/work/vvenc/source/Lib/EncoderLib/EncModeCtrl.cpp

Source
/* -----------------------------------------------------------------------------
The copyright in this software is being made available under the Clear BSD
License, included below. No patent rights, trademark rights and/or 
other Intellectual Property Rights other than the copyrights concerning 
the Software are granted under this license.

The Clear BSD License

Copyright (c) 2019-2026, Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. & The VVenC Authors.
All rights reserved.

Redistribution and use in source and binary forms, with or without modification,
are permitted (subject to the limitations in the disclaimer below) provided that
the following conditions are met:

     * Redistributions of source code must retain the above copyright notice,
     this list of conditions and the following disclaimer.

     * Redistributions in binary form must reproduce the above copyright
     notice, this list of conditions and the following disclaimer in the
     documentation and/or other materials provided with the distribution.

     * Neither the name of the copyright holder nor the names of its
     contributors may be used to endorse or promote products derived from this
     software without specific prior written permission.

NO EXPRESS OR IMPLIED LICENSES TO ANY PARTY'S PATENT RIGHTS ARE GRANTED BY
THIS LICENSE. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER
IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.


------------------------------------------------------------------------------------------- */


/** \file     EncModeCtrl.cpp
    \brief    Encoder controller for trying out specific modes
*/

#include "EncCu.h"

#include "EncModeCtrl.h"
#include "CommonLib/RdCost.h"
#include "CommonLib/CodingStructure.h"
#include "CommonLib/Picture.h"
#include "CommonLib/UnitTools.h"
#include "CommonLib/dtrace_next.h"

#include <cmath>

//! \ingroup EncoderLib
//! \{

namespace vvenc {

void CacheBlkInfoCtrl::create(int ctuSize)
{
  const unsigned numPos = MAX_CU_SIZE >> MIN_CU_LOG2;
  const int maxSizeIdx  = MAX_CU_SIZE_IDX - MIN_CU_LOG2;

  //static constexpr size_t numCu = 7921;

  size_t numCu = 0;
  
  for( int wIdx = 0; wIdx < maxSizeIdx; wIdx++ )
  {
    for( int hIdx = 0; hIdx < maxSizeIdx; hIdx++ )
    {
      for( unsigned y = 0; y < numPos; y++ )
      {
        for( unsigned x = 0; x < numPos; x++ )
        {
          // a block of width W might be offset of N * W + 1/2 W (bcs of TT), same for H
          // W = 1 << ( wIdx + 2 )
          // 1/2 W = 1 << ( wIdx + 1 )
          // remainder of (N+1/2)*W -> x & ( ( 1 << ( wIdx + 1 ) ) - 1 )
  
          if( (x + (1 << (wIdx)) <= (ctuSize >> MIN_CU_LOG2))
              && (y + (1 << (hIdx)) <= (ctuSize >> MIN_CU_LOG2))
              && (((x << MIN_CU_LOG2) & ((1 << (wIdx + MIN_CU_LOG2 - 1)) - 1)) == 0)
              && (((y << MIN_CU_LOG2) & ((1 << (hIdx + MIN_CU_LOG2 - 1)) - 1)) == 0) )
          {
            numCu++;
          }
        }
      }
    }
  }

  m_codedCUInfoBuf = new CodedCUInfo[numCu];
  CodedCUInfo* cuInfo = m_codedCUInfoBuf;

  for( int wIdx = 0; wIdx < maxSizeIdx; wIdx++ )
  {
    for( int hIdx = 0; hIdx < maxSizeIdx; hIdx++ )
    {
      for( unsigned y = 0; y < numPos; y++ )
      {
        for( unsigned x = 0; x < numPos; x++ )
        {
          // a block of width W might be offset of N * W + 1/2 W (bcs of TT), same for H
          // W = 1 << ( wIdx + 2 )
          // 1/2 W = 1 << ( wIdx + 1 )
          // remainder of (N+1/2)*W -> x & ( ( 1 << ( wIdx + 1 ) ) - 1 )

          if(( x + (1<<(wIdx)) <= ( ctuSize >> MIN_CU_LOG2 ) )
            && ( y + (1<<(hIdx)) <= ( ctuSize >> MIN_CU_LOG2 ) )
            && ( ( ( x << MIN_CU_LOG2 ) & ((1 << (wIdx + MIN_CU_LOG2 - 1)) - 1) ) == 0 )
            && ( ( ( y << MIN_CU_LOG2 ) & ((1 << (hIdx + MIN_CU_LOG2 - 1)) - 1) ) == 0 ) )
          {
            m_codedCUInfo[wIdx][hIdx][x][y] = cuInfo++;
            m_codedCUInfo[wIdx][hIdx][x][y]->poc       = -1;
            m_codedCUInfo[wIdx][hIdx][x][y]->ctuRsAddr = -1;
          }
          else
          {
            m_codedCUInfo[wIdx][hIdx][x][y] = nullptr;
          }
        }
      }
    }
  }
}

void CacheBlkInfoCtrl::destroy()
{
  delete[] m_codedCUInfoBuf;
  m_codedCUInfoBuf = nullptr;
}

void CacheBlkInfoCtrl::init( const Slice &slice )
{
  m_pcv = slice.pps->pcv;
}

void CacheBlkInfoCtrl::initBlk( const UnitArea& area, int poc )
{
  unsigned idx1, idx2, idx3, idx4;
  getAreaIdxNew( area.Y(), *m_pcv, idx1, idx2, idx3, idx4 );

  const int ctuRsAddr = getCtuAddr( area.lumaPos(), *m_pcv );
  CodedCUInfo* cuInfo = m_codedCUInfo[idx1][idx2][idx3][idx4];

  if( cuInfo->poc != poc || cuInfo->ctuRsAddr != ctuRsAddr )
  {
    GCC_WARNING_DISABLE_class_memaccess
    memset( cuInfo, 0, sizeof( CodedCUInfo ) );
    GCC_WARNING_RESET

    cuInfo->poc       = poc;
    cuInfo->ctuRsAddr = ctuRsAddr;
  }
}

CodedCUInfo& CacheBlkInfoCtrl::getBlkInfo( const UnitArea& area )
{
  unsigned idx1, idx2, idx3, idx4;
  getAreaIdxNew( area.Y(), *m_pcv, idx1, idx2, idx3, idx4 );
//  DTRACE( g_trace_ctx, D_TMP, "%d loc %d %d %d %d\n", g_trace_ctx->getChannelCounter(D_TMP), idx1, idx2, idx3, idx4);
  return *m_codedCUInfo[idx1][idx2][idx3][idx4];
}

void CodedCUInfo::setMv( const RefPicList refPicList, const int iRefIdx, const Mv& rMv )
{
  if( iRefIdx >= MAX_STORED_CU_INFO_REFS ) return;

  saveMv [refPicList][iRefIdx] = rMv;
  validMv[refPicList][iRefIdx] = true;
}

bool CodedCUInfo::getMv( const RefPicList refPicList, const int iRefIdx, Mv& rMv ) const
{
  if( iRefIdx >= MAX_STORED_CU_INFO_REFS )
  {
    rMv = saveMv[refPicList][0];
    return false;
  }

  rMv = saveMv[refPicList][iRefIdx];
  return validMv[refPicList][iRefIdx];
}

uint8_t CacheBlkInfoCtrl::findBestSbt( const UnitArea& area, const uint32_t curPuSse )
{
  unsigned idx1, idx2, idx3, idx4;
  getAreaIdxNew( area.Y(), *m_pcv, idx1, idx2, idx3, idx4 );
  CodedCUInfo* pSbtSave = m_codedCUInfo[idx1][idx2][idx3][idx4];

  for( int i = 0; i < pSbtSave->numPuInfoStored; i++ )
  {
    if( curPuSse == pSbtSave->puSse[i] )
    {
      return pSbtSave->puSbt[i];
    }
  }

  return MAX_UCHAR;
}

bool CacheBlkInfoCtrl::saveBestSbt( const UnitArea& area, const uint32_t curPuSse, const uint8_t curPuSbt )
{
  unsigned idx1, idx2, idx3, idx4;
  getAreaIdxNew( area.Y(), *m_pcv, idx1, idx2, idx3, idx4 );
  CodedCUInfo* pSbtSave = m_codedCUInfo[idx1][idx2][idx3][idx4];

  if( pSbtSave->numPuInfoStored == SBT_NUM_SL )
  {
    return false;
  }

  pSbtSave->puSse[pSbtSave->numPuInfoStored] = curPuSse;
  pSbtSave->puSbt[pSbtSave->numPuInfoStored] = curPuSbt;
  pSbtSave->numPuInfoStored++;

  return true;
}

static bool isTheSameNbHood( const CodingUnit &cu, const CodingStructure& cs, const Partitioner &partitioner, int picW, int picH )
{
  if( cu.chType != partitioner.chType )
  {
    return false;
  }

  const PartitioningStack &ps = partitioner.getPartStack();

  int i = 1;

  for( ; i < ps.size(); i++ )
  {
    if( ps[i].split != CU::getSplitAtDepth( cu, i - 1 ) )
    {
      break;
    }
  }

  const UnitArea& cmnAnc = ps[i - 1].parts[ps[i - 1].idx];
  const UnitArea cuArea  = CS::getArea( cs, cu, partitioner.chType, partitioner.treeType );

  for( int i = 0; i < cmnAnc.blocks.size(); i++ )
  {
    if( i < cuArea.blocks.size() && cuArea.blocks[i].valid() && cuArea.blocks[i].pos() != cmnAnc.blocks[i].pos() )
    {
      return false;
    }
  }

  return true;
}

void BestEncInfoCache::create( const bool reuseCuResults, const ChromaFormat chFmt, const int ctuSize )
{
  m_reuseCuResults = reuseCuResults;

  if( !m_reuseCuResults ) return;

  m_chFmt = chFmt;
  std::fill_n( &m_bestEncInfo[0][0][0][0], sizeof( m_bestEncInfo ) / sizeof( m_bestEncInfo[0][0][0][0] ), nullptr );
}

void BestEncInfoCache::destroy()
{
  if( !m_reuseCuResults ) return;

  std::fill_n( &m_bestEncInfo[0][0][0][0], sizeof( m_bestEncInfo ) / sizeof( m_bestEncInfo[0][0][0][0] ), nullptr );
  m_filledSlots.clear();
  m_entryArena.release();
  m_coeffArena.release();
  m_dmvrMvArena.release();

  m_pcv = nullptr;
}

void BestEncInfoCache::init( const Slice &slice )
{
  if( !m_reuseCuResults ) return;

  // A new CTU starts, and no result of the previous one can match its CUs.
  for( BestEncodingInfo** slot : m_filledSlots )
  {
    *slot = nullptr;
  }
  m_filledSlots.clear();
  m_entryArena.reset();
  m_coeffArena.reset();
  m_dmvrMvArena.reset();

  bool isInitialized = m_pcv;

  m_pcv = slice.pps->pcv;

  if( isInitialized ) return;

  m_dummyCS.pcv = m_pcv;
}

// Takes storage for the result of a CU with size indices wIdx and hIdx, set up the way the earlier
// fixed table of results was set up when it was created.
BestEncodingInfo* BestEncInfoCache::xNewEntry( const int wIdx, const int hIdx )
{
  BestEncodingInfo* encInfo = m_entryArena.take( 1 );
  encInfo->~BestEncodingInfo();
  new( encInfo ) BestEncodingInfo();

  const UnitArea area( m_chFmt, Area( 0, 0, 1 << ( wIdx + MIN_CU_LOG2 ), 1 << ( hIdx + MIN_CU_LOG2 ) ) );

  encInfo->cu.UnitArea::operator=( area );
  encInfo->tu.UnitArea::operator=( area );

  encInfo->cu.chType   = CH_L;
  encInfo->cu.treeType = TREE_D;
  encInfo->cu.modeType = MODE_TYPE_ALL;
  encInfo->cu.qp       = MAX_SCHAR;
  encInfo->tu.chType   = CH_L;

  if( hIdx >= 1 && wIdx >= 1 && ( wIdx + hIdx ) >= 3 )
  {
    const int dmvrSize = ( 1 << std::max( 0, ( wIdx + MIN_CU_LOG2 - DMVR_SUBCU_SIZE_LOG2 ) ) ) * ( 1 << std::max( 0, ( hIdx + MIN_CU_LOG2 - DMVR_SUBCU_SIZE_LOG2 ) ) );
    encInfo->cu.mvdL0SubPu = m_dmvrMvArena.take( dmvrSize );
  }

  TCoeffSig* coeff[MAX_NUM_TBLOCKS] = { 0, };
  for( int i = 0; i < area.blocks.size(); i++ )
  {
    coeff[i] = m_coeffArena.take( area.blocks[i].area() );
  }

  encInfo->tu.cs = &m_dummyCS;
  encInfo->tu.init( coeff );

  encInfo->poc      = -1;
  encInfo->testMode = EncTestMode();
  return encInfo;
}

bool BestEncInfoCache::setFromCs( const CodingStructure& cs, const EncTestMode& testMode, const Partitioner& partitioner )
{
  if( !m_reuseCuResults ) return false;

  if( cs.cus.size() != 1 || cs.tus.size() != 1 || partitioner.maxBTD <= 1 )
  {
    return false;
  }

  unsigned idx1, idx2, idx3, idx4;
  getAreaIdxNew( cs.area.Y(), *m_pcv, idx1, idx2, idx3, idx4 );

  BestEncodingInfo*& slot = m_bestEncInfo[idx1][idx2][idx3][idx4];
  if( !slot )
  {
    slot = xNewEntry( idx1, idx2 );
    m_filledSlots.push_back( &slot );
  }
  BestEncodingInfo& encInfo = *slot;

  encInfo.poc            =  cs.picture->poc;
  encInfo.cu.UnitArea::operator= ( *cs.cus.front() );
  encInfo.tu.UnitArea::operator= ( *cs.tus.front() );
  encInfo.cu             = *cs.cus.front();
  for( auto &blk : cs.tus.front()->blocks )
  {
    if( blk.valid() ) encInfo.tu.copyComponentFrom( *cs.tus.front(), blk.compID );
  }
  encInfo.testMode       = testMode;
  encInfo.dist           = cs.dist;
  encInfo.costEDO        = cs.costDbOffset;

  return true;
}

bool BestEncInfoCache::isReusingCuValid( const CodingStructure& cs, const Partitioner& partitioner, int qp )
{
  if( !m_reuseCuResults || partitioner.treeType == TREE_C || partitioner.maxBTD <= 1 )
  {
    return false; //if save & load is allowed for chroma CUs, we should check whether luma info (pred, recon, etc) is the same, which is quite complex
  }

  unsigned idx1, idx2, idx3, idx4;
  getAreaIdxNew( cs.area.Y(), *m_pcv, idx1, idx2, idx3, idx4 );

  const BestEncodingInfo* stored = m_bestEncInfo[idx1][idx2][idx3][idx4];
  if( !stored )
  {
    return false;
  }
  const BestEncodingInfo& encInfo = *stored;

  if( encInfo.cu.treeType != partitioner.treeType || encInfo.cu.modeType != partitioner.modeType )
  {
    return false;
  }
  if( encInfo.cu.qp != qp )
    return false;
  if( cs.picture->poc != encInfo.poc 
    || CS::getArea( cs, cs.area, partitioner.chType, partitioner.treeType ) != CS::getArea( cs, encInfo.cu, partitioner.chType, partitioner.treeType ) 
    || !isTheSameNbHood( encInfo.cu, cs, partitioner, (cs.picture->Y().width), (cs.picture->Y().height))
    || CU::isIBC(encInfo.cu)
    || partitioner.currQgEnable() || cs.currQP[partitioner.chType] != encInfo.cu.qp
    )
  {
    return false;
  }
  else
  {
    return true;
  }
}

bool BestEncInfoCache::setCsFrom( CodingStructure& cs, EncTestMode& testMode, const Partitioner& partitioner ) const
{
  if( !m_reuseCuResults ) return false;

  unsigned idx1, idx2, idx3, idx4;
  getAreaIdxNew( cs.area.Y(), *m_pcv, idx1, idx2, idx3, idx4 );

  const BestEncodingInfo* stored = m_bestEncInfo[idx1][idx2][idx3][idx4];
  if( !stored )
  {
    return false;
  }
  const BestEncodingInfo& encInfo = *stored;

  if( cs.picture->poc != encInfo.poc 
    || CS::getArea( cs, cs.area, partitioner.chType, partitioner.treeType ) != CS::getArea( cs, encInfo.cu, partitioner.chType, partitioner.treeType ) 
    || !isTheSameNbHood( encInfo.cu, cs, partitioner, (cs.picture->Y().width), (cs.picture->Y().height))
    || partitioner.currQgEnable() || cs.currQP[partitioner.chType] != encInfo.cu.qp
    )
  {
    return false;
  }

  const UnitArea ua = CS::getArea( cs, cs.area, partitioner.chType, partitioner.treeType );
  CodingUnit     &cu = cs.addCU( ua, partitioner.chType );
  cu.treeType = partitioner.treeType;
  cu.modeType = partitioner.modeType;
  cu.initPuData();
  TransformUnit  &tu = cs.addTU( ua, partitioner.chType, &cu );

  cu          .repositionTo( encInfo.cu );
  cu          .repositionTo( encInfo.cu );
  tu          .repositionTo( encInfo.tu );

  cu          = encInfo.cu;
  cu          = encInfo.cu;
  for( auto &blk : tu.blocks )
  {
    if( blk.valid() ) tu.copyComponentFrom( encInfo.tu, blk.compID );
  }

  testMode    = encInfo.testMode;
  cs.dist     = encInfo.dist;
  cs.costDbOffset = encInfo.costEDO;
  return true;
}


//////////////////////////////////////////////////////////////////////////
// EncModeCtrl
//////////////////////////////////////////////////////////////////////////
void EncModeCtrl::init( const VVEncCfg& encCfg, RdCost* pRdCost )
{
  m_pcEncCfg = &encCfg;
  m_pcRdCost = pRdCost;
  comprCUCtx = nullptr;

  CacheBlkInfoCtrl::create( encCfg.m_CTUSize );
  BestEncInfoCache::create( encCfg.m_reuseCuResults, encCfg.m_internChromaFormat, encCfg.m_CTUSize );
}

void EncModeCtrl::destroy()
{
  CacheBlkInfoCtrl::destroy();
  BestEncInfoCache::destroy();
}

void EncModeCtrl::initCTUEncoding( const Slice &slice, int tileIdx )
{
  CacheBlkInfoCtrl::init( slice );
  BestEncInfoCache::init( slice );

  CHECK( !m_ComprCUCtxList.empty(), "Mode list is not empty at the beginning of a CTU" );

  if( m_pcEncCfg->m_fastQtBtEnc )
  {
    m_skipThresholdE0023FastEnc = ((slice.getMinPictureDistance() <= PICTURE_DISTANCE_TH) ? FAST_SKIP_DEPTH : SKIP_DEPTH);
  }
  else
  {
    m_skipThresholdE0023FastEnc = SKIP_DEPTH;
  }

  m_tileIdx = tileIdx;
}

void EncModeCtrl::initCULevel( Partitioner &partitioner, const CodingStructure& cs, int  MergeSimpleFlag)
{
  // Min/max depth
  unsigned minDepth = 0;
  unsigned maxDepth = cs.pcv->getMaxDepth( cs.slice->sliceType, partitioner.chType );
  if( m_pcEncCfg->m_useFastLCTU )
  {
    partitioner.setMaxMinDepth(minDepth, maxDepth, cs, cs.picture->useQtbttSpeedUpMode, MergeSimpleFlag);
  }

  minDepth = std::max<unsigned>( minDepth, cs.pcv->getMinDepth( cs.slice->sliceType, partitioner.chType ) );
  maxDepth = std::min<unsigned>( maxDepth, cs.pcv->getMaxDepth( cs.slice->sliceType, partitioner.chType ) );

  m_ComprCUCtxList.push_back( ComprCUCtx( cs, minDepth, maxDepth ) );
  comprCUCtx = &m_ComprCUCtxList.back();

  const CodingUnit* cuLeft  = cs.getCURestricted( cs.area.blocks[partitioner.chType].pos().offset( -1, 0 ), cs.area.blocks[partitioner.chType].pos(), cs.slice->independentSliceIdx, m_tileIdx, partitioner.chType, partitioner.treeType );
  const CodingUnit* cuAbove = cs.getCURestricted( cs.area.blocks[partitioner.chType].pos().offset( 0, -1 ), cs.area.blocks[partitioner.chType].pos(), cs.slice->independentSliceIdx, m_tileIdx, partitioner.chType, partitioner.treeType );

  const bool qtBeforeBt = ( (  cuLeft  &&  cuAbove  && cuLeft ->qtDepth > partitioner.currQtDepth && cuAbove->qtDepth > partitioner.currQtDepth )
                         || (  cuLeft  && !cuAbove  && cuLeft ->qtDepth > partitioner.currQtDepth )
                         || ( !cuLeft  &&  cuAbove  && cuAbove->qtDepth > partitioner.currQtDepth )
                         || ( !cuAbove && !cuLeft   && cs.area.lwidth() >= ( 32 << cs.slice->TLayer ) )
                         || ( m_pcEncCfg->m_qtbttSpeedUp > 1 && partitioner.maxBTD < ( ( cs.slice->isIntra() && !cs.sps->IBC ) ? 3 : 2 ) ) )
                         && ( cs.area.lwidth() > ( cs.pcv->getMinQtSize( *cs.slice, partitioner.chType ) << 1 ) );

  // set features
  ComprCUCtx &cuECtx    = *comprCUCtx;
  cuECtx.qtBeforeBt     = qtBeforeBt;
  cuECtx.doTriHorzSplit = true;
  cuECtx.doTriVertSplit = true;
  cuECtx.doMoreSplits   = 3;
  cuECtx.isReusingCu    = isReusingCuValid( cs, partitioner, cs.baseQP );
  cuECtx.didHorzSplit   = partitioner.canSplit( CU_HORZ_SPLIT, cs );
  cuECtx.didVertSplit   = partitioner.canSplit( CU_VERT_SPLIT, cs );
  cuECtx.doHorChromaSplit = true;
  cuECtx.doVerChromaSplit = true;
  cuECtx.doQtChromaSplit  = true;

  if( m_pcEncCfg->m_contentBasedFastQtbt && cs.pcv->getMaxMTTDepth(*cs.slice, partitioner.chType))
  {
    const CompArea& currArea = partitioner.currArea().Y();
    int cuHeight  = currArea.height;
    int cuWidth   = currArea.width;

    const bool condIntraInter = m_pcEncCfg->m_IntraPeriod == 1 ? ( partitioner.currBtDepth == 0 ) : ( cuHeight > 32 && cuWidth > 32 );

    if( cuWidth == cuHeight && condIntraInter )
    {
      const CPelBuf bufCurrArea = cs.getOrgBuf( partitioner.currArea().block( COMP_Y ) );

      int grad[4];
      g_pelBufOP.absGradientSums( bufCurrArea.buf, bufCurrArea.stride, cuWidth, cuHeight, grad );

      cuECtx.grad_horVal = (double)grad[0];
      cuECtx.grad_verVal = (double)grad[1];
      cuECtx.grad_dowVal = (double)grad[2];
      cuECtx.grad_dupVal = (double)grad[3];
    }
  }
}

void EncModeCtrl::finishCULevel( Partitioner &partitioner )
{
  m_ComprCUCtxList.pop_back();
  comprCUCtx = m_ComprCUCtxList.size() ? &m_ComprCUCtxList.back() : nullptr;
}

bool EncModeCtrl::trySplit( const EncTestMode& encTestmode, const CodingStructure &cs, Partitioner& partitioner, const EncTestMode& lastTestmode )
{
  ComprCUCtx& cuECtx = *comprCUCtx;

  const PartSplit implicitSplit = partitioner.getImplicitSplit( cs );
  const bool isBoundary         = implicitSplit != CU_DONT_SPLIT;

  if( ( m_pcEncCfg->m_IntraPeriod == 1 ) && ( partitioner.chType == CH_C ) && ( !cuECtx.doQtChromaSplit ) )
  {
    cuECtx.maxDepth         = partitioner.currDepth;
  }

  if( isBoundary )
  {
    if( encTestmode.type != ETM_SPLIT_QT )
    {
      return getPartSplit( encTestmode ) == implicitSplit;
    }
    else
    {
      return partitioner.canSplit( CU_QUAD_SPLIT, cs );
    }
  }

  const Slice&           slice       = *cs.slice;
  const uint32_t         width       = partitioner.currArea().lumaSize().width;
  const CodingStructure *bestCS      = cuECtx.bestCS;
  const CodingUnit      *bestCU      = cuECtx.bestCU;

  if( cuECtx.minDepth > partitioner.currQtDepth && partitioner.canSplit( CU_QUAD_SPLIT, cs ) )
  {
    // enforce QT
    return encTestmode.type == ETM_SPLIT_QT;
  }
  else if( encTestmode.type == ETM_SPLIT_QT && cuECtx.maxDepth <= partitioner.currQtDepth )
  {
    // don't check this QT depth
    return false;
  }

  CHECK( ! isModeSplit( encTestmode ), "wrong method" );

  //////////////////////////////////////////////////////////////////////////
  // skip-history rule - don't split further if at least for three past levels
  //                     in the split tree it was found that skip is the best mode
  //////////////////////////////////////////////////////////////////////////
  int skipScore = 0;

  if ((!slice.isIntra() || slice.sps->IBC) && cuECtx.isBestNoSplitSkip )
  {
    for( int i = 2; i <= m_ComprCUCtxList.size(); i++ )
    {
      if( ( m_ComprCUCtxList.end() - i )->isBestNoSplitSkip )
      {
        skipScore += 1;
      }
      else
      {
        break;
      }
    }
  }

  const PartSplit split = getPartSplit( encTestmode );
  if( !partitioner.canSplit( split, cs ) || skipScore >= 2 || ( skipScore == 1 && m_ComprCUCtxList.size() == 2 ) )
  {
    if( split == CU_HORZ_SPLIT ) cuECtx.didHorzSplit = false;
    if( split == CU_VERT_SPLIT ) cuECtx.didVertSplit = false;
    if( split == CU_QUAD_SPLIT ) cuECtx.didQuadSplit = false;
    return false;
  }

  if( isChroma( partitioner.chType ) && !cuECtx.doHorChromaSplit && ( split == CU_HORZ_SPLIT || split == CU_TRIH_SPLIT ) )
  {
    if( split == CU_HORZ_SPLIT )
    {
      cuECtx.didHorzSplit = false;
    }

    return false;
  }

  if( isChroma( partitioner.chType ) && !cuECtx.doVerChromaSplit && ( split == CU_VERT_SPLIT || split == CU_TRIV_SPLIT ) )
  {
    if( split == CU_VERT_SPLIT )
    {
      cuECtx.didVertSplit = false;
    }

    return false;
  }

  if( m_pcEncCfg->m_contentBasedFastQtbt )
  {
    const CompArea& currArea = partitioner.currArea().Y();
    int cuHeight  = currArea.height;
    int cuWidth   = currArea.width;

    const bool condIntraInter = m_pcEncCfg->m_IntraPeriod == 1 ? ( partitioner.currBtDepth == 0 ) : ( cuHeight > 32 && cuWidth > 32 );

    if( cuWidth == cuHeight && condIntraInter && split != CU_QUAD_SPLIT )
    {
      const double th1 = m_pcEncCfg->m_IntraPeriod == 1 ?  1.2              :  1.0;
      const double th2 = m_pcEncCfg->m_IntraPeriod == 1 ? (1.2 / sqrt( 2 )) : (1.0 / sqrt( 2 ));

      if( cuECtx.grad_horVal > th1 * cuECtx.grad_verVal && cuECtx.grad_horVal > th2 * cuECtx.grad_dowVal && cuECtx.grad_horVal > th2 * cuECtx.grad_dupVal && ( split == CU_HORZ_SPLIT || split == CU_TRIH_SPLIT ) )
      {
        return false;
      }
      if( th2 * cuECtx.grad_dupVal < cuECtx.grad_verVal && th2 * cuECtx.grad_dowVal < cuECtx.grad_verVal && th1 * cuECtx.grad_horVal < cuECtx.grad_verVal && ( split == CU_VERT_SPLIT || split == CU_TRIV_SPLIT ) )
      {
        return false;
      }
    }

    if( m_pcEncCfg->m_IntraPeriod == 1 && cuWidth <= 32 && cuHeight <= 32 && bestCS && bestCS->tus.size() == 1 && bestCU && bestCU->depth == partitioner.currDepth && partitioner.currBtDepth > 1 && isLuma( partitioner.chType ) )
    {
      if( !bestCU->rootCbf )
      {
        return false;
      }
    }
  }

  if( m_pcEncCfg->m_qtbttSpeedUp > 1 )
  {
    const int availDepth = cs.pcv->getMaxMTTDepth( slice, partitioner.chType ) - partitioner.currMtDepth;
    if( bestCU && bestCU->skip && availDepth <= ( 3 - m_skipThresholdE0023FastEnc ) && !isModeSplit( lastTestmode ) && split != CU_QUAD_SPLIT )
    {
      return false;
    }
  }
  if( bestCU && bestCU->skip && bestCU->mtDepth >= m_skipThresholdE0023FastEnc && !isModeSplit( lastTestmode ) )
  {
    return false;
  }

  bool resetFeature = false;

  switch( split )
  {
    case CU_QUAD_SPLIT:
      {
        if( !cuECtx.qtBeforeBt && bestCU )
        {
          unsigned maxBTD        = cs.pcv->getMaxMTTDepth( slice, partitioner.chType );
          const CodingUnit *cuBR = bestCS->cus.back();
          unsigned height        = partitioner.currArea().lumaSize().height;

          if(((bestCU->btDepth == 0 && maxBTD >= ((slice.isIntra() && !slice.sps->IBC) ? 3 : 2))
            || (bestCU->btDepth == 1 && cuBR && cuBR->btDepth == 1 && maxBTD >= ((slice.isIntra() && !slice.sps->IBC) ? 4 : 3)))
            && (width <= MAX_TB_SIZEY && height <= MAX_TB_SIZEY)
            && cuECtx.didHorzSplit && cuECtx.didVertSplit )
          {
            return false;
          }
        }
        if( bestCS )
        {
          if( m_pcEncCfg->m_useEarlyCU == 2 && bestCS->cost != MAX_DOUBLE && bestCU && bestCU->skip && cuECtx.nonSkipWasTested && bestCS->cus.size() == 1 )
          {
            return false;
          }

          int stopSplit = (m_pcEncCfg->m_FastInferMerge >> 4) && (bestCS->slice->TLayer > 4);
          int limitBLsize = stopSplit ? 2048 : 1024;
          if((m_pcEncCfg->m_useEarlyCU == 1 || stopSplit) && bestCS->cost != MAX_DOUBLE && bestCU && bestCU->skip && partitioner.currArea().lumaSize().area() < limitBLsize )
          {
            return false;
          }
        }
      }
      break;
    case CU_HORZ_SPLIT:
    case CU_VERT_SPLIT:
      resetFeature = true;
      break;
    case CU_TRIH_SPLIT:
      if( cuECtx.didHorzSplit && bestCU && bestCU->btDepth == partitioner.currBtDepth && !bestCU->rootCbf )
      {
        return false;
      }

      if( m_pcEncCfg->m_qtbttSpeedUp > 1 )
      {
        //unsigned maxBTD = cs.pcv->getMaxMTTDepth( slice, partitioner.chType );
        if( /*maxBTD == 1 && */cuECtx.didHorzSplit && cuECtx.didVertSplit && cuECtx.bestCostHorzSplit > cuECtx.bestCostVertSplit )
        {
          return false;
        }
      }

      if( !cuECtx.doTriHorzSplit )
      {
        return false;
      }

      if (m_pcEncCfg->m_fastTTSplit && cuECtx.bestCostHorzSplit < MAX_DOUBLE)
      {
        if (cuECtx.bestCostBeforeSplit < MAX_DOUBLE && (cuECtx.bestCostHorzSplit > m_pcEncCfg->m_fastTT_th * cuECtx.bestCostBeforeSplit))
        {
          return false;
        }
        if (cuECtx.bestCostVertSplit < MAX_DOUBLE && (cuECtx.bestCostHorzSplit > m_pcEncCfg->m_fastTT_th * cuECtx.bestCostVertSplit))
        {
          return false;
        }
      }
      break;
    case CU_TRIV_SPLIT:
      if( cuECtx.didVertSplit && bestCU && bestCU->btDepth == partitioner.currBtDepth && !bestCU->rootCbf )
      {
        return false;
      }

      if( m_pcEncCfg->m_qtbttSpeedUp > 1 )
      {
        //unsigned maxBTD = cs.pcv->getMaxMTTDepth( slice, partitioner.chType );
        if( /*maxBTD == 1 && */cuECtx.didHorzSplit && cuECtx.didVertSplit && cuECtx.bestCostHorzSplit < cuECtx.bestCostVertSplit )
        {
          return false;
        }
      }

      if( !cuECtx.doTriVertSplit )
      {
        return false;
      }

      if (m_pcEncCfg->m_fastTTSplit && cuECtx.bestCostVertSplit < MAX_DOUBLE)
      {
        if (cuECtx.bestCostBeforeSplit < MAX_DOUBLE && (cuECtx.bestCostVertSplit > m_pcEncCfg->m_fastTT_th * cuECtx.bestCostBeforeSplit))
        {
          return false;
        }
        if (cuECtx.bestCostHorzSplit < MAX_DOUBLE && (cuECtx.bestCostVertSplit > m_pcEncCfg->m_fastTT_th * cuECtx.bestCostHorzSplit))
        {
          return false;
        }
      }
      break;
    default:
      THROW( "Only CU split modes are governed by the EncModeCtrl" );
      return false;
      break;
  }

  switch( split )
  {
    case CU_HORZ_SPLIT:
    case CU_TRIH_SPLIT:
      if( cuECtx.qtBeforeBt && cuECtx.didQuadSplit )
      {
        if( cuECtx.maxQtSubDepth > partitioner.currQtDepth + 1 )
        {
          if( resetFeature ) cuECtx.didHorzSplit = false;
          return false;
        }
      }
      break;
    case CU_VERT_SPLIT:
    case CU_TRIV_SPLIT:
      if( cuECtx.qtBeforeBt && cuECtx.didQuadSplit )
      {
        if( cuECtx.maxQtSubDepth > partitioner.currQtDepth + 1 )
        {
          if( resetFeature ) cuECtx.didVertSplit = false;
          return false;
        }
      }
      break;
    default:
      break;
  }

  if( split == CU_QUAD_SPLIT )
  {
    cuECtx.didQuadSplit = m_pcEncCfg->m_qtbttSpeedUp <= 1 || !!cuECtx.doMoreSplits;
  }

  return m_pcEncCfg->m_qtbttSpeedUp <= 1 || !!cuECtx.doMoreSplits;
}

bool EncModeCtrl::tryMode( const EncTestMode& encTestmode, const CodingStructure &cs, Partitioner& partitioner )
{
  CHECK( isModeSplit( encTestmode ), "wrong method");

  ComprCUCtx& cuECtx = m_ComprCUCtxList.back();

  if( cuECtx.minDepth > partitioner.currQtDepth && partitioner.canSplit( CU_QUAD_SPLIT, cs ) )
  {
    // enforce QT
    return false;
  }

  const Slice&           slice        = *cs.slice;
  const uint32_t         numComp      = getNumberValidComponents( slice.sps->chromaFormatIdc );
  const CodedCUInfo      &relatedCU   = getBlkInfo( partitioner.currArea() );
  const Area             &lumaArea    = partitioner.currArea().Y();
  const CodingStructure* bestCS       = cuECtx.bestCS;
  const EncTestMode      bestMode     = bestCS ?cuECtx.bestMode : EncTestMode();

  if( encTestmode.type == ETM_INTRA )
  {
    // if this is removed, the IntraSearch::xIntraCodingLumaQT needs to be adapted to support Intra TU split
    // also isXXAvailable in IntraPrediction.cpp need to be fixed to check availability within the same CU without isDecomp
    if (m_pcEncCfg->m_FastInferMerge && !slice.isIntra() && !slice.isIRAP() && !(cs.area.lwidth() == 4 && cs.area.lheight() == 4) && !partitioner.isConsIntra())
    {
      if (bestCS && (bestCS->slice->TLayer > (m_pcEncCfg->m_maxTLayer - (m_pcEncCfg->m_FastInferMerge & 7)))
        && (bestCS->bestParent != nullptr) && bestCS->bestParent->cus.size() && (bestCS->bestParent->cus[0]->skip))
      {
        return false;
      }
    }

    if( lumaArea.width > cs.sps->getMaxTbSize() || lumaArea.height > cs.sps->getMaxTbSize() )
    {
      return false;
    }

    if (m_pcEncCfg->m_usePbIntraFast && (!cs.slice->isIntra() || cs.slice->sps->IBC) && cuECtx.interHad == 0 && cuECtx.bestCU && !CU::isIntra(*cuECtx.bestCU))
    {
      return false;
    }

    // INTRA MODES
    if (cs.sps->IBC && !cuECtx.bestTU)
    {
     // return true;
    }
    else
    if( partitioner.isConsIntra() && !cuECtx.bestTU )
    {
      //return true;
    }
    else
    if ( lumaArea.width == 4 && lumaArea.height == 4 && !slice.isIntra() && !cuECtx.bestTU )
    {
      //return true;
    }
    else
    if (!(slice.isIRAP() || bestMode.type == ETM_INTRA || !cuECtx.bestTU ||
      ((!m_pcEncCfg->m_bDisableIntraCUsInInterSlices) && (!relatedCU.isInter || !relatedCU.isIBC) && (
                                    ( cuECtx.bestTU->cbf[0] != 0 ) ||
           ( ( numComp > COMP_Cb ) && cuECtx.bestTU->cbf[1] != 0 ) ||
           ( ( numComp > COMP_Cr ) && cuECtx.bestTU->cbf[2] != 0 )  // avoid very complex intra if it is unlikely
         ) ) ) )
    {
      return false;
    }
    else
    if( cuECtx.bestCS && cuECtx.bestCU && cuECtx.interHad )
    {
      // Get SATD threshold from best Inter-CU
      if( !cs.slice->isIRAP() && m_pcEncCfg->m_usePbIntraFast )
      {
        const DFunc dfunc = DF_HAD;
        DistParam distParam = m_pcRdCost->setDistParam( cs.getOrgBuf( COMP_Y ), cuECtx.bestCS->getPredBuf( COMP_Y ), cs.sps->bitDepths[ CH_L ], dfunc );
        cuECtx.interHad = distParam.distFunc( distParam );
      }
    }
    if ((m_pcEncCfg->m_IBCFastMethod > 1) && !cs.slice->isIntra()//IBC_FAST_METHOD_NOINTRA_IBCCBF0
      && (bestMode.type == ETM_IBC || bestMode.type == ETM_IBC_MERGE)
      && (!cuECtx.bestCU->Y().valid() || cuECtx.bestTU->cbf[0] == 0)
      && (!cuECtx.bestCU->Cb().valid() || cuECtx.bestTU->cbf[1] == 0)
      && (!cuECtx.bestCU->Cr().valid() || cuECtx.bestTU->cbf[2] == 0))
    {
      return false;
    }
    if (m_pcEncCfg->m_FastIntraTools)
    {
      if (relatedCU.relatedCuIsValid)
      {
        cuECtx.relatedCuIsValid = relatedCU.relatedCuIsValid;
      }
      if (relatedCU.isIntra)
      {
        cuECtx.isIntra = relatedCU.isIntra;
      }
    }
  }
  else if( isModeInter( encTestmode ) )
  {
    // INTER MODES (ME + MERGE/SKIP)
    CHECK( slice.isIntra(), "Inter-mode should not be in the I-Slice mode list!" );

    // --- Check if we can quit current mode using SAVE/LOAD coding history

    if( encTestmode.type == ETM_INTER_ME )
    {
      if( encTestmode.opts == ETO_STANDARD )
      {
        if (m_pcEncCfg->m_FastInferMerge)
        {
          if (bestCS && (bestCS->slice->TLayer > (m_pcEncCfg->m_maxTLayer - (m_pcEncCfg->m_FastInferMerge & 7)))
            && (bestCS->bestParent != nullptr) && bestCS->bestParent->cus.size() && (bestCS->bestParent->cus[0]->skip))
          {
            return false;
          }
        }
        if( relatedCU.isSkip || relatedCU.isIntra )
        {
          return false;
        }
      }
    }
  }
  else if (encTestmode.type == ETM_IBC || encTestmode.type == ETM_IBC_MERGE)
  {
    if ((m_pcEncCfg->m_IBCFastMethod > 1) && !(slice.isIRAP() || bestMode.type == ETM_INTRA || !cuECtx.bestTU ||
      ((!m_pcEncCfg->m_bDisableIntraCUsInInterSlices) && (!relatedCU.isInter || !relatedCU.isIBC) && (
        (cuECtx.bestTU->cbf[0] != 0) ||
        ((numComp > COMP_Cb) && cuECtx.bestTU->cbf[1] != 0) ||
        ((numComp > COMP_Cr) && cuECtx.bestTU->cbf[2] != 0)  // avoid very complex intra if it is unlikely
        ))))
    {
      return false;
    }
    if ((m_pcEncCfg->m_IBCFastMethod > 3) &&(lumaArea.width == 4 && lumaArea.height == 4 && !slice.isIntra()))
    {
      return false;
    }
    // IBC MODES
    return slice.sps->IBC && (partitioner.currArea().lumaSize().width < 128 && partitioner.currArea().lumaSize().height < 128);
  }
  else
  {
    THROW("problem");
    return false;
  }

  STAT_COUNT_CU_MODES( partitioner.chType == CH_L, g_cuCounters1D[CU_MODES_TRIED][0][!cs.slice->isIntra() + cs.slice->depth] );
  STAT_COUNT_CU_MODES( partitioner.chType == CH_L && !cs.slice->isIntra(), g_cuCounters2D[CU_MODES_TRIED][Log2( cs.area.lheight() )][Log2( cs.area.lwidth() )] );
  return true;
}


void EncModeCtrl::beforeSplit( Partitioner& partitioner )
{
  ComprCUCtx&           cuECtx  = m_ComprCUCtxList.back();

  if( ! cuECtx.bestCS )
    return;

  CodedCUInfo    &relatedCU   = getBlkInfo( partitioner.currArea() );
  const CodingUnit&  bestCU   = *cuECtx.bestCU;

  cuECtx.bestNsPredMode       = cuECtx.bestMode;
  cuECtx.bestCostBeforeSplit  = cuECtx.bestCS->cost;

  setFromCs( *cuECtx.bestCS, cuECtx.bestMode, partitioner );

  if( bestCU.skip )
  {
    cuECtx.doMoreSplits--;
  }

  if( partitioner.modeType == MODE_TYPE_INTRA && partitioner.chType == CH_L )
  {
    return; //not set best coding mode for intra coding pass
  }

  // assume the non-split modes are done and set the marks for the best found mode
  if( CU::isInter( bestCU ) )
  {
    relatedCU.isInter     = true;
    relatedCU.isSkip     |= bestCU.skip;
    relatedCU.isMMVDSkip |= bestCU.mmvdSkip;
    relatedCU.BcwIdx      = bestCU.BcwIdx;
  }
  else if (CU::isIBC(bestCU))
  {
    relatedCU.isIBC = true;
    relatedCU.isSkip |= bestCU.skip;
  }
  else if( CU::isIntra( bestCU ) )
  {
    relatedCU.isIntra     = true;
    if (m_pcEncCfg->m_FastIntraTools)
    {
      if ( cuECtx.intraWasTested && (!relatedCU.relatedCuIsValid || cuECtx.bestCS->cost < relatedCU.bestCost))
      {
        relatedCU.bestCost = cuECtx.bestCS->cost;
        relatedCU.relatedCuIsValid = true;
      }
    }
  }
  cuECtx.isBestNoSplitSkip = bestCU.skip;
}


bool EncModeCtrl::useModeResult( const EncTestMode& encTestmode, CodingStructure*& tempCS, Partitioner& partitioner, const bool useEDO )
{
  ComprCUCtx& cuECtx = m_ComprCUCtxList.back();


  if(      encTestmode.type == ETM_SPLIT_BT_H )
  {
    cuECtx.bestCostHorzSplit = tempCS->cost;
  }
  else if( encTestmode.type == ETM_SPLIT_BT_V )
  {
    cuECtx.bestCostVertSplit = tempCS->cost;
  }
  else if( !isModeSplit( encTestmode ) && isModeInter( encTestmode ) && tempCS->cus.size() == 1 )
  {
    cuECtx.nonSkipWasTested |= !tempCS->cus.front()->skip;
  }
  if (m_pcEncCfg->m_AMVRspeed && encTestmode.type == ETM_INTER_ME)
  {
    int imvMode = (encTestmode.opts & ETO_IMV) >> ETO_IMV_SHIFT;

    if (imvMode == 0)
    {
      if (tempCS->cost < cuECtx.bestCostNoImv )
      {
        cuECtx.bestCostNoImv = tempCS->cost;
      }
    }
  }

  if( encTestmode.type == ETM_SPLIT_QT )
  {
    int maxQtD = 0;
    for( const auto& cu : tempCS->cus )
    {
      maxQtD = std::max<int>( maxQtD, cu->qtDepth );
    }
    cuECtx.maxQtSubDepth = maxQtD;
  }

  int maxMtD = tempCS->pcv->getMaxMTTDepth( *tempCS->slice, partitioner.chType ) + partitioner.currImplicitBtDepth;

  if( encTestmode.type == ETM_SPLIT_BT_H )
  {
    if( tempCS->cus.size() > 2 )
    {
      int h_2   = tempCS->area.blocks[partitioner.chType].height / 2;
      int cu1_h = tempCS->cus.front()->blocks[partitioner.chType].height;
      int cu2_h = tempCS->cus.back() ->blocks[partitioner.chType].height;

      cuECtx.doTriHorzSplit = cu1_h < h_2 || cu2_h < h_2 || partitioner.currMtDepth + 1 == maxMtD;
    }
  }
  else if( encTestmode.type == ETM_SPLIT_BT_V )
  {
    if( tempCS->cus.size() > 2 )
    {
      int w_2   = tempCS->area.blocks[partitioner.chType].width / 2;
      int cu1_w = tempCS->cus.front()->blocks[partitioner.chType].width;
      int cu2_w = tempCS->cus.back() ->blocks[partitioner.chType].width;

      cuECtx.doTriVertSplit = cu1_w < w_2 || cu2_w < w_2 || partitioner.currMtDepth + 1 == maxMtD;
    }
  }

  if( encTestmode.type == ETM_SPLIT_BT_V || encTestmode.type == ETM_SPLIT_BT_H || encTestmode.type == ETM_SPLIT_QT )
  {
    bool isAllSkip = true;

    for( const auto& cu : tempCS->cus )
    {
      isAllSkip &= cu->skip;
    }

    if( isAllSkip ) cuECtx.doMoreSplits--;
    if( isAllSkip && encTestmode.type == ETM_SPLIT_QT )
                    cuECtx.doMoreSplits--;
  }

  // for now just a simple decision based on RD-cost or choose tempCS if bestCS is not yet coded
  if( tempCS->cost != MAX_DOUBLE && ( !cuECtx.bestCS || ( ( tempCS->cost + ( useEDO ? tempCS->costDbOffset : 0 ) ) < ( cuECtx.bestCS->cost + ( useEDO ? cuECtx.bestCS->costDbOffset : 0 ) ) ) ) )
  {
    cuECtx.bestCS   = tempCS;
    cuECtx.bestCU   = tempCS->cus[0];
    cuECtx.bestTU   = cuECtx.bestCU->firstTU;
    cuECtx.bestMode = encTestmode;

    return true;
  }
  else
  {
    return false;
  }
}

} // namespace vvenc

//! \}


Coverage Report

Created: 2026-10-28 06:18