chType == CHANNEL_TYPE_CHROMA ? 1 : 0;
#endif
  if( cu.ispMode != NOT_INTRA_SUBPARTITIONS ||
#if JVET_O0925_MIP_SIMPLIFICATIONS
      (cu.cs->sps->getUseLFNST() && CU::isIntra(cu) && cu.mipFlag && !allowLfnstWithMip(cu.firstPU->lumaSize())) ||
#else
      cu.mipFlag == true ||
#endif
#if JVET_O0050_LOCAL_DUAL_TREE
    ( cu.isSepTree() && cu.chType == CHANNEL_TYPE_CHROMA && std::min( cu.blocks[ 1 ].width, cu.blocks[ 1 ].height ) < 4 )
#else
    ( CS::isDualITree( *cu.cs ) && cu.chType == CHANNEL_TYPE_CHROMA && std::min( cu.blocks[ 1 ].width, cu.blocks[ 1 ].height ) < 4 )
#endif
#if JVET_O0213_RESTRICT_LFNST_TO_MAX_TB_SIZE
#if JVET_O0545_MAX_TB_SIGNALLING
    || ( cu.blocks[ chIdx ].lumaSize().width > cu.cs->sps->getMaxTbSize() || cu.blocks[ chIdx ].lumaSize().height > cu.cs->sps->getMaxTbSize() )
#else
    || ( cu.blocks[ chIdx ].lumaSize().width > MAX_TB_SIZEY || cu.blocks[ chIdx ].lumaSize().height > MAX_TB_SIZEY )
#endif
#endif
    )
  {
    return;
  }

  if( cu.cs->sps->getUseLFNST() && CU::isIntra( cu ) && !CU::isLosslessCoded( cu ) )
  {
#if JVET_O0050_LOCAL_DUAL_TREE
    const bool lumaFlag                   = cu.isSepTree() ? (   isLuma( cu.chType ) ? true : false ) : true;
    const bool chromaFlag                 = cu.isSepTree() ? ( isChroma( cu.chType ) ? true : false ) : true;
#else
    const bool lumaFlag                   = CS::isDualITree( *cu.cs ) ? (   isLuma( cu.chType ) ? true : false ) : true;
    const bool chromaFlag                 = CS::isDualITree( *cu.cs ) ? ( isChroma( cu.chType ) ? true : false ) : true;
#endif
#if !JVET_O0472_LFNST_SIGNALLING_LAST_SCAN_POS
          bool nonZeroCoeffNonTs;
#endif
#if JVET_O0094_LFNST_ZERO_PRIM_COEFFS
          bool nonZeroCoeffNonTsCorner8x8 = ( lumaFlag && cuCtx.violatesLfnstConstrained[CHANNEL_TYPE_LUMA] ) || (chromaFlag && cuCtx.violatesLfnstConstrained[CHANNEL_TYPE_CHROMA] );
#else
          bool nonZeroCoeffNonTsCorner8x8 = CU::getNumNonZeroCoeffNonTsCorner8x8( cu, lumaFlag, chromaFlag ) > 0;
#endif
#if !JVET_O0472_LFNST_SIGNALLING_LAST_SCAN_POS
#if JVET_O0050_LOCAL_DUAL_TREE
    const int  nonZeroCoeffThr            = cu.isSepTree() ? ( isLuma( cu.chType ) ? LFNST_SIG_NZ_LUMA : LFNST_SIG_NZ_CHROMA ) : LFNST_SIG_NZ_LUMA + LFNST_SIG_NZ_CHROMA;
#else
    const int  nonZeroCoeffThr            = CS::isDualITree( *cu.cs ) ? ( isLuma( cu.chType ) ? LFNST_SIG_NZ_LUMA : LFNST_SIG_NZ_CHROMA ) : LFNST_SIG_NZ_LUMA + LFNST_SIG_NZ_CHROMA;
#endif
    cuCtx.numNonZeroCoeffNonTs            = CU::getNumNonZeroCoeffNonTs( cu, lumaFlag, chromaFlag );
    nonZeroCoeffNonTs                     = cuCtx.numNonZeroCoeffNonTs > nonZeroCoeffThr;
#endif
#if JVET_O0368_LFNST_WITH_DCT2_ONLY
    const bool isNonDCT2 = (TU::getCbf(*cu.firstTU, ComponentID(COMPONENT_Y)) && cu.firstTU->mtsIdx != MTS_DCT2_DCT2);
#if JVET_O0472_LFNST_SIGNALLING_LAST_SCAN_POS
    if( !cuCtx.lfnstLastScanPos || nonZeroCoeffNonTsCorner8x8 || isNonDCT2 )
#else
    if (!nonZeroCoeffNonTs || nonZeroCoeffNonTsCorner8x8 || isNonDCT2 )
#endif
#else
#if JVET_O0472_LFNST_SIGNALLING_LAST_SCAN_POS
    if( !cuCtx.lfnstLastScanPos || nonZeroCoeffNonTsCorner8x8 )
#else
    if( !nonZeroCoeffNonTs || nonZeroCoeffNonTsCorner8x8 )
#endif
#endif
    {
      return;
    }
  }
  else
  {
    return;
  }


  unsigned cctx = 0;
#if JVET_O0368_LFNST_WITH_DCT2_ONLY
#if JVET_O0050_LOCAL_DUAL_TREE
  if ( cu.isSepTree() ) cctx++;
#else
  if ( CS::isDualITree(*cu.cs) ) cctx++;
#endif
#else
#if JVET_O0545_MAX_TB_SIGNALLING
#if JVET_O0050_LOCAL_DUAL_TREE
  if( ( cu.firstTU->mtsIdx < MTS_DST7_DST7 || !TU::getCbf(*cu.firstTU, COMPONENT_Y) ) && cu.isSepTree() ) cctx++;
#else
  if( ( cu.firstTU->mtsIdx < MTS_DST7_DST7 || !TU::getCbf(*cu.firstTU, COMPONENT_Y) ) && CS::isDualITree( *cu.cs ) ) cctx++;
#endif
#else
#if JVET_O0050_LOCAL_DUAL_TREE
  if( cu.firstTU->mtsIdx < MTS_DST7_DST7 && cu.isSepTree() ) cctx++;
#else
  if( cu.firstTU->mtsIdx < MTS_DST7_DST7 && CS::isDualITree( *cu.cs ) ) cctx++;
#endif
#endif
#endif

  const uint32_t idxLFNST = cu.lfnstIdx;
  assert( idxLFNST < 3 );
  m_BinEncoder.encodeBin( idxLFNST ? 1 : 0, Ctx::LFNSTIdx( cctx ) );

  if( idxLFNST )
  {
    m_BinEncoder.encodeBinEP( ( idxLFNST - 1 ) ? 1 : 0 );
  }

  DTRACE( g_trace_ctx, D_SYNTAX, "residual_lfnst_mode() etype=%d pos=(%d,%d) mode=%d\n", COMPONENT_Y, cu.lx(), cu.ly(), ( int ) cu.lfnstIdx );
}

void CABACWriter::last_sig_coeff( CoeffCodingContext& cctx, const TransformUnit& tu, ComponentID compID )
{
  unsigned blkPos = cctx.blockPos( cctx.scanPosLast() );
  unsigned posX, posY;
  {
    posY  = blkPos / cctx.width();
    posX  = blkPos - ( posY * cctx.width() );
  }

  unsigned CtxLast;
  unsigned GroupIdxX = g_uiGroupIdx[ posX ];
  unsigned GroupIdxY = g_uiGroupIdx[ posY ];

  unsigned maxLastPosX = cctx.maxLastPosX();
  unsigned maxLastPosY = cctx.maxLastPosY();

  if( ( tu.mtsIdx > MTS_SKIP || ( tu.cu->sbtInfo != 0 && tu.blocks[ compID ].width <= 32 && tu.blocks[ compID ].height <= 32 ) ) && !tu.cu->transQuantBypass && compID == COMPONENT_Y )
  {
    maxLastPosX = ( tu.blocks[compID].width  == 32 ) ? g_uiGroupIdx[ 15 ] : maxLastPosX;
    maxLastPosY = ( tu.blocks[compID].height == 32 ) ? g_uiGroupIdx[ 15 ] : maxLastPosY;
  }

  for( CtxLast = 0; CtxLast < GroupIdxX; CtxLast++ )
  {
    m_BinEncoder.encodeBin( 1, cctx.lastXCtxId( CtxLast ) );
  }
  if( GroupIdxX < maxLastPosX )
  {
    m_BinEncoder.encodeBin( 0, cctx.lastXCtxId( CtxLast ) );
  }
  for( CtxLast = 0; CtxLast < GroupIdxY; CtxLast++ )
  {
    m_BinEncoder.encodeBin( 1, cctx.lastYCtxId( CtxLast ) );
  }
  if( GroupIdxY < maxLastPosY )
  {
    m_BinEncoder.encodeBin( 0, cctx.lastYCtxId( CtxLast ) );
  }
  if( GroupIdxX > 3 )
  {
    posX -= g_uiMinInGroup[ GroupIdxX ];
    for (int i = ( ( GroupIdxX - 2 ) >> 1 ) - 1 ; i >= 0; i-- )
    {
      m_BinEncoder.encodeBinEP( ( posX >> i ) & 1 );
    }
  }
  if( GroupIdxY > 3 )
  {
    posY -= g_uiMinInGroup[ GroupIdxY ];
    for ( int i = ( ( GroupIdxY - 2 ) >> 1 ) - 1 ; i >= 0; i-- )
    {
      m_BinEncoder.encodeBinEP( ( posY >> i ) & 1 );
    }
  }
}



void CABACWriter::residual_coding_subblock( CoeffCodingContext& cctx, const TCoeff* coeff, const int stateTransTable, int& state )
{
  //===== init =====
  const int   minSubPos   = cctx.minSubPos();
  const bool  isLast      = cctx.isLast();
  int         firstSigPos = ( isLast ? cctx.scanPosLast() : cctx.maxSubPos() );
  int         nextSigPos  = firstSigPos;

  //===== encode significant_coeffgroup_flag =====
  if( !isLast && cctx.isNotFirst() )
  {
    if( cctx.isSigGroup() )
    {
      m_BinEncoder.encodeBin( 1, cctx.sigGroupCtxId() );
    }
    else
    {
      m_BinEncoder.encodeBin( 0, cctx.sigGroupCtxId() );
      return;
    }
  }

  uint8_t   ctxOffset[16];

  //===== encode absolute values =====
  const int inferSigPos   = nextSigPos != cctx.scanPosLast() ? ( cctx.isNotFirst() ? minSubPos : -1 ) : nextSigPos;
  int       firstNZPos    = nextSigPos;
  int       lastNZPos     = -1;
  int       remAbsLevel   = -1;
  int       numNonZero    =  0;
  unsigned  signPattern   =  0;
#if JVET_O0052_TU_LEVEL_CTX_CODED_BIN_CONSTRAINT
  int       remRegBins    = cctx.regBinLimit;
#else
  bool      is2x2subblock = ( cctx.log2CGSize() == 2 );
  int       remRegBins    = ( is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK );
#endif
  int       firstPosMode2 = minSubPos - 1;

  for( ; nextSigPos >= minSubPos && remRegBins >= 4; nextSigPos-- )
  {
    TCoeff    Coeff      = coeff[ cctx.blockPos( nextSigPos ) ];
    unsigned  sigFlag    = ( Coeff != 0 );
    if( numNonZero || nextSigPos != inferSigPos )
    {
      const unsigned sigCtxId = cctx.sigCtxIdAbs( nextSigPos, coeff, state );
      m_BinEncoder.encodeBin( sigFlag, sigCtxId );
      DTRACE( g_trace_ctx, D_SYNTAX_RESI, "sig_bin() bin=%d ctx=%d\n", sigFlag, sigCtxId );
      remRegBins--;
    }
    else if( nextSigPos != cctx.scanPosLast() )
    {
      cctx.sigCtxIdAbs( nextSigPos, coeff, state ); // required for setting variables that are needed for gtx/par context selection
    }

    if( sigFlag )
    {
      uint8_t&  ctxOff  = ctxOffset[ nextSigPos - minSubPos ];
      ctxOff            = cctx.ctxOffsetAbs();
      numNonZero++;
      firstNZPos  = nextSigPos;
      lastNZPos   = std::max<int>( lastNZPos, nextSigPos );
      remAbsLevel = abs( Coeff ) - 1;

      if( nextSigPos != cctx.scanPosLast() ) signPattern <<= 1;
      if( Coeff < 0 )                        signPattern++;

      unsigned gt1 = !!remAbsLevel;
      m_BinEncoder.encodeBin( gt1, cctx.greater1CtxIdAbs(ctxOff) );
      DTRACE( g_trace_ctx, D_SYNTAX_RESI, "gt1_flag() bin=%d ctx=%d\n", gt1, cctx.greater1CtxIdAbs(ctxOff) );
      remRegBins--;

      if( gt1 )
      {
        remAbsLevel  -= 1;
        m_BinEncoder.encodeBin( remAbsLevel&1, cctx.parityCtxIdAbs( ctxOff ) );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "par_flag() bin=%d ctx=%d\n", remAbsLevel&1, cctx.parityCtxIdAbs( ctxOff ) );
        remAbsLevel >>= 1;

        remRegBins--;
        unsigned gt2 = !!remAbsLevel;
        m_BinEncoder.encodeBin(gt2, cctx.greater2CtxIdAbs(ctxOff));
        DTRACE(g_trace_ctx, D_SYNTAX_RESI, "gt2_flag() bin=%d ctx=%d\n", gt2, cctx.greater2CtxIdAbs(ctxOff));
        remRegBins--;
      }
    }

    state = ( stateTransTable >> ((state<<2)+((Coeff&1)<<1)) ) & 3;
  }
  firstPosMode2 = nextSigPos;
#if JVET_O0052_TU_LEVEL_CTX_CODED_BIN_CONSTRAINT
  cctx.regBinLimit = remRegBins;
#endif


  //===== 2nd PASS: Go-rice codes =====
  unsigned ricePar = 0;
  for( int scanPos = firstSigPos; scanPos > firstPosMode2; scanPos-- )
  {
    int       sumAll = cctx.templateAbsSum(scanPos, coeff, 4);
    ricePar = g_auiGoRiceParsCoeff[sumAll];
    unsigned absLevel = abs( coeff[ cctx.blockPos( scanPos ) ] );
    if( absLevel >= 4 )
    {
      unsigned rem      = ( absLevel - 4 ) >> 1;
      m_BinEncoder.encodeRemAbsEP( rem, ricePar, cctx.extPrec(), cctx.maxLog2TrDRange() );
      DTRACE( g_trace_ctx, D_SYNTAX_RESI, "rem_val() bin=%d ctx=%d\n", rem, ricePar );
    }
  }

  //===== coeff bypass ====
  for( int scanPos = firstPosMode2; scanPos >= minSubPos; scanPos-- )
  {
    TCoeff    Coeff     = coeff[ cctx.blockPos( scanPos ) ];
    unsigned  absLevel  = abs( Coeff );
    int       sumAll = cctx.templateAbsSum(scanPos, coeff, 0);
    int       rice      = g_auiGoRiceParsCoeff                        [sumAll];
    int       pos0      = g_auiGoRicePosCoeff0[std::max(0, state - 1)][sumAll];
    unsigned  rem       = ( absLevel == 0 ? pos0 : absLevel <= pos0 ? absLevel-1 : absLevel );
    m_BinEncoder.encodeRemAbsEP( rem, rice, cctx.extPrec(), cctx.maxLog2TrDRange() );
    DTRACE( g_trace_ctx, D_SYNTAX_RESI, "rem_val() bin=%d ctx=%d\n", rem, rice );
    state = ( stateTransTable >> ((state<<2)+((absLevel&1)<<1)) ) & 3;
    if( absLevel )
    {
      numNonZero++;
      firstNZPos = scanPos;
      lastNZPos   = std::max<int>( lastNZPos, scanPos );
      signPattern <<= 1;
      if( Coeff < 0 ) signPattern++;
    }
  }

  //===== encode sign's =====
  unsigned numSigns = numNonZero;
  if( cctx.hideSign( firstNZPos, lastNZPos ) )
  {
    numSigns    --;
    signPattern >>= 1;
  }
  m_BinEncoder.encodeBinsEP( signPattern, numSigns );
}

void CABACWriter::residual_codingTS( const TransformUnit& tu, ComponentID compID )
{
  DTRACE( g_trace_ctx, D_SYNTAX, "residual_codingTS() etype=%d pos=(%d,%d) size=%dx%d\n", tu.blocks[compID].compID, tu.blocks[compID].x, tu.blocks[compID].y, tu.blocks[compID].width, tu.blocks[compID].height );

  // init coeff coding context
  CoeffCodingContext  cctx    ( tu, compID, false, tu.cu->bdpcmMode );
  const TCoeff*       coeff   = tu.getCoeffs( compID ).buf;

  cctx.setNumCtxBins( 2 * tu.lwidth()*tu.lheight() );

  // determine and set last coeff position and sig group flags
  std::bitset<MLS_GRP_NUM> sigGroupFlags;
  for( int scanPos = 0; scanPos < cctx.maxNumCoeff(); scanPos++)
  {
    unsigned blkPos = cctx.blockPos( scanPos );
    if( coeff[blkPos] )
    {
      sigGroupFlags.set( scanPos >> cctx.log2CGSize() );
    }
  }

  // code subblocks
  for( int subSetId = 0; subSetId <= ( cctx.maxNumCoeff() - 1 ) >> cctx.log2CGSize(); subSetId++ )
  {
    cctx.initSubblock         ( subSetId, sigGroupFlags[subSetId] );
    residual_coding_subblockTS( cctx, coeff );
  }
}

void CABACWriter::residual_coding_subblockTS( CoeffCodingContext& cctx, const TCoeff* coeff )
{
  //===== init =====
  const int   minSubPos   = cctx.maxSubPos();
  int         firstSigPos = cctx.minSubPos();
  int         nextSigPos  = firstSigPos;

  //===== encode significant_coeffgroup_flag =====
  if( !cctx.isLastSubSet() || !cctx.only1stSigGroup() )
  {
    if( cctx.isSigGroup() )
    {
#if !JVET_O0409_EXCLUDE_CODED_SUB_BLK_FLAG_FROM_COUNT
      if( cctx.isContextCoded() )
      {
#endif
        m_BinEncoder.encodeBin( 1, cctx.sigGroupCtxId( true ) );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_sigGroup() bin=%d ctx=%d\n", 1, cctx.sigGroupCtxId() );
#if !JVET_O0409_EXCLUDE_CODED_SUB_BLK_FLAG_FROM_COUNT
      }
      else
      {
        m_BinEncoder.encodeBinEP( 1 );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_sigGroup() EPbin=%d\n", 1 );
      }
#endif
    }
    else
    {
#if !JVET_O0409_EXCLUDE_CODED_SUB_BLK_FLAG_FROM_COUNT
      if( cctx.isContextCoded() )
      {
#endif
        m_BinEncoder.encodeBin( 0, cctx.sigGroupCtxId( true ) );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_sigGroup() bin=%d ctx=%d\n", 0, cctx.sigGroupCtxId() );
#if !JVET_O0409_EXCLUDE_CODED_SUB_BLK_FLAG_FROM_COUNT
      }
      else
      {
        m_BinEncoder.encodeBinEP( 0 );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_sigGroup() EPbin=%d\n", 0 );
      }
#endif
      return;
    }
  }

  //===== encode absolute values =====
  const int inferSigPos   = minSubPos;
  int       remAbsLevel   = -1;
  int       numNonZero    =  0;

#if JVET_O0122_TS_SIGN_LEVEL
  int rightPixel, belowPixel, modAbsCoeff;
#endif

  for( ; nextSigPos <= minSubPos; nextSigPos++ )
  {
    TCoeff    Coeff      = coeff[ cctx.blockPos( nextSigPos ) ];
    unsigned  sigFlag    = ( Coeff != 0 );
    if( numNonZero || nextSigPos != inferSigPos )
    {
      if( cctx.isContextCoded() )
      {
        const unsigned sigCtxId = cctx.sigCtxIdAbsTS( nextSigPos, coeff );
        m_BinEncoder.encodeBin( sigFlag, sigCtxId );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_sig_bin() bin=%d ctx=%d\n", sigFlag, sigCtxId );
      }
      else
      {
        m_BinEncoder.encodeBinEP( sigFlag );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_sig_bin() EPbin=%d\n", sigFlag );
      }
    }

    if( sigFlag )
    {
      //===== encode sign's =====
      int sign = Coeff < 0;
      if( cctx.isContextCoded() )
      {
#if JVET_O0122_TS_SIGN_LEVEL
        const unsigned signCtxId = cctx.signCtxIdAbsTS(nextSigPos, coeff, cctx.bdpcm());
        m_BinEncoder.encodeBin(sign, signCtxId);
#else
        m_BinEncoder.encodeBin( sign, Ctx::TsResidualSign( cctx.bdpcm() ? 1 : 0 ) );
#endif
      }
      else
      {
        m_BinEncoder.encodeBinEP( sign );
      }
      numNonZero++;
#if JVET_O0122_TS_SIGN_LEVEL
      cctx.neighTS(rightPixel, belowPixel, nextSigPos, coeff);
      modAbsCoeff = cctx.deriveModCoeff(rightPixel, belowPixel, abs(Coeff), cctx.bdpcm());
      remAbsLevel = modAbsCoeff - 1;
#else
      remAbsLevel = abs( Coeff ) - 1;
#endif

      unsigned gt1 = !!remAbsLevel;
#if JVET_O0122_TS_SIGN_LEVEL
      const unsigned gt1CtxId = cctx.lrg1CtxIdAbsTS(nextSigPos, coeff, cctx.bdpcm());
      if (cctx.isContextCoded())
      {
        m_BinEncoder.encodeBin(gt1, gt1CtxId);
        DTRACE(g_trace_ctx, D_SYNTAX_RESI, "ts_gt1_flag() bin=%d ctx=%d\n", gt1, gt1CtxId);
      }
      else
      {
        m_BinEncoder.encodeBinEP(gt1);
        DTRACE(g_trace_ctx, D_SYNTAX_RESI, "ts_gt1_flag() EPbin=%d\n", gt1);
      }
#else
      if( cctx.isContextCoded() )
      {
        m_BinEncoder.encodeBin( gt1, cctx.greaterXCtxIdAbsTS(0) );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_gt1_flag() bin=%d ctx=%d\n", gt1, cctx.greaterXCtxIdAbsTS(0) );
      }
      else
      {
        m_BinEncoder.encodeBinEP( gt1 );
        DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_gt1_flag() EPbin=%d\n", gt1 );
      }
#endif

      if( gt1 )
      {
        remAbsLevel  -= 1;
        if( cctx.isContextCoded() )
        {
          m_BinEncoder.encodeBin( remAbsLevel&1, cctx.parityCtxIdAbsTS() );
          DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_par_flag() bin=%d ctx=%d\n", remAbsLevel&1, cctx.parityCtxIdAbsTS() );
        }