QuantRDOQ.cpp

    for ( ctxId = 0; ctxId < g_uiGroupIdx[dim2-1]; ctxId++)
    {
      const BinFracBits fB = fracBits.getFracBitsArray( cctx.lastYCtxId(ctxId) );
      lastBitsY[ ctxId ]   = bitsY + fB.intBits[ 0 ];
      bitsY               +=         fB.intBits[ 1 ];
    }
    lastBitsY[ctxId] = bitsY;
  }


  bool bFoundLast = false;
  for (int iCGScanPos = iCGLastScanPos; iCGScanPos >= 0; iCGScanPos--)
  {
    d64BaseCost -= pdCostCoeffGroupSig [ iCGScanPos ];
    if (cctx.isSigGroup( iCGScanPos ) )
    {
      uint32_t maxNonZeroPosInCG = iCGSizeM1;
      if( lfnstIdx > 0 && ( ( uiWidth == 4 && uiHeight == 4 ) || ( uiWidth == 8 && uiHeight == 8 && cctx.cgPosX() == 0 && cctx.cgPosY() == 0 ) ) )
      {
        maxNonZeroPosInCG = 7;
      }
      for( int iScanPosinCG = maxNonZeroPosInCG; iScanPosinCG >= 0; iScanPosinCG-- )
      {
        iScanPos = iCGScanPos * (iCGSizeM1 + 1) + iScanPosinCG;

        if (iScanPos > iLastScanPos)
        {
          continue;
        }
        uint32_t   uiBlkPos     = cctx.blockPos( iScanPos );

        if( piDstCoeff[ uiBlkPos ] )
        {
          uint32_t   uiPosY = uiBlkPos >> uiLog2BlockWidth;
          uint32_t   uiPosX = uiBlkPos - ( uiPosY << uiLog2BlockWidth );
          double d64CostLast  = xGetRateLast( lastBitsX, lastBitsY, uiPosX, uiPosY );

          double totalCost = d64BaseCost + d64CostLast - pdCostSig[ iScanPos ];

          if( totalCost < d64BestCost )
          {
            iBestLastIdxP1  = iScanPos + 1;
            d64BestCost     = totalCost;
          }
          if( piDstCoeff[ uiBlkPos ] > 1 )
          {
            bFoundLast = true;
            break;
          }
          d64BaseCost      -= pdCostCoeff[ iScanPos ];
          d64BaseCost      += pdCostCoeff0[ iScanPos ];
        }
        else
        {
          d64BaseCost      -= pdCostSig[ iScanPos ];
        }
      } //end for
      if (bFoundLast)
      {
        break;
      }
    } // end if (uiSigCoeffGroupFlag[ uiCGBlkPos ])
    DTRACE( g_trace_ctx, D_RDOQ_COST, "%d: %3d, %3d, %dx%d, comp=%d\n", DTRACE_GET_COUNTER( g_trace_ctx, D_RDOQ_COST ), rect.x, rect.y, rect.width, rect.height, compID );
    DTRACE( g_trace_ctx, D_RDOQ_COST, "Uncoded=%d\n", (int64_t)( d64BlockUncodedCost ) );
    DTRACE( g_trace_ctx, D_RDOQ_COST, "Coded  =%d\n", (int64_t)( d64BaseCost ) );

  } // end for


  for ( int scanPos = 0; scanPos < iBestLastIdxP1; scanPos++ )
  {
    int blkPos = cctx.blockPos( scanPos );
    TCoeff level = piDstCoeff[ blkPos ];
    uiAbsSum += level;
    piDstCoeff[ blkPos ] = ( plSrcCoeff[ blkPos ] < 0 ) ? -level : level;
  }

  //===== clean uncoded coefficients =====
  for ( int scanPos = iBestLastIdxP1; scanPos <= iLastScanPos; scanPos++ )
  {
    piDstCoeff[ cctx.blockPos( scanPos ) ] = 0;
  }

  if( cctx.signHiding() && uiAbsSum>=2)
  {
    const double inverseQuantScale = double(g_invQuantScales[0][cQP.rem(isTransformSkip)]);
    int64_t rdFactor = (int64_t)(inverseQuantScale * inverseQuantScale * (1 << (2 * cQP.per(isTransformSkip))) / m_dLambda / 16
                                  / (1 << (2 * DISTORTION_PRECISION_ADJUSTMENT(channelBitDepth)))
                             + 0.5);

    int lastCG = -1;
    int absSum = 0 ;
    int n ;
    for (int subSet = iCGNum - 1; subSet >= 0; subSet--)
    {
      int  subPos         = subSet << cctx.log2CGSize();
      int  firstNZPosInCG = iCGSizeM1 + 1, lastNZPosInCG = -1;
      absSum = 0 ;

      for( n = iCGSizeM1; n >= 0; --n )
      {
        if( piDstCoeff[ cctx.blockPos( n + subPos )] )
        {
          lastNZPosInCG = n;
          break;
        }
      }

      for( n = 0; n <= iCGSizeM1; n++ )
      {
        if( piDstCoeff[ cctx.blockPos( n + subPos )] )
        {
          firstNZPosInCG = n;
          break;
        }
      }

      for( n = firstNZPosInCG; n <= lastNZPosInCG; n++ )
      {
        absSum += int(piDstCoeff[ cctx.blockPos( n + subPos )]);
      }

      if(lastNZPosInCG>=0 && lastCG==-1)
      {
        lastCG = 1;
      }

      if( lastNZPosInCG-firstNZPosInCG>=SBH_THRESHOLD )
      {
        uint32_t signbit = (piDstCoeff[cctx.blockPos(subPos+firstNZPosInCG)]>0?0:1);
        if( signbit!=(absSum&0x1) )  // hide but need tune
        {
          // calculate the cost
          int64_t minCostInc = std::numeric_limits<int64_t>::max(), curCost = std::numeric_limits<int64_t>::max();
          int minPos = -1, finalChange = 0, curChange = 0;

          for( n = (lastCG == 1 ? lastNZPosInCG : iCGSizeM1); n >= 0; --n )
          {
            uint32_t uiBlkPos   = cctx.blockPos( n + subPos );
            if(piDstCoeff[ uiBlkPos ] != 0 )
            {
              int64_t costUp   = rdFactor * ( - deltaU[uiBlkPos] ) + rateIncUp[uiBlkPos];
              int64_t costDown = rdFactor * (   deltaU[uiBlkPos] ) + rateIncDown[uiBlkPos]
                -   ((abs(piDstCoeff[uiBlkPos]) == 1) ? sigRateDelta[uiBlkPos] : 0);

              if(lastCG==1 && lastNZPosInCG==n && abs(piDstCoeff[uiBlkPos])==1)
              {
                costDown -= (4<<SCALE_BITS);
              }

              if(costUp<costDown)
              {
                curCost = costUp;
                curChange =  1;
              }
              else
              {
                curChange = -1;
                if(n==firstNZPosInCG && abs(piDstCoeff[uiBlkPos])==1)
                {
                  curCost = std::numeric_limits<int64_t>::max();
                }
                else
                {
                  curCost = costDown;
                }
              }
            }
            else
            {
              curCost = rdFactor * ( - (abs(deltaU[uiBlkPos])) ) + (1<<SCALE_BITS) + rateIncUp[uiBlkPos] + sigRateDelta[uiBlkPos] ;
              curChange = 1 ;

              if(n<firstNZPosInCG)
              {
                uint32_t thissignbit = (plSrcCoeff[uiBlkPos]>=0?0:1);
                if(thissignbit != signbit )
                {
                  curCost = std::numeric_limits<int64_t>::max();
                }
              }
            }

            if( curCost<minCostInc)
            {
              minCostInc = curCost;
              finalChange = curChange;
              minPos = uiBlkPos;
            }
          }

          if(piDstCoeff[minPos] == entropyCodingMaximum || piDstCoeff[minPos] == entropyCodingMinimum)
          {
            finalChange = -1;
          }

          if(plSrcCoeff[minPos]>=0)
          {
            piDstCoeff[minPos] += finalChange ;
          }
          else
          {
            piDstCoeff[minPos] -= finalChange ;
          }
        }
      }

      if(lastCG==1)
      {
        lastCG=0 ;
      }
    }
  }
}

void QuantRDOQ::xRateDistOptQuantTS( TransformUnit &tu, const ComponentID &compID, const CCoeffBuf &coeffs, TCoeff &absSum, const QpParam &qp, const Ctx &ctx )
{
  const FracBitsAccess& fracBits = ctx.getFracBitsAcess();

  const SPS &sps            = *tu.cs->sps;
  const CompArea &rect      = tu.blocks[compID];
  const uint32_t width      = rect.width;
  const uint32_t height     = rect.height;
  const ChannelType chType  = toChannelType(compID);
  const int channelBitDepth = sps.getBitDepth( chType );

  const bool extendedPrecision     = sps.getSpsRangeExtension().getExtendedPrecisionProcessingFlag();
  const int  maxLog2TrDynamicRange = sps.getMaxLog2TrDynamicRange(chType);

  int transformShift = getTransformShift( channelBitDepth, rect.size(), maxLog2TrDynamicRange );

  if( extendedPrecision )
  {
    transformShift = std::max<int>( 0, transformShift );
  }

        double   blockUncodedCost                   = 0;
  const uint32_t maxNumCoeff                        = rect.area();

  CHECK( compID >= MAX_NUM_TBLOCKS, "Invalid component ID" );

  int scalingListType = getScalingListType( tu.cu->predMode, compID );
  CHECK( scalingListType >= SCALING_LIST_NUM, "Invalid scaling list" );

  const TCoeff *srcCoeff = coeffs.buf;
        TCoeff *dstCoeff = tu.getCoeffs( compID ).buf;

  double *costCoeff  = m_pdCostCoeff;
  double *costSig    = m_pdCostSig;
  double *costCoeff0 = m_pdCostCoeff0;

  memset( m_pdCostCoeff,  0, sizeof( double ) *  maxNumCoeff );
  memset( m_pdCostSig,    0, sizeof( double ) *  maxNumCoeff );

  m_bdpcm = 0;

  const bool   needsSqrt2Scale = TU::needsSqrt2Scale( tu, compID );  // should always be false - transform-skipped blocks don't require sqrt(2) compensation.
#if JVET_P0058_CHROMA_TS
  const bool   isTransformSkip = (tu.mtsIdx[compID] == MTS_SKIP);
#else
#if JVET_P0059_CHROMA_BDPCM
  const bool   isTransformSkip = (tu.mtsIdx == MTS_SKIP && isLuma(compID)) || ( tu.cu->bdpcmModeChroma && isChroma(compID) );
#else
  const bool   isTransformSkip = tu.mtsIdx==MTS_SKIP && isLuma(compID);
#endif
#endif
#if JVET_P1000_REMOVE_TRANFORMSHIFT_IN_TS_MODE
  const int    qBits = QUANT_SHIFT + qp.per(isTransformSkip) + (isTransformSkip ? 0 : transformShift) + (needsSqrt2Scale ? -1 : 0);  // Right shift of non-RDOQ quantizer;  level = (coeff*uiQ + offset)>>q_bits
#else
  const int    qBits = QUANT_SHIFT + qp.per(isTransformSkip) + transformShift + ( needsSqrt2Scale ? -1 : 0 );  // Right shift of non-RDOQ quantizer;  level = (coeff*uiQ + offset)>>q_bits
#endif
  const int    quantisationCoefficient = g_quantScales[needsSqrt2Scale?1:0][qp.rem(isTransformSkip)];
#if JVET_P1000_REMOVE_TRANFORMSHIFT_IN_TS_MODE
  const double errorScale              = xGetErrScaleCoeff( TU::needsSqrt2Scale(tu, compID), width, height, qp.rem(isTransformSkip), maxLog2TrDynamicRange, channelBitDepth, isTransformSkip);
#else
  const double errorScale              = xGetErrScaleCoeff( TU::needsSqrt2Scale( tu, compID ), width, height, qp.rem(isTransformSkip), maxLog2TrDynamicRange, channelBitDepth );
#endif

  const TCoeff entropyCodingMaximum = ( 1 << maxLog2TrDynamicRange ) - 1;

  uint32_t coeffLevels[3];
  double   coeffLevelError[4];

  CoeffCodingContext cctx( tu, compID, tu.cs->slice->getSignDataHidingEnabledFlag() );
  const int sbSizeM1    = ( 1 << cctx.log2CGSize() ) - 1;
  double    baseCost    = 0;
  uint32_t  goRiceParam = 0;

  double *costSigSubBlock = m_pdCostCoeffGroupSig;
  memset( costSigSubBlock, 0, ( maxNumCoeff >> cctx.log2CGSize() ) * sizeof( double ) );

  const int sbNum = width * height >> cctx.log2CGSize();
  int scanPos;
  coeffGroupRDStats rdStats;

  bool anySigCG = false;

#if JVET_P0072_SIMPLIFIED_TSRC
  int maxCtxBins = (cctx.maxNumCoeff() * 7) >> 2;
  cctx.setNumCtxBins(maxCtxBins);
#endif

  for( int sbId = 0; sbId < sbNum; sbId++ )
  {
    cctx.initSubblock( sbId );

    int noCoeffCoded = 0;
    baseCost = 0.0;
    memset( &rdStats, 0, sizeof (coeffGroupRDStats));

#if JVET_P0072_SIMPLIFIED_TSRC
    rdStats.iNumSbbCtxBins = 0;
#endif

    for( int scanPosInSB = 0; scanPosInSB <= sbSizeM1; scanPosInSB++ )
    {
      int lastPosCoded = sbSizeM1;
      scanPos = cctx.minSubPos() + scanPosInSB;
      //===== quantization =====
      uint32_t blkPos = cctx.blockPos( scanPos );

      // set coeff
      const int64_t          tmpLevel    = int64_t( abs( srcCoeff[blkPos] ) ) * quantisationCoefficient;
      const Intermediate_Int levelDouble = (Intermediate_Int)std::min<int64_t>( tmpLevel, std::numeric_limits<Intermediate_Int>::max() - ( Intermediate_Int( 1 ) << ( qBits - 1 ) ) );

      uint32_t roundAbsLevel = std::min<uint32_t>(uint32_t(entropyCodingMaximum), uint32_t((levelDouble + (Intermediate_Int(1) << (qBits - 1))) >> qBits));
      uint32_t minAbsLevel = (roundAbsLevel > 1 ? roundAbsLevel - 1 : 1);

      uint32_t downAbsLevel = std::min<uint32_t>(uint32_t(entropyCodingMaximum), uint32_t(levelDouble >> qBits));
      uint32_t upAbsLevel = std::min<uint32_t>(uint32_t(entropyCodingMaximum), downAbsLevel + 1);

      m_testedLevels = 0;
      coeffLevels[m_testedLevels++] = roundAbsLevel;

      if (minAbsLevel != roundAbsLevel)
        coeffLevels[m_testedLevels++] = minAbsLevel;

      int rightPixel, belowPixel, predPixel;

      cctx.neighTS(rightPixel, belowPixel, scanPos, dstCoeff);
      predPixel = cctx.deriveModCoeff(rightPixel, belowPixel, upAbsLevel, 0);

      if (upAbsLevel != roundAbsLevel && upAbsLevel != minAbsLevel && predPixel == 1)
        coeffLevels[m_testedLevels++] = upAbsLevel;

      double dErr = double(levelDouble);
      coeffLevelError[0] = dErr * dErr * errorScale;

      costCoeff0[scanPos] = coeffLevelError[0];
      blockUncodedCost   += costCoeff0[ scanPos ];
      dstCoeff[blkPos]    = coeffLevels[0];

      //===== coefficient level estimation =====
            unsigned    ctxIdSig = cctx.sigCtxIdAbsTS( scanPos, dstCoeff );
            uint32_t    cLevel;
      const BinFracBits fracBitsPar = fracBits.getFracBitsArray( cctx.parityCtxIdAbsTS() );

      goRiceParam = cctx.templateAbsSumTS( scanPos, dstCoeff );
      unsigned ctxIdSign = cctx.signCtxIdAbsTS(scanPos, dstCoeff, 0);
      const BinFracBits fracBitsSign = fracBits.getFracBitsArray(ctxIdSign);
      const uint8_t     sign         = srcCoeff[ blkPos ] < 0 ? 1 : 0;

      DTRACE_COND( ( coeffLevels[0] != 0 ), g_trace_ctx, D_RDOQ_MORE, " uiCtxSig=%d", ctxIdSig );

      unsigned gt1CtxId = cctx.lrg1CtxIdAbsTS(scanPos, dstCoeff, 0);
      const BinFracBits fracBitsGr1 = fracBits.getFracBitsArray(gt1CtxId);

      const BinFracBits fracBitsSig = fracBits.getFracBitsArray( ctxIdSig );
      bool lastCoeff = false; //
      if (scanPosInSB == lastPosCoded && noCoeffCoded == 0)
      {
        lastCoeff = true;
      }
#if JVET_P0072_SIMPLIFIED_TSRC
      int numUsedCtxBins = 0;
      cLevel = xGetCodedLevelTSPred(costCoeff[scanPos], costCoeff0[scanPos], costSig[scanPos], levelDouble, qBits, errorScale, coeffLevels, coeffLevelError,
                                    &fracBitsSig, fracBitsPar, cctx, fracBits, fracBitsSign, fracBitsGr1, sign, rightPixel, belowPixel, goRiceParam, lastCoeff, extendedPrecision, maxLog2TrDynamicRange, numUsedCtxBins);

      cctx.decimateNumCtxBins(numUsedCtxBins);
      rdStats.iNumSbbCtxBins += numUsedCtxBins;

#else
      cLevel = xGetCodedLevelTSPred( costCoeff[ scanPos ], costCoeff0[ scanPos ], costSig[ scanPos ], levelDouble, qBits, errorScale, coeffLevels, coeffLevelError,
                                    &fracBitsSig, fracBitsPar, cctx, fracBits, fracBitsSign, fracBitsGr1, sign, rightPixel, belowPixel, goRiceParam, lastCoeff, extendedPrecision, maxLog2TrDynamicRange);
#endif

      if (cLevel > 0)
      {
        noCoeffCoded++;
      }

      TCoeff level = cLevel;
      dstCoeff[blkPos] = (level != 0 && srcCoeff[blkPos] < 0) ? -level : level;
      baseCost           += costCoeff[ scanPos ];
      rdStats.d64SigCost += costSig[ scanPos ];

      if( dstCoeff[ blkPos ] )
      {
        cctx.setSigGroup();
        rdStats.d64CodedLevelandDist += costCoeff [ scanPos ] - costSig[ scanPos ];
        rdStats.d64UncodedDist       += costCoeff0[ scanPos ];
      }
    } //end for (iScanPosinCG)

    if( !cctx.isSigGroup() )
    {
      const BinFracBits fracBitsSigGroup = fracBits.getFracBitsArray( cctx.sigGroupCtxId( true ) );
      baseCost += xGetRateSigCoeffGroup( fracBitsSigGroup, 0 ) - rdStats.d64SigCost;
      costSigSubBlock[cctx.subSetId()] = xGetRateSigCoeffGroup( fracBitsSigGroup, 0 );
#if JVET_P0072_SIMPLIFIED_TSRC
      cctx.increaseNumCtxBins(rdStats.iNumSbbCtxBins); // skip sub-block
#endif
    }
    else if( sbId != sbNum - 1 || anySigCG )
    {
      // rd-cost if SigCoeffGroupFlag = 0, initialization
      double costZeroSB = baseCost;

      const BinFracBits fracBitsSigGroup = fracBits.getFracBitsArray( cctx.sigGroupCtxId( true ) );

      baseCost   += xGetRateSigCoeffGroup( fracBitsSigGroup, 1 );
      costZeroSB += xGetRateSigCoeffGroup( fracBitsSigGroup, 0 );
      costSigSubBlock[ cctx.subSetId() ] = xGetRateSigCoeffGroup( fracBitsSigGroup, 1 );

      costZeroSB += rdStats.d64UncodedDist;         // distortion for resetting non-zero levels to zero levels
      costZeroSB -= rdStats.d64CodedLevelandDist;   // distortion and level cost for keeping all non-zero levels
      costZeroSB -= rdStats.d64SigCost;             // sig cost for all coeffs, including zero levels and non-zerl levels

      if( costZeroSB < baseCost )
      {
        cctx.resetSigGroup();
        baseCost = costZeroSB;
        costSigSubBlock[ cctx.subSetId() ] = xGetRateSigCoeffGroup( fracBitsSigGroup, 0 );
#if JVET_P0072_SIMPLIFIED_TSRC
        cctx.increaseNumCtxBins(rdStats.iNumSbbCtxBins); // skip sub-block
#endif
        for( int scanPosInSB = 0; scanPosInSB <= sbSizeM1; scanPosInSB++ )
        {
          scanPos = cctx.minSubPos() + scanPosInSB;
          uint32_t blkPos = cctx.blockPos( scanPos );

          if( dstCoeff[ blkPos ] )
          {
            dstCoeff[ blkPos ] = 0;
            costCoeff[ scanPos ] = costCoeff0[ scanPos ];
            costSig[ scanPos] = 0;
          }
        }
      }
      else
      {
        anySigCG = true;
      }
    }
  }

  //===== estimate last position =====
  for( int scanPos = 0; scanPos < maxNumCoeff; scanPos++ )
  {
    int blkPos = cctx.blockPos( scanPos );
    TCoeff level = dstCoeff[ blkPos ];
    absSum += abs(level);
  }
}

void QuantRDOQ::forwardRDPCM( TransformUnit &tu, const ComponentID &compID, const CCoeffBuf &coeffs, TCoeff &absSum, const QpParam &qp, const Ctx &ctx )
{
  const FracBitsAccess& fracBits = ctx.getFracBitsAcess();

  const SPS &sps = *tu.cs->sps;
  const CompArea &rect = tu.blocks[compID];
  const uint32_t width = rect.width;
  const uint32_t height = rect.height;
  const ChannelType chType = toChannelType(compID);
  const int channelBitDepth = sps.getBitDepth(chType);

  const bool extendedPrecision = sps.getSpsRangeExtension().getExtendedPrecisionProcessingFlag();
  const int  maxLog2TrDynamicRange = sps.getMaxLog2TrDynamicRange(chType);
#if JVET_P0059_CHROMA_BDPCM
  const int  dirMode = isLuma(compID) ? tu.cu->bdpcmMode : tu.cu->bdpcmModeChroma;
#else
  const int  dirMode = tu.cu->bdpcmMode;
#endif
  int transformShift = getTransformShift(channelBitDepth, rect.size(), maxLog2TrDynamicRange);

  if (extendedPrecision)
  {
    transformShift = std::max<int>(0, transformShift);
  }

  double   blockUncodedCost = 0;
  const uint32_t maxNumCoeff = rect.area();

  CHECK(compID >= MAX_NUM_TBLOCKS, "Invalid component ID");

  int scalingListType = getScalingListType(tu.cu->predMode, compID);
  CHECK(scalingListType >= SCALING_LIST_NUM, "Invalid scaling list");

  const TCoeff *srcCoeff = coeffs.buf;
  TCoeff *dstCoeff = tu.getCoeffs(compID).buf;

  double *costCoeff = m_pdCostCoeff;
  double *costSig = m_pdCostSig;
  double *costCoeff0 = m_pdCostCoeff0;

  memset(m_pdCostCoeff, 0, sizeof(double) *  maxNumCoeff);
  memset(m_pdCostSig, 0, sizeof(double) *  maxNumCoeff);
  memset(m_fullCoeff, 0, sizeof(TCoeff) * maxNumCoeff);

  m_bdpcm = dirMode;

  const bool   needsSqrt2Scale = TU::needsSqrt2Scale(tu, compID);  // should always be false - transform-skipped blocks don't require sqrt(2) compensation.
#if JVET_P0058_CHROMA_TS
  const bool   isTransformSkip = (tu.mtsIdx[compID] == MTS_SKIP);
#else
#if JVET_P0059_CHROMA_BDPCM
  const bool   isTransformSkip = (tu.mtsIdx == MTS_SKIP && isLuma(compID)) || (tu.cu->bdpcmModeChroma && isChroma(compID) );
#else
  const bool   isTransformSkip = tu.mtsIdx==MTS_SKIP && isLuma(compID);
#endif
#endif
#if JVET_P1000_REMOVE_TRANFORMSHIFT_IN_TS_MODE
  const int    qBits = QUANT_SHIFT + qp.per(isTransformSkip) + (isTransformSkip? 0 : transformShift) + ( needsSqrt2Scale ? -1 : 0);  // Right shift of non-RDOQ quantizer;  level = (coeff*uiQ + offset)>>q_bits
#else
  const int    qBits = QUANT_SHIFT + qp.per(isTransformSkip) + transformShift + ( needsSqrt2Scale ? -1 : 0 );  // Right shift of non-RDOQ quantizer;  level = (coeff*uiQ + offset)>>q_bits
#endif
  const int    quantisationCoefficient = g_quantScales[needsSqrt2Scale ? 1 : 0][qp.rem(isTransformSkip)];
#if JVET_P1000_REMOVE_TRANFORMSHIFT_IN_TS_MODE
  const double errorScale = xGetErrScaleCoeff(TU::needsSqrt2Scale(tu, compID), width, height, qp.rem(isTransformSkip), maxLog2TrDynamicRange, channelBitDepth, isTransformSkip);
#else
  const double errorScale = xGetErrScaleCoeff(TU::needsSqrt2Scale(tu, compID), width, height, qp.rem(isTransformSkip), maxLog2TrDynamicRange, channelBitDepth);
#endif
  TrQuantParams trQuantParams;
#if JVET_P1000_REMOVE_TRANFORMSHIFT_IN_TS_MODE
  trQuantParams.rightShift = (IQUANT_SHIFT - ((isTransformSkip ? 0 : transformShift) + qp.per(isTransformSkip)));
#else
  trQuantParams.rightShift = (IQUANT_SHIFT - (transformShift + qp.per(isTransformSkip)));
#endif
  trQuantParams.qScale = g_invQuantScales[needsSqrt2Scale ? 1 : 0][qp.rem(isTransformSkip)];

  const TCoeff entropyCodingMaximum = (1 << maxLog2TrDynamicRange) - 1;

  uint32_t coeffLevels[3];
  double   coeffLevelError[4];

  CoeffCodingContext cctx(tu, compID, tu.cs->slice->getSignDataHidingEnabledFlag());
  const int sbSizeM1 = (1 << cctx.log2CGSize()) - 1;
  double    baseCost = 0;
  uint32_t  goRiceParam = 0;

  double *costSigSubBlock = m_pdCostCoeffGroupSig;
  memset(costSigSubBlock, 0, (maxNumCoeff >> cctx.log2CGSize()) * sizeof(double));

  const int sbNum = width * height >> cctx.log2CGSize();
  int scanPos;
  coeffGroupRDStats rdStats;

  bool anySigCG = false;

#if JVET_P0072_SIMPLIFIED_TSRC
  int maxCtxBins = (cctx.maxNumCoeff() * 7) >> 2;
  cctx.setNumCtxBins(maxCtxBins);
#endif

  for (int sbId = 0; sbId < sbNum; sbId++)
  {
    cctx.initSubblock(sbId);

    int noCoeffCoded = 0;
    baseCost = 0.0;
    memset(&rdStats, 0, sizeof(coeffGroupRDStats));
#if JVET_P0072_SIMPLIFIED_TSRC
    rdStats.iNumSbbCtxBins = 0;
#endif

    for (int scanPosInSB = 0; scanPosInSB <= sbSizeM1; scanPosInSB++)
    {
      int lastPosCoded = sbSizeM1;
      scanPos = cctx.minSubPos() + scanPosInSB;
      //===== quantization =====
      uint32_t blkPos = cctx.blockPos(scanPos);

      const int posX = cctx.posX(scanPos);
      const int posY = cctx.posY(scanPos);
      const int posS = (1 == dirMode) ? posX : posY;
      const int posNb = (1 == dirMode) ? (posX - 1) + posY * coeffs.stride : posX + (posY - 1) * coeffs.stride;
      TCoeff predCoeff = (0 != posS) ? m_fullCoeff[posNb] : 0;

      // set coeff
      const int64_t          tmpLevel = int64_t(abs(srcCoeff[blkPos] - predCoeff)) * quantisationCoefficient;
      const Intermediate_Int levelDouble = (Intermediate_Int)std::min<int64_t>(tmpLevel, std::numeric_limits<Intermediate_Int>::max() - (Intermediate_Int(1) << (qBits - 1)));
      uint32_t roundAbsLevel = std::min<uint32_t>(uint32_t(entropyCodingMaximum), uint32_t((levelDouble + (Intermediate_Int(1) << (qBits - 1))) >> qBits));
      uint32_t minAbsLevel = (roundAbsLevel > 1 ? roundAbsLevel - 1 : 1);

      m_testedLevels = 0;
      coeffLevels[m_testedLevels++] = roundAbsLevel;

      if (minAbsLevel != roundAbsLevel)
        coeffLevels[m_testedLevels++] = minAbsLevel;

      double dErr = double(levelDouble);
      coeffLevelError[0]  = dErr * dErr * errorScale;

      costCoeff0[scanPos] = coeffLevelError[0];
      blockUncodedCost   += costCoeff0[scanPos];
      dstCoeff[blkPos]    = coeffLevels[0];

      //===== coefficient level estimation =====
      unsigned    ctxIdSig = cctx.sigCtxIdAbsTS(scanPos, dstCoeff);
      uint32_t    cLevel;
      const BinFracBits fracBitsPar = fracBits.getFracBitsArray(cctx.parityCtxIdAbsTS());

      goRiceParam = cctx.templateAbsSumTS(scanPos, dstCoeff);
      unsigned ctxIdSign = cctx.signCtxIdAbsTS(scanPos, dstCoeff, dirMode);
      const BinFracBits fracBitsSign = fracBits.getFracBitsArray(ctxIdSign);
      const uint8_t     sign = srcCoeff[blkPos] - predCoeff < 0 ? 1 : 0;
      unsigned gt1CtxId = cctx.lrg1CtxIdAbsTS(scanPos, dstCoeff, dirMode);
      const BinFracBits fracBitsGr1 = fracBits.getFracBitsArray(gt1CtxId);

      DTRACE_COND((dstCoeff[blkPos] != 0), g_trace_ctx, D_RDOQ_MORE, " uiCtxSig=%d", ctxIdSig);

      const BinFracBits fracBitsSig = fracBits.getFracBitsArray(ctxIdSig);
      bool lastCoeff = false; //
      if (scanPosInSB == lastPosCoded && noCoeffCoded == 0)
      {
        lastCoeff = true;
      }
      int rightPixel, belowPixel;
      cctx.neighTS(rightPixel, belowPixel, scanPos, dstCoeff);
#if JVET_P0072_SIMPLIFIED_TSRC
      int numUsedCtxBins = 0;
      cLevel = xGetCodedLevelTSPred(costCoeff[scanPos], costCoeff0[scanPos], costSig[scanPos], levelDouble, qBits, errorScale, coeffLevels, coeffLevelError,
        &fracBitsSig, fracBitsPar, cctx, fracBits, fracBitsSign, fracBitsGr1, sign, rightPixel, belowPixel, goRiceParam, lastCoeff, extendedPrecision, maxLog2TrDynamicRange, numUsedCtxBins);
      cctx.decimateNumCtxBins(numUsedCtxBins);
      rdStats.iNumSbbCtxBins += numUsedCtxBins;
#else
      cLevel = xGetCodedLevelTSPred(costCoeff[scanPos], costCoeff0[scanPos], costSig[scanPos], levelDouble, qBits, errorScale, coeffLevels, coeffLevelError,
        &fracBitsSig, fracBitsPar, cctx, fracBits, fracBitsSign, fracBitsGr1, sign, rightPixel, belowPixel, goRiceParam, lastCoeff, extendedPrecision, maxLog2TrDynamicRange);
#endif
      if (cLevel > 0)
      {
        noCoeffCoded++;
      }
      dstCoeff[blkPos] = cLevel;

      if (sign)
      {
        dstCoeff[blkPos] = -dstCoeff[blkPos];
      }
      xDequantSample( m_fullCoeff[blkPos], dstCoeff[blkPos], trQuantParams );
      m_fullCoeff[blkPos] += predCoeff;

      baseCost += costCoeff[scanPos];
      rdStats.d64SigCost += costSig[scanPos];

      if (dstCoeff[blkPos])
      {
        cctx.setSigGroup();
        rdStats.d64CodedLevelandDist += costCoeff[scanPos] - costSig[scanPos];
        rdStats.d64UncodedDist += costCoeff0[scanPos];
      }
    } //end for (iScanPosinCG)

    if (!cctx.isSigGroup())
    {
      const BinFracBits fracBitsSigGroup = fracBits.getFracBitsArray(cctx.sigGroupCtxId(true));
      baseCost += xGetRateSigCoeffGroup(fracBitsSigGroup, 0) - rdStats.d64SigCost;
      costSigSubBlock[cctx.subSetId()] = xGetRateSigCoeffGroup(fracBitsSigGroup, 0);
#if JVET_P0072_SIMPLIFIED_TSRC
      cctx.increaseNumCtxBins(rdStats.iNumSbbCtxBins); // skip sub-block
#endif
    }
    else if (sbId != sbNum - 1 || anySigCG)
    {
      // rd-cost if SigCoeffGroupFlag = 0, initialization
      double costZeroSB = baseCost;

      const BinFracBits fracBitsSigGroup = fracBits.getFracBitsArray(cctx.sigGroupCtxId(true));

      baseCost += xGetRateSigCoeffGroup(fracBitsSigGroup, 1);
      costZeroSB += xGetRateSigCoeffGroup(fracBitsSigGroup, 0);
      costSigSubBlock[cctx.subSetId()] = xGetRateSigCoeffGroup(fracBitsSigGroup, 1);

      costZeroSB += rdStats.d64UncodedDist;         // distortion for resetting non-zero levels to zero levels
      costZeroSB -= rdStats.d64CodedLevelandDist;   // distortion and level cost for keeping all non-zero levels
      costZeroSB -= rdStats.d64SigCost;             // sig cost for all coeffs, including zero levels and non-zerl levels

      if (costZeroSB < baseCost)
      {
        cctx.resetSigGroup();
        baseCost = costZeroSB;
        costSigSubBlock[cctx.subSetId()] = xGetRateSigCoeffGroup(fracBitsSigGroup, 0);
#if JVET_P0072_SIMPLIFIED_TSRC
        cctx.increaseNumCtxBins(rdStats.iNumSbbCtxBins); // skip sub-block
#endif

        for (int scanPosInSB = 0; scanPosInSB <= sbSizeM1; scanPosInSB++)
        {
          scanPos = cctx.minSubPos() + scanPosInSB;
          uint32_t blkPos = cctx.blockPos(scanPos);

          const int posX = cctx.posX(scanPos);
          const int posY = cctx.posY(scanPos);
          const int posS = (1 == dirMode) ? posX : posY;
          const int posNb = (1 == dirMode) ? (posX - 1) + posY * coeffs.stride : posX + (posY - 1) * coeffs.stride;
          m_fullCoeff[scanPos] = (0 != posS) ? m_fullCoeff[posNb] : 0;

          if (dstCoeff[blkPos])
          {
            dstCoeff[blkPos] = 0;
            costCoeff[scanPos] = costCoeff0[scanPos];
            costSig[scanPos] = 0;
          }
        }
      }
      else
      {
        anySigCG = true;
      }
    }
  }

  //===== estimate last position =====
  for (int scanPos = 0; scanPos < maxNumCoeff; scanPos++)
  {
    int blkPos = cctx.blockPos(scanPos);
    TCoeff level = dstCoeff[blkPos];
    absSum += abs(level);
  }
}

void QuantRDOQ::xDequantSample(TCoeff& pRes, TCoeff& coeff, const TrQuantParams& trQuantParams)
{
  // xDequant
  if (trQuantParams.rightShift > 0)
  {
    const Intermediate_Int qAdd = Intermediate_Int(1) << (trQuantParams.rightShift - 1);
    pRes = TCoeff((Intermediate_Int(coeff) * trQuantParams.qScale + qAdd) >> trQuantParams.rightShift);
  }
  else
  {
    pRes = TCoeff((Intermediate_Int(coeff) * trQuantParams.qScale) << -trQuantParams.rightShift);
  }
}

inline uint32_t QuantRDOQ::xGetCodedLevelTSPred(double&            rd64CodedCost,
  double&            rd64CodedCost0,
  double&            rd64CodedCostSig,
  Intermediate_Int    levelDouble,
  int                 qBits,
  double              errorScale,
  uint32_t coeffLevels[],
  double coeffLevelError[],
  const BinFracBits* fracBitsSig,
  const BinFracBits& fracBitsPar,
  CoeffCodingContext& cctx,
  const FracBitsAccess& fracBitsAccess,
  const BinFracBits& fracBitsSign,
  const BinFracBits& fracBitsGt1,
  const uint8_t      sign,
  int                rightPixel,
  int                belowPixel,
  uint16_t           ricePar,
  bool               isLast,
  bool               useLimitedPrefixLength,
  const int          maxLog2TrDynamicRange
#if JVET_P0072_SIMPLIFIED_TSRC
  , int&               numUsedCtxBins
#endif
) const
{
  double currCostSig = 0;
  uint32_t   bestAbsLevel = 0;
#if JVET_P0072_SIMPLIFIED_TSRC
  numUsedCtxBins = 0;
  int numBestCtxBin = 0;
#endif
  if (!isLast && coeffLevels[0] < 3)
  {
#if JVET_P0072_SIMPLIFIED_TSRC
    if (cctx.numCtxBins() >= 4)
      rd64CodedCostSig = xGetRateSigCoef(*fracBitsSig, 0);
    else
      rd64CodedCostSig = xGetICost(1 << SCALE_BITS);
#else
    rd64CodedCostSig = xGetRateSigCoef(*fracBitsSig, 0);
#endif
    rd64CodedCost = rd64CodedCost0 + rd64CodedCostSig;
#if JVET_P0072_SIMPLIFIED_TSRC
    if (cctx.numCtxBins() >= 4)
      numUsedCtxBins++;
#endif
    if (coeffLevels[0] == 0)
    {
      return bestAbsLevel;
    }
  }
  else
  {
    rd64CodedCost = MAX_DOUBLE;
  }

  if (!isLast)
  {
#if JVET_P0072_SIMPLIFIED_TSRC
    if (cctx.numCtxBins() >= 4)
      currCostSig = xGetRateSigCoef(*fracBitsSig, 1);
    else
      currCostSig = xGetICost(1 << SCALE_BITS);
#else
    currCostSig = xGetRateSigCoef(*fracBitsSig, 1);
#endif
#if JVET_P0072_SIMPLIFIED_TSRC
    if (coeffLevels[0] >= 3 && cctx.numCtxBins() >= 4)
      numUsedCtxBins++;
#endif
  }

  for (int errorInd = 1; errorInd <= m_testedLevels; errorInd++)
  {
    int absLevel = coeffLevels[errorInd - 1];
    double dErr = 0.0;
    dErr = double(levelDouble - (Intermediate_Int(absLevel) << qBits));
    coeffLevelError[errorInd] = dErr * dErr * errorScale;
#if JVET_P0298_DISABLE_LEVELMAPPING_IN_BYPASS
    int modAbsLevel = absLevel;
    if (cctx.numCtxBins() >= 4) 
    {
      modAbsLevel = cctx.deriveModCoeff(rightPixel, belowPixel, absLevel, m_bdpcm);
    }
#else
    int modAbsLevel = cctx.deriveModCoeff(rightPixel, belowPixel, absLevel, m_bdpcm);
#endif
#if JVET_P0072_SIMPLIFIED_TSRC
    int numCtxBins = 0;
    double dCurrCost = coeffLevelError[errorInd] + xGetICost(xGetICRateTS(modAbsLevel, fracBitsPar, cctx, fracBitsAccess, fracBitsSign, fracBitsGt1, numCtxBins, sign, ricePar, useLimitedPrefixLength, maxLog2TrDynamicRange));
#else
    double dCurrCost = coeffLevelError[errorInd] + xGetICost(xGetICRateTS(modAbsLevel, fracBitsPar, cctx, fracBitsAccess, fracBitsSign, fracBitsGt1, sign, ricePar, useLimitedPrefixLength, maxLog2TrDynamicRange));
#endif

#if JVET_P0072_SIMPLIFIED_TSRC
    if (cctx.numCtxBins() >= 4)
      dCurrCost += currCostSig; // if cctx.numCtxBins < 4, xGetICRateTS return rate including sign cost. dont need to add any more
#else
    dCurrCost += currCostSig;
#endif

    if (dCurrCost < rd64CodedCost)
    {
      bestAbsLevel = absLevel;
      rd64CodedCost = dCurrCost;
      rd64CodedCostSig = currCostSig;
#if JVET_P0072_SIMPLIFIED_TSRC
      numBestCtxBin = numCtxBins;
#endif
    }
  }
#if JVET_P0072_SIMPLIFIED_TSRC
  numUsedCtxBins += numBestCtxBin;
#endif
  return bestAbsLevel;
}

inline int QuantRDOQ::xGetICRateTS( const uint32_t            absLevel,
                                    const BinFracBits&        fracBitsPar,
                                    const CoeffCodingContext& cctx,
                                    const FracBitsAccess&     fracBitsAccess,
                                    const BinFracBits&        fracBitsSign,
                                    const BinFracBits&        fracBitsGt1,
#if JVET_P0072_SIMPLIFIED_TSRC
                                    int&                      numCtxBins,
#endif
                                    const uint8_t             sign,
                                    const uint16_t            ricePar,
                                    const bool                useLimitedPrefixLength,
                                    const int                 maxLog2TrDynamicRange  ) const
{
 
  
#if JVET_P0072_SIMPLIFIED_TSRC
  if (cctx.numCtxBins() < 4) // Full by-pass coding 
  {
    int rate = absLevel ? (1 << SCALE_BITS) : 0; // 1 bit to signal sign of non-zero 

    uint32_t symbol = absLevel;

    uint32_t length;
    const int threshold = COEF_REMAIN_BIN_REDUCTION;
    if (symbol < (threshold << ricePar))
    {
      length = symbol >> ricePar;
      rate += (length + 1 + ricePar) << SCALE_BITS;
    }
    else if (useLimitedPrefixLength)
    {
      const uint32_t maximumPrefixLength = (32 - (COEF_REMAIN_BIN_REDUCTION + maxLog2TrDynamicRange));

      uint32_t prefixLength = 0;
      uint32_t suffix = (symbol >> ricePar) - COEF_REMAIN_BIN_REDUCTION;

      while ((prefixLength < maximumPrefixLength) && (suffix > ((2 << prefixLength) - 2)))
      {
        prefixLength++;
      }

      const uint32_t suffixLength = (prefixLength == maximumPrefixLength) ? (maxLog2TrDynamicRange - ricePar) : (prefixLength + 1/*separator*/);

      rate += (COEF_REMAIN_BIN_REDUCTION + prefixLength + suffixLength + ricePar) << SCALE_BITS;
    }
    else
    {
      length = ricePar;
      symbol = symbol - (threshold << ricePar);
      while (symbol >= (1 << length))
      {
        symbol -= (1 << (length++));
      }
      rate += (threshold + length + 1 - ricePar + length) << SCALE_BITS;
    }

    return rate;
  }

  else if (cctx.numCtxBins() >= 4 && cctx.numCtxBins() < 8) // First pass context coding and all by-pass coding ( Sign flag is not counted here)
  {
    int rate = fracBitsSign.intBits[sign]; // sign bits
    if (absLevel)
      numCtxBins++;

    if (absLevel > 1)
    {
      rate += fracBitsGt1.intBits[1];
      rate += fracBitsPar.intBits[(absLevel - 2) & 1];

      numCtxBins += 2;

      int cutoffVal = 2;

      if (absLevel >= cutoffVal)
      {
        uint32_t symbol = (absLevel - cutoffVal) >> 1;
        uint32_t length;
        const int threshold = COEF_REMAIN_BIN_REDUCTION;
        if (symbol < (threshold << ricePar))
        {
          length = symbol >> ricePar;
          rate += (length + 1 + ricePar) << SCALE_BITS;
        }
        else if (useLimitedPrefixLength)
        {
          const uint32_t maximumPrefixLength = (32 - (COEF_REMAIN_BIN_REDUCTION + maxLog2TrDynamicRange));

          uint32_t prefixLength = 0;
          uint32_t suffix = (symbol >> ricePar) - COEF_REMAIN_BIN_REDUCTION;

          while ((prefixLength < maximumPrefixLength) && (suffix > ((2 << prefixLength) - 2)))
          {
            prefixLength++;
          }

          const uint32_t suffixLength = (prefixLength == maximumPrefixLength) ? (maxLog2TrDynamicRange - ricePar) : (prefixLength + 1/*separator*/);

          rate += (COEF_REMAIN_BIN_REDUCTION + prefixLength + suffixLength + ricePar) << SCALE_BITS;
        }
        else
        {
          length = ricePar;
          symbol = symbol - (threshold << ricePar);
          while (symbol >= (1 << length))
          {
            symbol -= (1 << (length++));
          }
          rate += (threshold + length + 1 - ricePar + length) << SCALE_BITS;
        }
      }
    }
    else if (absLevel == 1)
    {
      rate += fracBitsGt1.intBits[0];
      numCtxBins++;
    }
    else
    {
      rate = 0;
    }
    return rate;

  }

#endif
  
  
  int rate = fracBitsSign.intBits[sign];

#if JVET_P0072_SIMPLIFIED_TSRC
  if (absLevel)
    numCtxBins++;
#endif

  if( absLevel > 1 )