lzmaencoder.cpp

sleuthit-2.09 一个磁盘的工具集

    _isRepG1[i].Init();
    _isRepG2[i].Init();
  }

  _literalEncoder.Init();

  {
    for(UInt32 i = 0; i < kNumLenToPosStates; i++)
      _posSlotEncoder[i].Init();
  }
  {
    for(UInt32 i = 0; i < kNumFullDistances - kEndPosModelIndex; i++)
      _posEncoders[i].Init();
  }

  _lenEncoder.Init(1 << _posStateBits);
  _repMatchLenEncoder.Init(1 << _posStateBits);

  _posAlignEncoder.Init();

  _longestMatchWasFound = false;
  _optimumEndIndex = 0;
  _optimumCurrentIndex = 0;
  _additionalOffset = 0;

  return S_OK;
}

HRESULT CEncoder::MovePos(UInt32 num)
{
  if (num == 0)
    return S_OK;
  _additionalOffset += num;
  return _matchFinder->Skip(num);
}

UInt32 CEncoder::Backward(UInt32 &backRes, UInt32 cur)
{
  _optimumEndIndex = cur;
  UInt32 posMem = _optimum[cur].PosPrev;
  UInt32 backMem = _optimum[cur].BackPrev;
  do
  {
    if (_optimum[cur].Prev1IsChar)
    {
      _optimum[posMem].MakeAsChar();
      _optimum[posMem].PosPrev = posMem - 1;
      if (_optimum[cur].Prev2)
      {
        _optimum[posMem - 1].Prev1IsChar = false;
        _optimum[posMem - 1].PosPrev = _optimum[cur].PosPrev2;
        _optimum[posMem - 1].BackPrev = _optimum[cur].BackPrev2;
      }
    }
    UInt32 posPrev = posMem;
    UInt32 backCur = backMem;

    backMem = _optimum[posPrev].BackPrev;
    posMem = _optimum[posPrev].PosPrev;

    _optimum[posPrev].BackPrev = backCur;
    _optimum[posPrev].PosPrev = cur;
    cur = posPrev;
  }
  while(cur != 0);
  backRes = _optimum[0].BackPrev;
  _optimumCurrentIndex  = _optimum[0].PosPrev;
  return _optimumCurrentIndex; 
}

/*
Out:
  (lenRes == 1) && (backRes == 0xFFFFFFFF) means Literal
*/

HRESULT CEncoder::GetOptimum(UInt32 position, UInt32 &backRes, UInt32 &lenRes)
{
  if(_optimumEndIndex != _optimumCurrentIndex)
  {
    const COptimal &optimum = _optimum[_optimumCurrentIndex];
    lenRes = optimum.PosPrev - _optimumCurrentIndex;
    backRes = optimum.BackPrev;
    _optimumCurrentIndex = optimum.PosPrev;
    return S_OK;
  }
  _optimumCurrentIndex = _optimumEndIndex = 0;
  
  UInt32 lenMain, numDistancePairs;
  if (!_longestMatchWasFound)
  {
    RINOK(ReadMatchDistances(lenMain, numDistancePairs));
  }
  else
  {
    lenMain = _longestMatchLength;
    numDistancePairs = _numDistancePairs;
    _longestMatchWasFound = false;
  }

  const Byte *data = _matchFinder->GetPointerToCurrentPos() - 1;
  UInt32 numAvailableBytes = _matchFinder->GetNumAvailableBytes() + 1;
  if (numAvailableBytes < 2)
  {
    backRes = (UInt32)(-1);
    lenRes = 1;
    return S_OK;
  }
  if (numAvailableBytes > kMatchMaxLen)
    numAvailableBytes = kMatchMaxLen;

  UInt32 reps[kNumRepDistances];
  UInt32 repLens[kNumRepDistances];
  UInt32 repMaxIndex = 0;
  UInt32 i;
  for(i = 0; i < kNumRepDistances; i++)
  {
    reps[i] = _repDistances[i];
    UInt32 backOffset = reps[i] + 1;
    if (data[0] != data[(size_t)0 - backOffset] || data[1] != data[(size_t)1 - backOffset])
    {
      repLens[i] = 0;
      continue;
    }
    UInt32 lenTest;
    for (lenTest = 2; lenTest < numAvailableBytes && 
        data[lenTest] == data[(size_t)lenTest - backOffset]; lenTest++);
    repLens[i] = lenTest;
    if (lenTest > repLens[repMaxIndex])
      repMaxIndex = i;
  }
  if(repLens[repMaxIndex] >= _numFastBytes)
  {
    backRes = repMaxIndex;
    lenRes = repLens[repMaxIndex];
    return MovePos(lenRes - 1);
  }

  UInt32 *matchDistances = _matchDistances + 1;
  if(lenMain >= _numFastBytes)
  {
    backRes = matchDistances[numDistancePairs - 1] + kNumRepDistances; 
    lenRes = lenMain;
    return MovePos(lenMain - 1);
  }
  Byte currentByte = *data;
  Byte matchByte = data[(size_t)0 - reps[0] - 1];

  if(lenMain < 2 && currentByte != matchByte && repLens[repMaxIndex] < 2)
  {
    backRes = (UInt32)-1;
    lenRes = 1;
    return S_OK;
  }

  _optimum[0].State = _state;

  UInt32 posState = (position & _posStateMask);

  _optimum[1].Price = _isMatch[_state.Index][posState].GetPrice0() + 
      _literalEncoder.GetSubCoder(position, _previousByte)->GetPrice(!_state.IsCharState(), matchByte, currentByte);
  _optimum[1].MakeAsChar();

  UInt32 matchPrice = _isMatch[_state.Index][posState].GetPrice1();
  UInt32 repMatchPrice = matchPrice + _isRep[_state.Index].GetPrice1();

  if(matchByte == currentByte)
  {
    UInt32 shortRepPrice = repMatchPrice + GetRepLen1Price(_state, posState);
    if(shortRepPrice < _optimum[1].Price)
    {
      _optimum[1].Price = shortRepPrice;
      _optimum[1].MakeAsShortRep();
    }
  }
  UInt32 lenEnd = ((lenMain >= repLens[repMaxIndex]) ? lenMain : repLens[repMaxIndex]);

  if(lenEnd < 2)
  {
    backRes = _optimum[1].BackPrev;
    lenRes = 1;
    return S_OK;
  }

  _optimum[1].PosPrev = 0;
  for (i = 0; i < kNumRepDistances; i++)
    _optimum[0].Backs[i] = reps[i];

  UInt32 len = lenEnd;
  do
    _optimum[len--].Price = kIfinityPrice;
  while (len >= 2);

  for(i = 0; i < kNumRepDistances; i++)
  {
    UInt32 repLen = repLens[i];
    if (repLen < 2)
      continue;
    UInt32 price = repMatchPrice + GetPureRepPrice(i, _state, posState);
    do
    {
      UInt32 curAndLenPrice = price + _repMatchLenEncoder.GetPrice(repLen - 2, posState);
      COptimal &optimum = _optimum[repLen];
      if (curAndLenPrice < optimum.Price) 
      {
        optimum.Price = curAndLenPrice;
        optimum.PosPrev = 0;
        optimum.BackPrev = i;
        optimum.Prev1IsChar = false;
      }
    }
    while(--repLen >= 2);
  }

  UInt32 normalMatchPrice = matchPrice + _isRep[_state.Index].GetPrice0();

  len = ((repLens[0] >= 2) ? repLens[0] + 1 : 2);
  if (len <= lenMain)
  {
    UInt32 offs = 0;
    while (len > matchDistances[offs])
      offs += 2;
    for(; ; len++)
    {
      UInt32 distance = matchDistances[offs + 1];
      UInt32 curAndLenPrice = normalMatchPrice + GetPosLenPrice(distance, len, posState);
      COptimal &optimum = _optimum[len];
      if (curAndLenPrice < optimum.Price) 
      {
        optimum.Price = curAndLenPrice;
        optimum.PosPrev = 0;
        optimum.BackPrev = distance + kNumRepDistances;
        optimum.Prev1IsChar = false;
      }
      if (len == matchDistances[offs])
      {
        offs += 2;
        if (offs == numDistancePairs)
          break;
      }
    }
  }

  UInt32 cur = 0;

  while(true)
  {
    cur++;
    if(cur == lenEnd)
    {
      lenRes = Backward(backRes, cur);
      return S_OK;
    }
    UInt32 newLen, numDistancePairs;
    RINOK(ReadMatchDistances(newLen, numDistancePairs));
    if(newLen >= _numFastBytes)
    {
      _numDistancePairs = numDistancePairs;
      _longestMatchLength = newLen;
      _longestMatchWasFound = true;
      lenRes = Backward(backRes, cur);
      return S_OK;
    }
    position++;
    COptimal &curOptimum = _optimum[cur];
    UInt32 posPrev = curOptimum.PosPrev;
    CState state;
    if (curOptimum.Prev1IsChar)
    {
      posPrev--;
      if (curOptimum.Prev2)
      {
        state = _optimum[curOptimum.PosPrev2].State;
        if (curOptimum.BackPrev2 < kNumRepDistances)
          state.UpdateRep();
        else
          state.UpdateMatch();
      }
      else
        state = _optimum[posPrev].State;
      state.UpdateChar();
    }
    else
      state = _optimum[posPrev].State;
    if (posPrev == cur - 1)
    {
      if (curOptimum.IsShortRep())
        state.UpdateShortRep();
      else
        state.UpdateChar();
    }
    else
    {
      UInt32 pos;
      if (curOptimum.Prev1IsChar && curOptimum.Prev2)
      {
        posPrev = curOptimum.PosPrev2;
        pos = curOptimum.BackPrev2;
        state.UpdateRep();
      }
      else
      {
        pos = curOptimum.BackPrev;
        if (pos < kNumRepDistances)
          state.UpdateRep();
        else
          state.UpdateMatch();
      }
      const COptimal &prevOptimum = _optimum[posPrev];
      if (pos < kNumRepDistances)
      {
        reps[0] = prevOptimum.Backs[pos];
    		UInt32 i;
        for(i = 1; i <= pos; i++)
          reps[i] = prevOptimum.Backs[i - 1];
        for(; i < kNumRepDistances; i++)
          reps[i] = prevOptimum.Backs[i];
      }
      else
      {
        reps[0] = (pos - kNumRepDistances);
        for(UInt32 i = 1; i < kNumRepDistances; i++)
          reps[i] = prevOptimum.Backs[i - 1];
      }
    }
    curOptimum.State = state;
    for(UInt32 i = 0; i < kNumRepDistances; i++)
      curOptimum.Backs[i] = reps[i];
    UInt32 curPrice = curOptimum.Price; 
    const Byte *data = _matchFinder->GetPointerToCurrentPos() - 1;
    const Byte currentByte = *data;
    const Byte matchByte = data[(size_t)0 - reps[0] - 1];

    UInt32 posState = (position & _posStateMask);

    UInt32 curAnd1Price = curPrice +
        _isMatch[state.Index][posState].GetPrice0() +
        _literalEncoder.GetSubCoder(position, data[(size_t)0 - 1])->GetPrice(!state.IsCharState(), matchByte, currentByte);

    COptimal &nextOptimum = _optimum[cur + 1];

    bool nextIsChar = false;
    if (curAnd1Price < nextOptimum.Price) 
    {
      nextOptimum.Price = curAnd1Price;
      nextOptimum.PosPrev = cur;
      nextOptimum.MakeAsChar();
      nextIsChar = true;
    }

    UInt32 matchPrice = curPrice + _isMatch[state.Index][posState].GetPrice1();
    UInt32 repMatchPrice = matchPrice + _isRep[state.Index].GetPrice1();
    
    if(matchByte == currentByte &&
        !(nextOptimum.PosPrev < cur && nextOptimum.BackPrev == 0))
    {
      UInt32 shortRepPrice = repMatchPrice + GetRepLen1Price(state, posState);
      if(shortRepPrice <= nextOptimum.Price)
      {
        nextOptimum.Price = shortRepPrice;
        nextOptimum.PosPrev = cur;
        nextOptimum.MakeAsShortRep();
        nextIsChar = true;
      }
    }
    /*
    if(newLen == 2 && matchDistances[2] >= kDistLimit2) // test it maybe set 2000 ?
      continue;
    */

    UInt32 numAvailableBytesFull = _matchFinder->GetNumAvailableBytes() + 1;
    numAvailableBytesFull = MyMin(kNumOpts - 1 - cur, numAvailableBytesFull);
    UInt32 numAvailableBytes = numAvailableBytesFull;

    if (numAvailableBytes < 2)
      continue;
    if (numAvailableBytes > _numFastBytes)
      numAvailableBytes = _numFastBytes;
    if (!nextIsChar && matchByte != currentByte) // speed optimization
    {
      // try Literal + rep0
      UInt32 backOffset = reps[0] + 1;
      UInt32 limit = MyMin(numAvailableBytesFull, _numFastBytes + 1);
      UInt32 temp;
      for (temp = 1; temp < limit && 
          data[temp] == data[(size_t)temp - backOffset]; temp++);
      UInt32 lenTest2 = temp - 1;
      if (lenTest2 >= 2)
      {
        CState state2 = state;
        state2.UpdateChar();
        UInt32 posStateNext = (position + 1) & _posStateMask;
        UInt32 nextRepMatchPrice = curAnd1Price + 
            _isMatch[state2.Index][posStateNext].GetPrice1() +
            _isRep[state2.Index].GetPrice1();
        // for (; lenTest2 >= 2; lenTest2--)
        {
          UInt32 offset = cur + 1 + lenTest2;
          while(lenEnd < offset)
            _optimum[++lenEnd].Price = kIfinityPrice;
          UInt32 curAndLenPrice = nextRepMatchPrice + GetRepPrice(
              0, lenTest2, state2, posStateNext);
          COptimal &optimum = _optimum[offset];
          if (curAndLenPrice < optimum.Price) 
          {
            optimum.Price = curAndLenPrice;
            optimum.PosPrev = cur + 1;
            optimum.BackPrev = 0;
            optimum.Prev1IsChar = true;
            optimum.Prev2 = false;
          }
        }
      }
    }
    
    UInt32 startLen = 2; // speed optimization 
    for(UInt32 repIndex = 0; repIndex < kNumRepDistances; repIndex++)
    {
      // UInt32 repLen = _matchFinder->GetMatchLen(0 - 1, reps[repIndex], newLen); // test it;
      UInt32 backOffset = reps[repIndex] + 1;
      if (data[0] != data[(size_t)0 - backOffset] ||
          data[1] != data[(size_t)1 - backOffset])
        continue;
      UInt32 lenTest;
      for (lenTest = 2; lenTest < numAvailableBytes && 
          data[lenTest] == data[(size_t)lenTest - backOffset]; lenTest++);
      while(lenEnd < cur + lenTest)
        _optimum[++lenEnd].Price = kIfinityPrice;
      UInt32 lenTestTemp = lenTest;
      UInt32 price = repMatchPrice + GetPureRepPrice(repIndex, state, posState);
      do
      {
        UInt32 curAndLenPrice = price + _repMatchLenEncoder.GetPrice(lenTest - 2, posState);
        COptimal &optimum = _optimum[cur + lenTest];
        if (curAndLenPrice < optimum.Price) 
        {
          optimum.Price = curAndLenPrice;
          optimum.PosPrev = cur;
          optimum.BackPrev = repIndex;
          optimum.Prev1IsChar = false;
        }
      }
      while(--lenTest >= 2);
      lenTest = lenTestTemp;
      
      if (repIndex == 0)
        startLen = lenTest + 1;
        
      // if (_maxMode)
        {
          UInt32 lenTest2 = lenTest + 1;
          UInt32 limit = MyMin(numAvailableBytesFull, lenTest2 + _numFastBytes);
          for (; lenTest2 < limit && 
              data[lenTest2] == data[(size_t)lenTest2 - backOffset]; lenTest2++);
          lenTest2 -= lenTest + 1;
          if (lenTest2 >= 2)
          {
            CState state2 = state;
            state2.UpdateRep();
            UInt32 posStateNext = (position + lenTest) & _posStateMask;
            UInt32 curAndLenCharPrice = 
                price + _repMatchLenEncoder.GetPrice(lenTest - 2, posState) + 
                _isMatch[state2.Index][posStateNext].GetPrice0() +
                _literalEncoder.GetSubCoder(position + lenTest, data[(size_t)lenTest - 1])->GetPrice(
                true, data[(size_t)lenTest - backOffset], data[lenTest]);
            state2.UpdateChar();
            posStateNext = (position + lenTest + 1) & _posStateMask;
            UInt32 nextRepMatchPrice = curAndLenCharPrice + 
                _isMatch[state2.Index][posStateNext].GetPrice1() +
                _isRep[state2.Index].GetPrice1();
            
            // for(; lenTest2 >= 2; lenTest2--)
            {
              UInt32 offset = cur + lenTest + 1 + lenTest2;
              while(lenEnd < offset)
                _optimum[++lenEnd].Price = kIfinityPrice;
              UInt32 curAndLenPrice = nextRepMatchPrice + GetRepPrice(
                  0, lenTest2, state2, posStateNext);
              COptimal &optimum = _optimum[offset];
              if (curAndLenPrice < optimum.Price) 
              {
                optimum.Price = curAndLenPrice;
                optimum.PosPrev = cur + lenTest + 1;
                optimum.BackPrev = 0;
                optimum.Prev1IsChar = true;
                optimum.Prev2 = true;
                optimum.PosPrev2 = cur;
                optimum.BackPrev2 = repIndex;
              }
            }
          }
        }
      }
    
    //    for(UInt32 lenTest = 2; lenTest <= newLen; lenTest++)
    if (newLen > numAvailableBytes)
    {
      newLen = numAvailableBytes;
      for (numDistancePairs = 0; newLen > matchDistances[numDistancePairs]; numDistancePairs += 2);
      matchDistances[numDistancePairs] = newLen;
      numDistancePairs += 2;
    }
    if (newLen >= startLen)
    {
      UInt32 normalMatchPrice = matchPrice + _isRep[state.Index].GetPrice0();
      while(lenEnd < cur + newLen)
        _optimum[++lenEnd].Price = kIfinityPrice;

      UInt32 offs = 0;
      while(startLen > matchDistances[offs])
        offs += 2;
      UInt32 curBack = matchDistances[offs + 1];
      UInt32 posSlot = GetPosSlot2(curBack);
      for(UInt32 lenTest = /*2*/ startLen; ; lenTest++)
      {
        UInt32 curAndLenPrice = normalMatchPrice;
        UInt32 lenToPosState = GetLenToPosState(lenTest);
        if (curBack < kNumFullDistances)
          curAndLenPrice += _distancesPrices[lenToPosState][curBack];
        else
          curAndLenPrice += _posSlotPrices[lenToPosState][posSlot] + _alignPrices[curBack & kAlignMask];
  
        curAndLenPrice += _lenEncoder.GetPrice(lenTest - kMatchMinLen, posState);