Prepare v1.9.1-rc1 (#318)

* Preparing v1.9.1 with fixes to RC, IF and 8-bit encoding

* Set version to 1.9.1-rc1

CMakeLists.txt

@@ -10,10 +10,10 @@ if( NOT CMAKE_VERSION VERSION_LESS 3.13.0 )
 endif()
 
 # project name
-project( vvenc VERSION 1.9.0 )
+project( vvenc VERSION 1.9.1 )
 
 # set alternative version numbering for release candidates
-#set( PROJECT_VERSION_RC rc1 )
+set( PROJECT_VERSION_RC rc1 )
 if( PROJECT_VERSION_RC )
     set( PROJECT_VERSION "${PROJECT_VERSION}-${PROJECT_VERSION_RC}" )
 endif()

changelog.txt

@@ -1,3 +1,21 @@
+/////////////////////////////////////////
+tag 1.9.1-rc1
+* libvvenc:
+  - added library parameters:
+    - vvenc_config::m_maxPicWidth, vvenc_config::m_maxPicHeight to signal maximal possible picture size when
+      encoding for resolution change (e.g. using CRA_CRE decoder refresh)
+  - added medium_lowDecEnergy as a built-in preset (in C-API: VVENC_MEDIUM_LOWDECNRG)
+  - refactoring and improvements of affine search and interpolation filtering
+  - fixed 8-bit encoding from 10-bit input and proper handling of 8-bit video in BIM
+  - added proper setting of conformance window
+  - refactoring and improvements to rate control (RC)
+    - fixed potential integer overflow
+    - finalized spatial subsampling implementation for 2-pass RC
+
+* vvencFFapp:
+  - added parameter: MaxPicSize/MaxPicWidth/MaxPicHeight to control the signaled maximal picture size during
+                     resolution change
+
 /////////////////////////////////////////
 tag 1.9.0
 

include/vvenc/vvencCfg.h

@@ -114,6 +114,7 @@ typedef enum
  VVENC_MEDIUM    = 2,
  VVENC_SLOW      = 3,
  VVENC_SLOWER    = 4,
+ VVENC_MEDIUM_LOWDECNRG = 130,       // medium based low decoding energy
  VVENC_FIRSTPASS = 254,
  VVENC_TOOLTEST  = 255,
 }vvencPresetMode;
@@ -848,7 +849,7 @@ VVENC_DECL bool vvenc_init_config_parameter( vvenc_config *cfg );
 #define VVENC_OPT_TICKSPERSEC          "tickspersec"          // m_TicksPerSecond
 #define VVENC_OPT_INPUTBITDEPTH        "inputbitdepth"        // m_inputBitDepth
 #define VVENC_OPT_FRAMES               "framestobeencoded"    // m_framesToBeEncoded
-#define VVENC_OPT_PRESET               "preset"               // set preset like "faster,fast,medium,slow,slower
+#define VVENC_OPT_PRESET               "preset"               // set preset like "faster,fast,medium,slow,slower,medium_lowDecEnergy
 #define VVENC_OPT_THREADS              "threads"              // m_numThreads
 #define VVENC_OPT_BITRATE              "bitrate"              // m_RCTargetBitrate
 #define VVENC_OPT_QP                   "qp"                   // m_QP

source/Lib/EncoderLib/BitAllocation.cpp

@@ -114,7 +114,7 @@ double filterAndCalculateAverageActivity (const Pel* pSrc, const int iSrcStride,
   }
   if (spVisAct)
   {
-    *spVisAct = unsigned (0.5 + spatAct * double(bitDepth < 12 ? 1 << (12 - bitDepth) : 1));
+    *spVisAct = unsigned (0.5 + spatAct * double (bitDepth < 12 ? 1 << (12 - bitDepth) : 1));
   }
 
   // skip first row as there may be a black border frame
@@ -474,7 +474,7 @@ int BitAllocation::applyQPAdaptationSlice (const Slice* slice, const VVEncCfg* e
       {
         const uint32_t nCtu = ctuBoundingAddr - ctuStartAddr;
 
-        pic->picSpVisAct = ClipBD (uint16_t ((picSpVisAct + (nCtu >> 1)) / nCtu), bitDepth);
+        pic->picSpVisAct = ClipBD (uint16_t ((picSpVisAct + (nCtu >> 1)) / nCtu), 12);
       }
       if (encCfg->m_usePerceptQPATempFiltISlice && slice->isIntra() && pic->getOrigBuf (compID).buf != pic->getOrigBufPrev (compID, PREV_FRAME_1).buf && zeroMinActCTUs * 2 > ctuBoundingAddr - ctuStartAddr)
       {

source/Lib/EncoderLib/PreProcess.cpp

@@ -356,7 +356,7 @@ uint16_t PreProcess::xGetPicVisualActivity(Picture* curPic, const Picture* refPi
       &picSpVisAct);
 
   uint16_t ret = ClipBD( (uint16_t)( 0.5 + visActY ), bitDepth );
-  curPic->picSpVisAct = ClipBD( (uint16_t) picSpVisAct, bitDepth );
+  curPic->picSpVisAct = ClipBD( (uint16_t) picSpVisAct, 12 );
 
   return ret;
 }

source/Lib/EncoderLib/RateCtrl.cpp

@@ -635,7 +635,7 @@ void RateCtrl::readStatsFile()
 
 void RateCtrl::adjustStatsDownsample()
 {
-  int meanValue = 0; //MCTF or Activity
+  int64_t meanValue = 0; //MCTF or Activity
   int amount = 0;
   auto itr = m_listRCFirstPassStats.begin();
   for (; itr != m_listRCFirstPassStats.end(); itr++)
@@ -652,7 +652,7 @@ void RateCtrl::adjustStatsDownsample()
   if (meanValue != 0)
   {
     meanValue = meanValue / amount;
-    int sumVar = 0;
+    int64_t sumVar = 0;
     int numVar = 0;
     auto itrv = m_listRCFirstPassStats.begin();
     for (; itrv != m_listRCFirstPassStats.end(); itrv++)
@@ -666,8 +666,8 @@ void RateCtrl::adjustStatsDownsample()
     }
     if (numVar)
     {
-      sumVar = sumVar / (numVar - 1);
-      sumVar = (int) sqrt((double) sumVar);
+      sumVar = sumVar / std::max(1, numVar - 1);
+      sumVar = int64_t (0.5 + sqrt((double) sumVar));
     }
     int value_gopbefore = 0;
     int value_gopcur = 0;
@@ -701,7 +701,7 @@ void RateCtrl::adjustStatsDownsample()
         doChangeBits = false;
         if (stat.gopNum != 0)
         {
-          int var_cur = abs(statValue - meanValue);
+          const int64_t var_cur = abs(statValue - meanValue);
           if (var_cur > (sumVar << 1))
           {
             doChangeBits = true;

source/Lib/apputils/VVEncAppCfg.h

@@ -121,6 +121,8 @@ const std::vector<SVPair<vvencPresetMode>> PresetToEnumMap =
   { "medium",    vvencPresetMode::VVENC_MEDIUM },
   { "slow",      vvencPresetMode::VVENC_SLOW },
   { "slower",    vvencPresetMode::VVENC_SLOWER },
+  { "medium_lowDecEnergy", vvencPresetMode::VVENC_MEDIUM_LOWDECNRG },
+  { "medium_lowdecenergy", vvencPresetMode::VVENC_MEDIUM_LOWDECNRG },
   { "firstpass", vvencPresetMode::VVENC_FIRSTPASS },
   { "tooltest",  vvencPresetMode::VVENC_TOOLTEST },
 };
@@ -612,7 +614,7 @@ int parse( int argc, char* argv[], vvenc_config* c, std::ostream& rcOstr )
   {
     opts.setSubSection("Encoder Options");
     opts.addOptions()
-    ("preset",                                          toPreset,                                            "preset for detailed parameter configuration (faster, fast, medium, slow, slower)")
+    ("preset",                                          toPreset,                                            "preset for detailed parameter configuration (faster, fast, medium, slow, slower, medium_lowDecEnergy)")
     ("bitrate,b",                                       toBitrate,                                           "bitrate for rate control (0: constant-QP encoding without rate control; otherwise\n"
                                                                                                              "bits/second; use e.g. 1.5M, 1.5Mbps, 1500k, 1500kbps, 1500000bps, 1500000)")
     ("maxrate,m",                                       toMaxRate,                                           "approximate maximum instantaneous bitrate for constrained VBR in rate control (0:\n"
@@ -634,7 +636,7 @@ int parse( int argc, char* argv[], vvenc_config* c, std::ostream& rcOstr )
     opts.setSubSection("Threading, performance");
     opts.addOptions()
     ("Threads,t",                                       c->m_numThreads,                                     "Number of threads")
-    ("preset",                                          toPreset,                                            "select preset for specific encoding setting (faster, fast, medium, slow, slower)")
+    ("preset",                                          toPreset,                                            "select preset for specific encoding setting (faster, fast, medium, slow, slower, medium_lowDecEnergy)")
     ("Tiles",                                           toNumTiles,                                          "Set number of tile columns and rows")
     ;
 
@@ -1081,8 +1083,8 @@ int parse( int argc, char* argv[], vvenc_config* c, std::ostream& rcOstr )
     ("ReduceIntraChromaModesFullRD",                    c->m_reduceIntraChromaModesFullRD,                   "Reduce modes for chroma full RD intra search")
     ("FirstPassMode",                                   c->m_FirstPassMode,                                  "Mode for first encoding pass when using rate control "
                                                                                                                "(0: default, 1: faster, 2: faster with temporal downsampling, "
-                                                                                                                "3: (experimental) faster with resolution downsampling, "
-                                                                                                                "4: (experimental) faster with temporal and resolution downsampling)" )
+                                                                                                                "3: faster with resolution downsampling, "
+                                                                                                                "4: faster with temporal and resolution downsampling)" )
     ;
 
     opts.setSubSection("Input Options");
@@ -1370,7 +1372,7 @@ virtual std::string getAppConfigAsString( vvenc_config* c, vvencMsgLevel eMsgLev
   if( eMsgLevel >= VVENC_INFO )
   {
     if( format != ext )
-      css << loglvl << "Input File                             : " << m_inputFileName << " (" << format << ")\n";
+      css << loglvl << "Input File                             : " << m_inputFileName << "  (" << format << ")\n";
     else
       css << loglvl << "Input File                             : " << m_inputFileName << "\n";
     css << loglvl << "Bitstream File                         : " << m_bitstreamFileName << "\n";

source/Lib/vvenc/vvencCfg.cpp

@@ -741,7 +741,7 @@ VVENC_DECL bool vvenc_init_config_parameter( vvenc_config *c )
   {
     css << "TicksPerSecond must be a multiple of FrameRate/Framescale (" << c->m_FrameRate << "/" << c->m_FrameScale << "). Use 27000000 for NTSC content"  << std::endl;
   }
-  vvenc_confirmParameter( c, ( c->m_TicksPerSecond > 0 ) && ((int64_t)c->m_TicksPerSecond*(int64_t)c->m_FrameScale)%c->m_FrameRate, css.str().c_str() );
+  vvenc_confirmParameter( c, ( c->m_TicksPerSecond > 0 && c->m_FrameRate > 0) && ((int64_t)c->m_TicksPerSecond*(int64_t)c->m_FrameScale)%c->m_FrameRate, css.str().c_str() );
 
 
   vvenc_confirmParameter( c, c->m_numThreads < -1 || c->m_numThreads > 256,                              "Number of threads out of range (-1 <= t <= 256)");
@@ -788,6 +788,16 @@ VVENC_DECL bool vvenc_init_config_parameter( vvenc_config *c )
 
   vvenc::MsgLog msg(c->m_msgCtx, c->m_msgFnc);
 
+  if( c->m_FirstPassMode > 2 && c->m_RCTargetBitrate != 0 )
+  {
+    // lookahead will only be set to 0 later
+    if( c->m_LookAhead > 0 || c->m_RCNumPasses == 1 || std::min( c->m_SourceWidth, c->m_SourceHeight ) < 720 ) 
+    {
+      c->m_FirstPassMode -= 2;  
+      msg.log( VVENC_NOTICE, "First pass spatial downsampling (FirstPassMode=3/4) cannot be used for single pass encoding and videos with resolution lower than 720p, changing FirstPassMode to %d!\n", c->m_FirstPassMode );
+    }
+  }
+
   const double d = (3840.0 * 2160.0) / double (c->m_SourceWidth * c->m_SourceHeight);
   const int rcQP = (c->m_RCInitialQP > 0 ? std::min (vvenc::MAX_QP, c->m_RCInitialQP) : std::max (0, vvenc::MAX_QP_PERCEPT_QPA - (c->m_FirstPassMode > 2 ? 4 : 2) - int (0.5 + sqrt ((d * std::max (0, c->m_RCTargetBitrate)) / 500000.0))));
 
@@ -808,7 +818,6 @@ VVENC_DECL bool vvenc_init_config_parameter( vvenc_config *c )
   vvenc_confirmParameter( c, c->m_RCTargetBitrate != VVENC_RC_OFF && ( c->m_RCTargetBitrate < 0 || c->m_RCTargetBitrate > 800000000 ),  "TargetBitrate must be between 0 and 800000000" );
   vvenc_confirmParameter( c, c->m_RCTargetBitrate != VVENC_RC_OFF && (int64_t) c->m_RCMaxBitrate * 2 < (int64_t) c->m_RCTargetBitrate * 3, "MaxBitrate must be at least 1.5*TargetBitrate" );
   vvenc_confirmParameter( c, c->m_RCTargetBitrate != VVENC_RC_OFF && ( c->m_FirstPassMode < 0 || c->m_FirstPassMode > 4 ), "FirstPassMode must be 0, 1, 2, 3, or 4" );
-  vvenc_confirmParameter( c, std::min( c->m_SourceWidth, c->m_SourceHeight ) < 720 && ( c->m_FirstPassMode == 3 || c->m_FirstPassMode == 4 ), "Spatial downsampling cannot be applied to small-resolution videos (i.e. below 720p), use FirstPassMode values up to '2'");
 
   if ( c->m_internChromaFormat < 0 || c->m_internChromaFormat >= VVENC_NUM_CHROMA_FORMAT )
   {
@@ -1215,7 +1224,7 @@ VVENC_DECL bool vvenc_init_config_parameter( vvenc_config *c )
 
   if( c->m_maxPicWidth > 0 && c->m_maxPicHeight > 0 )
   {
-    vvenc_confirmParameter( c, !c->m_rprEnabledFlag || !c->m_resChangeInClvsEnabled, "if a maxSize is set, both RPR and resChangeInClvsEnabled have to enabled" );
+    vvenc_confirmParameter( c, !c->m_rprEnabledFlag || !c->m_resChangeInClvsEnabled, "if max picture size is set, both RPR and resChangeInClvsEnabled have to be enabled" );
   }
 
   if( c->m_IntraPeriod == 0 && c->m_IntraPeriodSec > 0 )
@@ -1895,8 +1904,6 @@ static bool checkCfgParameter( vvenc_config *c )
   vvenc_confirmParameter( c, c->m_LookAhead && c->m_RCNumPasses != 1,       "Look-ahead encoding is not supported for two-pass rate control" );
   vvenc_confirmParameter( c, !c->m_LookAhead && c->m_RCNumPasses == 1 && c->m_RCTargetBitrate > 0, "Look-ahead encoding must be used with one-pass rate control" );
   vvenc_confirmParameter( c, c->m_LookAhead && c->m_RCTargetBitrate == 0,   "Look-ahead encoding is not supported when rate control is disabled" );
-  vvenc_confirmParameter( c, (c->m_LookAhead || c->m_RCNumPasses == 1) && c->m_FirstPassMode > 2, "Look-ahead or one pass RC can not be used with resolution downsampling" );
-
 
   vvenc_confirmParameter(c, !((c->m_level==VVENC_LEVEL1)
     || (c->m_level==VVENC_LEVEL2) || (c->m_level==VVENC_LEVEL2_1)
@@ -2588,6 +2595,7 @@ VVENC_DECL int vvenc_init_preset( vvenc_config *c, vvencPresetMode preset )
       break;
 
     case vvencPresetMode::VVENC_MEDIUM:
+    case vvencPresetMode::VVENC_MEDIUM_LOWDECNRG:
 
       // motion estimation
       c->m_SearchRange                     = 384;
@@ -2658,6 +2666,28 @@ VVENC_DECL int vvenc_init_preset( vvenc_config *c, vvencPresetMode preset )
       c->m_IBCFastMethod                   = 3;
       c->m_TSsize                          = 4;
 
+      if( preset == vvencPresetMode::VVENC_MEDIUM_LOWDECNRG )
+      {
+        c->m_numRefPics                    = 2;
+        c->m_deblockLastTLayers            = 1;
+        c->m_maxMTTDepth                   = 332222;
+        c->m_maxMTTDepthI                  = 3;
+        c->m_Affine                        = 3;
+        c->m_alfSpeed                      = 1;
+        c->m_BCW                           = 2;
+        c->m_BDOF                          = 0;
+        c->m_DMVR                          = 0;
+        c->m_ISP                           = 0;
+        c->m_LFNST                         = 0;
+        c->m_lumaReshapeEnable             = 0;
+        c->m_MIP                           = 0;
+        c->m_useFastMIP                    = 0;
+        c->m_bUseSAO                       = true;
+        c->m_saoScc                        = true;
+        c->m_SbTMVP                        = 0;
+        c->m_useFastMrg                    = 2;
+      }
+
       break;
 
     case vvencPresetMode::VVENC_SLOW:
@@ -2918,7 +2948,9 @@ VVENC_DECL const char* vvenc_get_config_as_string( vvenc_config *c, vvencMsgLeve
 
   if( eMsgLevel >= VVENC_INFO )
   {
-    css << loglvl << "Internal format                        : " << c->m_PadSourceWidth << "x" << c->m_PadSourceHeight << " " <<  (double)c->m_FrameRate/c->m_FrameScale << "Hz "  << getDynamicRangeStr(c->m_HdrMode) << "\n";
+    css << loglvl << "Internal format                        : " << c->m_PadSourceWidth << "x" << c->m_PadSourceHeight << "  " <<  (double)c->m_FrameRate/c->m_FrameScale << " Hz  "  << getDynamicRangeStr(c->m_HdrMode) << "\n";
+    css << loglvl << "Threads                                : " << c->m_numThreads << "  (parallel frames: " << c->m_maxParallelFrames <<")\n";
+
     css << loglvl << "Rate control                           : ";
 
     if ( c->m_RCTargetBitrate > 0 )
@@ -3157,6 +3189,7 @@ VVENC_DECL const char* vvenc_get_config_as_string( vvenc_config *c, vvencMsgLeve
     }
 
     css << "LookAhead:" << c->m_LookAhead << " ";
+    css << "FirstPassMode:" << c->m_FirstPassMode << " ";
 
     css << "\n" << loglvl << "PARALLEL PROCESSING CFG: ";
     css << "NumThreads:" << c->m_numThreads << " ";