Skip to content
Snippets Groups Projects
EncCu.cpp 175 KiB
Newer Older
/* The copyright in this software is being made available under the BSD
 * License, included below. This software may be subject to other third party
 * and contributor rights, including patent rights, and no such rights are
 * granted under this license.
 *
 * Copyright (c) 2010-2019, ITU/ISO/IEC
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted 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 ITU/ISO/IEC nor the names of its contributors may
 *    be used to endorse or promote products derived from this software without
 *    specific prior written permission.
 *
 * 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     EncCu.cpp
    \brief    Coding Unit (CU) encoder class
*/

#include "EncCu.h"

#include "EncLib.h"
#include "Analyze.h"
#include "AQp.h"

#include "CommonLib/dtrace_codingstruct.h"
#include "CommonLib/Picture.h"
#include "CommonLib/UnitTools.h"


#include "CommonLib/dtrace_buffer.h"

#include <stdio.h>
#include <cmath>
#include <algorithm>
#if ENABLE_WPP_PARALLELISM
#include <mutex>
extern std::recursive_mutex g_cache_mutex;
#endif



//! \ingroup EncoderLib
//! \{

// ====================================================================================================================
// Constructor / destructor / create / destroy
// ====================================================================================================================
#if JVET_M0883_TRIANGLE_SIGNALING
const TriangleMotionInfo  EncCu::m_triangleModeTest[TRIANGLE_MAX_NUM_CANDS] =
{
  TriangleMotionInfo( 0, 1, 0 ), TriangleMotionInfo( 1, 0, 1 ), TriangleMotionInfo( 1, 0, 2 ), TriangleMotionInfo( 0, 0, 1 ), TriangleMotionInfo( 0, 2, 0 ),
  TriangleMotionInfo( 1, 0, 3 ), TriangleMotionInfo( 1, 0, 4 ), TriangleMotionInfo( 1, 1, 0 ), TriangleMotionInfo( 0, 3, 0 ), TriangleMotionInfo( 0, 4, 0 ),
  TriangleMotionInfo( 0, 0, 2 ), TriangleMotionInfo( 0, 1, 2 ), TriangleMotionInfo( 1, 1, 2 ), TriangleMotionInfo( 0, 0, 4 ), TriangleMotionInfo( 0, 0, 3 ),
  TriangleMotionInfo( 0, 1, 3 ), TriangleMotionInfo( 0, 1, 4 ), TriangleMotionInfo( 1, 1, 4 ), TriangleMotionInfo( 1, 1, 3 ), TriangleMotionInfo( 1, 2, 1 ),
  TriangleMotionInfo( 1, 2, 0 ), TriangleMotionInfo( 0, 2, 1 ), TriangleMotionInfo( 0, 4, 3 ), TriangleMotionInfo( 1, 3, 0 ), TriangleMotionInfo( 1, 3, 2 ),
  TriangleMotionInfo( 1, 3, 4 ), TriangleMotionInfo( 1, 4, 0 ), TriangleMotionInfo( 1, 3, 1 ), TriangleMotionInfo( 1, 2, 3 ), TriangleMotionInfo( 1, 4, 1 ),
  TriangleMotionInfo( 0, 4, 1 ), TriangleMotionInfo( 0, 2, 3 ), TriangleMotionInfo( 1, 4, 2 ), TriangleMotionInfo( 0, 3, 2 ), TriangleMotionInfo( 1, 4, 3 ),
  TriangleMotionInfo( 0, 3, 1 ), TriangleMotionInfo( 0, 2, 4 ), TriangleMotionInfo( 1, 2, 4 ), TriangleMotionInfo( 0, 4, 2 ), TriangleMotionInfo( 0, 3, 4 ),
};
#endif

void EncCu::create( EncCfg* encCfg )
{
  unsigned      uiMaxWidth    = encCfg->getMaxCUWidth();
  unsigned      uiMaxHeight   = encCfg->getMaxCUHeight();
  ChromaFormat  chromaFormat  = encCfg->getChromaFormatIdc();

  unsigned      numWidths     = gp_sizeIdxInfo->numWidths();
  unsigned      numHeights    = gp_sizeIdxInfo->numHeights();
  m_pTempCS = new CodingStructure**  [numWidths];
  m_pBestCS = new CodingStructure**  [numWidths];

Li's avatar
Li committed
  m_pTempMotLUTs = new LutMotionCand**[numWidths];
  m_pBestMotLUTs = new LutMotionCand**[numWidths];
  m_pSplitTempMotLUTs = new LutMotionCand**[numWidths];
  for( unsigned w = 0; w < numWidths; w++ )
  {
    m_pTempCS[w] = new CodingStructure*  [numHeights];
    m_pBestCS[w] = new CodingStructure*  [numHeights];
Li's avatar
Li committed
    m_pTempMotLUTs[w] = new LutMotionCand*[numHeights];
    m_pBestMotLUTs[w] = new LutMotionCand*[numHeights];
    m_pSplitTempMotLUTs[w] = new LutMotionCand*[numHeights];

    for( unsigned h = 0; h < numHeights; h++ )
    {
      unsigned width  = gp_sizeIdxInfo->sizeFrom( w );
      unsigned height = gp_sizeIdxInfo->sizeFrom( h );

Karsten Suehring's avatar
Karsten Suehring committed
      if( gp_sizeIdxInfo->isCuSize( width ) && gp_sizeIdxInfo->isCuSize( height ) )
      {
        m_pTempCS[w][h] = new CodingStructure( m_unitCache.cuCache, m_unitCache.puCache, m_unitCache.tuCache );
        m_pBestCS[w][h] = new CodingStructure( m_unitCache.cuCache, m_unitCache.puCache, m_unitCache.tuCache );

        m_pTempCS[w][h]->create( chromaFormat, Area( 0, 0, width, height ), false );
        m_pBestCS[w][h]->create( chromaFormat, Area( 0, 0, width, height ), false );
Li's avatar
Li committed
        m_pTempMotLUTs[w][h] = new LutMotionCand ;
        m_pBestMotLUTs[w][h] = new LutMotionCand ;
        m_pSplitTempMotLUTs[w][h] = new LutMotionCand;
Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
        m_pSplitTempMotLUTs[w][h]->currCnt = 0;
        m_pSplitTempMotLUTs[w][h]->currCntIBC = 0;
        m_pSplitTempMotLUTs[w][h]->motionCand = nullptr;
        m_pSplitTempMotLUTs[w][h]->motionCand = new MotionInfo[MAX_NUM_HMVP_CANDS * 2];

        m_pTempMotLUTs[w][h]->currCnt = 0;
        m_pTempMotLUTs[w][h]->currCntIBC = 0;
        m_pTempMotLUTs[w][h]->motionCand = nullptr;
        m_pTempMotLUTs[w][h]->motionCand = new MotionInfo[MAX_NUM_HMVP_CANDS * 2];

        m_pBestMotLUTs[w][h]->currCnt = 0;
        m_pBestMotLUTs[w][h]->currCntIBC = 0;
        m_pBestMotLUTs[w][h]->motionCand = nullptr;
        m_pBestMotLUTs[w][h]->motionCand = new MotionInfo[MAX_NUM_HMVP_CANDS * 2];
#else
        m_pSplitTempMotLUTs[w][h]->currCnt = 0;
        m_pSplitTempMotLUTs[w][h]->motionCand = nullptr;
        m_pSplitTempMotLUTs[w][h]->motionCand = new MotionInfo[MAX_NUM_HMVP_CANDS];
        m_pTempMotLUTs[w][h]->currCnt = 0;
        m_pTempMotLUTs[w][h]->motionCand = nullptr;
        m_pTempMotLUTs[w][h]->motionCand = new MotionInfo[MAX_NUM_HMVP_CANDS];
        m_pBestMotLUTs[w][h]->currCnt = 0;
        m_pBestMotLUTs[w][h]->motionCand = nullptr;
        m_pBestMotLUTs[w][h]->motionCand = new MotionInfo[MAX_NUM_HMVP_CANDS];
Yu Han's avatar
Yu Han committed
#endif
      }
      else
      {
        m_pTempCS[w][h] = nullptr;
        m_pBestCS[w][h] = nullptr;
        m_pTempMotLUTs[w][h] = nullptr;
        m_pBestMotLUTs[w][h] = nullptr;
        m_pSplitTempMotLUTs[w][h] = nullptr;
      }
    }
  }

  // WIA: only the weight==height case is relevant without QTBT
  m_pImvTempCS = nullptr;

  m_cuChromaQpOffsetIdxPlus1 = 0;

  unsigned maxDepth = numWidths + numHeights;

Karsten Suehring's avatar
Karsten Suehring committed
  m_modeCtrl = new EncModeCtrlMTnoRQT();
  for (unsigned ui = 0; ui < MMVD_MRG_MAX_RD_BUF_NUM; ui++)
  {
    m_acMergeBuffer[ui].create( chromaFormat, Area( 0, 0, uiMaxWidth, uiMaxHeight ) );
  }
  for (unsigned ui = 0; ui < MRG_MAX_NUM_CANDS; ui++)
  {
    m_acRealMergeBuffer[ui].create(chromaFormat, Area(0, 0, uiMaxWidth, uiMaxHeight));
  }
#if JVET_M0883_TRIANGLE_SIGNALING
  for( unsigned ui = 0; ui < TRIANGLE_MAX_NUM_UNI_CANDS; ui++ )
  {
    for( unsigned uj = 0; uj < TRIANGLE_MAX_NUM_UNI_CANDS; uj++ )
    {
      if(ui == uj)
        continue;
      uint8_t idxBits0 = ui + (ui == TRIANGLE_MAX_NUM_UNI_CANDS - 1 ? 0 : 1);
      uint8_t candIdx1Enc = uj - (uj > ui ? 1 : 0);
      uint8_t idxBits1 = candIdx1Enc + (candIdx1Enc == TRIANGLE_MAX_NUM_UNI_CANDS - 2 ? 0 : 1);
      m_triangleIdxBins[1][ui][uj] = m_triangleIdxBins[0][ui][uj] = 1 + idxBits0 + idxBits1;
    }
  }
#endif
  for( unsigned ui = 0; ui < TRIANGLE_MAX_NUM_CANDS; ui++ )
  {
rlliao's avatar
rlliao committed
    m_acTriangleWeightedBuffer[ui].create( chromaFormat, Area( 0, 0, uiMaxWidth, uiMaxHeight ) );

void EncCu::destroy()
{
  unsigned numWidths  = gp_sizeIdxInfo->numWidths();
  unsigned numHeights = gp_sizeIdxInfo->numHeights();

  for( unsigned w = 0; w < numWidths; w++ )
  {
    for( unsigned h = 0; h < numHeights; h++ )
    {
Karsten Suehring's avatar
Karsten Suehring committed
      if( m_pBestCS[w][h] ) m_pBestCS[w][h]->destroy();
      if( m_pTempCS[w][h] ) m_pTempCS[w][h]->destroy();
Karsten Suehring's avatar
Karsten Suehring committed
      delete m_pBestCS[w][h];
      delete m_pTempCS[w][h];
      if (m_pTempMotLUTs[w][h])
      {
        delete[] m_pTempMotLUTs[w][h]->motionCand;
        m_pTempMotLUTs[w][h]->motionCand = nullptr;
Xiang Li's avatar
Xiang Li committed
        delete m_pTempMotLUTs[w][h];
Karsten Suehring's avatar
Karsten Suehring committed
      }
      if (m_pBestMotLUTs[w][h])
      {
        delete[] m_pBestMotLUTs[w][h]->motionCand;
        m_pBestMotLUTs[w][h]->motionCand = nullptr;
Xiang Li's avatar
Xiang Li committed
        delete m_pBestMotLUTs[w][h];
Karsten Suehring's avatar
Karsten Suehring committed
      }
Karsten Suehring's avatar
Karsten Suehring committed
      if (m_pSplitTempMotLUTs[w][h])
      {
        delete[] m_pSplitTempMotLUTs[w][h]->motionCand;
        m_pSplitTempMotLUTs[w][h]->motionCand = nullptr;
Xiang Li's avatar
Xiang Li committed
        delete m_pSplitTempMotLUTs[w][h];
    delete[] m_pBestMotLUTs[w];
    delete[] m_pTempMotLUTs[w];
    delete[] m_pSplitTempMotLUTs[w];
  }

  delete[] m_pBestCS; m_pBestCS = nullptr;
  delete[] m_pTempCS; m_pTempCS = nullptr;
  delete[] m_pSplitTempMotLUTs; m_pSplitTempMotLUTs = nullptr;
  delete[] m_pBestMotLUTs; m_pBestMotLUTs = nullptr;
  delete[] m_pTempMotLUTs; m_pTempMotLUTs = nullptr;
Taoran Lu's avatar
Taoran Lu committed
#if JVET_M0427_INLOOP_RESHAPER && REUSE_CU_RESULTS
  if (m_tmpStorageLCU)
  {
    m_tmpStorageLCU->destroy();
    delete m_tmpStorageLCU;  m_tmpStorageLCU = nullptr;
  }
#endif

#if REUSE_CU_RESULTS
  m_modeCtrl->destroy();

#endif
  delete m_modeCtrl;
  m_modeCtrl = nullptr;

  // WIA: only the weight==height case is relevant without QTBT
  if( m_pImvTempCS )
  {
    for( unsigned w = 0; w < numWidths; w++ )
    {
Karsten Suehring's avatar
Karsten Suehring committed
      if( m_pImvTempCS[w] )
Karsten Suehring's avatar
Karsten Suehring committed
        m_pImvTempCS[w]->destroy();
        delete[] m_pImvTempCS[w];
  for (unsigned ui = 0; ui < MMVD_MRG_MAX_RD_BUF_NUM; ui++)
  for (unsigned ui = 0; ui < MRG_MAX_NUM_CANDS; ui++)
  {
    m_acRealMergeBuffer[ui].destroy();
  }
  for( unsigned ui = 0; ui < TRIANGLE_MAX_NUM_CANDS; ui++ )
  {
rlliao's avatar
rlliao committed
    m_acTriangleWeightedBuffer[ui].destroy();
}



EncCu::~EncCu()
{
}



/** \param    pcEncLib      pointer of encoder class
 */
void EncCu::init( EncLib* pcEncLib, const SPS& sps PARL_PARAM( const int tId ) )
{
  m_pcEncCfg           = pcEncLib;
  m_pcIntraSearch      = pcEncLib->getIntraSearch( PARL_PARAM0( tId ) );
  m_pcInterSearch      = pcEncLib->getInterSearch( PARL_PARAM0( tId ) );
  m_pcTrQuant          = pcEncLib->getTrQuant( PARL_PARAM0( tId ) );
  m_pcRdCost           = pcEncLib->getRdCost ( PARL_PARAM0( tId ) );
  m_CABACEstimator     = pcEncLib->getCABACEncoder( PARL_PARAM0( tId ) )->getCABACEstimator( &sps );
  m_CABACEstimator->setEncCu(this);
  m_CtxCache           = pcEncLib->getCtxCache( PARL_PARAM0( tId ) );
  m_pcRateCtrl         = pcEncLib->getRateCtrl();
  m_pcSliceEncoder     = pcEncLib->getSliceEncoder();
#if ENABLE_SPLIT_PARALLELISM || ENABLE_WPP_PARALLELISM
  m_pcEncLib           = pcEncLib;
  m_dataId             = tId;
#endif
Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
  m_pcLoopFilter       = pcEncLib->getLoopFilter();
#endif
#if JVET_M0170_MRG_SHARELIST
  m_shareState = NO_SHARE;
  m_pcInterSearch->setShareState(0);
  setShareStateDec(0);
#endif

#if JVET_M0170_MRG_SHARELIST
  m_shareBndPosX = -1;
  m_shareBndPosY = -1;
  m_shareBndSizeW = 0;
  m_shareBndSizeH = 0;
#endif
Adam Wieckowski's avatar
Adam Wieckowski committed
#if REUSE_CU_RESULTS || JVET_M0170_MRG_SHARELIST || JVET_M0427_INLOOP_RESHAPER
  DecCu::init( m_pcTrQuant, m_pcIntraSearch, m_pcInterSearch );

#endif
  m_modeCtrl->init( m_pcEncCfg, m_pcRateCtrl, m_pcRdCost );

  m_pcInterSearch->setModeCtrl( m_modeCtrl );
#if JVET_M0102_INTRA_SUBPARTITIONS
  m_pcIntraSearch->setModeCtrl( m_modeCtrl );
#endif
  ::memset(m_subMergeBlkSize, 0, sizeof(m_subMergeBlkSize));
  ::memset(m_subMergeBlkNum, 0, sizeof(m_subMergeBlkNum));
  m_prevPOC = MAX_UINT;
Xiaozhong Xu's avatar
Xiaozhong Xu committed

#if  JVET_M0255_FRACMMVD_SWITCH
  if ( ( m_pcEncCfg->getIBCHashSearch() && m_pcEncCfg->getIBCMode() ) || m_pcEncCfg->getAllowDisFracMMVD() )
Yu Han's avatar
Yu Han committed
  if (m_pcEncCfg->getIBCHashSearch() && m_pcEncCfg->getIBCMode())
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  {
Yu Han's avatar
Yu Han committed
    m_ibcHashMap.init(m_pcEncCfg->getSourceWidth(), m_pcEncCfg->getSourceHeight());
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  }
}

// ====================================================================================================================
// Public member functions
// ====================================================================================================================

void EncCu::compressCtu( CodingStructure& cs, const UnitArea& area, const unsigned ctuRsAddr, const int prevQP[], const int currQP[] )
{
#if !JVET_M0255_FRACMMVD_SWITCH
  if (m_pcEncCfg->getIBCHashSearch() && ctuRsAddr == 0 && cs.slice->getSPS()->getIBCMode())
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  {
Taoran Lu's avatar
Taoran Lu committed
#if JVET_M0427_INLOOP_RESHAPER
    if (cs.slice->getSPS()->getUseReshaper() && m_pcReshape->getCTUFlag())
      cs.picture->getOrigBuf(COMPONENT_Y).rspSignal(m_pcReshape->getFwdLUT());
#endif
Yu Han's avatar
Yu Han committed
    m_ibcHashMap.rebuildPicHashMap(cs.picture->getOrigBuf());
Taoran Lu's avatar
Taoran Lu committed
#if JVET_M0427_INLOOP_RESHAPER
    if (cs.slice->getSPS()->getUseReshaper() && m_pcReshape->getCTUFlag())
      cs.picture->getOrigBuf().copyFrom(cs.picture->getTrueOrigBuf());
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  }
  m_modeCtrl->initCTUEncoding( *cs.slice );

#if ENABLE_SPLIT_PARALLELISM
  if( m_pcEncCfg->getNumSplitThreads() > 1 )
  {
    for( int jId = 1; jId < NUM_RESERVERD_SPLIT_JOBS; jId++ )
    {
      EncCu*            jobEncCu  = m_pcEncLib->getCuEncoder( cs.picture->scheduler.getSplitDataId( jId ) );
      CacheBlkInfoCtrl* cacheCtrl = dynamic_cast< CacheBlkInfoCtrl* >( jobEncCu->m_modeCtrl );
      if( cacheCtrl )
      {
        cacheCtrl->init( *cs.slice );
      }
    }
  }

  if( auto* cacheCtrl = dynamic_cast<CacheBlkInfoCtrl*>( m_modeCtrl ) ) { cacheCtrl->tick(); }
#endif
  // init the partitioning manager
  Partitioner *partitioner = PartitionerFactory::get( *cs.slice );
  partitioner->initCtu( area, CH_L, *cs.slice );
Yu Han's avatar
Yu Han committed
  if (m_pcEncCfg->getIBCMode())
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  {
    if (area.lx() == 0 && area.ly() == 0)
    {
      m_pcInterSearch->resetIbcSearch();
    }
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    m_pcInterSearch->resetCtuRecord();
Yu Han's avatar
Yu Han committed
    m_ctuIbcSearchRangeX = m_pcEncCfg->getIBCLocalSearchRangeX();
    m_ctuIbcSearchRangeY = m_pcEncCfg->getIBCLocalSearchRangeY();
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  }
Yu Han's avatar
Yu Han committed
  if (m_pcEncCfg->getIBCMode() && m_pcEncCfg->getIBCHashSearch() && (m_pcEncCfg->getIBCFastMethod() & IBC_FAST_METHOD_ADAPTIVE_SEARCHRANGE))
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  {
Yu Han's avatar
Yu Han committed
    const int hashHitRatio = m_ibcHashMap.getHashHitRatio(area.Y()); // in percent
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    if (hashHitRatio < 5) // 5%
    {
Yu Han's avatar
Yu Han committed
      m_ctuIbcSearchRangeX >>= 1;
      m_ctuIbcSearchRangeY >>= 1;
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    }
Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
    if (cs.slice->getNumRefIdx(REF_PIC_LIST_0) > 0)
#else
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    if (cs.slice->getNumRefIdx(REF_PIC_LIST_0) > 1)
Yu Han's avatar
Yu Han committed
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    {
Yu Han's avatar
Yu Han committed
      m_ctuIbcSearchRangeX >>= 1;
      m_ctuIbcSearchRangeY >>= 1;
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    }
  }
  // init current context pointer
  m_CurrCtx = m_CtxBuffer.data();

  CodingStructure *tempCS = m_pTempCS[gp_sizeIdxInfo->idxFrom( area.lumaSize().width )][gp_sizeIdxInfo->idxFrom( area.lumaSize().height )];
  CodingStructure *bestCS = m_pBestCS[gp_sizeIdxInfo->idxFrom( area.lumaSize().width )][gp_sizeIdxInfo->idxFrom( area.lumaSize().height )];
Li's avatar
Li committed
  LutMotionCand *tempMotCandLUTs = m_pTempMotLUTs[gp_sizeIdxInfo->idxFrom(area.lumaSize().width)][gp_sizeIdxInfo->idxFrom(area.lumaSize().height)];
  LutMotionCand *bestMotCandLUTs = m_pBestMotLUTs[gp_sizeIdxInfo->idxFrom(area.lumaSize().width)][gp_sizeIdxInfo->idxFrom(area.lumaSize().height)];
  cs.slice->copyMotionLUTs(cs.slice->getMotionLUTs(), tempMotCandLUTs);
  cs.slice->copyMotionLUTs(cs.slice->getMotionLUTs(), bestMotCandLUTs);

  cs.initSubStructure( *tempCS, partitioner->chType, partitioner->currArea(), false );
  cs.initSubStructure( *bestCS, partitioner->chType, partitioner->currArea(), false );
  tempCS->currQP[CH_L] = bestCS->currQP[CH_L] =
  tempCS->baseQP       = bestCS->baseQP       = currQP[CH_L];
  tempCS->prevQP[CH_L] = bestCS->prevQP[CH_L] = prevQP[CH_L];

  xCompressCU( tempCS, bestCS, *partitioner
    , tempMotCandLUTs
    , bestMotCandLUTs
  );

  // all signals were already copied during compression if the CTU was split - at this point only the structures are copied to the top level CS
Taoran Lu's avatar
Taoran Lu committed
#if JVET_M0427_INLOOP_RESHAPER
  const bool copyUnsplitCTUSignals = bestCS->cus.size() == 1;
#else
  const bool copyUnsplitCTUSignals = bestCS->cus.size() == 1 && KEEP_PRED_AND_RESI_SIGNALS;
Taoran Lu's avatar
Taoran Lu committed
#endif
  cs.useSubStructure( *bestCS, partitioner->chType, CS::getArea( *bestCS, area, partitioner->chType ), copyUnsplitCTUSignals, false, false, copyUnsplitCTUSignals );
  cs.slice->copyMotionLUTs(bestMotCandLUTs, cs.slice->getMotionLUTs());
  if (CS::isDualITree (cs) && isChromaEnabled (cs.pcv->chrFormat))
  {
    m_CABACEstimator->getCtx() = m_CurrCtx->start;

    partitioner->initCtu( area, CH_C, *cs.slice );

    cs.initSubStructure( *tempCS, partitioner->chType, partitioner->currArea(), false );
    cs.initSubStructure( *bestCS, partitioner->chType, partitioner->currArea(), false );
    tempCS->currQP[CH_C] = bestCS->currQP[CH_C] =
    tempCS->baseQP       = bestCS->baseQP       = currQP[CH_C];
    tempCS->prevQP[CH_C] = bestCS->prevQP[CH_C] = prevQP[CH_C];

    xCompressCU( tempCS, bestCS, *partitioner
      , tempMotCandLUTs
      , bestMotCandLUTs
    );
Taoran Lu's avatar
Taoran Lu committed
#if JVET_M0427_INLOOP_RESHAPER
    const bool copyUnsplitCTUSignals = bestCS->cus.size() == 1;
#else
    const bool copyUnsplitCTUSignals = bestCS->cus.size() == 1 && KEEP_PRED_AND_RESI_SIGNALS;
Taoran Lu's avatar
Taoran Lu committed
#endif
    cs.useSubStructure( *bestCS, partitioner->chType, CS::getArea( *bestCS, area, partitioner->chType ), copyUnsplitCTUSignals, false, false, copyUnsplitCTUSignals );
  }

  if (m_pcEncCfg->getUseRateCtrl())
  {
    (m_pcRateCtrl->getRCPic()->getLCU(ctuRsAddr)).m_actualMSE = (double)bestCS->dist / (double)m_pcRateCtrl->getRCPic()->getLCU(ctuRsAddr).m_numberOfPixel;
  }
  // reset context states and uninit context pointer
  m_CABACEstimator->getCtx() = m_CurrCtx->start;
  m_CurrCtx                  = 0;
  delete partitioner;

#if ENABLE_SPLIT_PARALLELISM && ENABLE_WPP_PARALLELISM
  if( m_pcEncCfg->getNumSplitThreads() > 1 && m_pcEncCfg->getNumWppThreads() > 1 )
  {
    cs.picture->finishCtuPart( area );
  }
#endif

  // Ensure that a coding was found
  // Selected mode's RD-cost must be not MAX_DOUBLE.
  CHECK( bestCS->cus.empty()                                   , "No possible encoding found" );
  CHECK( bestCS->cus[0]->predMode == NUMBER_OF_PREDICTION_MODES, "No possible encoding found" );
  CHECK( bestCS->cost             == MAX_DOUBLE                , "No possible encoding found" );
}

// ====================================================================================================================
// Protected member functions
// ====================================================================================================================

static int xCalcHADs8x8_ISlice(const Pel *piOrg, const int iStrideOrg)
{
  int k, i, j, jj;
  int diff[64], m1[8][8], m2[8][8], m3[8][8], iSumHad = 0;

  for (k = 0; k < 64; k += 8)
  {
    diff[k + 0] = piOrg[0];
    diff[k + 1] = piOrg[1];
    diff[k + 2] = piOrg[2];
    diff[k + 3] = piOrg[3];
    diff[k + 4] = piOrg[4];
    diff[k + 5] = piOrg[5];
    diff[k + 6] = piOrg[6];
    diff[k + 7] = piOrg[7];

    piOrg += iStrideOrg;
  }

  //horizontal
  for (j = 0; j < 8; j++)
  {
    jj = j << 3;
    m2[j][0] = diff[jj    ] + diff[jj + 4];
    m2[j][1] = diff[jj + 1] + diff[jj + 5];
    m2[j][2] = diff[jj + 2] + diff[jj + 6];
    m2[j][3] = diff[jj + 3] + diff[jj + 7];
    m2[j][4] = diff[jj    ] - diff[jj + 4];
    m2[j][5] = diff[jj + 1] - diff[jj + 5];
    m2[j][6] = diff[jj + 2] - diff[jj + 6];
    m2[j][7] = diff[jj + 3] - diff[jj + 7];

    m1[j][0] = m2[j][0] + m2[j][2];
    m1[j][1] = m2[j][1] + m2[j][3];
    m1[j][2] = m2[j][0] - m2[j][2];
    m1[j][3] = m2[j][1] - m2[j][3];
    m1[j][4] = m2[j][4] + m2[j][6];
    m1[j][5] = m2[j][5] + m2[j][7];
    m1[j][6] = m2[j][4] - m2[j][6];
    m1[j][7] = m2[j][5] - m2[j][7];

    m2[j][0] = m1[j][0] + m1[j][1];
    m2[j][1] = m1[j][0] - m1[j][1];
    m2[j][2] = m1[j][2] + m1[j][3];
    m2[j][3] = m1[j][2] - m1[j][3];
    m2[j][4] = m1[j][4] + m1[j][5];
    m2[j][5] = m1[j][4] - m1[j][5];
    m2[j][6] = m1[j][6] + m1[j][7];
    m2[j][7] = m1[j][6] - m1[j][7];
  }

  //vertical
  for (i = 0; i < 8; i++)
  {
    m3[0][i] = m2[0][i] + m2[4][i];
    m3[1][i] = m2[1][i] + m2[5][i];
    m3[2][i] = m2[2][i] + m2[6][i];
    m3[3][i] = m2[3][i] + m2[7][i];
    m3[4][i] = m2[0][i] - m2[4][i];
    m3[5][i] = m2[1][i] - m2[5][i];
    m3[6][i] = m2[2][i] - m2[6][i];
    m3[7][i] = m2[3][i] - m2[7][i];

    m1[0][i] = m3[0][i] + m3[2][i];
    m1[1][i] = m3[1][i] + m3[3][i];
    m1[2][i] = m3[0][i] - m3[2][i];
    m1[3][i] = m3[1][i] - m3[3][i];
    m1[4][i] = m3[4][i] + m3[6][i];
    m1[5][i] = m3[5][i] + m3[7][i];
    m1[6][i] = m3[4][i] - m3[6][i];
    m1[7][i] = m3[5][i] - m3[7][i];

    m2[0][i] = m1[0][i] + m1[1][i];
    m2[1][i] = m1[0][i] - m1[1][i];
    m2[2][i] = m1[2][i] + m1[3][i];
    m2[3][i] = m1[2][i] - m1[3][i];
    m2[4][i] = m1[4][i] + m1[5][i];
    m2[5][i] = m1[4][i] - m1[5][i];
    m2[6][i] = m1[6][i] + m1[7][i];
    m2[7][i] = m1[6][i] - m1[7][i];
  }

  for (i = 0; i < 8; i++)
  {
    for (j = 0; j < 8; j++)
    {
      iSumHad += abs(m2[i][j]);
    }
  }
  iSumHad -= abs(m2[0][0]);
  iSumHad = (iSumHad + 2) >> 2;
  return(iSumHad);
}

int  EncCu::updateCtuDataISlice(const CPelBuf buf)
{
  int  xBl, yBl;
  const int iBlkSize = 8;
  const Pel* pOrgInit = buf.buf;
  int  iStrideOrig = buf.stride;

  int iSumHad = 0;
  for( yBl = 0; ( yBl + iBlkSize ) <= buf.height; yBl += iBlkSize )
  {
    for( xBl = 0; ( xBl + iBlkSize ) <= buf.width; xBl += iBlkSize )
    {
      const Pel* pOrg = pOrgInit + iStrideOrig*yBl + xBl;
      iSumHad += xCalcHADs8x8_ISlice( pOrg, iStrideOrig );
    }
  }
  return( iSumHad );
}

bool EncCu::xCheckBestMode( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
  bool bestCSUpdated = false;
  if( !tempCS->cus.empty() )
  {
    if( tempCS->cus.size() == 1 )
    {
      const CodingUnit& cu = *tempCS->cus.front();
      CHECK( cu.skip && !cu.firstPU->mergeFlag, "Skip flag without a merge flag is not allowed!" );
    }

#if WCG_EXT
    DTRACE_BEST_MODE( tempCS, bestCS, m_pcRdCost->getLambda( true ) );
#else
    DTRACE_BEST_MODE( tempCS, bestCS, m_pcRdCost->getLambda() );
#endif

    if( m_modeCtrl->useModeResult( encTestMode, tempCS, partitioner ) )
    {
      if( tempCS->cus.size() == 1 )
      {
        // if tempCS is not a split-mode
        CodingUnit &cu = *tempCS->cus.front();

        if( CU::isLosslessCoded( cu ) && !cu.ipcm )
        {
          xFillPCMBuffer( cu );
        }
      }

      std::swap( tempCS, bestCS );
      // store temp best CI for next CU coding
      m_CurrCtx->best = m_CABACEstimator->getCtx();
Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
      m_bestModeUpdated = true;
#endif
      bestCSUpdated = true;
    }
  }

  // reset context states
  m_CABACEstimator->getCtx() = m_CurrCtx->start;
  return bestCSUpdated;
void EncCu::xCompressCU( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner
  , LutMotionCand *&tempMotCandLUTs
  , LutMotionCand *&bestMotCandLUTs
#if JVET_M0170_MRG_SHARELIST
  if (m_shareState == NO_SHARE)
  {
    tempCS->sharedBndPos = tempCS->area.Y().lumaPos();
    tempCS->sharedBndSize.width = tempCS->area.lwidth();
    tempCS->sharedBndSize.height = tempCS->area.lheight();
    bestCS->sharedBndPos = bestCS->area.Y().lumaPos();
    bestCS->sharedBndSize.width = bestCS->area.lwidth();
    bestCS->sharedBndSize.height = bestCS->area.lheight();
  }
#endif
#if ENABLE_SPLIT_PARALLELISM
  CHECK( m_dataId != tempCS->picture->scheduler.getDataId(), "Working in the wrong dataId!" );

  if( m_pcEncCfg->getNumSplitThreads() != 1 && tempCS->picture->scheduler.getSplitJobId() == 0 )
  {
    if( m_modeCtrl->isParallelSplit( *tempCS, partitioner ) )
    {
      m_modeCtrl->setParallelSplit( true );
      xCompressCUParallel( tempCS, bestCS, partitioner );
      return;
    }
  }

#endif

  Slice&   slice      = *tempCS->slice;
  const PPS &pps      = *tempCS->pps;
  const SPS &sps      = *tempCS->sps;
  const uint32_t uiLPelX  = tempCS->area.Y().lumaPos().x;
  const uint32_t uiTPelY  = tempCS->area.Y().lumaPos().y;

  const unsigned wIdx = gp_sizeIdxInfo->idxFrom( partitioner.currArea().lwidth()  );

  const UnitArea currCsArea = clipArea( CS::getArea( *bestCS, bestCS->area, partitioner.chType ), *tempCS->picture );
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
  if (m_pImvTempCS && (!slice.isIntra() || slice.getSPS()->getIBCFlag()))
Yu Han's avatar
Yu Han committed
#else
Yu Han's avatar
Yu Han committed
#endif
Karsten Suehring's avatar
Karsten Suehring committed
    tempCS->initSubStructure( *m_pImvTempCS[wIdx], partitioner.chType, partitioner.currArea(), false );
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  tempCS->chType = partitioner.chType;
  bestCS->chType = partitioner.chType;
  m_modeCtrl->initCULevel( partitioner, *tempCS );
  if( partitioner.currQtDepth == 0 && partitioner.currMtDepth == 0 && !tempCS->slice->isIntra() && ( sps.getUseSBT() || sps.getUseInterMTS() ) )
#else
  if( partitioner.currQtDepth == 0 && partitioner.currMtDepth == 0 && !tempCS->slice->isIntra() && ( sps.getUseSBT() || sps.getUseInterEMT() ) )
#endif
  {
    auto slsSbt = dynamic_cast<SaveLoadEncInfoSbt*>( m_modeCtrl );
    int maxSLSize = sps.getUseSBT() ? tempCS->slice->getSPS()->getMaxSbtSize() : MTS_INTER_MAX_CU_SIZE;
#else
    int maxSLSize = sps.getUseSBT() ? tempCS->slice->getSPS()->getMaxSbtSize() : EMT_INTER_MAX_CU_WITH_QTBT;
#endif
    slsSbt->resetSaveloadSbt( maxSLSize );
  }
  m_sbtCostSave[0] = m_sbtCostSave[1] = MAX_DOUBLE;
#endif

  m_CurrCtx->start = m_CABACEstimator->getCtx();

  m_cuChromaQpOffsetIdxPlus1 = 0;

  if( slice.getUseChromaQpAdj() )
  {
    int lgMinCuSize = sps.getLog2MinCodingBlockSize() +
      std::max<int>( 0, sps.getLog2DiffMaxMinCodingBlockSize() - int( pps.getPpsRangeExtension().getDiffCuChromaQpOffsetDepth() ) );
    m_cuChromaQpOffsetIdxPlus1 = ( ( uiLPelX >> lgMinCuSize ) + ( uiTPelY >> lgMinCuSize ) ) % ( pps.getPpsRangeExtension().getChromaQpOffsetListLen() + 1 );
  }

  if( !m_modeCtrl->anyMode() )
  {
    m_modeCtrl->finishCULevel( partitioner );
    return;
  }
Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
  if ((!slice.isIntra() || slice.getSPS()->getIBCFlag())
Yu Han's avatar
Yu Han committed
#else
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  if (!slice.isIntra()
Yu Han's avatar
Yu Han committed
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    && tempCS->chType == CHANNEL_TYPE_LUMA
    )
Li's avatar
Li committed
  {
    tempCS->slice->copyMotionLUTs(tempMotCandLUTs, tempCS->slice->getMotionLUTs());
  }

  DTRACE_UPDATE( g_trace_ctx, std::make_pair( "cux", uiLPelX ) );
  DTRACE_UPDATE( g_trace_ctx, std::make_pair( "cuy", uiTPelY ) );
  DTRACE_UPDATE( g_trace_ctx, std::make_pair( "cuw", tempCS->area.lwidth() ) );
  DTRACE_UPDATE( g_trace_ctx, std::make_pair( "cuh", tempCS->area.lheight() ) );
  DTRACE( g_trace_ctx, D_COMMON, "@(%4d,%4d) [%2dx%2d]\n", tempCS->area.lx(), tempCS->area.ly(), tempCS->area.lwidth(), tempCS->area.lheight() );


#if JVET_M0170_MRG_SHARELIST
  int startShareThisLevel = 0;
#endif
#if JVET_M0246_AFFINE_AMVR
  m_pcInterSearch->resetSavedAffineMotion();
#endif
    EncTestMode currTestMode = m_modeCtrl->currTestMode();

    if (pps.getUseDQP() && CS::isDualITree(*tempCS) && isChroma(partitioner.chType))
    {
      const Position chromaCentral(tempCS->area.Cb().chromaPos().offset(tempCS->area.Cb().chromaSize().width >> 1, tempCS->area.Cb().chromaSize().height >> 1));
      const Position lumaRefPos(chromaCentral.x << getComponentScaleX(COMPONENT_Cb, tempCS->area.chromaFormat), chromaCentral.y << getComponentScaleY(COMPONENT_Cb, tempCS->area.chromaFormat));
      const CodingStructure* baseCS = bestCS->picture->cs;
      const CodingUnit* colLumaCu = baseCS->getCU(lumaRefPos, CHANNEL_TYPE_LUMA);

#if SHARP_LUMA_DELTA_QP || ENABLE_QPA_SUB_CTU
    if (partitioner.currDepth <= pps.getMaxCuDQPDepth() && (
        (m_pcEncCfg->getLumaLevelToDeltaQPMapping().isEnabled()) ||
#endif
#if ENABLE_QPA_SUB_CTU
        (m_pcEncCfg->getUsePerceptQPA() && !m_pcEncCfg->getUseRateCtrl() && pps.getUseDQP())
#else
        false
#endif
      ))
    {
#if ENABLE_SPLIT_PARALLELISM
      CHECK( tempCS->picture->scheduler.getSplitJobId() > 0, "Changing lambda is only allowed in the master thread!" );
#endif
      if (currTestMode.qp >= 0)
      {
        updateLambda (&slice, currTestMode.qp, CS::isDualITree (*tempCS) || (partitioner.currDepth == 0));
      }
    }
#endif

    if( currTestMode.type == ETM_INTER_ME )
    {
      if( ( currTestMode.opts & ETO_IMV ) != 0 )
      {
#if JVET_M0246_AFFINE_AMVR
        tempCS->bestCS = bestCS;
        xCheckRDCostInterIMV( tempCS, bestCS, partitioner, currTestMode );
        tempCS->bestCS = nullptr;
#else
        xCheckRDCostInterIMV(tempCS, bestCS, partitioner, currTestMode);
#if JVET_M0246_AFFINE_AMVR
        tempCS->bestCS = bestCS;
        xCheckRDCostInter( tempCS, bestCS, partitioner, currTestMode );
        tempCS->bestCS = nullptr;
#else
        xCheckRDCostInter( tempCS, bestCS, partitioner, currTestMode );
#endif
#if JVET_M0253_HASH_ME
    else if (currTestMode.type == ETM_HASH_INTER)
    {
      xCheckRDCostHashInter( tempCS, bestCS, partitioner, currTestMode );
    }
#endif
    else if( currTestMode.type == ETM_AFFINE )
    {
      xCheckRDCostAffineMerge2Nx2N( tempCS, bestCS, partitioner, currTestMode );
    }
#if REUSE_CU_RESULTS
    else if( currTestMode.type == ETM_RECO_CACHED )
    {
      xReuseCachedResult( tempCS, bestCS, partitioner );
    }
#endif
    else if( currTestMode.type == ETM_MERGE_SKIP )
    {
      xCheckRDCostMerge2Nx2N( tempCS, bestCS, partitioner, currTestMode );
      CodingUnit* cu = bestCS->getCU(partitioner.chType);
Xiaozhong Xu's avatar
Xiaozhong Xu committed
      if (cu)
      cu->mmvdSkip = cu->skip == false ? false : cu->mmvdSkip;
    else if( currTestMode.type == ETM_MERGE_TRIANGLE )
    {
      xCheckRDCostMergeTriangle2Nx2N( tempCS, bestCS, partitioner, currTestMode );
    }
    else if( currTestMode.type == ETM_INTRA )
    {
      xCheckRDCostIntra( tempCS, bestCS, partitioner, currTestMode );
    }
    else if( currTestMode.type == ETM_IPCM )
    {
      xCheckIntraPCM( tempCS, bestCS, partitioner, currTestMode );
    }
Yu Han's avatar
Yu Han committed
    else if (currTestMode.type == ETM_IBC)
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    {
Yu Han's avatar
Yu Han committed
      xCheckRDCostIBCMode(tempCS, bestCS, partitioner, currTestMode);
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    }
Yu Han's avatar
Yu Han committed
    else if (currTestMode.type == ETM_IBC_MERGE)
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    {
Yu Han's avatar
Yu Han committed
      xCheckRDCostIBCModeMerge2Nx2N(tempCS, bestCS, partitioner, currTestMode);
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    }
      xCheckModeSplit( tempCS, bestCS, partitioner, currTestMode
        , tempMotCandLUTs
        , bestMotCandLUTs
        , partitioner.currArea()
      );
Karsten Suehring's avatar
Karsten Suehring committed
      THROW( "Don't know how to handle mode: type = " << currTestMode.type << ", options = " << currTestMode.opts );
    }
  } while( m_modeCtrl->nextMode( *tempCS, partitioner ) );

#if JVET_M0170_MRG_SHARELIST
  if(startShareThisLevel == 1)
  {
    m_shareState = NO_SHARE;
    m_pcInterSearch->setShareState(m_shareState);
    setShareStateDec(m_shareState);
  }
#endif

  //////////////////////////////////////////////////////////////////////////
  // Finishing CU
#if ENABLE_SPLIT_PARALLELISM
  if( bestCS->cus.empty() )
  {
    CHECK( bestCS->cost != MAX_DOUBLE, "Cost should be maximal if no encoding found" );
    CHECK( bestCS->picture->scheduler.getSplitJobId() == 0, "Should always get a result in serial case" );

    m_modeCtrl->finishCULevel( partitioner );
    return;
  }

#endif
  // set context states
  m_CABACEstimator->getCtx() = m_CurrCtx->best;

  // QP from last processed CU for further processing
  bestCS->prevQP[partitioner.chType] = bestCS->cus.back()->qp;
Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
  if ((!slice.isIntra() || slice.getSPS()->getIBCFlag())
Yu Han's avatar
Yu Han committed
#else
  if (!slice.isIntra()
Yu Han's avatar
Yu Han committed
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    && bestCS->chType == CHANNEL_TYPE_LUMA
Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
    && bestCS->cus.size() == 1 && (bestCS->cus.back()->predMode == MODE_INTER || bestCS->cus.back()->predMode == MODE_IBC)
#else
    && bestCS->cus.size() == 1 && bestCS->cus.back()->predMode == MODE_INTER
Yu Han's avatar
Yu Han committed
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    && bestCS->area.Y() == (*bestCS->cus.back()).Y()
    )
  {
    bestCS->slice->updateMotionLUTs(bestMotCandLUTs, (*bestCS->cus.back()));
  }
Taoran Lu's avatar
Taoran Lu committed
#if JVET_M0427_INLOOP_RESHAPER
  bestCS->picture->getPredBuf(currCsArea).copyFrom(bestCS->getPredBuf(currCsArea));
#endif
  bestCS->picture->getRecoBuf( currCsArea ).copyFrom( bestCS->getRecoBuf( currCsArea ) );
  m_modeCtrl->finishCULevel( partitioner );

#if ENABLE_SPLIT_PARALLELISM
  if( tempCS->picture->scheduler.getSplitJobId() == 0 && m_pcEncCfg->getNumSplitThreads() != 1 )
  {
    tempCS->picture->finishParallelPart( currCsArea );
  }

#endif
  // Assert if Best prediction mode is NONE
  // Selected mode's RD-cost must be not MAX_DOUBLE.
  CHECK( bestCS->cus.empty()                                   , "No possible encoding found" );
  CHECK( bestCS->cus[0]->predMode == NUMBER_OF_PREDICTION_MODES, "No possible encoding found" );
  CHECK( bestCS->cost             == MAX_DOUBLE                , "No possible encoding found" );
}

#if SHARP_LUMA_DELTA_QP || ENABLE_QPA_SUB_CTU
void EncCu::updateLambda (Slice* slice, const int dQP, const bool updateRdCostLambda)
{
#if WCG_EXT
  int    NumberBFrames = ( m_pcEncCfg->getGOPSize() - 1 );
  int    SHIFT_QP = 12;
  double dLambda_scale = 1.0 - Clip3( 0.0, 0.5, 0.05*(double)(slice->getPic()->fieldPic ? NumberBFrames/2 : NumberBFrames) );

  int bitdepth_luma_qp_scale = 6
                               * (slice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8
                                  - DISTORTION_PRECISION_ADJUSTMENT(slice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA)));
  double qp_temp = (double) dQP + bitdepth_luma_qp_scale - SHIFT_QP;
  double dQPFactor = m_pcEncCfg->getGOPEntry( m_pcSliceEncoder->getGopId() ).m_QPFactor;
  if( slice->getSliceType() == I_SLICE )
  {
    if( m_pcEncCfg->getIntraQpFactor() >= 0.0 /*&& m_pcEncCfg->getGOPEntry( m_pcSliceEncoder->getGopId() ).m_sliceType != I_SLICE*/ )
    {
      dQPFactor = m_pcEncCfg->getIntraQpFactor();
    }
    else
    {
      if( m_pcEncCfg->getLambdaFromQPEnable() )
      {
        dQPFactor = 0.57;
      }
      else
      {
        dQPFactor = 0.57*dLambda_scale;
      }
    }
  }
  else if( m_pcEncCfg->getLambdaFromQPEnable() )
  {
    dQPFactor = 0.57*dQPFactor;
  }

  double dLambda = dQPFactor*pow( 2.0, qp_temp/3.0 );
  int depth = slice->getDepth();

  if( !m_pcEncCfg->getLambdaFromQPEnable() && depth>0 )
  {
    int qp_temp_slice = slice->getSliceQp() + bitdepth_luma_qp_scale - SHIFT_QP; // avoid lambda  over adjustment,  use slice_qp here
    dLambda *= Clip3( 2.00, 4.00, (qp_temp_slice / 6.0) ); // (j == B_SLICE && p_cur_frm->layer != 0 )
  }
  if( !m_pcEncCfg->getUseHADME() && slice->getSliceType( ) != I_SLICE )
  {
    dLambda *= 0.95;
  }

  const int temporalId = m_pcEncCfg->getGOPEntry( m_pcSliceEncoder->getGopId() ).m_temporalId;
  const std::vector<double> &intraLambdaModifiers = m_pcEncCfg->getIntraLambdaModifier();
  double lambdaModifier;
  if( slice->getSliceType( ) != I_SLICE || intraLambdaModifiers.empty())
  {
    lambdaModifier = m_pcEncCfg->getLambdaModifier(temporalId);
  }
  else
  {
    lambdaModifier = intraLambdaModifiers[(temporalId < intraLambdaModifiers.size()) ? temporalId : (intraLambdaModifiers.size() - 1)];
  }
  dLambda *= lambdaModifier;

  int qpBDoffset = slice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA);
  int iQP = Clip3(-qpBDoffset, MAX_QP, (int)floor((double)dQP + 0.5));
  m_pcSliceEncoder->setUpLambda(slice, dLambda, iQP);

#else
  const double oldQP     = (double)slice->getSliceQpBase();
#if ENABLE_QPA_SUB_CTU
  const double oldLambda = (m_pcEncCfg->getUsePerceptQPA() && !m_pcEncCfg->getUseRateCtrl() && slice->getPPS()->getUseDQP()) ? slice->getLambdas()[0] :
                           m_pcSliceEncoder->calculateLambda (slice, m_pcSliceEncoder->getGopId(), slice->getDepth(), oldQP, oldQP, iQP);
#else
  const double oldLambda = m_pcSliceEncoder->calculateLambda (slice, m_pcSliceEncoder->getGopId(), slice->getDepth(), oldQP, oldQP, iQP);
#endif
  const double newLambda = oldLambda * pow (2.0, ((double)dQP - oldQP) / 3.0);
#if RDOQ_CHROMA_LAMBDA
  const double chromaLambda = newLambda / m_pcRdCost->getChromaWeight();
  const double lambdaArray[MAX_NUM_COMPONENT] = {newLambda, chromaLambda, chromaLambda};
  m_pcTrQuant->setLambdas (lambdaArray);
#else
  m_pcTrQuant->setLambda (newLambda);
#endif
  if (updateRdCostLambda)
  {
    m_pcRdCost->setLambda (newLambda, slice->getSPS()->getBitDepths());
  }
#endif
}
#endif

#if ENABLE_SPLIT_PARALLELISM
//#undef DEBUG_PARALLEL_TIMINGS
//#define DEBUG_PARALLEL_TIMINGS 1
void EncCu::xCompressCUParallel( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner )
{
  const unsigned wIdx = gp_sizeIdxInfo->idxFrom( partitioner.currArea().lwidth() );
  const unsigned hIdx = gp_sizeIdxInfo->idxFrom( partitioner.currArea().lheight() );

  Picture* picture = tempCS->picture;

  int numJobs = m_modeCtrl->getNumParallelJobs( *bestCS, partitioner );

  bool    jobUsed                            [NUM_RESERVERD_SPLIT_JOBS];
  std::fill( jobUsed, jobUsed + NUM_RESERVERD_SPLIT_JOBS, false );

  const UnitArea currArea = CS::getArea( *tempCS, partitioner.currArea(), partitioner.chType );
#if ENABLE_WPP_PARALLELISM
  const int      wppTId   = picture->scheduler.getWppThreadId();
#endif
  const bool doParallel   = !m_pcEncCfg->getForceSingleSplitThread();
#if _MSC_VER && ENABLE_WPP_PARALLELISM
#pragma omp parallel for schedule(dynamic,1) num_threads(NUM_SPLIT_THREADS_IF_MSVC) if(doParallel)
#else
  omp_set_num_threads( m_pcEncCfg->getNumSplitThreads() );

#pragma omp parallel for schedule(dynamic,1) if(doParallel)
#endif
  for( int jId = 1; jId <= numJobs; jId++ )
  {
    // thread start
#if ENABLE_WPP_PARALLELISM
    picture->scheduler.setWppThreadId( wppTId );
#endif
    picture->scheduler.setSplitThreadId();
    picture->scheduler.setSplitJobId( jId );

    Partitioner* jobPartitioner = PartitionerFactory::get( *tempCS->slice );
    EncCu*       jobCuEnc       = m_pcEncLib->getCuEncoder( picture->scheduler.getSplitDataId( jId ) );
    auto*        jobBlkCache    = dynamic_cast<CacheBlkInfoCtrl*>( jobCuEnc->m_modeCtrl );

    jobPartitioner->copyState( partitioner );
    jobCuEnc      ->copyState( this, *jobPartitioner, currArea, true );

    if( jobBlkCache )
    {
      jobBlkCache->tick();
    }

    CodingStructure *&jobBest = jobCuEnc->m_pBestCS[wIdx][hIdx];
    CodingStructure *&jobTemp = jobCuEnc->m_pTempCS[wIdx][hIdx];

    jobUsed[jId] = true;

    jobCuEnc->xCompressCU( jobTemp, jobBest, *jobPartitioner );

    delete jobPartitioner;

    picture->scheduler.setSplitJobId( 0 );
    // thread stop
  }
  picture->scheduler.setSplitThreadId( 0 );

  int    bestJId  = 0;
  double bestCost = bestCS->cost;
  for( int jId = 1; jId <= numJobs; jId++ )
  {
    EncCu* jobCuEnc = m_pcEncLib->getCuEncoder( picture->scheduler.getSplitDataId( jId ) );

    if( jobUsed[jId] && jobCuEnc->m_pBestCS[wIdx][hIdx]->cost < bestCost )
    {
      bestCost = jobCuEnc->m_pBestCS[wIdx][hIdx]->cost;
      bestJId  = jId;
    }
  }

  if( bestJId > 0 )
  {
    copyState( m_pcEncLib->getCuEncoder( picture->scheduler.getSplitDataId( bestJId ) ), partitioner, currArea, false );
    m_CurrCtx->best = m_CABACEstimator->getCtx();

    tempCS = m_pTempCS[wIdx][hIdx];
    bestCS = m_pBestCS[wIdx][hIdx];
  }

  const int      bitDepthY = tempCS->sps->getBitDepth( CH_L );
  const UnitArea clipdArea = clipArea( currArea, *picture );

  CHECK( calcCheckSum( picture->getRecoBuf( clipdArea.Y() ), bitDepthY ) != calcCheckSum( bestCS->getRecoBuf( clipdArea.Y() ), bitDepthY ), "Data copied incorrectly!" );

  picture->finishParallelPart( currArea );

  if( auto *blkCache = dynamic_cast<CacheBlkInfoCtrl*>( m_modeCtrl ) )
  {
    for( int jId = 1; jId <= numJobs; jId++ )
    {
      if( !jobUsed[jId] || jId == bestJId ) continue;

      auto *jobBlkCache = dynamic_cast<CacheBlkInfoCtrl*>( m_pcEncLib->getCuEncoder( picture->scheduler.getSplitDataId( jId ) )->m_modeCtrl );
      CHECK( !jobBlkCache, "If own mode controller has blk info cache capability so should all other mode controllers!" );
      blkCache->CacheBlkInfoCtrl::copyState( *jobBlkCache, partitioner.currArea() );
    }

    blkCache->tick();
  }

}

void EncCu::copyState( EncCu* other, Partitioner& partitioner, const UnitArea& currArea, const bool isDist )
{
  const unsigned wIdx = gp_sizeIdxInfo->idxFrom( partitioner.currArea().lwidth () );
  const unsigned hIdx = gp_sizeIdxInfo->idxFrom( partitioner.currArea().lheight() );

  if( isDist )
  {
    other->m_pBestCS[wIdx][hIdx]->initSubStructure( *m_pBestCS[wIdx][hIdx], partitioner.chType, partitioner.currArea(), false );
    other->m_pTempCS[wIdx][hIdx]->initSubStructure( *m_pTempCS[wIdx][hIdx], partitioner.chType, partitioner.currArea(), false );
  }
  else
  {
          CodingStructure* dst =        m_pBestCS[wIdx][hIdx];
    const CodingStructure *src = other->m_pBestCS[wIdx][hIdx];
    bool keepResi = KEEP_PRED_AND_RESI_SIGNALS;

    dst->useSubStructure( *src, partitioner.chType, currArea, KEEP_PRED_AND_RESI_SIGNALS, true, keepResi, keepResi );
    dst->cost           =  src->cost;
    dst->dist           =  src->dist;
    dst->fracBits       =  src->fracBits;
    dst->features       =  src->features;
  }

  if( isDist )
  {
    m_CurrCtx = m_CtxBuffer.data();
  }

  m_pcInterSearch->copyState( *other->m_pcInterSearch );
  m_modeCtrl     ->copyState( *other->m_modeCtrl, partitioner.currArea() );
  m_pcRdCost     ->copyState( *other->m_pcRdCost );
  m_pcTrQuant    ->copyState( *other->m_pcTrQuant );

  m_CABACEstimator->getCtx() = other->m_CABACEstimator->getCtx();
}
#endif

void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode
Li's avatar
Li committed
  , LutMotionCand* &tempMotCandLUTs
  , LutMotionCand* &bestMotCandLUTs
{
  const int qp                = encTestMode.qp;
  const PPS &pps              = *tempCS->pps;
  const Slice &slice          = *tempCS->slice;
  const bool bIsLosslessMode  = false; // False at this level. Next level down may set it to true.
  const int oldPrevQp         = tempCS->prevQP[partitioner.chType];
  const uint32_t currDepth        = partitioner.currDepth;

  const unsigned wParIdx = gp_sizeIdxInfo->idxFrom(parArea.lwidth());
  const unsigned hParIdx = gp_sizeIdxInfo->idxFrom(parArea.lheight());
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  if (tempCS->chType == CHANNEL_TYPE_LUMA)
  tempCS->slice->copyMotionLUTs(tempMotCandLUTs, m_pSplitTempMotLUTs[wParIdx][hParIdx]);

  const PartSplit split = getPartSplit( encTestMode );

  CHECK( split == CU_DONT_SPLIT, "No proper split provided!" );

  tempCS->initStructData( qp, bIsLosslessMode );

  m_CABACEstimator->getCtx() = m_CurrCtx->start;

Karsten Suehring's avatar
Karsten Suehring committed
  const TempCtx ctxStartSP( m_CtxCache, SubCtx( Ctx::SplitFlag,   m_CABACEstimator->getCtx() ) );
Adam Wieckowski's avatar
Adam Wieckowski committed
#if JVET_M0421_SPLIT_SIG
  const TempCtx ctxStartQt( m_CtxCache, SubCtx( Ctx::SplitQtFlag, m_CABACEstimator->getCtx() ) );
  const TempCtx ctxStartHv( m_CtxCache, SubCtx( Ctx::SplitHvFlag, m_CABACEstimator->getCtx() ) );
  const TempCtx ctxStart12( m_CtxCache, SubCtx( Ctx::Split12Flag, m_CABACEstimator->getCtx() ) );
#else
Karsten Suehring's avatar
Karsten Suehring committed
  const TempCtx ctxStartBT( m_CtxCache, SubCtx( Ctx::BTSplitFlag, m_CABACEstimator->getCtx() ) );
Adam Wieckowski's avatar
Adam Wieckowski committed
#endif
Karsten Suehring's avatar
Karsten Suehring committed
  m_CABACEstimator->resetBits();
Adam Wieckowski's avatar
Adam Wieckowski committed
#if JVET_M0421_SPLIT_SIG
  m_CABACEstimator->split_cu_mode( split, *tempCS, partitioner );
#else
Karsten Suehring's avatar
Karsten Suehring committed
  if( partitioner.getImplicitSplit( *tempCS ) != CU_QUAD_SPLIT )
  {
    if( partitioner.canSplit( CU_QUAD_SPLIT, *tempCS ) )
Karsten Suehring's avatar
Karsten Suehring committed
      m_CABACEstimator->split_cu_flag( split == CU_QUAD_SPLIT, *tempCS, partitioner );
    }
    if( split != CU_QUAD_SPLIT )
    {
      m_CABACEstimator->split_cu_mode_mt( split, *tempCS, partitioner );
Karsten Suehring's avatar
Karsten Suehring committed
  }
Adam Wieckowski's avatar
Adam Wieckowski committed
#endif
Karsten Suehring's avatar
Karsten Suehring committed
  const double factor = ( tempCS->currQP[partitioner.chType] > 30 ? 1.1 : 1.075 );
Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
  tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();
  if (!tempCS->useDbCost)
    CHECK(bestCS->costDbOffset != 0, "error");
  const double cost   = m_pcRdCost->calcRdCost( uint64_t( m_CABACEstimator->getEstFracBits() + ( ( bestCS->fracBits ) / factor ) ), Distortion( bestCS->dist / factor ) ) + bestCS->costDbOffset / factor;
#else
Karsten Suehring's avatar
Karsten Suehring committed
  const double cost   = m_pcRdCost->calcRdCost( uint64_t( m_CABACEstimator->getEstFracBits() + ( ( bestCS->fracBits ) / factor ) ), Distortion( bestCS->dist / factor ) );
Nan Hu's avatar
Nan Hu committed
#endif
Karsten Suehring's avatar
Karsten Suehring committed
  m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitFlag,   ctxStartSP );
Adam Wieckowski's avatar
Adam Wieckowski committed
#if JVET_M0421_SPLIT_SIG
  m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitQtFlag, ctxStartQt );
  m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitHvFlag, ctxStartHv );
  m_CABACEstimator->getCtx() = SubCtx( Ctx::Split12Flag, ctxStart12 );
#else
Karsten Suehring's avatar
Karsten Suehring committed
  m_CABACEstimator->getCtx() = SubCtx( Ctx::BTSplitFlag, ctxStartBT );
Adam Wieckowski's avatar
Adam Wieckowski committed
#endif
Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
  if (cost > bestCS->cost + bestCS->costDbOffset
Nan Hu's avatar
Nan Hu committed
#else
  if (cost > bestCS->cost
#endif
#if ENABLE_QPA_SUB_CTU
    || (m_pcEncCfg->getUsePerceptQPA() && !m_pcEncCfg->getUseRateCtrl() && pps.getUseDQP() && (pps.getMaxCuDQPDepth() > 0) && (split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT) &&
        (partitioner.currArea().lwidth() == tempCS->pcv->maxCUWidth) && (partitioner.currArea().lheight() == tempCS->pcv->maxCUHeight)) // force quad-split or no split at CTU level
#endif
    )
Karsten Suehring's avatar
Karsten Suehring committed
  {
    xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
    return;
#if JVET_M0170_MRG_SHARELIST
  if (!slice.isIntra()
    && tempCS->chType == CHANNEL_TYPE_LUMA
    )
  {
    tempCS->slice->copyMotionLUTs(tempMotCandLUTs, tempCS->slice->getMotionLUTs());
  }

  int startShareThisLevel = 0;
  const uint32_t uiLPelX = tempCS->area.Y().lumaPos().x;
  const uint32_t uiTPelY = tempCS->area.Y().lumaPos().y;

  int splitRatio = 1;
  CHECK(!(split == CU_QUAD_SPLIT || split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT
    || split == CU_TRIH_SPLIT || split == CU_TRIV_SPLIT), "invalid split type");
  splitRatio = (split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT) ? 1 : 2;

  bool isOneChildSmall = ((tempCS->area.lwidth())*(tempCS->area.lheight()) >> splitRatio) < MRG_SHARELIST_SHARSIZE;

  if ((((tempCS->area.lwidth())*(tempCS->area.lheight())) > (MRG_SHARELIST_SHARSIZE * 1)))
  {
    m_shareState = NO_SHARE;
  }

  if (m_shareState == NO_SHARE)//init state
  {
    if (isOneChildSmall)
    {
      m_shareState = GEN_ON_SHARED_BOUND;//share start state
      startShareThisLevel = 1;
    }
  }
#if JVET_M0483_IBC
  if ((m_shareState == GEN_ON_SHARED_BOUND) && (!slice.isIntra() || slice.getSPS()->getIBCFlag()))
#else
  if ((m_shareState == GEN_ON_SHARED_BOUND) && (!slice.isIntra()))
  {
#if JVET_M0170_MRG_SHARELIST
    tempCS->slice->copyMotionLUTs(tempCS->slice->getMotionLUTs(), tempCS->slice->m_MotionCandLuTsBkup);
    m_shareBndPosX = uiLPelX;
    m_shareBndPosY = uiTPelY;
    m_shareBndSizeW = tempCS->area.lwidth();
    m_shareBndSizeH = tempCS->area.lheight();
    m_shareState = SHARING;
#endif
  }


  m_pcInterSearch->setShareState(m_shareState);
  setShareStateDec(m_shareState);
#endif

  partitioner.splitCurrArea( split, *tempCS );

  m_CurrCtx++;

  tempCS->getRecoBuf().fill( 0 );
Taoran Lu's avatar
Taoran Lu committed

#if JVET_M0427_INLOOP_RESHAPER
  tempCS->getPredBuf().fill(0);
#endif
  AffineMVInfo tmpMVInfo;
  bool isAffMVInfoSaved;
  m_pcInterSearch->savePrevAffMVInfo(0, tmpMVInfo, isAffMVInfoSaved);

  do
  {
    const auto &subCUArea  = partitioner.currArea();

    if( tempCS->picture->Y().contains( subCUArea.lumaPos() ) )
    {
      const unsigned wIdx    = gp_sizeIdxInfo->idxFrom( subCUArea.lwidth () );
      const unsigned hIdx    = gp_sizeIdxInfo->idxFrom( subCUArea.lheight() );

      CodingStructure *tempSubCS = m_pTempCS[wIdx][hIdx];
      CodingStructure *bestSubCS = m_pBestCS[wIdx][hIdx];

      tempCS->initSubStructure( *tempSubCS, partitioner.chType, subCUArea, false );
      tempCS->initSubStructure( *bestSubCS, partitioner.chType, subCUArea, false );
Li's avatar
Li committed
      LutMotionCand *tempSubMotCandLUTs = m_pTempMotLUTs[wIdx][hIdx];
      LutMotionCand *bestSubMotCandLUTs = m_pBestMotLUTs[wIdx][hIdx];
Xiaozhong Xu's avatar
Xiaozhong Xu committed
      if (tempCS->chType == CHANNEL_TYPE_LUMA)
      {
        tempCS->slice->copyMotionLUTs(tempMotCandLUTs, tempSubMotCandLUTs);
        tempCS->slice->copyMotionLUTs(tempMotCandLUTs, bestSubMotCandLUTs);
      }
#if JVET_M0170_MRG_SHARELIST
      tempSubCS->sharedBndPos.x = (m_shareState == SHARING) ? m_shareBndPosX : tempSubCS->area.Y().lumaPos().x;
      tempSubCS->sharedBndPos.y = (m_shareState == SHARING) ? m_shareBndPosY : tempSubCS->area.Y().lumaPos().y;
      tempSubCS->sharedBndSize.width = (m_shareState == SHARING) ? m_shareBndSizeW : tempSubCS->area.lwidth();
      tempSubCS->sharedBndSize.height = (m_shareState == SHARING) ? m_shareBndSizeH : tempSubCS->area.lheight();
      bestSubCS->sharedBndPos.x = (m_shareState == SHARING) ? m_shareBndPosX : tempSubCS->area.Y().lumaPos().x;
      bestSubCS->sharedBndPos.y = (m_shareState == SHARING) ? m_shareBndPosY : tempSubCS->area.Y().lumaPos().y;
      bestSubCS->sharedBndSize.width = (m_shareState == SHARING) ? m_shareBndSizeW : tempSubCS->area.lwidth();
      bestSubCS->sharedBndSize.height = (m_shareState == SHARING) ? m_shareBndSizeH : tempSubCS->area.lheight();
#endif
      xCompressCU( tempSubCS, bestSubCS, partitioner
        , tempSubMotCandLUTs
        , bestSubMotCandLUTs
      );

      if( bestSubCS->cost == MAX_DOUBLE )
      {
        CHECK( split == CU_QUAD_SPLIT, "Split decision reusing cannot skip quad split" );
        tempCS->cost = MAX_DOUBLE;
Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
        tempCS->costDbOffset = 0;
        tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();
#endif
        bool bestCSUpdated =
        xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
Xiaozhong Xu's avatar
Xiaozhong Xu committed
        if (tempCS->chType == CHANNEL_TYPE_LUMA)
        if (bestCSUpdated)
        {
          std::swap(tempMotCandLUTs, bestMotCandLUTs);
        }
        return;
      }

      bool keepResi = KEEP_PRED_AND_RESI_SIGNALS;
      tempCS->useSubStructure( *bestSubCS, partitioner.chType, CS::getArea( *tempCS, subCUArea, partitioner.chType ), KEEP_PRED_AND_RESI_SIGNALS, true, keepResi, keepResi );
Xiaozhong Xu's avatar
Xiaozhong Xu committed
      if (tempCS->chType == CHANNEL_TYPE_LUMA)
      tempCS->slice->copyMotionLUTs(bestSubMotCandLUTs, tempMotCandLUTs);

      if(currDepth < pps.getMaxCuDQPDepth())
      {
        tempCS->prevQP[partitioner.chType] = bestSubCS->prevQP[partitioner.chType];
      }

      tempSubCS->releaseIntermediateData();
      bestSubCS->releaseIntermediateData();
    }
  } while( partitioner.nextPart( *tempCS ) );

  partitioner.exitCurrSplit();

#if JVET_M0170_MRG_SHARELIST
  if (startShareThisLevel == 1)
  {
    m_shareState = NO_SHARE;
    m_pcInterSearch->setShareState(m_shareState);
    setShareStateDec(m_shareState);
  }
#endif

  m_CurrCtx--;

  // Finally, generate split-signaling bits for RD-cost check
  const PartSplit implicitSplit = partitioner.getImplicitSplit( *tempCS );

  {
    bool enforceQT = implicitSplit == CU_QUAD_SPLIT;
#if HM_QTBT_REPRODUCE_FAST_LCTU_BUG

    // LARGE CTU bug
Karsten Suehring's avatar
Karsten Suehring committed
    if( m_pcEncCfg->getUseFastLCTU() )
      unsigned maxDepth = g_aucLog2[tempCS->sps->getCTUSize()] - g_aucLog2[tempCS->sps->getMinQTSize(slice.getSliceType(), partitioner.chType)];

      if( auto ad = dynamic_cast<AdaptiveDepthPartitioner*>( &partitioner ) )
      {
        ad->setMaxMinDepth( minDepth, maxDepth, *tempCS );
      }

      if( minDepth > partitioner.currQtDepth )
      {
        // enforce QT
        enforceQT = true;
      }
    }
#endif

    if( !enforceQT )
    {
      m_CABACEstimator->resetBits();

Adam Wieckowski's avatar
Adam Wieckowski committed
#if JVET_M0421_SPLIT_SIG
      m_CABACEstimator->split_cu_mode( split, *tempCS, partitioner );
#else
      if( partitioner.canSplit( CU_QUAD_SPLIT, *tempCS ) )
      {
        m_CABACEstimator->split_cu_flag( split == CU_QUAD_SPLIT, *tempCS, partitioner );
      }
      if( split != CU_QUAD_SPLIT )
      {
        m_CABACEstimator->split_cu_mode_mt( split, *tempCS, partitioner );
      }
Adam Wieckowski's avatar
Adam Wieckowski committed
#endif

      tempCS->fracBits += m_CABACEstimator->getEstFracBits(); // split bits
    }
  }

  tempCS->cost = m_pcRdCost->calcRdCost( tempCS->fracBits, tempCS->dist );

  // Check Delta QP bits for splitted structure
  xCheckDQP( *tempCS, partitioner, true );

  // If the configuration being tested exceeds the maximum number of bytes for a slice / slice-segment, then
  // a proper RD evaluation cannot be performed. Therefore, termination of the
  // slice/slice-segment must be made prior to this CTU.
  // This can be achieved by forcing the decision to be that of the rpcTempCU.
  // The exception is each slice / slice-segment must have at least one CTU.
  if (bestCS->cost != MAX_DOUBLE)
  {
#if HEVC_TILES_WPP
    const TileMap& tileMap = *tempCS->picture->tileMap;
#endif
#if HEVC_TILES_WPP || HEVC_DEPENDENT_SLICES
    const uint32_t CtuAddr             = CU::getCtuAddr( *bestCS->getCU( partitioner.chType ) );
#endif
    const bool isEndOfSlice        =    slice.getSliceMode() == FIXED_NUMBER_OF_BYTES
                                      && ((slice.getSliceBits() + CS::getEstBits(*bestCS)) > slice.getSliceArgument() << 3)
#if HEVC_TILES_WPP
                                      && CtuAddr != tileMap.getCtuTsToRsAddrMap(slice.getSliceCurStartCtuTsAddr())
#endif
#if HEVC_DEPENDENT_SLICES
                                      && CtuAddr != tileMap.getCtuTsToRsAddrMap(slice.getSliceSegmentCurStartCtuTsAddr());
#else
                                      ;
#endif

#if HEVC_DEPENDENT_SLICES
    const bool isEndOfSliceSegment =    slice.getSliceSegmentMode() == FIXED_NUMBER_OF_BYTES
                                      && ((slice.getSliceSegmentBits() + CS::getEstBits(*bestCS)) > slice.getSliceSegmentArgument() << 3)
                                      && CtuAddr != tileMap.getCtuTsToRsAddrMap(slice.getSliceSegmentCurStartCtuTsAddr());
                                          // Do not need to check slice condition for slice-segment since a slice-segment is a subset of a slice.
    if (isEndOfSlice || isEndOfSliceSegment)
#else
    if(isEndOfSlice)
#endif
    {
      bestCS->cost = MAX_DOUBLE;
Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
      bestCS->costDbOffset = 0;
#endif
Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
  else
  {
    bestCS->costDbOffset = 0;
  }
  tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();
#endif

  // RD check for sub partitioned coding structure.
  bool bestCSUpdated =
  xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );

  if (isAffMVInfoSaved)
    m_pcInterSearch->addAffMVInfo(tmpMVInfo);

Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
  if ((!slice.isIntra() || slice.getSPS()->getIBCFlag())
Yu Han's avatar
Yu Han committed
#else
Xiaozhong Xu's avatar
Xiaozhong Xu committed
  if (!slice.isIntra()
Yu Han's avatar
Yu Han committed
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    && tempCS->chType == CHANNEL_TYPE_LUMA
    )
Li's avatar
Li committed
  {
    if (bestCSUpdated)
Li's avatar
Li committed
    {
      std::swap(tempMotCandLUTs, bestMotCandLUTs);
    }
    tempCS->slice->copyMotionLUTs(m_pSplitTempMotLUTs[wParIdx][hParIdx], tempMotCandLUTs);
  }
  tempCS->releaseIntermediateData();

  tempCS->prevQP[partitioner.chType] = oldPrevQp;
}


void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
Tung Nguyen's avatar
Tung Nguyen committed
#if !JVET_M0464_UNI_MTS
  double bestInterCost             = m_modeCtrl->getBestInterCost();
  double costSize2Nx2NemtFirstPass = m_modeCtrl->getEmtSize2Nx2NFirstPassCost();
  bool skipSecondEmtPass           = m_modeCtrl->getSkipSecondEMTPass();
  const SPS &sps                   = *tempCS->sps;
Tung Nguyen's avatar
Tung Nguyen committed
#endif
  const PPS &pps      = *tempCS->pps;
#if !JVET_M0464_UNI_MTS
  const CodingUnit *bestCU    = bestCS->getCU( partitioner.chType );
Karsten Suehring's avatar
Karsten Suehring committed
  const int maxSizeEMT        = EMT_INTRA_MAX_CU_WITH_QTBT;
  uint8_t considerEmtSecondPass = ( sps.getUseIntraEMT() && isLuma( partitioner.chType ) && partitioner.currArea().lwidth() <= maxSizeEMT && partitioner.currArea().lheight() <= maxSizeEMT ) ? 1 : 0;
Tung Nguyen's avatar
Tung Nguyen committed
#endif
#if JVET_M0102_INTRA_SUBPARTITIONS
  bool   useIntraSubPartitions   = false;
  double maxCostAllowedForChroma = MAX_DOUBLE;
#if JVET_M0464_UNI_MTS
  const  CodingUnit *bestCU      = bestCS->getCU( partitioner.chType );
#endif
#endif
Tung Nguyen's avatar
Tung Nguyen committed
#if JVET_M0464_UNI_MTS
  {
#else
  for( uint8_t emtCuFlag = 0; emtCuFlag <= considerEmtSecondPass; emtCuFlag++ )
  {
    //Possible early EMT tests interruptions
    //2) Second EMT pass. This "if clause" is necessary because of the NSST and PDPC "for loops".
    if( emtCuFlag && skipSecondEmtPass )
    {
      continue;
    }
    //3) if interHad is 0, only try further modes if some intra mode was already better than inter
    if( m_pcEncCfg->getUsePbIntraFast() && !tempCS->slice->isIntra() && bestCU && CU::isInter( *bestCS->getCU( partitioner.chType ) ) && interHad == 0 )
    {
      continue;
    }
Tung Nguyen's avatar
Tung Nguyen committed
#endif

    tempCS->initStructData( encTestMode.qp, encTestMode.lossless );

    CodingUnit &cu      = tempCS->addCU( CS::getArea( *tempCS, tempCS->area, partitioner.chType ), partitioner.chType );

    partitioner.setCUData( cu );
    cu.slice            = tempCS->slice;
#if HEVC_TILES_WPP
    cu.tileIdx          = tempCS->picture->tileMap->getTileIdxMap( tempCS->area.lumaPos() );
#endif
    cu.skip             = false;
    cu.predMode         = MODE_INTRA;
    cu.transQuantBypass = encTestMode.lossless;
    cu.chromaQpAdj      = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
    cu.qp               = encTestMode.qp;
  //cu.ipcm             = false;
Tung Nguyen's avatar
Tung Nguyen committed
#if !JVET_M0464_UNI_MTS
Tung Nguyen's avatar
Tung Nguyen committed
#endif
#if JVET_M0102_INTRA_SUBPARTITIONS
    cu.ispMode          = NOT_INTRA_SUBPARTITIONS;
#endif
Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
    m_bestModeUpdated = tempCS->useDbCost = bestCS->useDbCost = false;
#endif

#if JVET_M0102_INTRA_SUBPARTITIONS
      //the Intra SubPartitions mode uses the value of the best cost so far (luma if it is the fast version) to avoid test non-necessary lines
      const double bestCostSoFar = CS::isDualITree( *tempCS ) ? m_modeCtrl->getBestCostWithoutSplitFlags() : bestCU && bestCU->predMode == MODE_INTRA ? bestCS->lumaCost : bestCS->cost;
      m_pcIntraSearch->estIntraPredLumaQT( cu, partitioner, bestCostSoFar );

      useIntraSubPartitions = cu.ispMode != NOT_INTRA_SUBPARTITIONS;
      if( !CS::isDualITree( *tempCS ) )
      {
        tempCS->lumaCost = m_pcRdCost->calcRdCost( tempCS->fracBits, tempCS->dist );
        if( useIntraSubPartitions )
        {
          //the difference between the best cost so far and the current luma cost is stored to avoid testing the Cr component if the cost of luma + Cb is larger than the best cost
          maxCostAllowedForChroma = bestCS->cost < MAX_DOUBLE ? bestCS->cost - tempCS->lumaCost : MAX_DOUBLE;
        }
      }
#else
      m_pcIntraSearch->estIntraPredLumaQT( cu, partitioner );

      if (m_pcEncCfg->getUsePbIntraFast() && tempCS->dist == std::numeric_limits<Distortion>::max()
          && tempCS->interHad == 0)
      {
        interHad = 0;
        // JEM assumes only perfect reconstructions can from now on beat the inter mode
        m_modeCtrl->enforceInterHad( 0 );
Tung Nguyen's avatar
Tung Nguyen committed
#if JVET_M0464_UNI_MTS
        return;
#else
Tung Nguyen's avatar
Tung Nguyen committed
#endif
      }

      if( !CS::isDualITree( *tempCS ) )
      {
        cu.cs->picture->getRecoBuf( cu.Y() ).copyFrom( cu.cs->getRecoBuf( COMPONENT_Y ) );
Taoran Lu's avatar
Taoran Lu committed
#if JVET_M0427_INLOOP_RESHAPER
        cu.cs->picture->getPredBuf(cu.Y()).copyFrom(cu.cs->getPredBuf(COMPONENT_Y));
#endif
      }
    }

    if( tempCS->area.chromaFormat != CHROMA_400 && ( partitioner.chType == CHANNEL_TYPE_CHROMA || !CS::isDualITree( *tempCS ) ) )
    {
#if JVET_M0102_INTRA_SUBPARTITIONS
      TUIntraSubPartitioner subTuPartitioner( partitioner );
      m_pcIntraSearch->estIntraPredChromaQT( cu, ( !useIntraSubPartitions || ( CS::isDualITree( *cu.cs ) && !isLuma( CHANNEL_TYPE_CHROMA ) ) ) ? partitioner : subTuPartitioner, maxCostAllowedForChroma );
      if( useIntraSubPartitions && !cu.ispMode )
      {
        //At this point the temp cost is larger than the best cost. Therefore, we can already skip the remaining calculations
#if JVET_M0464_UNI_MTS
        return;
#else
        continue;
#endif
      }
#else
      m_pcIntraSearch->estIntraPredChromaQT( cu, partitioner );
    }

    cu.rootCbf = false;

    for( uint32_t t = 0; t < getNumberValidTBlocks( *cu.cs->pcv ); t++ )
    {
      cu.rootCbf |= cu.firstTU->cbf[t] != 0;
    }

    // Get total bits for current mode: encode CU
    m_CABACEstimator->resetBits();

    if( pps.getTransquantBypassEnabledFlag() )
    {
      m_CABACEstimator->cu_transquant_bypass_flag( cu );
    }

Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
    if ((!cu.cs->slice->isIntra() || cu.cs->slice->getSPS()->getIBCFlag())
Yu Han's avatar
Yu Han committed
#else
    if( !cu.cs->slice->isIntra()
Yu Han's avatar
Yu Han committed
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
      && cu.Y().valid()
    {
      m_CABACEstimator->cu_skip_flag ( cu );
    }
    m_CABACEstimator->pred_mode      ( cu );
    m_CABACEstimator->extend_ref_line( cu );
#if JVET_M0102_INTRA_SUBPARTITIONS
    m_CABACEstimator->isp_mode       ( cu );
#endif
    m_CABACEstimator->pcm_data       ( cu, partitioner );

    // Encode Coefficients
    CUCtx cuCtx;
    cuCtx.isDQPCoded = true;
    cuCtx.isChromaQpAdjCoded = true;
    m_CABACEstimator->cu_residual( cu, partitioner, cuCtx );

    tempCS->fracBits = m_CABACEstimator->getEstFracBits();
    tempCS->cost     = m_pcRdCost->calcRdCost(tempCS->fracBits, tempCS->dist);

#if JVET_M0102_INTRA_SUBPARTITIONS
#if !JVET_M0464_UNI_MTS
    double bestIspCost = cu.ispMode ? CS::isDualITree(*tempCS) ? tempCS->cost : tempCS->lumaCost : MAX_DOUBLE;
#endif
    const double tmpCostWithoutSplitFlags = tempCS->cost;
#endif
    xEncodeDontSplit( *tempCS, partitioner );

    xCheckDQP( *tempCS, partitioner );

#if JVET_M0102_INTRA_SUBPARTITIONS
    if( tempCS->cost < bestCS->cost )
    {
      m_modeCtrl->setBestCostWithoutSplitFlags( tmpCostWithoutSplitFlags );
    }
#endif
Nan Hu's avatar
Nan Hu committed

#if JVET_M0428_ENC_DB_OPT
Nan Hu's avatar
Nan Hu committed
    xCalDebCost( *tempCS, partitioner );
Nan Hu's avatar
Nan Hu committed
    tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();
#endif

Tung Nguyen's avatar
Tung Nguyen committed
#if !JVET_M0464_UNI_MTS
    // we save the cost of the modes for the first EMT pass
Karsten Suehring's avatar
Karsten Suehring committed
    if( !emtCuFlag ) static_cast< double& >( costSize2Nx2NemtFirstPass ) = tempCS->cost;
Tung Nguyen's avatar
Tung Nguyen committed
#endif

#if WCG_EXT
    DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda( true ) );
#else
    DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda() );
#endif
    xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );

Tung Nguyen's avatar
Tung Nguyen committed
#if !JVET_M0464_UNI_MTS
#if JVET_M0102_INTRA_SUBPARTITIONS
    //we decide to skip the second emt pass or not according to the ISP results
    if (considerEmtSecondPass && cu.ispMode && !emtCuFlag && tempCS->slice->isIntra())
    {
      double bestCostDct2NoIsp = m_modeCtrl->getEmtFirstPassNoIspCost();
      CHECKD(bestCostDct2NoIsp <= bestIspCost, "wrong cost!");
      double nSamples = (double)(cu.lwidth() << g_aucLog2[cu.lheight()]);
      double threshold = 1 + 1.4 / sqrt(nSamples);
      if (bestCostDct2NoIsp > bestIspCost*threshold)
      {
        skipSecondEmtPass = true;
        m_modeCtrl->setSkipSecondEMTPass(true);
        break;
      }
    }
#endif
    //now we check whether the second pass of SIZE_2Nx2N and the whole Intra SIZE_NxN should be skipped or not
Karsten Suehring's avatar
Karsten Suehring committed
    if( !emtCuFlag && !tempCS->slice->isIntra() && bestCU && bestCU->predMode != MODE_INTRA && m_pcEncCfg->getFastInterEMT() )
    {
      const double thEmtInterFastSkipIntra = 1.4; // Skip checking Intra if "2Nx2N using DCT2" is worse than best Inter mode
      if( costSize2Nx2NemtFirstPass > thEmtInterFastSkipIntra * bestInterCost )
      {
        skipSecondEmtPass = true;
        m_modeCtrl->setSkipSecondEMTPass( true );
        break;
      }
    }
Tung Nguyen's avatar
Tung Nguyen committed
#endif
  } //for emtCuFlag
}

void EncCu::xCheckIntraPCM(CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
  tempCS->initStructData( encTestMode.qp, encTestMode.lossless );

  CodingUnit &cu      = tempCS->addCU( CS::getArea( *tempCS, tempCS->area, partitioner.chType ), partitioner.chType );

  partitioner.setCUData( cu );
  cu.slice            = tempCS->slice;
#if HEVC_TILES_WPP
  cu.tileIdx          = tempCS->picture->tileMap->getTileIdxMap( tempCS->area.lumaPos() );
#endif
  cu.skip             = false;
  cu.predMode         = MODE_INTRA;
  cu.transQuantBypass = encTestMode.lossless;
  cu.chromaQpAdj      = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
  cu.qp               = encTestMode.qp;
  cu.ipcm             = true;

  tempCS->addPU( CS::getArea( *tempCS, tempCS->area, partitioner.chType ), partitioner.chType );

  tempCS->addTU( CS::getArea( *tempCS, tempCS->area, partitioner.chType ), partitioner.chType );

  m_pcIntraSearch->IPCMSearch(*tempCS, partitioner);

  m_CABACEstimator->getCtx() = m_CurrCtx->start;

  m_CABACEstimator->resetBits();

  if( tempCS->pps->getTransquantBypassEnabledFlag() )
  {
    m_CABACEstimator->cu_transquant_bypass_flag( cu );
  }

Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
Yu Han's avatar
Yu Han committed
  if ((!cu.cs->slice->isIntra() || cu.cs->slice->getSPS()->getIBCFlag())
Yu Han's avatar
Yu Han committed
#else
  if( !cu.cs->slice->isIntra()
Yu Han's avatar
Yu Han committed
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
    && cu.Y().valid()
  {
    m_CABACEstimator->cu_skip_flag ( cu );
  }
  m_CABACEstimator->pred_mode      ( cu );
  m_CABACEstimator->pcm_data       ( cu, partitioner );


  tempCS->fracBits = m_CABACEstimator->getEstFracBits();
  tempCS->cost     = m_pcRdCost->calcRdCost(tempCS->fracBits, tempCS->dist);

  xEncodeDontSplit( *tempCS, partitioner );

  xCheckDQP( *tempCS, partitioner );

Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
Nan Hu's avatar
Nan Hu committed
  xCalDebCost( *tempCS, partitioner );
Nan Hu's avatar
Nan Hu committed
  tempCS->useDbCost = m_pcEncCfg->getUseEncDbOpt();
#endif

#if WCG_EXT
  DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda( true ) );
#else
  DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda() );
#endif
  xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
}

void EncCu::xCheckDQP( CodingStructure& cs, Partitioner& partitioner, bool bKeepCtx )
{
  CHECK( bKeepCtx && cs.cus.size() <= 1 && partitioner.getImplicitSplit( cs ) == CU_DONT_SPLIT, "bKeepCtx should only be set in split case" );
  CHECK( !bKeepCtx && cs.cus.size() > 1, "bKeepCtx should never be set for non-split case" );

  if( !cs.pps->getUseDQP() )
  {
    return;
  }

  if (CS::isDualITree(cs) && isChroma(partitioner.chType))
  if( bKeepCtx && partitioner.currDepth != cs.pps->getMaxCuDQPDepth() )
  {
    return;
  }

  if( !bKeepCtx && partitioner.currDepth > cs.pps->getMaxCuDQPDepth() )
  {
    return;
  }

  CodingUnit* cuFirst = cs.getCU( partitioner.chType );

  CHECK( !cuFirst, "No CU available" );

  bool hasResidual = false;
  for( const auto &cu : cs.cus )
  {
    if( cu->rootCbf )
    {
      hasResidual = true;
      break;
    }
  }

  int predQP = CU::predictQP( *cuFirst, cs.prevQP[partitioner.chType] );

  if( hasResidual )
  {
    TempCtx ctxTemp( m_CtxCache );
    if( !bKeepCtx ) ctxTemp = SubCtx( Ctx::DeltaQP, m_CABACEstimator->getCtx() );

    m_CABACEstimator->resetBits();
    m_CABACEstimator->cu_qp_delta( *cuFirst, predQP, cuFirst->qp );

    cs.fracBits += m_CABACEstimator->getEstFracBits(); // dQP bits
    cs.cost      = m_pcRdCost->calcRdCost(cs.fracBits, cs.dist);


    if( !bKeepCtx ) m_CABACEstimator->getCtx() = SubCtx( Ctx::DeltaQP, ctxTemp );

    // NOTE: reset QPs for CUs without residuals up to first coded CU
    for( const auto &cu : cs.cus )
    {
      if( cu->rootCbf )
      {
        break;
      }
      cu->qp = predQP;
    }
  }
  else
  {
    // No residuals: reset CU QP to predicted value
    for( const auto &cu : cs.cus )
    {
      cu->qp = predQP;
    }
  }
}

void EncCu::xFillPCMBuffer( CodingUnit &cu )
{
  const ChromaFormat format        = cu.chromaFormat;
  const uint32_t numberValidComponents = getNumberValidComponents(format);

  for( auto &tu : CU::traverseTUs( cu ) )
  {
    for( uint32_t ch = 0; ch < numberValidComponents; ch++ )
    {
      const ComponentID compID = ComponentID( ch );

      const CompArea &compArea = tu.blocks[ compID ];

      const CPelBuf source      = tu.cs->getOrgBuf( compArea );
             PelBuf destination = tu.getPcmbuf( compID );

      destination.copyFrom( source );
    }
  }
}
#if JVET_M0253_HASH_ME
void EncCu::xCheckRDCostHashInter( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
  bool isPerfectMatch = false;

  tempCS->initStructData(encTestMode.qp, encTestMode.lossless);
  m_pcInterSearch->resetBufferedUniMotions();
  m_pcInterSearch->setAffineModeSelected(false);
  CodingUnit &cu = tempCS->addCU(tempCS->area, partitioner.chType);
  partitioner.setCUData(cu);
  cu.slice = tempCS->slice;
  cu.skip = false;
  cu.predMode = MODE_INTER;
  cu.transQuantBypass = encTestMode.lossless;
  cu.chromaQpAdj = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
  cu.qp = encTestMode.qp;
#if !JVET_M0483_IBC
  cu.ibc = false;
  CU::addPUs(cu);
  cu.mmvdSkip = false;
  cu.firstPU->mmvdMergeFlag = false;

  if (m_pcInterSearch->predInterHashSearch(cu, partitioner, isPerfectMatch))
  {
    const unsigned wIdx = gp_sizeIdxInfo->idxFrom(tempCS->area.lwidth());
    double equGBiCost = MAX_DOUBLE;

Nan Hu's avatar
Nan Hu committed
#if JVET_M0428_ENC_DB_OPT
    m_bestModeUpdated = tempCS->useDbCost = bestCS->useDbCost = false;
#endif

#if JVET_M0464_UNI_MTS
    xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 0
      , m_pImvTempCS ? m_pImvTempCS[wIdx] : NULL
      , 0
      , &equGBiCost
#else
    xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 0
      , m_pImvTempCS ? m_pImvTempCS[wIdx] : NULL
      , 1
      , 0
      , &equGBiCost
#endif
    );
Nan Hu's avatar
Nan Hu committed

#if JVET_M0428_ENC_DB_OPT
Nan Hu's avatar
Nan Hu committed
    if ( m_bestModeUpdated && bestCS->cost != MAX_DOUBLE )
    {
      xCalDebCost( *bestCS, partitioner );
    }
Nan Hu's avatar
Nan Hu committed
#endif
  }
  tempCS->initStructData(encTestMode.qp, encTestMode.lossless);
  if (cu.lwidth() != 64)
  {
    isPerfectMatch = false;
  }
  m_modeCtrl->setIsHashPerfectMatch(isPerfectMatch);
}
#endif

void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
  const Slice &slice = *tempCS->slice;

  CHECK( slice.getSliceType() == I_SLICE, "Merge modes not available for I-slices" );

  tempCS->initStructData( encTestMode.qp, encTestMode.lossless );

  MergeCtx mergeCtx;
  const SPS &sps = *tempCS->sps;

  if( sps.getSBTMVPEnabledFlag() )
  {
    Size bufSize = g_miScaling.scale( tempCS->area.lumaSize() );
    mergeCtx.subPuMvpMiBuf    = MotionBuf( m_SubPuMiBuf,    bufSize );
  }

#if JVET_M0147_DMVR
  Mv   refinedMvdL0[MAX_NUM_PARTS_IN_CTU][MRG_MAX_NUM_CANDS];
#endif
  setMergeBestSATDCost( MAX_DOUBLE );

  {
    // first get merge candidates
    CodingUnit cu( tempCS->area );
    cu.cs       = tempCS;
    cu.predMode = MODE_INTER;
    cu.slice    = tempCS->slice;
#if HEVC_TILES_WPP
    cu.tileIdx  = tempCS->picture->tileMap->getTileIdxMap(tempCS->area.lumaPos());
#endif

    PredictionUnit pu( tempCS->area );
    pu.cu = &cu;
    pu.cs = tempCS;
#if JVET_M0170_MRG_SHARELIST
    pu.shareParentPos = tempCS->sharedBndPos;
    pu.shareParentSize = tempCS->sharedBndSize;
#endif
    PU::getInterMergeCandidates(pu, mergeCtx
      , 0
    );
#if !JVET_M0068_M0171_MMVD_CLEANUP
    PU::restrictBiPredMergeCands(pu, mergeCtx);
    PU::getInterMMVDMergeCandidates(pu, mergeCtx);
  bool candHasNoResidual[MRG_MAX_NUM_CANDS + MMVD_ADD_NUM];
  for (uint32_t ui = 0; ui < MRG_MAX_NUM_CANDS + MMVD_ADD_NUM; ui++)
  {
    candHasNoResidual[ui] = false;
  }
  bool                                        bestIsSkip = false;
  bool                                        bestIsMMVDSkip = true;
  PelUnitBuf                                  acMergeBuffer[MRG_MAX_NUM_CANDS];
  PelUnitBuf                                  acMergeRealBuffer[MMVD_MRG_MAX_RD_BUF_NUM];
  PelUnitBuf *                                acMergeTempBuffer[MMVD_MRG_MAX_RD_NUM];
  PelUnitBuf *                                singleMergeTempBuffer;
  int                                         insertPos;
  unsigned                                    uiNumMrgSATDCand = mergeCtx.numValidMergeCand + MMVD_ADD_NUM;

  static_vector<unsigned, MRG_MAX_NUM_CANDS + MMVD_ADD_NUM>  RdModeList;
  bool                                        mrgTempBufSet = false;

  for (unsigned i = 0; i < MRG_MAX_NUM_CANDS + MMVD_ADD_NUM; i++)
  {
    RdModeList.push_back(i);
  }
  const UnitArea localUnitArea(tempCS->area.chromaFormat, Area(0, 0, tempCS->area.Y().width, tempCS->area.Y().height));
  for (unsigned i = 0; i < MMVD_MRG_MAX_RD_BUF_NUM; i++)
  {
    acMergeRealBuffer[i] = m_acMergeBuffer[i].getBuf(localUnitArea);
    if (i < MMVD_MRG_MAX_RD_NUM)
    {
      acMergeTempBuffer[i] = acMergeRealBuffer + i;
    }
    else
    {
      singleMergeTempBuffer = acMergeRealBuffer + i;
    }
  }

  static_vector<unsigned, MRG_MAX_NUM_CANDS + MMVD_ADD_NUM>  RdModeList2; // store the Intra mode for Intrainter
  RdModeList2.clear();
  bool isIntrainterEnabled = sps.getUseMHIntra();
  if (bestCS->area.lwidth() * bestCS->area.lheight() < 64 || bestCS->area.lwidth() >= MAX_CU_SIZE || bestCS->area.lheight() >= MAX_CU_SIZE)
  {
    isIntrainterEnabled = false;
  }
  bool isTestSkipMerge[MRG_MAX_NUM_CANDS]; // record if the merge candidate has tried skip mode
  for (uint32_t idx = 0; idx < MRG_MAX_NUM_CANDS; idx++)
  {
    isTestSkipMerge[idx] = false;
  }
  if( m_pcEncCfg->getUseFastMerge() || isIntrainterEnabled)
    if (isIntrainterEnabled)
    {
      uiNumMrgSATDCand += 1;
    }
    bestIsSkip       = false;

    if( auto blkCache = dynamic_cast< CacheBlkInfoCtrl* >( m_modeCtrl ) )
    {
Yu Han's avatar
Yu Han committed
#if JVET_M0483_IBC
      if (slice.getSPS()->getIBCFlag())
#else
      if (slice.getSPS()->getIBCMode())
Yu Han's avatar
Yu Han committed
#endif
Xiaozhong Xu's avatar
Xiaozhong Xu committed
      {
        ComprCUCtx cuECtx = m_modeCtrl->getComprCUCtx();
        bestIsSkip = blkCache->isSkip(tempCS->area) && cuECtx.bestCU;
      }
      else
      bestIsSkip = blkCache->isSkip( tempCS->area );
      bestIsMMVDSkip = blkCache->isMMVDSkip(tempCS->area);

    if (isIntrainterEnabled) // always perform low complexity check
    {
      bestIsSkip = false;
    }

    static_vector<double, MRG_MAX_NUM_CANDS + MMVD_ADD_NUM> candCostList;
    // 1. Pass: get SATD-cost for selected candidates and reduce their count
    if( !bestIsSkip )
    {
      RdModeList.clear();
      mrgTempBufSet       = true;
      const double sqrtLambdaForFirstPass = m_pcRdCost->getMotionLambda( encTestMode.lossless );

      CodingUnit &cu      = tempCS->addCU( tempCS->area, partitioner.chType );
      const double sqrtLambdaForFirstPassIntra = m_pcRdCost->getMotionLambda(cu.transQuantBypass) / double(1 << SCALE_BITS);

      partitioner.setCUData( cu );
      cu.slice            = tempCS->slice;
#if HEVC_TILES_WPP
      cu.tileIdx          = tempCS->picture->tileMap->getTileIdxMap( tempCS->area.lumaPos() );
#endif
      cu.skip             = false;
    //cu.affine
      cu.predMode         = MODE_INTER;
    //cu.LICFlag
      cu.transQuantBypass = encTestMode.lossless;
      cu.chromaQpAdj      = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
      cu.qp               = encTestMode.qp;
    //cu.emtFlag  is set below

      PredictionUnit &pu  = tempCS->addPU( cu, partitioner.chType );

      DistParam distParam;
      const bool bUseHadamard= !encTestMode.lossless;
      m_pcRdCost->setDistParam (distParam, tempCS->getOrgBuf().Y(), m_acMergeBuffer[0].Y(), sps.getBitDepth (CHANNEL_TYPE_LUMA), COMPONENT_Y, bUseHadamard);

Loading
Loading full blame...