JVET-M0255: MMVD without Fractional Distances for SCC.

cfg/encoder_intra_vtm.cfg

@@ -112,6 +112,7 @@ DepQuant                     : 1
 IMV                          : 2
 ALF                          : 1
 CPR                          : 0      # turned off in CTC 
+AllowDisFracMMVD             : 1
 
 # Fast tools
 PBIntraFast                  : 1

cfg/encoder_lowdelay_P_vtm.cfg

@@ -129,6 +129,7 @@ IMV                          : 2
 ALF                          : 1
 MHIntra                      : 1
 CPR                          : 0      # turned off in CTC
+AllowDisFracMMVD             : 1
 
 # Fast tools
 PBIntraFast                  : 1

cfg/encoder_lowdelay_vtm.cfg

@@ -132,6 +132,7 @@ GBiFast                      : 1
 MHIntra                      : 1
 Triangle                     : 1
 CPR                          : 0      # turned off in CTC
+AllowDisFracMMVD             : 1
 
 # Fast tools
 PBIntraFast                  : 1

cfg/encoder_randomaccess_vtm.cfg

@@ -147,6 +147,7 @@ BIO                          : 1
 MHIntra                      : 1
 Triangle                     : 1
 CPR                          : 0      # turned off in CTC
+AllowDisFracMMVD             : 1
 
 # Fast tools
 PBIntraFast                  : 1

source/App/EncoderApp/EncApp.cpp

@@ -257,7 +257,9 @@ void EncApp::xInitLibCfg()
 #endif  
   m_cEncLib.setUseMHIntra                                        ( m_MHIntra );
   m_cEncLib.setUseTriangle                                       ( m_Triangle );
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  m_cEncLib.setAllowDisFracMMVD                                  ( m_allowDisFracMMVD );
+#endif
   m_cEncLib.setCPRMode                                           ( m_CPRMode );
   m_cEncLib.setCPRLocalSearchRangeX                              ( m_CPRLocalSearchRangeX );
   m_cEncLib.setCPRLocalSearchRangeY                              ( m_CPRLocalSearchRangeY );

source/App/EncoderApp/EncAppCfg.cpp

@@ -864,7 +864,9 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
 #endif
   ("MHIntra",                                         m_MHIntra,                                        false, "Enable MHIntra mode")
   ("Triangle",                                        m_Triangle,                                       false, "Enable triangular shape motion vector prediction (0:off, 1:on)")
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  ("AllowDisFracMMVD",                                m_allowDisFracMMVD,                               false, "Disable fractional MVD in MMVD mode adaptively")
+#endif
   ( "CPR",                                            m_CPRMode,                                           0u, "CPRMode (0x1:enabled, 0x0:disabled)  [default: disabled]")
   ( "CPRLocalSearchRangeX",                           m_CPRLocalSearchRangeX,                            128u, "Search range of CPR local search in x direction")
   ( "CPRLocalSearchRangeY",                           m_CPRLocalSearchRangeY,                            128u, "Search range of CPR local search in y direction")
@@ -3145,6 +3147,9 @@ void EncAppCfg::xPrintParameter()
 #endif
     msg(VERBOSE, "MHIntra:%d ", m_MHIntra);
     msg( VERBOSE, "Triangle:%d ", m_Triangle );
+#if JVET_M0255_FRACMMVD_SWITCH
+    msg( VERBOSE, "AllowDisFracMMVD:%d ", m_allowDisFracMMVD );
+#endif
   }
     msg(VERBOSE, "CPR:%d ", m_CPRMode);
   msg( VERBOSE, "WrapAround:%d ", m_wrapAround);

source/App/EncoderApp/EncAppCfg.h

@@ -237,7 +237,9 @@ protected:
 
   bool      m_MHIntra;
   bool      m_Triangle;
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  bool      m_allowDisFracMMVD;
+#endif
 
   unsigned  m_CPRMode;
   unsigned  m_CPRLocalSearchRangeX;

source/Lib/CommonLib/ContextModelling.cpp

@@ -501,8 +501,15 @@ void MergeCtx::setMmvdMergeCandiInfo(PredictionUnit& pu, int candIdx)
   tempIdx = tempIdx - fPosBaseIdx * (MMVD_MAX_REFINE_NUM);
   fPosStep = tempIdx / 4;
   fPosPosition = tempIdx - fPosStep * (4);
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  int offset = refMvdCands[fPosStep];
+  if ( pu.cu->slice->getDisFracMMVD() )
+  {
+    offset <<= 2;
+  }
+#else
   const int offset = refMvdCands[fPosStep];
+#endif
   const int refList0 = mmvdBaseMv[fPosBaseIdx][0].refIdx;
   const int refList1 = mmvdBaseMv[fPosBaseIdx][1].refIdx;
 

source/Lib/CommonLib/Slice.cpp

@@ -92,6 +92,9 @@ Slice::Slice()
 , m_colRefIdx                     ( 0 )
 , m_maxNumMergeCand               ( 0 )
 , m_maxNumAffineMergeCand         ( 0 )
+#if JVET_M0255_FRACMMVD_SWITCH
+, m_disFracMMVD                   ( false )
+#endif
 , m_uiTLayer                      ( 0 )
 , m_bTLayerSwitchingFlag          ( false )
 , m_sliceMode                     ( NO_SLICES )
@@ -211,6 +214,9 @@ void Slice::initSlice()
 
   m_bFinalized=false;
 
+#if JVET_M0255_FRACMMVD_SWITCH
+  m_disFracMMVD          = false;
+#endif
   m_substreamSizes.clear();
   m_cabacInitFlag        = false;
   m_cabacWinUpdateMode   = 0;
@@ -861,6 +867,9 @@ void Slice::copySliceInfo(Slice *pSrc, bool cpyAlmostAll)
   m_enableTMVPFlag                = pSrc->m_enableTMVPFlag;
   m_maxNumMergeCand               = pSrc->m_maxNumMergeCand;
   m_maxNumAffineMergeCand         = pSrc->m_maxNumAffineMergeCand;
+#if JVET_M0255_FRACMMVD_SWITCH
+  m_disFracMMVD                   = pSrc->m_disFracMMVD;
+#endif
   if( cpyAlmostAll ) m_encCABACTableIdx  = pSrc->m_encCABACTableIdx;
   m_splitConsOverrideFlag         = pSrc->m_splitConsOverrideFlag;
   m_uiMinQTSize                   = pSrc->m_uiMinQTSize;
@@ -1767,6 +1776,9 @@ SPSNext::SPSNext( SPS& sps )
   , m_MTTEnabled                ( false )
   , m_MHIntra                   ( false )
   , m_Triangle                  ( false )
+#if JVET_M0255_FRACMMVD_SWITCH
+  , allowDisFracMMVD            ( false )
+#endif
 #if ENABLE_WPP_PARALLELISM
   , m_NextDQP                   ( false )
 #endif

source/Lib/CommonLib/Slice.h

@@ -820,6 +820,9 @@ private:
   bool              m_MTTEnabled;                 //
   bool              m_MHIntra;
   bool              m_Triangle;
+#if JVET_M0255_FRACMMVD_SWITCH
+  bool              allowDisFracMMVD;
+#endif
 #if ENABLE_WPP_PARALLELISM
   bool              m_NextDQP;
 #endif
@@ -916,6 +919,10 @@ public:
   bool      getUseTriangle        ()                                      const     { return m_Triangle; }
   void      setCPRMode            (unsigned CPRMode)                                { m_CPRMode = CPRMode; }
   unsigned  getCPRMode            ()                                      const     { return m_CPRMode; }
+#if JVET_M0255_FRACMMVD_SWITCH
+  void      setAllowDisFracMMVD   ( bool b )                                        { allowDisFracMMVD = b;    }
+  bool      getAllowDisFracMMVD   ()                                      const     { return allowDisFracMMVD; }
+#endif
 };
 
 
@@ -1602,7 +1609,9 @@ private:
   uint32_t                       m_colRefIdx;
   uint32_t                       m_maxNumMergeCand;
   uint32_t                   m_maxNumAffineMergeCand;
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  bool                       m_disFracMMVD;
+#endif
   double                     m_lambdas[MAX_NUM_COMPONENT];
 
   bool                       m_abEqualRef  [NUM_REF_PIC_LIST_01][MAX_NUM_REF][MAX_NUM_REF];
@@ -1843,7 +1852,10 @@ public:
   uint32_t                    getMaxNumMergeCand() const                             { return m_maxNumMergeCand;                                     }
   void                        setMaxNumAffineMergeCand( uint32_t val )               { m_maxNumAffineMergeCand = val;  }
   uint32_t                    getMaxNumAffineMergeCand() const                       { return m_maxNumAffineMergeCand; }
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  void                        setDisFracMMVD( bool val )                             { m_disFracMMVD = val;                                          }
+  bool                        getDisFracMMVD() const                                 { return m_disFracMMVD;                                         }
+#endif
   void                        setNoOutputPriorPicsFlag( bool val )                   { m_noOutputPriorPicsFlag = val;                                }
   bool                        getNoOutputPriorPicsFlag() const                       { return m_noOutputPriorPicsFlag;                               }
 

source/Lib/CommonLib/TypeDef.h

@@ -64,6 +64,7 @@
 
 #define JVET_M0464_UNI_MTS                                1
 #define JVET_M0068_M0171_MMVD_CLEANUP                     1 // MMVD cleanup with 1) flip removal, 2) L1 zero vector fix, 3) bi-pred restriction after merge/MMVD
+#define JVET_M0255_FRACMMVD_SWITCH                        1 // disable fractional MVD in MMVD adaptively
 
 #if JVET_M0464_UNI_MTS
 typedef std::pair<int, bool> TrMode;

source/Lib/DecoderLib/VLCReader.cpp

@@ -818,6 +818,9 @@ void HLSyntaxReader::parseSPSNext( SPSNext& spsNext, const bool usePCM )
   }
   READ_FLAG( symbol,  "mtt_enabled_flag" );                       spsNext.setMTTMode                ( symbol );
   READ_FLAG( symbol,  "mhintra_flag" );                           spsNext.setUseMHIntra             ( symbol != 0 );
+#if JVET_M0255_FRACMMVD_SWITCH
+  READ_FLAG( symbol,  "sps_fracmmvd_disabled_flag" );             spsNext.setAllowDisFracMMVD       ( symbol != 0 );
+#endif
   READ_FLAG( symbol,    "triangle_flag" );                          spsNext.setUseTriangle            ( symbol != 0 );
 #if ENABLE_WPP_PARALLELISM
   READ_FLAG( symbol,  "next_dqp_enabled_flag" );                  spsNext.setUseNextDQP             ( symbol != 0 );
@@ -1767,6 +1770,13 @@ void HLSyntaxReader::parseSliceHeader (Slice* pcSlice, ParameterSetManager *para
         READ_UVLC( uiCode, "five_minus_max_num_affine_merge_cand" );
         pcSlice->setMaxNumAffineMergeCand( AFFINE_MRG_MAX_NUM_CANDS - uiCode );
       }
+#if JVET_M0255_FRACMMVD_SWITCH
+      if ( sps->getSpsNext().getAllowDisFracMMVD() )
+      {
+        READ_FLAG( uiCode, "tile_group_fracmmvd_disabled_flag" );
+        pcSlice->setDisFracMMVD( uiCode ? true : false );
+      }
+#endif
     }
 
     READ_SVLC( iCode, "slice_qp_delta" );

source/Lib/EncoderLib/EncCfg.h

@@ -238,7 +238,9 @@ protected:
 
   bool      m_MHIntra;
   bool      m_Triangle;
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  bool      m_allowDisFracMMVD;
+#endif
   unsigned  m_CPRMode;
   unsigned  m_CPRLocalSearchRangeX;
   unsigned  m_CPRLocalSearchRangeY;
@@ -754,7 +756,10 @@ public:
   bool      getUseMHIntra                   ()         const { return m_MHIntra; }
   void      setUseTriangle                  ( bool b )       { m_Triangle = b; }
   bool      getUseTriangle                  ()         const { return m_Triangle; }
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  void      setAllowDisFracMMVD             ( bool b )       { m_allowDisFracMMVD = b;    }
+  bool      getAllowDisFracMMVD             ()         const { return m_allowDisFracMMVD; }
+#endif
 
   void      setCPRMode                      (unsigned n)     { m_CPRMode = n; }
   unsigned  getCPRMode                      ()         const { return m_CPRMode; }

source/Lib/EncoderLib/EncCu.cpp

@@ -280,7 +280,11 @@ void EncCu::init( EncLib* pcEncLib, const SPS& sps PARL_PARAM( const int tId ) )
   ::memset(m_subMergeBlkNum, 0, sizeof(m_subMergeBlkNum));
   m_prevPOC = MAX_UINT;
 
+#if  JVET_M0255_FRACMMVD_SWITCH
+  if ( ( m_pcEncCfg->getCPRHashSearch() && m_pcEncCfg->getCPRMode() ) || m_pcEncCfg->getAllowDisFracMMVD() )
+#else
   if (m_pcEncCfg->getCPRHashSearch() && m_pcEncCfg->getCPRMode())
+#endif
   {
     m_cprHashMap.init(m_pcEncCfg->getSourceWidth(), m_pcEncCfg->getSourceHeight());
   }
@@ -292,10 +296,12 @@ void EncCu::init( EncLib* pcEncLib, const SPS& sps PARL_PARAM( const int tId ) )
 
 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->getCPRHashSearch() && ctuRsAddr == 0 && cs.slice->getSPS()->getSpsNext().getCPRMode())
   {
     m_cprHashMap.rebuildPicHashMap(cs.picture->getOrigBuf());
   }
+#endif
   m_modeCtrl->initCTUEncoding( *cs.slice );
 
 #if ENABLE_SPLIT_PARALLELISM

source/Lib/EncoderLib/EncCu.h

@@ -151,6 +151,10 @@ public:
 
   void   setMergeBestSATDCost(double cost) { m_mergeBestSATDCost = cost; }
   double getMergeBestSATDCost()            { return m_mergeBestSATDCost; }
+#if JVET_M0255_FRACMMVD_SWITCH
+  CprHashMap& getCprHashMap()              { return m_cprHashMap;        }
+  EncCfg*     getEncCfg()            const { return m_pcEncCfg;          }
+#endif
 
   ~EncCu();
 

source/Lib/EncoderLib/EncLib.cpp

@@ -891,7 +891,9 @@ void EncLib::xInitSPS(SPS &sps)
 
   sps.getSpsNext().setUseMHIntra            ( m_MHIntra );
   sps.getSpsNext().setUseTriangle           ( m_Triangle );
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  sps.getSpsNext().setAllowDisFracMMVD      ( m_allowDisFracMMVD );
+#endif
   sps.getSpsNext().setCPRMode               ( m_CPRMode );
 
   sps.setWrapAroundEnabledFlag                      ( m_wrapAround );

source/Lib/EncoderLib/EncSlice.cpp

@@ -1387,8 +1387,48 @@ void EncSlice::compressSlice( Picture* pcPic, const bool bCompressEntireSlice, c
 
 }
 
+#if JVET_M0255_FRACMMVD_SWITCH
+void EncSlice::checkDisFracMmvd( Picture* pcPic, uint32_t startCtuTsAddr, uint32_t boundingCtuTsAddr )
+{
+  CodingStructure&  cs            = *pcPic->cs;
+  Slice* pcSlice                  = cs.slice;
+  const PreCalcValues& pcv        = *cs.pcv;
+  const uint32_t    widthInCtus   = pcv.widthInCtus;
+#if HEVC_TILES_WPP
+  const TileMap&  tileMap         = *pcPic->tileMap;
+#endif
+  const uint32_t hashThreshold    = 20;
+  uint32_t totalCtu               = 0;
+  uint32_t hashRatio              = 0;
+
+  if ( !pcSlice->getSPS()->getSpsNext().getAllowDisFracMMVD() )
+  {
+    return;
+  }
+
+  for ( uint32_t ctuTsAddr = startCtuTsAddr; ctuTsAddr < boundingCtuTsAddr; ctuTsAddr++ )
+  {
+#if HEVC_TILES_WPP
+    const uint32_t ctuRsAddr = tileMap.getCtuTsToRsAddrMap( ctuTsAddr );
+#else
+    const uint32_t ctuRsAddr = ctuTsAddr;
+#endif
+    const uint32_t ctuXPosInCtus        = ctuRsAddr % widthInCtus;
+    const uint32_t ctuYPosInCtus        = ctuRsAddr / widthInCtus;
 
+    const Position pos ( ctuXPosInCtus * pcv.maxCUWidth, ctuYPosInCtus * pcv.maxCUHeight );
+    const UnitArea ctuArea( cs.area.chromaFormat, Area( pos.x, pos.y, pcv.maxCUWidth, pcv.maxCUHeight ) );
 
+    hashRatio += m_pcCuEncoder->getCprHashMap().getHashHitRatio( ctuArea.Y() );
+    totalCtu++;
+  }
+
+  if ( hashRatio > totalCtu * hashThreshold )
+  {
+    pcSlice->setDisFracMMVD( true );
+  }
+}
+#endif
 
 void EncSlice::encodeCtus( Picture* pcPic, const bool bCompressEntireSlice, const bool bFastDeltaQP, uint32_t startCtuTsAddr, uint32_t boundingCtuTsAddr, EncLib* pEncLib )
 {
@@ -1439,7 +1479,14 @@ void EncSlice::encodeCtus( Picture* pcPic, const bool bCompressEntireSlice, cons
 #if HEVC_DEPENDENT_SLICES
   }
 #endif
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  if ( pcSlice->getSPS()->getSpsNext().getAllowDisFracMMVD() || 
+      ( pcSlice->getSPS()->getSpsNext().getCPRMode() && m_pcCuEncoder->getEncCfg()->getCPRHashSearch() ) )
+  {
+    m_pcCuEncoder->getCprHashMap().rebuildPicHashMap( cs.picture->getOrigBuf() );
+  }
+  checkDisFracMmvd( pcPic, startCtuTsAddr, boundingCtuTsAddr );
+#endif
   // for every CTU in the slice segment (may terminate sooner if there is a byte limit on the slice-segment)
   for( uint32_t ctuTsAddr = startCtuTsAddr; ctuTsAddr < boundingCtuTsAddr; ctuTsAddr++ )
   {

source/Lib/EncoderLib/EncSlice.h

@@ -146,7 +146,9 @@ public:
   static
 #endif
   void    encodeCtus          ( Picture* pcPic, const bool bCompressEntireSlice, const bool bFastDeltaQP, uint32_t startCtuTsAddr, uint32_t boundingCtuTsAddr, EncLib* pcEncLib );
-
+#if JVET_M0255_FRACMMVD_SWITCH
+  void    checkDisFracMmvd    ( Picture* pcPic, uint32_t startCtuTsAddr, uint32_t boundingCtuTsAddr );
+#endif
 
   // misc. functions
   void    setSearchRange      ( Slice* pcSlice  );                                  ///< set ME range adaptively

source/Lib/EncoderLib/VLCWriter.cpp

@@ -559,6 +559,9 @@ void HLSWriter::codeSPSNext( const SPSNext& spsNext, const bool usePCM )
 
   WRITE_FLAG( spsNext.getMTTEnabled() ? 1 : 0,                                                  "mtt_enabled_flag" );
   WRITE_FLAG( spsNext.getUseMHIntra() ? 1 : 0,                                                  "mhintra_flag" );
+#if JVET_M0255_FRACMMVD_SWITCH
+  WRITE_FLAG( spsNext.getAllowDisFracMMVD() ? 1 : 0,                                            "sps_fracmmvd_disabled_flag" );
+#endif
   WRITE_FLAG( spsNext.getUseTriangle() ? 1: 0,                                                  "triangle_flag" );
 #if ENABLE_WPP_PARALLELISM
   WRITE_FLAG( spsNext.getUseNextDQP(),                                                          "next_dqp_enabled_flag" );
@@ -1285,6 +1288,12 @@ void HLSWriter::codeSliceHeader         ( Slice* pcSlice )
         CHECK( pcSlice->getMaxNumAffineMergeCand() > AFFINE_MRG_MAX_NUM_CANDS, "More affine merge candidates signalled than supported" );
         WRITE_UVLC( AFFINE_MRG_MAX_NUM_CANDS - pcSlice->getMaxNumAffineMergeCand(), "five_minus_max_num_affine_merge_cand" );
       }
+#if JVET_M0255_FRACMMVD_SWITCH
+      if ( pcSlice->getSPS()->getSpsNext().getAllowDisFracMMVD() )
+      {
+        WRITE_FLAG( pcSlice->getDisFracMMVD(), "tile_group_fracmmvd_disabled_flag" );
+      }
+#endif
     }
     int iCode = pcSlice->getSliceQp() - ( pcSlice->getPPS()->getPicInitQPMinus26() + 26 );
     WRITE_SVLC( iCode, "slice_qp_delta" );