diff --git a/CMakeLists.txt b/CMakeLists.txt
index 282de51154e9e96256d775b965180f65a940da2f..f052c0828bbe80df7f7cf0cded99ecc16e742924 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -70,9 +70,11 @@ if( OpenMP_FOUND )
endif()
# Enable warnings for some generators and toolsets.
-bb_enable_warnings( gcc warnings-as-errors -Wno-sign-compare )
+# bb_enable_warnings( gcc warnings-as-errors -Wno-sign-compare )
# bb_enable_warnings( gcc -Wno-unused-variable )
# bb_enable_warnings( gcc-4.8 warnings-as-errors -Wno-unused-variable )
+# for gcc 8.2:
+bb_enable_warnings( gcc warnings-as-errors -Wno-sign-compare -Wno-class-memaccess)
if( XCODE )
bb_enable_warnings( clang warnings-as-errors
@@ -104,6 +106,7 @@ endif()
if( MSVC )
add_compile_options( "/MP" )
add_compile_options( "/EHsc" )
+ add_compile_options( "/MT" )
endif()
# set address sanitizer compiler arguments
diff --git a/cfg/encoder_intra_vtm.cfg b/cfg/encoder_intra_vtm.cfg
index f6665ccda58e9a5830521de0c992eecac99758eb..9ebe108b223591b97bdf19d0aa9fedf453d9ec95 100644
--- a/cfg/encoder_intra_vtm.cfg
+++ b/cfg/encoder_intra_vtm.cfg
@@ -106,13 +106,11 @@ EMT : 1
EMTFast : 1
Affine : 1
HighPrecMv : 1
-
SubPuMvp : 1
MaxNumMergeCand : 6
LMChroma : 1 # use CCLM only
DepQuant : 1
IMV : 2
-
ALF : 1
# Fast tools
diff --git a/cfg/encoder_lowdelay_P_vtm.cfg b/cfg/encoder_lowdelay_P_vtm.cfg
index 5f588af822131c8a0c58e7658c4006f23bd51aa0..d83bed2d615bbe62a8bed6139c9daf01129f36f8 100644
--- a/cfg/encoder_lowdelay_P_vtm.cfg
+++ b/cfg/encoder_lowdelay_P_vtm.cfg
@@ -122,13 +122,11 @@ EMT : 1
EMTFast : 1
Affine : 1
HighPrecMv : 1
-
SubPuMvp : 1
MaxNumMergeCand : 6
LMChroma : 1 # use CCLM only
DepQuant : 1
IMV : 2
-
ALF : 1
# Fast tools
diff --git a/cfg/encoder_lowdelay_vtm.cfg b/cfg/encoder_lowdelay_vtm.cfg
index 281ec093f234ae30f099c1059bb7cd2e3e2851b2..3bcde02d4138a1f6e484d528c77ba67f77d138b4 100644
--- a/cfg/encoder_lowdelay_vtm.cfg
+++ b/cfg/encoder_lowdelay_vtm.cfg
@@ -122,13 +122,11 @@ EMT : 1
EMTFast : 1
Affine : 1
HighPrecMv : 1
-
SubPuMvp : 1
MaxNumMergeCand : 6
LMChroma : 1 # use CCLM only
DepQuant : 1
IMV : 2
-
ALF : 1
# Fast tools
diff --git a/cfg/encoder_randomaccess_vtm.cfg b/cfg/encoder_randomaccess_vtm.cfg
index 7e391142bb295f82824e5364bc8ffb143f424218..46d02af14ed62aa539294bc39e7b69ca46d7018a 100644
--- a/cfg/encoder_randomaccess_vtm.cfg
+++ b/cfg/encoder_randomaccess_vtm.cfg
@@ -136,13 +136,11 @@ EMT : 1
EMTFast : 1
Affine : 1
HighPrecMv : 1
-
SubPuMvp : 1
MaxNumMergeCand : 6
LMChroma : 1 # use CCLM only
DepQuant : 1
IMV : 2
-
ALF : 1
# Fast tools
@@ -150,7 +148,6 @@ PBIntraFast : 1
FastMrg : 1
AMaxBT : 1
-
### DO NOT ADD ANYTHING BELOW THIS LINE ###
### DO NOT DELETE THE EMPTY LINE BELOW ###
diff --git a/doc/.gitignore b/doc/.gitignore
new file mode 100644
index 0000000000000000000000000000000000000000..d58111897c8449a88a7e83a74f74c44a8ede7055
--- /dev/null
+++ b/doc/.gitignore
@@ -0,0 +1,8 @@
+# ignore these Latex files
+jctvcdoc.cls
+*.aux
+*.log
+*.lot
+*.out
+*.toc
+*~
diff --git a/doc/pyuv_format.pdf b/doc/pyuv_format.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..62c669e0bb4c52d5756b428ce0838eeb57a6d06f
Binary files /dev/null and b/doc/pyuv_format.pdf differ
diff --git a/doc/software-manual.pdf b/doc/software-manual.pdf
index 91bcc5a9f172ec570b4312133e74d3b867e71e52..0b4925cf32a94097c096cdc47707a5dbbebe3637 100644
Binary files a/doc/software-manual.pdf and b/doc/software-manual.pdf differ
diff --git a/doc/software-manual.tex b/doc/software-manual.tex
index ceb6b39c5d5b4736d3d7a96e94ee224d2d847cc2..0cd31316c6225c0c1924a52b909b8e5b26588341 100644
--- a/doc/software-manual.tex
+++ b/doc/software-manual.tex
@@ -3021,4 +3021,226 @@ If the decoder is compiled with the macro RExt__DECODER_DEBUG_BIT_STATISTICS def
The Linux makefile will compile both the analyser and standard version when the `all' or `everything' target is used (where the latter will also build high-bit-depth executables).
+
+\section{Block statistics extension}
+\label{sec:block-stat-extens}
+
+The block statistics extension enables straightforward visualization and statistical analysis of coding tool
+usage in encoded bitstreams. The extension enables the reference
+software encoder and decoder to write out statistics files in a configurable
+way, which in turn can be loaded into a suitable YUV player for overlay of the
+reconstructed YUV sequence, or can be used for statistical analysis at a
+selectable scope (e.g. block/picture/sequence level). An example implementation
+for such visualization is available with the open-source YUView player
+(https://github.com/IENT/YUView).
+
+
+\subsection{Usage}
+\label{sec:usage}
+
+The software has to be compiled with the macros ENABLE_TRACING and
+K0149_BLOCK_STATISTICS defined as 1. The statistics can be written by either
+encoder or decoder.
+
+The extension adds additional trace channels to the ``dtrace'' functionality of
+the software. The following trace channels were added:
+\begin{description}
+\item[D_BLOCK_STATISTICS_ALL] All syntax elements are written, no matter whether
+ they are actually encoded or derived.
+\item[D_BLOCK_STATISTICS_CODED] Tries to write only syntax elements, which have
+ also been encoded.
+\end{description}
+
+The following additional encoder options are available (part of ``dtrace''). See
+the file dtrace_next.h for more details.
+
+\begin{OptionTableNoShorthand}{Decoder options}{tab:decoder-block-statistics}
+\Option{TraceFile} &
+%\ShortOption{\None} &
+\Default{\None} &
+File name of the produced trace file.
+\\
+
+\Option{TraceRule} &
+%\ShortOption{-b} &
+\Default{\NotSet} &
+Specifies which traces should be saved, and for which POCs.
+\\
+
+\end{OptionTableNoShorthand}
+
+Concrete examples of calls for generating a block statistics file are:
+\begin{verbatim}
+bin/DecoderAppStatic -b str/BasketballDrive_1920x1080_QP37.vvc \
+ --TraceFile="stats/BasketballDrive_1920x1080_QP37_coded.vtmbmsstats" \
+ --TraceRule="D_BLOCK_STATISTICS_CODED:poc>=0"
+
+bin/DecoderAppStatic -b str/BasketballDrive_1920x1080_QP37.vvc \
+ --TraceFile="stats/BasketballDrive_1920x1080_QP37_all.vtmbmsstats" \
+ --TraceRule="D_BLOCK_STATISTICS_ALL:poc>=0"
+\end{verbatim}
+
+
+\subsection{Block statistics file formats}
+\label{sec:block-stat-file}
+The trace file will contain a header listing information of all available block
+statistics. For each statistic it lists a type and a scale for vectors or range
+for integers if applicable:
+\begin{verbatim}
+# VTMBMS Block Statistics
+# Sequence size: [832x 480]
+# Block Statistic Type: PredMode; Flag;
+# Block Statistic Type: MergeFlag; Flag;
+# Block Statistic Type: MVL0; Vector; Scale: 4
+# Block Statistic Type: MVL1; Vector; Scale: 4
+# Block Statistic Type: IPCM; Flag;
+# Block Statistic Type: Y_IntraMode; Integer; [0, 73]
+# Block Statistic Type: Cb_IntraMode; Integer; [0, 73]
+\end{verbatim}
+
+Two formats are available for the statistics for each block, a human readable
+format and a CSV based format. The header remains the same for both cases.
+
+For both formats each row contains the information for one block statistic. The
+order of the data is: picture order count (POC), location of top left corner of
+the block, size of the block, name of the statistic, and value of the
+statistic.
+The macro BLOCK_STATS_AS_CSV is available in order to choose the required format.
+The human readable format can also be easily processed with other software, for
+example YUView, using regular expressions. The CSV based formats provides the
+universal interface required by spreadsheet applications.
+
+The human readable format is based on the format used for the other dtrace
+statistics. Some examples for this format are:
+\begin{verbatim}
+BlockStat: POC 16 @( 112, 0) [ 8x 8] SkipFlag=1
+BlockStat: POC 16 @( 112, 0) [ 8x 8] InterDir=1
+BlockStat: POC 16 @( 112, 0) [ 8x 8] MergeFlag=1
+BlockStat: POC 16 @( 112, 0) [ 8x 8] MergeIdx=0
+BlockStat: POC 16 @( 112, 0) [ 8x 8] MergeType=0
+BlockStat: POC 16 @( 112, 0) [ 8x 8] MVPIdxL0=255
+BlockStat: POC 16 @( 112, 0) [ 8x 8] MVPNumL0=255
+BlockStat: POC 16 @( 112, 0) [ 8x 8] RefIdxL0=0
+BlockStat: POC 16 @( 112, 0) [ 8x 8] MVDL0={ 0, 0}
+BlockStat: POC 16 @( 112, 0) [ 8x 8] MVL0={ -70, 18}
+BlockStat: POC 16 @( 112, 8) [ 8x 8] PredMode=0
+BlockStat: POC 16 @( 112, 8) [ 8x 8] PartSize=0
+\end{verbatim}
+
+Some examples of the CSV based format are:
+\begin{verbatim}
+BlockStat;16; 112; 0; 8; 8;SkipFlag;1
+BlockStat;16; 112; 0; 8; 8;InterDir;1
+BlockStat;16; 112; 0; 8; 8;MergeFlag;1
+BlockStat;16; 112; 0; 8; 8;MergeIdx;0
+BlockStat;16; 112; 0; 8; 8;MergeType;0
+BlockStat;16; 112; 0; 8; 8;MVPIdxL0;255
+BlockStat;16; 112; 0; 8; 8;MVPNumL0;255
+BlockStat;16; 112; 0; 8; 8;RefIdxL0;0
+BlockStat;16; 112; 0; 8; 8;MVDL0; 0; 0
+BlockStat;16; 112; 0; 8; 8;MVL0; -70; 18
+BlockStat;16; 112; 8; 8; 8;PredMode;0
+BlockStat;16; 112; 8; 8; 8;PartSize;0
+\end{verbatim}
+
+\subsection{Visualization}
+\label{sec:visualization}
+
+The block statistics can be viewed with YUView, which is freely available under
+GPLv3: \url{https://github.com/IENT/YUView}. The latest releases and the master
+branch have the functionality required for viewing the block statistics. YUView
+assumes that the file extension of block statistics file is
+“.vtmbmsstatsâ€. However, if a file is not recognized you can choose from a list
+of supported file formats.
+
+
+Statistics can be overlaid with YUV sequences. Some example snapshots are:
+
+\begin{figure}[htpb]
+ \centering
+ \includegraphics[width=0.8\linewidth]{figures/YUView}
+ \caption{YUView}
+ \label{fig:yuview}
+\end{figure}
+
+\begin{figure}[htpb]
+ \centering
+ \includegraphics[width=0.5\linewidth]{figures/raceHorsesShot2MotionVectors}
+ \caption{Motion vectors}
+ \label{fig:motion-vectors}
+\end{figure}
+
+
+\begin{figure}[htpb]
+ \centering
+ \includegraphics[width=0.5\linewidth]{figures/raceHorsesShot3SkipFlag}
+ \caption{Skip flag}
+ \label{fig:skip-flag}
+\end{figure}
+
+\subsection{Adding statistics}
+\label{sec:adding-statistics}
+
+In order to add further block statistics, do the following:
+
+
+\begin{description}
+\item[source/Lib/CommonLib/dtrace_blockstatistics.h]
+ Add your statistic to the BlockStatistic enum:
+\begin{verbatim}
+enum class BlockStatistic {
+ // general
+ PredMode,
+ PartSize,
+ Depth,
+\end{verbatim}
+
+Further, add your statistic to the map blockstatistic2description:
+\begin{verbatim}
+static const std::map<BlockStatistic,
+ std::tuple<std::string, BlockStatisticType, std::string>>
+ blockstatistic2description =
+{
+ { BlockStatistic::PredMode,
+ std::tuple<std::string, BlockStatisticType, std::string>
+ {"PredMode", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::MergeFlag,
+ std::tuple<std::string, BlockStatisticType, std::string>
+ {"MergeFlag", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::MVL0,
+ std::tuple<std::string, BlockStatisticType, std::string>
+ {"MVL0", BlockStatisticType::Vector, "Scale: 4"}},
+ YOURS
+\end{verbatim}
+
+
+\item[source/Lib/CommonLib/dtrace_blockstatistics.cpp] All code for
+ writing syntax elements is kept in this file in
+ getAndStoreBlockStatistics. This function is called once for each
+ CTU, after it has been en/decoded. The following macros have been
+ defined to facilitate writing of block statistics:
+\begin{verbatim}
+DTRACE_BLOCK_SCALAR(ctx,channel,cs_cu_pu,stat_type,val)
+DTRACE_BLOCK_SCALAR_CHROMA(ctx,channel,cs_cu_pu,stat_type,val)
+DTRACE_BLOCK_VECTOR(ctx,channel,cu_pu,stat_type,v_x,v_y)
+DTRACE_BLOCK_AFFINETF(ctx,channel,pu,stat_type,v_x0,v_y0,v_x1,v_y1,v_x2,v_y2)
+\end{verbatim}
+
+An example:
+\begin{verbatim}
+DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL,
+ cu, GetBlockStatisticName(BlockStatistic::PredMode), cu.predMode);
+\end{verbatim}
+
+
+\item[Block statistics for debugging] The statistics can also be used
+ to write out other data, not just syntax elements. Add your
+ statistics to dtrace_blockstatistics.h. Where it should be used the
+ following headers have to be included:
+\begin{verbatim}
+#include "dtrace_next.h"
+#include "dtrace_blockstatistics.h"
+\end{verbatim}
+\end{description}
+
\end{document}
diff --git a/source/App/DecoderAnalyserApp/CMakeLists.txt b/source/App/DecoderAnalyserApp/CMakeLists.txt
index 46f0e44363824ac28b541c4e8576272936c854c6..bd4975afa64248238e18513ebcd6de594c42093c 100644
--- a/source/App/DecoderAnalyserApp/CMakeLists.txt
+++ b/source/App/DecoderAnalyserApp/CMakeLists.txt
@@ -29,7 +29,7 @@ target_compile_definitions( ${EXE_NAME} PUBLIC RExt__DECODER_DEBUG_BIT_STATISTIC
target_compile_definitions( ${EXE_NAME} PUBLIC RExt__DECODER_DEBUG_TOOL_STATISTICS=1 )
if( ENABLE_VTM )
- target_compile_definitions( ${EXE_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${EXE_NAME} PUBLIC BMS_TOOLS=0 )
endif()
if( SET_ENABLE_TRACING )
diff --git a/source/App/DecoderApp/CMakeLists.txt b/source/App/DecoderApp/CMakeLists.txt
index 8bd727935494d1eb74379616cd882ebf75986065..dec7c6e1ec22c9734e8b4cb33cae2c3862c0f8f5 100644
--- a/source/App/DecoderApp/CMakeLists.txt
+++ b/source/App/DecoderApp/CMakeLists.txt
@@ -27,7 +27,7 @@ add_executable( ${EXE_NAME} ${SRC_FILES} ${INC_FILES} ${NATVIS_FILES} ${CMAKE_CU
include_directories(${CMAKE_CURRENT_BINARY_DIR})
if( ENABLE_VTM )
- target_compile_definitions( ${EXE_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${EXE_NAME} PUBLIC BMS_TOOLS=0 )
endif()
diff --git a/source/App/DecoderApp/DecApp.cpp b/source/App/DecoderApp/DecApp.cpp
index a7f26e83303ff5b107c9152fc424d4d427f04b93..3c2f719c0959159f7a4280c15e362513d2b0fe7e 100644
--- a/source/App/DecoderApp/DecApp.cpp
+++ b/source/App/DecoderApp/DecApp.cpp
@@ -121,6 +121,10 @@ uint32_t DecApp::decode()
* requires the DecApp::decode() method to be called again with the same
* nal unit. */
#if RExt__DECODER_DEBUG_STATISTICS
+ CodingStatistics& stat = CodingStatistics::GetSingletonInstance();
+ CHECK(m_statMode < STATS__MODE_NONE || m_statMode > STATS__MODE_ALL, "Wrong coding statistics output mode");
+ stat.m_mode = m_statMode;
+
CodingStatistics::CodingStatisticsData* backupStats = new CodingStatistics::CodingStatisticsData(CodingStatistics::GetStatistics());
#endif
@@ -210,6 +214,11 @@ uint32_t DecApp::decode()
}
}
+ if (m_packedYUVMode && (m_outputBitDepth[CH_L] != 10 && m_outputBitDepth[CH_L] != 12))
+ {
+ EXIT ("Invalid output bit-depth for packed YUV output, aborting\n");
+ }
+
m_cVideoIOYuvReconFile.open( m_reconFileName, true, m_outputBitDepth, m_outputBitDepth, bitDepths.recon ); // write mode
openedReconFile = true;
}
@@ -279,9 +288,9 @@ void DecApp::xCreateDecLib()
m_cDecLib.create();
// initialize decoder class
- m_cDecLib.init(
+ m_cDecLib.init(
#if JVET_J0090_MEMORY_BANDWITH_MEASURE
- m_cacheCfgFile
+ m_cacheCfgFile
#endif
);
m_cDecLib.setDecodedPictureHashSEIEnabled(m_decodedPictureHashSEIEnabled);
@@ -394,11 +403,13 @@ void DecApp::xWriteOutput( PicList* pcListPic, uint32_t tId )
if (display)
{
m_cVideoIOYuvReconFile.write( pcPicTop->getRecoBuf(), pcPicBottom->getRecoBuf(),
- m_outputColourSpaceConvert,
- conf.getWindowLeftOffset() + defDisp.getWindowLeftOffset(),
- conf.getWindowRightOffset() + defDisp.getWindowRightOffset(),
- conf.getWindowTopOffset() + defDisp.getWindowTopOffset(),
- conf.getWindowBottomOffset() + defDisp.getWindowBottomOffset(), NUM_CHROMA_FORMAT, isTff );
+ m_outputColourSpaceConvert,
+ false, // TODO: m_packedYUVMode,
+ conf.getWindowLeftOffset() + defDisp.getWindowLeftOffset(),
+ conf.getWindowRightOffset() + defDisp.getWindowRightOffset(),
+ conf.getWindowTopOffset() + defDisp.getWindowTopOffset(),
+ conf.getWindowBottomOffset() + defDisp.getWindowBottomOffset(),
+ NUM_CHROMA_FORMAT, isTff );
}
}
@@ -445,6 +456,7 @@ void DecApp::xWriteOutput( PicList* pcListPic, uint32_t tId )
m_cVideoIOYuvReconFile.write( pcPic->getRecoBuf(),
m_outputColourSpaceConvert,
+ m_packedYUVMode,
conf.getWindowLeftOffset() + defDisp.getWindowLeftOffset(),
conf.getWindowRightOffset() + defDisp.getWindowRightOffset(),
conf.getWindowTopOffset() + defDisp.getWindowTopOffset(),
@@ -502,15 +514,18 @@ void DecApp::xFlushOutput( PicList* pcListPic )
// write to file
if ( !m_reconFileName.empty() )
{
- const Window &conf = pcPicTop->cs->sps->getConformanceWindow();
+ const Window &conf = pcPicTop->cs->sps->getConformanceWindow();
const Window defDisp = (m_respectDefDispWindow && pcPicTop->cs->sps->getVuiParametersPresentFlag()) ? pcPicTop->cs->sps->getVuiParameters()->getDefaultDisplayWindow() : Window();
- const bool isTff = pcPicTop->topField;
+ const bool isTff = pcPicTop->topField;
+
m_cVideoIOYuvReconFile.write( pcPicTop->getRecoBuf(), pcPicBottom->getRecoBuf(),
- m_outputColourSpaceConvert,
- conf.getWindowLeftOffset() + defDisp.getWindowLeftOffset(),
- conf.getWindowRightOffset() + defDisp.getWindowRightOffset(),
- conf.getWindowTopOffset() + defDisp.getWindowTopOffset(),
- conf.getWindowBottomOffset() + defDisp.getWindowBottomOffset(), NUM_CHROMA_FORMAT, isTff );
+ m_outputColourSpaceConvert,
+ false, // TODO: m_packedYUVMode,
+ conf.getWindowLeftOffset() + defDisp.getWindowLeftOffset(),
+ conf.getWindowRightOffset() + defDisp.getWindowRightOffset(),
+ conf.getWindowTopOffset() + defDisp.getWindowTopOffset(),
+ conf.getWindowBottomOffset() + defDisp.getWindowBottomOffset(),
+ NUM_CHROMA_FORMAT, isTff );
}
// update POC of display order
@@ -560,6 +575,7 @@ void DecApp::xFlushOutput( PicList* pcListPic )
m_cVideoIOYuvReconFile.write( pcPic->getRecoBuf(),
m_outputColourSpaceConvert,
+ m_packedYUVMode,
conf.getWindowLeftOffset() + defDisp.getWindowLeftOffset(),
conf.getWindowRightOffset() + defDisp.getWindowRightOffset(),
conf.getWindowTopOffset() + defDisp.getWindowTopOffset(),
diff --git a/source/App/DecoderApp/DecAppCfg.cpp b/source/App/DecoderApp/DecAppCfg.cpp
index 6e520b49eeca61d15991ddc1552d89fa9e637276..ca55e2493a7e9f28c0d619666abed5c477cfde9f 100644
--- a/source/App/DecoderApp/DecAppCfg.cpp
+++ b/source/App/DecoderApp/DecAppCfg.cpp
@@ -96,6 +96,7 @@ bool DecAppCfg::parseCfg( int argc, char* argv[] )
("SEIColourRemappingInfoFilename", m_colourRemapSEIFileName, string(""), "Colour Remapping YUV output file name. If empty, no remapping is applied (ignore SEI message)\n")
("OutputDecodedSEIMessagesFilename", m_outputDecodedSEIMessagesFilename, string(""), "When non empty, output decoded SEI messages to the indicated file. If file is '-', then output to stdout\n")
("ClipOutputVideoToRec709Range", m_bClipOutputVideoToRec709Range, false, "If true then clip output video to the Rec. 709 Range on saving")
+ ("PYUV", m_packedYUVMode, false, "If true then output 10-bit and 12-bit YUV data as 5-byte and 3-byte (respectively) packed YUV data. Ignored for interlaced output.")
#if ENABLE_TRACING
("TraceChannelsList", bTracingChannelsList, false, "List all available tracing channels" )
("TraceRule", sTracingRule, string( "" ), "Tracing rule (ex: \"D_CABAC:poc==8\" or \"D_REC_CB_LUMA:poc==8\")" )
@@ -103,6 +104,13 @@ bool DecAppCfg::parseCfg( int argc, char* argv[] )
#endif
#if JVET_J0090_MEMORY_BANDWITH_MEASURE
("CacheCfg", m_cacheCfgFile, string( "" ), "CacheCfg File" )
+#endif
+#if RExt__DECODER_DEBUG_STATISTICS
+ ("Stats", m_statMode, 3, "Control decoder debugging statistic output mode\n"
+ "\t0: disable statistic\n"
+ "\t1: enable bit statistic\n"
+ "\t2: enable tool statistic\n"
+ "\t3: enable bit and tool statistic\n")
#endif
;
@@ -220,6 +228,8 @@ DecAppCfg::DecAppCfg()
, m_respectDefDispWindow(0)
, m_outputDecodedSEIMessagesFilename()
, m_bClipOutputVideoToRec709Range(false)
+, m_packedYUVMode(false)
+, m_statMode(0)
{
for (uint32_t channelTypeIndex = 0; channelTypeIndex < MAX_NUM_CHANNEL_TYPE; channelTypeIndex++)
{
diff --git a/source/App/DecoderApp/DecAppCfg.h b/source/App/DecoderApp/DecAppCfg.h
index b721849a26366a6fe5b6209036b4db922ff77b92..7b370eb83804f11b678afaa513226d2a10d512da 100644
--- a/source/App/DecoderApp/DecAppCfg.h
+++ b/source/App/DecoderApp/DecAppCfg.h
@@ -70,7 +70,9 @@ protected:
int m_respectDefDispWindow; ///< Only output content inside the default display window
std::string m_outputDecodedSEIMessagesFilename; ///< filename to output decoded SEI messages to. If '-', then use stdout. If empty, do not output details.
bool m_bClipOutputVideoToRec709Range; ///< If true, clip the output video to the Rec 709 range on saving.
+ bool m_packedYUVMode; ///< If true, output 10-bit and 12-bit YUV data as 5-byte and 3-byte (respectively) packed YUV data
std::string m_cacheCfgFile; ///< Config file of cache model
+ int m_statMode; ///< Config statistic mode (0 - bit stat, 1 - tool stat, 3 - both)
public:
DecAppCfg();
diff --git a/source/App/DecoderApp/decmain.cpp b/source/App/DecoderApp/decmain.cpp
index 73c44ea0223c122f58fd44720751b6edefcf79d3..61db2a5e36a52e5465b5e265000d069771859d49 100644
--- a/source/App/DecoderApp/decmain.cpp
+++ b/source/App/DecoderApp/decmain.cpp
@@ -57,9 +57,9 @@ int main(int argc, char* argv[])
// print information
fprintf( stdout, "\n" );
#ifdef SVNREVISION
- fprintf( stdout, "VVCSoftware: VTM Decoder Version %s (%s@r%s) ", NEXT_SOFTWARE_VERSION, SVNRELATIVEURL, SVNREVISION /*NV_VERSION*/ );
+ fprintf( stdout, "VVCSoftware: BMS Decoder Version %s (%s@r%s) ", NEXT_SOFTWARE_VERSION, SVNRELATIVEURL, SVNREVISION /*NV_VERSION*/ );
#else
- fprintf( stdout, "VVCSoftware: VTM Decoder Version %s ", NEXT_SOFTWARE_VERSION /*NV_VERSION*/ );
+ fprintf( stdout, "VVCSoftware: BMS Decoder Version %s ", NEXT_SOFTWARE_VERSION /*NV_VERSION*/ );
#endif
fprintf( stdout, NVM_ONOS );
fprintf( stdout, NVM_COMPILEDBY );
diff --git a/source/App/EncoderApp/CMakeLists.txt b/source/App/EncoderApp/CMakeLists.txt
index 49ba1259f7bd21fa18ce71ceec10798eb3eb1ada..d5322a6b6f7740b35c7666be2ec0847f52e50709 100644
--- a/source/App/EncoderApp/CMakeLists.txt
+++ b/source/App/EncoderApp/CMakeLists.txt
@@ -29,7 +29,7 @@ add_executable( ${EXE_NAME} ${SRC_FILES} ${INC_FILES} ${NATVIS_FILES} ${CMAKE_CU
include_directories(${CMAKE_CURRENT_BINARY_DIR})
if( ENABLE_VTM )
- target_compile_definitions( ${EXE_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${EXE_NAME} PUBLIC BMS_TOOLS=0 )
endif()
if( SET_ENABLE_TRACING )
diff --git a/source/App/EncoderApp/EncApp.cpp b/source/App/EncoderApp/EncApp.cpp
index b48d190874b0fa1066265e4e19e957b232ce894c..59a2e070d57c2a80b45449b634ea5991a5f11768 100644
--- a/source/App/EncoderApp/EncApp.cpp
+++ b/source/App/EncoderApp/EncApp.cpp
@@ -247,6 +247,9 @@ void EncApp::xInitLibCfg()
m_cEncLib.setFastIntraEMT ( m_FastEMT & m_EMT & 1 );
m_cEncLib.setInterEMT ( ( m_EMT >> 1 ) & 1 );
m_cEncLib.setFastInterEMT ( ( m_FastEMT >> 1 ) & ( m_EMT >> 1 ) & 1 );
+#endif
+#if JVET_K0157
+ m_cEncLib.setUseCompositeRef ( m_compositeRefEnabled );
#endif
// ADD_NEW_TOOL : (encoder app) add setting of tool enabling flags and associated parameters here
@@ -549,6 +552,17 @@ void EncApp::xCreateLib( std::list<PelUnitBuf*>& recBufList
#endif
if (!m_reconFileName.empty())
{
+ if (m_packedYUVMode && ((m_outputBitDepth[CH_L] != 10 && m_outputBitDepth[CH_L] != 12)
+ || ((m_iSourceWidth & (1 + (m_outputBitDepth[CH_L] & 3))) != 0)))
+ {
+ EXIT ("Invalid output bit-depth or image width for packed YUV output, aborting\n");
+ }
+ if (m_packedYUVMode && (m_chromaFormatIDC != CHROMA_400) && ((m_outputBitDepth[CH_C] != 10 && m_outputBitDepth[CH_C] != 12)
+ || (((m_iSourceWidth / SPS::getWinUnitX (m_chromaFormatIDC)) & (1 + (m_outputBitDepth[CH_C] & 3))) != 0)))
+ {
+ EXIT ("Invalid chroma output bit-depth or image width for packed YUV output, aborting\n");
+ }
+
m_cVideoIOYuvReconFile.open(m_reconFileName, true, m_outputBitDepth, m_outputBitDepth, m_internalBitDepth); // write mode
}
@@ -737,7 +751,10 @@ void EncApp::xWriteOutput( int iNumEncoded, std::list<PelUnitBuf*>& recBufList
if (!m_reconFileName.empty())
{
- m_cVideoIOYuvReconFile.write( *pcPicYuvRecTop, *pcPicYuvRecBottom, ipCSC, m_confWinLeft, m_confWinRight, m_confWinTop, m_confWinBottom, NUM_CHROMA_FORMAT, m_isTopFieldFirst );
+ m_cVideoIOYuvReconFile.write( *pcPicYuvRecTop, *pcPicYuvRecBottom,
+ ipCSC,
+ false, // TODO: m_packedYUVMode,
+ m_confWinLeft, m_confWinRight, m_confWinTop, m_confWinBottom, NUM_CHROMA_FORMAT, m_isTopFieldFirst );
}
}
}
@@ -749,7 +766,9 @@ void EncApp::xWriteOutput( int iNumEncoded, std::list<PelUnitBuf*>& recBufList
if (!m_reconFileName.empty())
{
m_cVideoIOYuvReconFile.write( *pcPicYuvRec,
- ipCSC, m_confWinLeft, m_confWinRight, m_confWinTop, m_confWinBottom, NUM_CHROMA_FORMAT, m_bClipOutputVideoToRec709Range );
+ ipCSC,
+ m_packedYUVMode,
+ m_confWinLeft, m_confWinRight, m_confWinTop, m_confWinBottom, NUM_CHROMA_FORMAT, m_bClipOutputVideoToRec709Range );
}
}
}
diff --git a/source/App/EncoderApp/EncAppCfg.cpp b/source/App/EncoderApp/EncAppCfg.cpp
index 4070fcc13a3ca88005a324ed4cc6d256df767ed3..5149ada2620ed43d54ee65ffd6e1c3f57547ef8b 100644
--- a/source/App/EncoderApp/EncAppCfg.cpp
+++ b/source/App/EncoderApp/EncAppCfg.cpp
@@ -107,6 +107,7 @@ EncAppCfg::EncAppCfg()
: m_inputColourSpaceConvert(IPCOLOURSPACE_UNCHANGED)
, m_snrInternalColourSpace(false)
, m_outputInternalColourSpace(false)
+, m_packedYUVMode(false)
#if EXTENSION_360_VIDEO
, m_ext360(*this)
#endif
@@ -766,6 +767,7 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
("FramesToBeEncoded,f", m_framesToBeEncoded, 0, "Number of frames to be encoded (default=all)")
("ClipInputVideoToRec709Range", m_bClipInputVideoToRec709Range, false, "If true then clip input video to the Rec. 709 Range on loading when InternalBitDepth is less than MSBExtendedBitDepth")
("ClipOutputVideoToRec709Range", m_bClipOutputVideoToRec709Range, false, "If true then clip output video to the Rec. 709 Range on saving when OutputBitDepth is less than InternalBitDepth")
+ ("PYUV", m_packedYUVMode, false, "If true then output 10-bit and 12-bit YUV data as 5-byte and 3-byte (respectively) packed YUV data. Ignored for interlaced output.")
("SummaryOutFilename", m_summaryOutFilename, string(), "Filename to use for producing summary output file. If empty, do not produce a file.")
("SummaryPicFilenameBase", m_summaryPicFilenameBase, string(), "Base filename to use for producing summary picture output files. The actual filenames used will have I.txt, P.txt and B.txt appended. If empty, do not produce a file.")
("SummaryVerboseness", m_summaryVerboseness, 0u, "Specifies the level of the verboseness of the text output")
@@ -849,6 +851,9 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
"\t1: Enable fast methods only for Intra EMT\n"
"\t2: Enable fast methods only for Inter EMT\n"
"\t3: Enable fast methods for both Intra & Inter EMT\n")
+#endif
+#if JVET_K0157
+ ("CompositeLTReference", m_compositeRefEnabled, false, "Enable Composite Long Term Reference Frame")
#endif
// ADD_NEW_TOOL : (encoder app) add parsing parameters here
@@ -912,7 +917,7 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
/* Quantization parameters */
#if QP_SWITCHING_FOR_PARALLEL
("QP,q", m_iQP, 30, "Qp value")
- ("QPIncrementFrame,-qpif", m_qpIncrementAtSourceFrame, OptionalValue<uint32_t>(), "If a source file frame number is specified, the internal QP will be incremented for all POCs associated with source frames >= frame number. If empty, do not increment.")
+ ("QPIncrementFrame,-qpif", m_qpIncrementAtSourceFrame, OptionalValue<uint32_t>(), "If a source file frame number is specified, the internal QP will be incremented for all POCs associated with source frames >= frame number. If empty, do not increment.")
#else
("QP,q", m_fQP, 30.0, "Qp value, if value is float, QP is switched once during encoding")
#endif
@@ -1291,6 +1296,21 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
po::ErrorReporter err;
const list<const char*>& argv_unhandled = po::scanArgv(opts, argc, (const char**) argv, err);
+#if JVET_K0157
+ if (m_compositeRefEnabled)
+ {
+ for (int i = 0; i < m_iGOPSize; i++)
+ {
+ m_GOPList[i].m_POC *= 2;
+ m_GOPList[i].m_deltaRPS *= 2;
+ for (int j = 0; j < m_GOPList[i].m_numRefPics; j++)
+ {
+ m_GOPList[i].m_referencePics[j] *= 2;
+ }
+ }
+ }
+#endif
+
for (list<const char*>::const_iterator it = argv_unhandled.begin(); it != argv_unhandled.end(); it++)
{
msg( ERROR, "Unhandled argument ignored: `%s'\n", *it);
@@ -1797,14 +1817,25 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
#endif
#if ENABLE_QPA
- if( m_LargeCTU && m_bUsePerceptQPA && !m_bUseAdaptiveQP && ( m_iSourceHeight <= 1280 ) && ( m_iSourceWidth <= 2048 ) )
+ if (m_bUsePerceptQPA && !m_bUseAdaptiveQP && m_dualTree && (m_cbQpOffsetDualTree != 0 || m_crQpOffsetDualTree != 0))
+ {
+ msg( WARNING, "*************************************************************************\n" );
+ msg( WARNING, "* WARNING: chroma QPA on, ignoring nonzero dual-tree chroma QP offsets! *\n" );
+ msg( WARNING, "*************************************************************************\n" );
+ }
+
+ #if QP_SWITCHING_FOR_PARALLEL
+ if( m_LargeCTU && ( m_iQP < 38 ) && ( m_iGOPSize > 4 ) && m_bUsePerceptQPA && !m_bUseAdaptiveQP && ( m_iSourceHeight <= 1280 ) && ( m_iSourceWidth <= 2048 ) )
+ #else
+ if( m_LargeCTU && ( ( int ) m_fQP < 38 ) && ( m_iGOPSize > 4 ) && m_bUsePerceptQPA && !m_bUseAdaptiveQP && ( m_iSourceHeight <= 1280 ) && ( m_iSourceWidth <= 2048 ) )
+ #endif
#else
if( false )
#endif
{
- msg( WARNING, "***************************************************************************\n" );
- msg( WARNING, "* WARNING: QPA on with LargeCTU for incompatible size, limiting CTU size! *\n" );
- msg( WARNING, "***************************************************************************\n" );
+ msg( WARNING, "*************************************************************************\n" );
+ msg( WARNING, "* WARNING: QPA on with large CTU for <=HD sequences, limiting CTU size! *\n" );
+ msg( WARNING, "*************************************************************************\n" );
m_uiCTUSize = m_uiMaxCUWidth;
if( ( 1u << m_quadtreeTULog2MaxSize ) > m_uiCTUSize ) m_quadtreeTULog2MaxSize--;
@@ -1924,6 +1955,9 @@ bool EncAppCfg::xCheckParameter()
#if JVET_K1000_SIMPLIFIED_EMT
xConfirmPara( m_EMT, "EMT only allowed with NEXT profile" );
xConfirmPara( m_FastEMT, "EMT only allowed with NEXT profile" );
+#endif
+#if JVET_K0157
+ xConfirmPara(m_compositeRefEnabled, "Composite Reference Frame is only allowed with NEXT profile");
#endif
// ADD_NEW_TOOL : (parameter check) add a check for next tools here
}
@@ -2217,6 +2251,9 @@ bool EncAppCfg::xCheckParameter()
xConfirmPara( m_bipredSearchRange < 0 , "Bi-prediction refinement search range must be more than 0" );
xConfirmPara( m_minSearchWindow < 0, "Minimum motion search window size for the adaptive window ME must be greater than or equal to 0" );
xConfirmPara( m_iMaxDeltaQP > MAX_DELTA_QP, "Absolute Delta QP exceeds supported range (0 to 7)" );
+#if ENABLE_QPA
+ xConfirmPara( m_bUsePerceptQPA && m_uiDeltaQpRD > 0, "Perceptual QPA cannot be used together with slice-level multiple-QP optimization" );
+#endif
#if SHARP_LUMA_DELTA_QP
xConfirmPara( m_lumaLevelToDeltaQPMapping.mode && m_uiDeltaQpRD > 0, "Luma-level-based Delta QP cannot be used together with slice level multiple-QP optimization\n" );
#endif
@@ -2391,6 +2428,9 @@ bool EncAppCfg::xCheckParameter()
xConfirmPara( m_intraConstraintFlag, "IntraConstraintFlag cannot be 1 for inter sequences");
}
+#if JVET_K0157
+ int multipleFactor = m_compositeRefEnabled ? 2 : 1;
+#endif
bool verifiedGOP=false;
bool errorGOP=false;
int checkGOP=1;
@@ -2411,7 +2451,11 @@ bool EncAppCfg::xCheckParameter()
for(int i=0; i<m_iGOPSize; i++)
{
+#if JVET_K0157
+ if (m_GOPList[i].m_POC == m_iGOPSize * multipleFactor)
+#else
if(m_GOPList[i].m_POC==m_iGOPSize)
+#endif
{
xConfirmPara( m_GOPList[i].m_temporalId!=0 , "The last frame in each GOP must have temporal ID = 0 " );
}
@@ -2445,7 +2489,11 @@ bool EncAppCfg::xCheckParameter()
while(!verifiedGOP&&!errorGOP)
{
int curGOP = (checkGOP-1)%m_iGOPSize;
+#if JVET_K0157
+ int curPOC = ((checkGOP - 1) / m_iGOPSize)*m_iGOPSize * multipleFactor + m_GOPList[curGOP].m_POC;
+#else
int curPOC = ((checkGOP-1)/m_iGOPSize)*m_iGOPSize + m_GOPList[curGOP].m_POC;
+#endif
if(m_GOPList[curGOP].m_POC<0)
{
msg( WARNING, "\nError: found fewer Reference Picture Sets than GOPSize\n");
@@ -2472,7 +2520,11 @@ bool EncAppCfg::xCheckParameter()
found=true;
for(int k=0; k<m_iGOPSize; k++)
{
+#if JVET_K0157
+ if (absPOC % (m_iGOPSize * multipleFactor) == m_GOPList[k].m_POC % (m_iGOPSize * multipleFactor))
+#else
if(absPOC%m_iGOPSize == m_GOPList[k].m_POC%m_iGOPSize)
+#endif
{
if(m_GOPList[k].m_temporalId==m_GOPList[curGOP].m_temporalId)
{
@@ -2524,7 +2576,11 @@ bool EncAppCfg::xCheckParameter()
{
//step backwards in coding order and include any extra available pictures we might find useful to replace the ones with POC < 0.
int offGOP = (checkGOP-1+offset)%m_iGOPSize;
+#if JVET_K0157
+ int offPOC = ((checkGOP - 1 + offset) / m_iGOPSize)*(m_iGOPSize * multipleFactor) + m_GOPList[offGOP].m_POC;
+#else
int offPOC = ((checkGOP-1+offset)/m_iGOPSize)*m_iGOPSize + m_GOPList[offGOP].m_POC;
+#endif
if(offPOC>=0&&m_GOPList[offGOP].m_temporalId<=m_GOPList[curGOP].m_temporalId)
{
bool newRef=false;
@@ -2926,7 +2982,7 @@ void EncAppCfg::xPrintParameter()
msg( DETAILS, "Real Format : %dx%d %gHz\n", m_iSourceWidth - m_confWinLeft - m_confWinRight, m_iSourceHeight - m_confWinTop - m_confWinBottom, (double)m_iFrameRate / m_temporalSubsampleRatio );
msg( DETAILS, "Internal Format : %dx%d %gHz\n", m_iSourceWidth, m_iSourceHeight, (double)m_iFrameRate / m_temporalSubsampleRatio );
msg( DETAILS, "Sequence PSNR output : %s\n", ( m_printMSEBasedSequencePSNR ? "Linear average, MSE-based" : "Linear average only" ) );
- msg(DETAILS, "Hexadecimal PSNR output : %s\n", ( m_printHexPsnr ? "Enabled" : "Disabled" ) );
+ msg( DETAILS, "Hexadecimal PSNR output : %s\n", ( m_printHexPsnr ? "Enabled" : "Disabled" ) );
msg( DETAILS, "Sequence MSE output : %s\n", ( m_printSequenceMSE ? "Enabled" : "Disabled" ) );
msg( DETAILS, "Frame MSE output : %s\n", ( m_printFrameMSE ? "Enabled" : "Disabled" ) );
msg( DETAILS, "Cabac-zero-word-padding : %s\n", ( m_cabacZeroWordPaddingEnabled ? "Enabled" : "Disabled" ) );
@@ -3164,6 +3220,9 @@ void EncAppCfg::xPrintParameter()
#endif
#if JVET_K1000_SIMPLIFIED_EMT
msg( VERBOSE, "EMT: %1d(intra) %1d(inter) ", m_EMT & 1, ( m_EMT >> 1 ) & 1 );
+#endif
+#if JVET_K0157
+ msg(VERBOSE, "CompositeLTReference:%d ", m_compositeRefEnabled);
#endif
}
// ADD_NEW_TOOL (add some output indicating the usage of tools)
diff --git a/source/App/EncoderApp/EncAppCfg.h b/source/App/EncoderApp/EncAppCfg.h
index 10a0554a5566dda4e2157f8e84a4f11b136eb0a1..2ec9c86e06aa867ee60c1bc34e0936f64a488404 100644
--- a/source/App/EncoderApp/EncAppCfg.h
+++ b/source/App/EncoderApp/EncAppCfg.h
@@ -115,6 +115,7 @@ protected:
bool m_cabacZeroWordPaddingEnabled;
bool m_bClipInputVideoToRec709Range;
bool m_bClipOutputVideoToRec709Range;
+ bool m_packedYUVMode; ///< If true, output 10-bit and 12-bit YUV data as 5-byte and 3-byte (respectively) packed YUV data
// profile/level
Profile::Name m_profile;
@@ -229,6 +230,9 @@ protected:
int m_FastEMT; ///< XZ: Fast Methods of Enhanced Multiple Transform
#endif
+#if JVET_K0157
+ bool m_compositeRefEnabled;
+#endif
// ADD_NEW_TOOL : (encoder app) add tool enabling flags and associated parameters here
unsigned m_uiMaxCUWidth; ///< max. CU width in pixel
@@ -285,7 +289,7 @@ protected:
int m_maxNumOffsetsPerPic; ///< SAO maximun number of offset per picture
bool m_saoCtuBoundary; ///< SAO parameter estimation using non-deblocked pixels for CTU bottom and right boundary areas
#if K0238_SAO_GREEDY_MERGE_ENCODING
- bool m_saoGreedyMergeEnc; ///< SAO greedy merge encoding algorithm
+ bool m_saoGreedyMergeEnc; ///< SAO greedy merge encoding algorithm
#endif
// coding tools (loop filter)
bool m_bLoopFilterDisable; ///< flag for using deblocking filter
diff --git a/source/App/EncoderApp/encmain.cpp b/source/App/EncoderApp/encmain.cpp
index 7c34b5d14e793dab2ce557f07cfc6dbbb090e9f4..6d5bdf109d5ebcfee1e85c19cb14b0b4601d0597 100644
--- a/source/App/EncoderApp/encmain.cpp
+++ b/source/App/EncoderApp/encmain.cpp
@@ -87,9 +87,9 @@ int main(int argc, char* argv[])
// print information
fprintf( stdout, "\n" );
#ifdef SVNREVISION
- fprintf( stdout, "VVCSoftware: VTM Encoder Version %s (%s@r%s) ", NEXT_SOFTWARE_VERSION, SVNRELATIVEURL, SVNREVISION /*NV_VERSION*/ );
+ fprintf( stdout, "VVCSoftware: BMS Encoder Version %s (%s@r%s) ", NEXT_SOFTWARE_VERSION, SVNRELATIVEURL, SVNREVISION /*NV_VERSION*/ );
#else
- fprintf( stdout, "VVCSoftware: VTM Encoder Version %s ", NEXT_SOFTWARE_VERSION /*NV_VERSION*/ );
+ fprintf( stdout, "VVCSoftware: BMS Encoder Version %s ", NEXT_SOFTWARE_VERSION /*NV_VERSION*/ );
#endif
fprintf( stdout, NVM_ONOS );
fprintf( stdout, NVM_COMPILEDBY );
diff --git a/source/App/SEIRemovalApp/seiremovalmain.cpp b/source/App/SEIRemovalApp/seiremovalmain.cpp
index 4a4d79fa62554f374ca7b3359896630432756987..96eb667d378cd6bca665dd9f589b8d96e2b4f37b 100644
--- a/source/App/SEIRemovalApp/seiremovalmain.cpp
+++ b/source/App/SEIRemovalApp/seiremovalmain.cpp
@@ -57,9 +57,9 @@ int main(int argc, char* argv[])
// print information
fprintf( stdout, "\n" );
#ifdef SVNREVISION
- fprintf( stdout, "VVCSoftware: VTM Decoder Version %s (%s@r%s) ", NEXT_SOFTWARE_VERSION, SVNRELATIVEURL, SVNREVISION /*NV_VERSION*/ );
+ fprintf( stdout, "VVCSoftware: BMS Decoder Version %s (%s@r%s) ", NEXT_SOFTWARE_VERSION, SVNRELATIVEURL, SVNREVISION /*NV_VERSION*/ );
#else
- fprintf( stdout, "VVCSoftware: VTM Decoder Version %s ", NEXT_SOFTWARE_VERSION /*NV_VERSION*/ );
+ fprintf( stdout, "VVCSoftware: BMS Decoder Version %s ", NEXT_SOFTWARE_VERSION /*NV_VERSION*/ );
#endif
fprintf( stdout, NVM_ONOS );
fprintf( stdout, NVM_COMPILEDBY );
diff --git a/source/Lib/CommonAnalyserLib/CMakeLists.txt b/source/Lib/CommonAnalyserLib/CMakeLists.txt
index ff996a3391d016844131439ed8d495671ec2982c..e0cdf5fa1e65f7bf849ba5296d9a0cd897260f68 100644
--- a/source/Lib/CommonAnalyserLib/CMakeLists.txt
+++ b/source/Lib/CommonAnalyserLib/CMakeLists.txt
@@ -27,6 +27,9 @@ file( GLOB AVX2_SRC_FILES "../CommonLib/x86/avx2/*.cpp" )
# get sse4.1 source files
file( GLOB SSE41_SRC_FILES "../CommonLib/x86/sse41/*.cpp" )
+# get sse4.2 source files
+file( GLOB SSE42_SRC_FILES "../CommonLib/x86/sse42/*.cpp" )
+
# get libmd5 source files
file( GLOB MD5_SRC_FILES "../libmd5/*.cpp" )
@@ -35,7 +38,7 @@ file( GLOB MD5_INC_FILES "../libmd5/*.h" )
# get all source files
-set( SRC_FILES ${BASE_SRC_FILES} ${X86_SRC_FILES} ${SSE41_SRC_FILES} ${AVX_SRC_FILES} ${AVX2_SRC_FILES} ${MD5_SRC_FILES} )
+set( SRC_FILES ${BASE_SRC_FILES} ${X86_SRC_FILES} ${SSE41_SRC_FILES} ${SSE42_SRC_FILES} ${AVX_SRC_FILES} ${AVX2_SRC_FILES} ${MD5_SRC_FILES} )
# get all include files
set( INC_FILES ${BASE_INC_FILES} ${X86_INC_FILES} ${MD5_INC_FILES} )
@@ -46,7 +49,7 @@ add_library( ${LIB_NAME} STATIC ${SRC_FILES} ${INC_FILES} ${NATVIS_FILES} )
target_compile_definitions( ${LIB_NAME} PUBLIC RExt__DECODER_DEBUG_TOOL_STATISTICS=1 )
if( ENABLE_VTM )
- target_compile_definitions( ${LIB_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${LIB_NAME} PUBLIC BMS_TOOLS=0 )
endif()
if( EXTENSION_360_VIDEO )
@@ -86,6 +89,7 @@ target_link_libraries( ${LIB_NAME} Threads::Threads )
# set needed compile definitions
set_property( SOURCE ${SSE41_SRC_FILES} APPEND PROPERTY COMPILE_DEFINITIONS USE_SSE41 )
+set_property( SOURCE ${SSE42_SRC_FILES} APPEND PROPERTY COMPILE_DEFINITIONS USE_SSE42 )
set_property( SOURCE ${AVX_SRC_FILES} APPEND PROPERTY COMPILE_DEFINITIONS USE_AVX )
set_property( SOURCE ${AVX2_SRC_FILES} APPEND PROPERTY COMPILE_DEFINITIONS USE_AVX2 )
# set needed compile flags
@@ -94,6 +98,7 @@ if( MSVC )
set_property( SOURCE ${AVX2_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "/arch:AVX2" )
elseif( UNIX )
set_property( SOURCE ${SSE41_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "-msse4.1" )
+ set_property( SOURCE ${SSE42_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "-msse4.2" )
set_property( SOURCE ${AVX_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "-mavx" )
set_property( SOURCE ${AVX2_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "-mavx2" )
endif()
diff --git a/source/Lib/CommonLib/AdaptiveLoopFilter.cpp b/source/Lib/CommonLib/AdaptiveLoopFilter.cpp
index 8d5420fefb0c314ef8ead60648be042b9f1a4cad..d00fd667beaf955fb4fcfc7cb3f3a7986528419f 100644
--- a/source/Lib/CommonLib/AdaptiveLoopFilter.cpp
+++ b/source/Lib/CommonLib/AdaptiveLoopFilter.cpp
@@ -217,19 +217,25 @@ void AdaptiveLoopFilter::create( const int picWidth, const int picHeight, const
// Laplacian based activity
for( int i = 0; i < NUM_DIRECTIONS; i++ )
{
- m_laplacian[i] = new int*[m_CLASSIFICATION_BLK_SIZE + 5];
-
- for( int y = 0; y < m_CLASSIFICATION_BLK_SIZE + 5; y++ )
+ if ( m_laplacian[i] == nullptr )
{
- m_laplacian[i][y] = new int[m_CLASSIFICATION_BLK_SIZE + 5];
+ m_laplacian[i] = new int*[m_CLASSIFICATION_BLK_SIZE + 5];
+
+ for( int y = 0; y < m_CLASSIFICATION_BLK_SIZE + 5; y++ )
+ {
+ m_laplacian[i][y] = new int[m_CLASSIFICATION_BLK_SIZE + 5];
+ }
}
}
// Classification
- m_classifier = new AlfClassifier*[picHeight];
- for( int i = 0; i < picHeight; i++ )
+ if ( m_classifier == nullptr )
{
- m_classifier[i] = new AlfClassifier[picWidth];
+ m_classifier = new AlfClassifier*[picHeight];
+ for( int i = 0; i < picHeight; i++ )
+ {
+ m_classifier[i] = new AlfClassifier[picWidth];
+ }
}
}
diff --git a/source/Lib/CommonLib/Buffer.h b/source/Lib/CommonLib/Buffer.h
index 565a121fc4bf2392057cc8b2295cd2c7e90886cc..1763242e3dc0af2fd30165b09045a5028c460863 100644
--- a/source/Lib/CommonLib/Buffer.h
+++ b/source/Lib/CommonLib/Buffer.h
@@ -366,7 +366,6 @@ void AreaBuf<T>::addAvg( const AreaBuf<const T> &other1, const AreaBuf<const T>
template<>
void AreaBuf<Pel>::addAvg( const AreaBuf<const Pel> &other1, const AreaBuf<const Pel> &other2, const ClpRng& clpRng );
-
template<typename T>
void AreaBuf<T>::linearTransform( const int scale, const int shift, const int offset, bool bClip, const ClpRng& clpRng )
{
@@ -385,6 +384,7 @@ void AreaBuf<T>::toLast( const ClpRng& clpRng )
template<>
void AreaBuf<Pel>::toLast( const ClpRng& clpRng );
+
template<typename T>
void AreaBuf<T>::removeHighFreq( const AreaBuf<T>& other, const bool bClip, const ClpRng& clpRng )
{
@@ -579,7 +579,8 @@ struct UnitBuf
void addAvg ( const UnitBuf<const T> &other1, const UnitBuf<const T> &other2, const ClpRngs& clpRngs, const bool chromaOnly = false, const bool lumaOnly = false);
void extendSingleBorderPel();
void extendBorderPel ( unsigned margin );
- void removeHighFreq ( const UnitBuf<T>& other, const bool bClip, const ClpRngs& clpRngs);
+ void removeHighFreq ( const UnitBuf<T>& other, const bool bClip, const ClpRngs& clpRngs
+ );
UnitBuf< T> subBuf (const UnitArea& subArea);
const UnitBuf<const T> subBuf (const UnitArea& subArea) const;
@@ -648,6 +649,7 @@ void UnitBuf<T>::reconstruct(const UnitBuf<const T> &pred, const UnitBuf<const T
}
}
+
template<typename T>
void UnitBuf<T>::addAvg(const UnitBuf<const T> &other1, const UnitBuf<const T> &other2, const ClpRngs& clpRngs, const bool chromaOnly /* = false */, const bool lumaOnly /* = false */)
{
@@ -681,7 +683,8 @@ void UnitBuf<T>::extendBorderPel( unsigned margin )
}
template<typename T>
-void UnitBuf<T>::removeHighFreq( const UnitBuf<T>& other, const bool bClip, const ClpRngs& clpRngs)
+void UnitBuf<T>::removeHighFreq( const UnitBuf<T>& other, const bool bClip, const ClpRngs& clpRngs
+ )
{
for( unsigned i = 0; i < bufs.size(); i++ )
{
diff --git a/source/Lib/CommonLib/CMakeLists.txt b/source/Lib/CommonLib/CMakeLists.txt
index c633fec92feffc7661ac904de1c2632a7b4e3224..7a91672c034457ed88f3bf059fed4a539d1e7a7e 100644
--- a/source/Lib/CommonLib/CMakeLists.txt
+++ b/source/Lib/CommonLib/CMakeLists.txt
@@ -24,6 +24,9 @@ file( GLOB AVX_SRC_FILES "x86/avx/*.cpp" )
# get avx2 source files
file( GLOB AVX2_SRC_FILES "x86/avx2/*.cpp" )
+# get sse4.2 source files
+file( GLOB SSE42_SRC_FILES "x86/sse42/*.cpp" )
+
# get sse4.1 source files
file( GLOB SSE41_SRC_FILES "x86/sse41/*.cpp" )
@@ -35,7 +38,7 @@ file( GLOB MD5_INC_FILES "../libmd5/*.h" )
# get all source files
-set( SRC_FILES ${BASE_SRC_FILES} ${X86_SRC_FILES} ${SSE41_SRC_FILES} ${AVX_SRC_FILES} ${AVX2_SRC_FILES} ${MD5_SRC_FILES} )
+set( SRC_FILES ${BASE_SRC_FILES} ${X86_SRC_FILES} ${SSE41_SRC_FILES} ${SSE42_SRC_FILES} ${AVX_SRC_FILES} ${AVX2_SRC_FILES} ${MD5_SRC_FILES} )
# get all include files
set( INC_FILES ${BASE_INC_FILES} ${X86_INC_FILES} ${MD5_INC_FILES} )
@@ -45,7 +48,7 @@ set( INC_FILES ${BASE_INC_FILES} ${X86_INC_FILES} ${MD5_INC_FILES} )
add_library( ${LIB_NAME} STATIC ${SRC_FILES} ${INC_FILES} ${NATVIS_FILES} )
if( ENABLE_VTM )
- target_compile_definitions( ${LIB_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${LIB_NAME} PUBLIC BMS_TOOLS=0 )
endif()
if( EXTENSION_360_VIDEO )
@@ -85,6 +88,7 @@ target_link_libraries( ${LIB_NAME} Threads::Threads )
# set needed compile definitions
set_property( SOURCE ${SSE41_SRC_FILES} APPEND PROPERTY COMPILE_DEFINITIONS USE_SSE41 )
+set_property( SOURCE ${SSE42_SRC_FILES} APPEND PROPERTY COMPILE_DEFINITIONS USE_SSE42 )
set_property( SOURCE ${AVX_SRC_FILES} APPEND PROPERTY COMPILE_DEFINITIONS USE_AVX )
set_property( SOURCE ${AVX2_SRC_FILES} APPEND PROPERTY COMPILE_DEFINITIONS USE_AVX2 )
# set needed compile flags
@@ -93,6 +97,7 @@ if( MSVC )
set_property( SOURCE ${AVX2_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "/arch:AVX2" )
elseif( UNIX )
set_property( SOURCE ${SSE41_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "-msse4.1" )
+ set_property( SOURCE ${SSE42_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "-msse4.2" )
set_property( SOURCE ${AVX_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "-mavx" )
set_property( SOURCE ${AVX2_SRC_FILES} APPEND PROPERTY COMPILE_FLAGS "-mavx2" )
endif()
diff --git a/source/Lib/CommonLib/CodingStatistics.h b/source/Lib/CommonLib/CodingStatistics.h
index 619a896553eacc8df47fd3308ea91e1dba6a5e87..43496c0356fad85d3ef477cd0dcc62cd4dd5b41b 100644
--- a/source/Lib/CommonLib/CodingStatistics.h
+++ b/source/Lib/CommonLib/CodingStatistics.h
@@ -66,12 +66,6 @@ enum CodingStatisticsType
STATS__CABAC_BITS__REF_FRM_IDX,
STATS__CABAC_BITS__MVD,
STATS__CABAC_BITS__MVD_EP,
-#if JVET_K_AFFINE
- STATS__CABAC_BITS__AFFINE_FLAG,
-#if JVET_K0337_AFFINE_6PARA
- STATS__CABAC_BITS__AFFINE_TYPE,
-#endif
-#endif
STATS__CABAC_BITS__TRANSFORM_SUBDIV_FLAG,
STATS__CABAC_BITS__QT_ROOT_CBF,
STATS__CABAC_BITS__DELTA_QP_EP,
@@ -79,21 +73,14 @@ enum CodingStatisticsType
STATS__CABAC_BITS__QT_CBF,
STATS__CABAC_BITS__CROSS_COMPONENT_PREDICTION,
STATS__CABAC_BITS__TRANSFORM_SKIP_FLAGS,
-
STATS__CABAC_BITS__LAST_SIG_X_Y,
STATS__CABAC_BITS__SIG_COEFF_GROUP_FLAG,
STATS__CABAC_BITS__SIG_COEFF_MAP_FLAG,
-#if JVET_K0072
- STATS__CABAC_BITS__PAR_FLAG,
-#endif
STATS__CABAC_BITS__GT1_FLAG,
STATS__CABAC_BITS__GT2_FLAG,
STATS__CABAC_BITS__SIGN_BIT,
STATS__CABAC_BITS__ESCAPE_BITS,
STATS__CABAC_BITS__SAO,
-#if JVET_K0371_ALF
- STATS__CABAC_BITS__ALF,
-#endif
STATS__CABAC_TRM_BITS,
STATS__CABAC_FIXED_BITS,
STATS__CABAC_PCM_ALIGN_BITS,
@@ -104,6 +91,22 @@ enum CodingStatisticsType
STATS__CABAC_EP_BIT_ALIGNMENT,
STATS__CABAC_BITS__ALIGNED_SIGN_BIT,
STATS__CABAC_BITS__ALIGNED_ESCAPE_BITS,
+ STATS__CABAC_BITS__OTHER,
+ STATS__CABAC_BITS__INVALID,
+ STATS__TOOL_TOTAL_FRAME,// This is a special case and is not included in the report.
+#if JVET_K_AFFINE
+ STATS__CABAC_BITS__AFFINE_FLAG,
+ STATS__TOOL_AFF,
+#if JVET_K0337_AFFINE_6PARA
+ STATS__CABAC_BITS__AFFINE_TYPE,
+#endif
+#endif
+#if JVET_K0072
+ STATS__CABAC_BITS__PAR_FLAG,
+#endif
+#if JVET_K0371_ALF
+ STATS__CABAC_BITS__ALF,
+#endif
#if JVET_K0357_AMVR
STATS__CABAC_BITS__IMV_FLAG,
#endif
@@ -111,13 +114,23 @@ enum CodingStatisticsType
STATS__CABAC_BITS__EMT_CU_FLAG,
STATS__CABAC_BITS__EMT_TU_INDEX,
#endif
- STATS__CABAC_BITS__OTHER,
- STATS__CABAC_BITS__INVALID,
- STATS__TOOL_TOTAL_FRAME,// This is a special case and is not included in the report.
+#if JVET_K1000_SIMPLIFIED_EMT
+ STATS__TOOL_EMT,
+#endif
STATS__TOOL_TOTAL,
STATS__NUM_STATS
};
+enum CodingStatisticsMode
+{
+ STATS__MODE_NONE = 0,
+ STATS__MODE_BITS = 1,
+ STATS__MODE_TOOLS = 2,
+
+ STATS__MODE_ALL =
+ STATS__MODE_BITS | STATS__MODE_TOOLS
+};
+
static inline const char* getName(CodingStatisticsType name)
{
static const char *statNames[]=
@@ -326,11 +339,13 @@ public:
friend class CodingStatistics;
};
+ int m_mode;
+
private:
CodingStatisticsData data;
- CodingStatistics() : data()
+ CodingStatistics() : m_mode(STATS__MODE_ALL), data()
{
}
@@ -408,11 +423,8 @@ private:
printf( "\n" );
}
-public:
-
- ~CodingStatistics()
+ void OutputBitStats()
{
-#if RExt__DECODER_DEBUG_BIT_STATISTICS
const int64_t es = CODINGSTATISTICS_ENTROPYSCALE;
int64_t countTotal = 0;
@@ -594,9 +606,82 @@ public:
OutputDashedLine( "GRAND TOTAL" );
epTotalBits += cavlcTotalBits;
OutputLine ( "TOTAL", '~', "~~GT~~", "~~GT~~", "~~GT~~", cabacTotalBits, epTotalBits );
+ }
+
+ void OutputToolStats()
+ {
+ printf("\n");
+ printf( " %-45s- Width Height Type Count Impacted pixels %% Impacted pixels\n", "Tools statistics" );
+ OutputDashedLine( "" );
+
+ const uint64_t toolCount = STATS__TOOL_TOTAL - (STATS__TOOL_TOTAL_FRAME + 1);
+ StatTool subTotalTool[toolCount];
+ StatTool statTotalTool[toolCount][CODING_STATS_NUM_SUBCLASSES];
+ uint64_t totalPixels = GetStatisticTool( STATS__TOOL_TOTAL_FRAME ).pixels;
+ for( int i = 0; i < toolCount; i++ )
+ {
+ const int type = i + (STATS__TOOL_TOTAL_FRAME + 1);
+ const char *pName = getName( CodingStatisticsType( type ) );
+
+ for( uint32_t c = 0; c < CODING_STATS_NUM_SUBCLASSES; c++ )
+ {
+ StatTool &sTool = data.statistics_tool[type][c];
+ if( sTool.count == 0 )
+ {
+ continue;
+ }
+
+ uint32_t wIdx = CodingStatisticsClassType::GetSubClassWidth( c );
+ uint32_t hIdx = CodingStatisticsClassType::GetSubClassHeight( c );
+ OutputLine( pName, ':', wIdx, hIdx, CodingStatisticsClassType::GetSubClassString( c ), sTool, totalPixels );
+
+ statTotalTool[i][c] += sTool;
+ subTotalTool[i] += sTool;
+ }
+
+ if (subTotalTool[i].count != 0)
+ {
+ OutputLine( pName, '~', "~~ST~~", "~~ST~~", "~~ST~~", subTotalTool[i], totalPixels );
+ }
+ }
+
+ for( int i = 0; i < toolCount; i++ )
+ {
+ const int type = i + (STATS__TOOL_TOTAL_FRAME + 1);
+ const char *pName = getName( CodingStatisticsType( type ) );
+
+ if (subTotalTool[i].count != 0)
+ OutputDashedLine( "Break down by tool/Channel type" );
+
+ for( uint32_t c = 0; c < CODING_STATS_NUM_SUBCLASSES; c += CODING_STATS_NUM_SIZES )
+ {
+ StatTool typeTotalTool;
+ for( uint32_t w = 0; w < CODING_STATS_NUM_WIDTHS; w++ )
+ {
+ for( uint32_t h = 0; h < CODING_STATS_NUM_HEIGHTS; h++ )
+ typeTotalTool += statTotalTool[i][c + h * CODING_STATS_NUM_WIDTHS + w];
+ }
+
+ if( typeTotalTool.count != 0 )
+ {
+ OutputLine( pName, '=', "-", "-", CodingStatisticsClassType::GetSubClassString( c ), typeTotalTool, totalPixels );
+ }
+ }
+ }
+ }
+
+public:
+
+ ~CodingStatistics()
+ {
+#if RExt__DECODER_DEBUG_BIT_STATISTICS
+ if (m_mode & STATS__MODE_BITS)
+ OutputBitStats();
#endif //RExt__DECODER_DEBUG_BIT_STATISTICS
#ifdef RExt__DECODER_DEBUG_TOOL_STATISTICS
+ if (m_mode & STATS__MODE_TOOLS)
+ OutputToolStats();
#endif //RExt__DECODER_DEBUG_TOOL_STATISTICS
}
diff --git a/source/Lib/CommonLib/CodingStructure.cpp b/source/Lib/CommonLib/CodingStructure.cpp
index b6475c12c99705eab6e18a262dbe20964db70fe9..55c7eebec481a2c82e238f989bd3af1d8886681c 100644
--- a/source/Lib/CommonLib/CodingStructure.cpp
+++ b/source/Lib/CommonLib/CodingStructure.cpp
@@ -557,7 +557,9 @@ cTUTraverser CodingStructure::traverseTUs( const UnitArea& unit, const ChannelTy
void CodingStructure::allocateVectorsAtPicLevel()
{
- const int twice = ( !pcv->ISingleTree && slice->isIntra() && pcv->chrFormat != CHROMA_400 ) ? 2 : 1;
+ const int twice = (
+ !pcv->ISingleTree && slice->isIntra()
+ && pcv->chrFormat != CHROMA_400 ) ? 2 : 1;
size_t allocSize = twice * unitScale[0].scale( area.blocks[0].size() ).area();
cus.reserve( allocSize );
diff --git a/source/Lib/CommonLib/CodingStructure.h b/source/Lib/CommonLib/CodingStructure.h
index 15e5ea4dfd717f85164fc88ded69d70ce83ff0cf..df7337a1cad487139cf6ab0b2dfeef57cb619028 100644
--- a/source/Lib/CommonLib/CodingStructure.h
+++ b/source/Lib/CommonLib/CodingStructure.h
@@ -58,7 +58,6 @@ enum PictureType
PIC_ORG_RESI,
NUM_PIC_TYPES
};
-
extern XUCache g_globalUnitCache;
// ---------------------------------------------------------------------------
@@ -147,7 +146,6 @@ public:
cCUTraverser traverseCUs(const UnitArea& _unit, const ChannelType _chType) const;
cPUTraverser traversePUs(const UnitArea& _unit, const ChannelType _chType) const;
cTUTraverser traverseTUs(const UnitArea& _unit, const ChannelType _chType) const;
-
// ---------------------------------------------------------------------------
// encoding search utilities
// ---------------------------------------------------------------------------
diff --git a/source/Lib/CommonLib/Common.h b/source/Lib/CommonLib/Common.h
index a10194a5dffab672a5a45d59c80531d0d813737d..ad0e5ec79adb4872eef350b34eac84e9ce427bbc 100644
--- a/source/Lib/CommonLib/Common.h
+++ b/source/Lib/CommonLib/Common.h
@@ -114,7 +114,6 @@ struct UnitScale
Size scale( const Size &size ) const { return { size.width >> posx, size.height >> posy }; }
Area scale( const Area &_area ) const { return Area( scale( _area.pos() ), scale( _area.size() ) ); }
};
-
inline size_t rsAddr(const Position &pos, const uint32_t stride, const UnitScale &unitScale )
{
return (size_t)(stride >> unitScale.posx) * (size_t)(pos.y >> unitScale.posy) + (size_t)(pos.x >> unitScale.posx);
diff --git a/source/Lib/CommonLib/CommonDef.h b/source/Lib/CommonLib/CommonDef.h
index 7c8e0c1d131dfdf15e61305bf6aa46907fcd0417..d878f43d5bd1a0489a65f8b5aafc010739d7bd89 100644
--- a/source/Lib/CommonLib/CommonDef.h
+++ b/source/Lib/CommonLib/CommonDef.h
@@ -331,6 +331,7 @@ static const int AFFINE_MAX_NUM_COMB = 12; ///< max
static const int AFFINE_MIN_BLOCK_SIZE = 4; ///< Minimum affine MC block size
#endif
+
#if W0038_DB_OPT
static const int MAX_ENCODER_DEBLOCKING_QUALITY_LAYERS = 8 ;
#endif
@@ -401,7 +402,6 @@ static const unsigned C806_ALF_TEMPPRED_NUM = 6;
static const int NTAPS_LUMA = 8; ///< Number of taps for luma
static const int NTAPS_CHROMA = 4; ///< Number of taps for chroma
-
// ====================================================================================================================
// Macro functions
// ====================================================================================================================
diff --git a/source/Lib/CommonLib/Contexts.cpp b/source/Lib/CommonLib/Contexts.cpp
index d7d21d2fce751fc8d80eb670e6cec8687021e779..b387c12caabc004df24bc3a9bb36e0313f6cf210 100644
--- a/source/Lib/CommonLib/Contexts.cpp
+++ b/source/Lib/CommonLib/Contexts.cpp
@@ -383,6 +383,7 @@ const CtxSet ContextSetCfg::AffineType = ContextSetCfg::addCtxSet
#endif
#endif
+
const CtxSet ContextSetCfg::Mvd = ContextSetCfg::addCtxSet
({
{ 169, 198,},
@@ -804,7 +805,7 @@ void CtxStore<BinProbModel>::init( int qp, int initId )
const std::vector<uint8_t>& initTable = ContextSetCfg::getInitTable( initId );
CHECK( m_CtxBuffer.size() != initTable.size(),
"Size of init table (" << initTable.size() << ") does not match size of context buffer (" << m_CtxBuffer.size() << ")." );
- int clippedQP = std::min( std::max( 0, qp ), MAX_QP );
+ int clippedQP = Clip3( 0, MAX_QP, qp );
for( std::size_t k = 0; k < m_CtxBuffer.size(); k++ )
{
m_CtxBuffer[k].init( clippedQP, initTable[k] );
diff --git a/source/Lib/CommonLib/InterPrediction.cpp b/source/Lib/CommonLib/InterPrediction.cpp
index b5e4bc9dd77d8e79b3e7d0776214029da55155a3..89899afb677fe2f02a92f5037ae8298172c68421 100644
--- a/source/Lib/CommonLib/InterPrediction.cpp
+++ b/source/Lib/CommonLib/InterPrediction.cpp
@@ -109,7 +109,6 @@ void InterPrediction::destroy()
m_filteredBlockTmp[i][c] = nullptr;
}
}
-
}
void InterPrediction::init( RdCost* pcRdCost, ChromaFormat chromaFormatIDC )
@@ -124,7 +123,6 @@ void InterPrediction::init( RdCost* pcRdCost, ChromaFormat chromaFormatIDC )
}
m_currChromaFormat = chromaFormatIDC;
-
if( m_acYuvPred[REF_PIC_LIST_0][COMPONENT_Y] == nullptr ) // check if first is null (in which case, nothing initialised yet)
{
for( uint32_t c = 0; c < MAX_NUM_COMPONENT; c++ )
@@ -148,6 +146,7 @@ void InterPrediction::init( RdCost* pcRdCost, ChromaFormat chromaFormatIDC )
}
}
+
m_iRefListIdx = -1;
}
@@ -365,7 +364,9 @@ void InterPrediction::xSubPuMC( PredictionUnit& pu, PelUnitBuf& predBuf, const R
}
#endif
-void InterPrediction::xPredInterUni(const PredictionUnit& pu, const RefPicList& eRefPicList, PelUnitBuf& pcYuvPred, const bool& bi )
+
+void InterPrediction::xPredInterUni(const PredictionUnit& pu, const RefPicList& eRefPicList, PelUnitBuf& pcYuvPred, const bool& bi
+)
{
const SPS &sps = *pu.cs->sps;
@@ -400,7 +401,6 @@ void InterPrediction::xPredInterUni(const PredictionUnit& pu, const RefPicList&
for( uint32_t comp = COMPONENT_Y; comp < pcYuvPred.bufs.size() && comp <= m_maxCompIDToPred; comp++ )
{
const ComponentID compID = ComponentID( comp );
-
#if JVET_K_AFFINE
if ( pu.cu->affine )
{
@@ -526,7 +526,6 @@ void InterPrediction::xPredInterBlk ( const ComponentID& compID, const Predictio
refBuf = refPic->getRecoBuf( CompArea( compID, chFmt, offset, pu.blocks[compID].size() ) );
}
-
if( yFrac == 0 )
{
m_if.filterHor(compID, (Pel*) refBuf.buf, refBuf.stride, dstBuf.buf, dstBuf.stride, width, height, xFrac, rndRes, chFmt, clpRng);
@@ -537,7 +536,7 @@ void InterPrediction::xPredInterBlk ( const ComponentID& compID, const Predictio
}
else
{
- PelBuf tmpBuf = PelBuf(m_filteredBlockTmp[0][compID], pu.blocks[compID]);
+ PelBuf tmpBuf = PelBuf(m_filteredBlockTmp[0][compID], pu.blocks[compID]);
int vFilterSize = isLuma(compID) ? NTAPS_LUMA : NTAPS_CHROMA;
m_if.filterHor(compID, (Pel*) refBuf.buf - ((vFilterSize >> 1) - 1) * refBuf.stride, refBuf.stride, tmpBuf.buf, tmpBuf.stride, width, height + vFilterSize - 1, xFrac, false, chFmt, clpRng);
@@ -545,7 +544,6 @@ void InterPrediction::xPredInterBlk ( const ComponentID& compID, const Predictio
m_if.filterVer(compID, (Pel*) tmpBuf.buf + ((vFilterSize >> 1) - 1) * tmpBuf.stride, tmpBuf.stride, dstBuf.buf, dstBuf.stride, width, height, yFrac, false, rndRes, chFmt, clpRng);
JVET_J0090_SET_CACHE_ENABLE( true );
}
-
}
#if JVET_K_AFFINE
@@ -761,7 +759,8 @@ void InterPrediction::xWeightedAverage( const PredictionUnit& pu, const CPelUnit
}
}
-void InterPrediction::motionCompensation( PredictionUnit &pu, PelUnitBuf &predBuf, const RefPicList &eRefPicList )
+void InterPrediction::motionCompensation( PredictionUnit &pu, PelUnitBuf &predBuf, const RefPicList &eRefPicList
+)
{
CodingStructure &cs = *pu.cs;
const PPS &pps = *cs.pps;
@@ -801,7 +800,8 @@ void InterPrediction::motionCompensation( PredictionUnit &pu, PelUnitBuf &predBu
return;
}
-void InterPrediction::motionCompensation( CodingUnit &cu, const RefPicList &eRefPicList )
+void InterPrediction::motionCompensation( CodingUnit &cu, const RefPicList &eRefPicList
+)
{
for( auto &pu : CU::traversePUs( cu ) )
{
@@ -810,10 +810,12 @@ void InterPrediction::motionCompensation( CodingUnit &cu, const RefPicList &eRef
}
}
-void InterPrediction::motionCompensation( PredictionUnit &pu, const RefPicList &eRefPicList /*= REF_PIC_LIST_X*/ )
+void InterPrediction::motionCompensation( PredictionUnit &pu, const RefPicList &eRefPicList /*= REF_PIC_LIST_X*/
+)
{
PelUnitBuf predBuf = pu.cs->getPredBuf( pu );
- motionCompensation( pu, predBuf, eRefPicList );
+ motionCompensation( pu, predBuf, eRefPicList
+ );
}
diff --git a/source/Lib/CommonLib/InterPrediction.h b/source/Lib/CommonLib/InterPrediction.h
index a895322cead55ccc2ae4e85a53d3fe50bca856ee..cbe87a5165bd8774d59e4c1d92a3971ba862a6a4 100644
--- a/source/Lib/CommonLib/InterPrediction.h
+++ b/source/Lib/CommonLib/InterPrediction.h
@@ -49,7 +49,6 @@
#include "RdCost.h"
#include "ContextModelling.h"
-
// forward declaration
class Mv;
@@ -61,7 +60,6 @@ class Mv;
// Class definition
// ====================================================================================================================
-
class InterPrediction : public WeightPrediction
{
private:
@@ -84,7 +82,8 @@ protected:
int m_iRefListIdx;
- void xPredInterUni ( const PredictionUnit& pu, const RefPicList& eRefPicList, PelUnitBuf& pcYuvPred, const bool& bi );
+ void xPredInterUni ( const PredictionUnit& pu, const RefPicList& eRefPicList, PelUnitBuf& pcYuvPred, const bool& bi
+ );
void xPredInterBi ( PredictionUnit& pu, PelUnitBuf &pcYuvPred );
void xPredInterBlk ( const ComponentID& compID, const PredictionUnit& pu, const Picture* refPic, const Mv& _mv, PelUnitBuf& dstPic, const bool& bi, const ClpRng& clpRng
);
@@ -99,7 +98,6 @@ protected:
#if JVET_K0346
void xSubPuMC(PredictionUnit& pu, PelUnitBuf& predBuf, const RefPicList &eRefPicList = REF_PIC_LIST_X);
#endif
-
void destroy();
@@ -114,9 +112,12 @@ public:
void init (RdCost* pcRdCost, ChromaFormat chromaFormatIDC);
// inter
- void motionCompensation (PredictionUnit &pu, PelUnitBuf& predBuf, const RefPicList &eRefPicList = REF_PIC_LIST_X);
- void motionCompensation (PredictionUnit &pu, const RefPicList &eRefPicList = REF_PIC_LIST_X);
- void motionCompensation (CodingUnit &cu, const RefPicList &eRefPicList = REF_PIC_LIST_X);
+ void motionCompensation (PredictionUnit &pu, PelUnitBuf& predBuf, const RefPicList &eRefPicList = REF_PIC_LIST_X
+ );
+ void motionCompensation (PredictionUnit &pu, const RefPicList &eRefPicList = REF_PIC_LIST_X
+ );
+ void motionCompensation (CodingUnit &cu, const RefPicList &eRefPicList = REF_PIC_LIST_X
+ );
#if JVET_J0090_MEMORY_BANDWITH_MEASURE
void cacheAssign( CacheModel *cache );
diff --git a/source/Lib/CommonLib/LoopFilter.cpp b/source/Lib/CommonLib/LoopFilter.cpp
index 96a0dda5b7157dd6be99d3d00bb6f53af05cfcb4..926240e1e5253f7daf95a2ce8d5b62853d35682d 100644
--- a/source/Lib/CommonLib/LoopFilter.cpp
+++ b/source/Lib/CommonLib/LoopFilter.cpp
@@ -421,7 +421,6 @@ unsigned LoopFilter::xGetBoundaryStrengthSingle ( const CodingUnit& cu, const De
const Slice& sliceQ = *cu.slice;
const Position& cuPosLuma = cu.lumaPos();
-
const Position& posQ = localPos;
const Position posP = ( edgeDir == EDGE_VER ) ? posQ.offset( -1, 0 ) : posQ.offset( 0, -1 );
diff --git a/source/Lib/CommonLib/MotionInfo.h b/source/Lib/CommonLib/MotionInfo.h
index 001b84ad427718adf2ae4404ff1c4f7c8c2cb897..d90715d127b887d1027ac638717921697392a53a 100644
--- a/source/Lib/CommonLib/MotionInfo.h
+++ b/source/Lib/CommonLib/MotionInfo.h
@@ -108,7 +108,6 @@ struct MotionInfo
Mv mv [ NUM_REF_PIC_LIST_01 ];
int16_t refIdx [ NUM_REF_PIC_LIST_01 ];
-
MotionInfo() : isInter( false ), interDir( 0 ), sliceIdx( 0 ), refIdx{ NOT_VALID, NOT_VALID } { }
// ensure that MotionInfo(0) produces '\x000....' bit pattern - needed to work with AreaBuf - don't use this constructor for anything else
MotionInfo( int i ) : isInter( i != 0 ), interDir( 0 ), sliceIdx( 0 ), refIdx{ 0, 0 } { CHECKD( i != 0, "The argument for this constructor has to be '0'" ); }
@@ -143,4 +142,5 @@ struct MotionInfo
}
};
+
#endif // __MOTIONINFO__
diff --git a/source/Lib/CommonLib/Mv.h b/source/Lib/CommonLib/Mv.h
index e74c7fce2a17e3e5aad232edc6e17c9681fe1d3c..001d68ebe7c785a78f0df789d5d7ff68262498df 100644
--- a/source/Lib/CommonLib/Mv.h
+++ b/source/Lib/CommonLib/Mv.h
@@ -265,7 +265,6 @@ public:
}
#endif
};// END CLASS DEFINITION MV
-
#if JVET_K0357_AMVR
void roundMV( Mv& rcMv, unsigned imvShift );
#endif
diff --git a/source/Lib/CommonLib/Picture.cpp b/source/Lib/CommonLib/Picture.cpp
index a90cf4aaeb9371c3d10895fc60f8c6d2a9530ad3..525b8cc5fd7f134105e491908a4942e9551a86f1 100644
--- a/source/Lib/CommonLib/Picture.cpp
+++ b/source/Lib/CommonLib/Picture.cpp
@@ -728,6 +728,10 @@ Picture::Picture()
{
m_prevQP[i] = -1;
}
+#if JVET_K0157
+ m_spliceIdx = NULL;
+ m_ctuNums = 0;
+#endif
}
void Picture::create(const ChromaFormat &_chromaFormat, const Size &size, const unsigned _maxCUSize, const unsigned _margin, const bool _decoder)
@@ -788,6 +792,13 @@ void Picture::destroy()
tileMap = nullptr;
}
#endif
+#if JVET_K0157
+ if (m_spliceIdx)
+ {
+ delete[] m_spliceIdx;
+ m_spliceIdx = NULL;
+ }
+#endif
}
void Picture::createTempBuffers( const unsigned _maxCUSize )
@@ -903,6 +914,14 @@ void Picture::finalInit( const SPS& sps, const PPS& pps )
tileMap = new TileMap;
tileMap->create( sps, pps );
#endif
+#if JVET_K0157
+ if (m_spliceIdx == NULL)
+ {
+ m_ctuNums = cs->pcv->sizeInCtus;
+ m_spliceIdx = new int[m_ctuNums];
+ memset(m_spliceIdx, 0, m_ctuNums * sizeof(int));
+ }
+#endif
}
void Picture::allocateNewSlice()
@@ -1113,3 +1132,25 @@ Pel* Picture::getOrigin( const PictureType &type, const ComponentID compID ) con
return M_BUFS( jId, type ).getOrigin( compID );
}
+
+#if JVET_K0157
+void Picture::createSpliceIdx(int nums)
+{
+ m_ctuNums = nums;
+ m_spliceIdx = new int[m_ctuNums];
+ memset(m_spliceIdx, 0, m_ctuNums * sizeof(int));
+}
+
+bool Picture::getSpliceFull()
+{
+ int count = 0;
+ for (int i = 0; i < m_ctuNums; i++)
+ {
+ if (m_spliceIdx[i] != 0)
+ count++;
+ }
+ if (count < m_ctuNums * 0.25)
+ return false;
+ return true;
+}
+#endif
diff --git a/source/Lib/CommonLib/Picture.h b/source/Lib/CommonLib/Picture.h
index f9b360b987dccf6c160fb7063b6e350e56d8d5f6..d742e85cd8a0c7277c44472ab49918d74a3006d8 100644
--- a/source/Lib/CommonLib/Picture.h
+++ b/source/Lib/CommonLib/Picture.h
@@ -225,6 +225,14 @@ struct Picture : public UnitArea
int getPOC() const { return poc; }
void setBorderExtension( bool bFlag) { m_bIsBorderExtended = bFlag;}
Pel* getOrigin( const PictureType &type, const ComponentID compID ) const;
+
+#if JVET_K0157
+ int getSpliceIdx(uint32_t idx) const { return m_spliceIdx[idx]; }
+ void setSpliceIdx(uint32_t idx, int poc) { m_spliceIdx[idx] = poc; }
+ void createSpliceIdx(int nums);
+ bool getSpliceFull();
+#endif
+
public:
bool m_bIsBorderExtended;
bool referenced;
@@ -240,6 +248,11 @@ public:
uint32_t layer;
uint32_t depth;
+#if JVET_K0157
+ int* m_spliceIdx;
+ int m_ctuNums;
+#endif
+
#if ENABLE_SPLIT_PARALLELISM
#if ENABLE_WPP_PARALLELISM
PelStorage m_bufs[( PARL_SPLIT_MAX_NUM_JOBS * PARL_WPP_MAX_NUM_THREADS )][NUM_PIC_TYPES];
diff --git a/source/Lib/CommonLib/Quant.cpp b/source/Lib/CommonLib/Quant.cpp
index fddd9d2291ba5cd4d2dd7a9ecf9a17f43cb1c75c..61ebccbf0a90aa0bbc79003642a72c5ba31e8b5f 100644
--- a/source/Lib/CommonLib/Quant.cpp
+++ b/source/Lib/CommonLib/Quant.cpp
@@ -793,7 +793,8 @@ void Quant::quant(TransformUnit &tu, const ComponentID &compID, const CCoeffBuf
const int iQBits = QUANT_SHIFT + cQP.per + iTransformShift;
// QBits will be OK for any internal bit depth as the reduction in transform shift is balanced by an increase in Qp_per due to QpBDOffset
- const int64_t iAdd = int64_t(tu.cs->slice->getSliceType() == I_SLICE ? 171 : 85) << int64_t(iQBits - 9);
+ const int64_t iAdd = int64_t(tu.cs->slice->getSliceType() == I_SLICE
+ ? 171 : 85) << int64_t(iQBits - 9);
#if HEVC_USE_SIGN_HIDING
const int qBits8 = iQBits - 8;
#endif
diff --git a/source/Lib/CommonLib/RdCost.h b/source/Lib/CommonLib/RdCost.h
index 766816757c3bd86b2f9cb5a27d903303ccfaff81..c33ac7457a500d00821ce02c7feded2a7b3dd1ca 100644
--- a/source/Lib/CommonLib/RdCost.h
+++ b/source/Lib/CommonLib/RdCost.h
@@ -89,7 +89,10 @@ public:
// - 0 = no subsampling, 1 = even rows, 2 = every 4th, etc.
int subShift;
- DistParam() : org(), cur(), step( 1 ), bitDepth( 0 ), useMR( false ), applyWeight( false ), isBiPred( false ), wpCur( nullptr ), compID( MAX_NUM_COMPONENT ), maximumDistortionForEarlyExit( std::numeric_limits<Distortion>::max() ), subShift( 0 ) { }
+ DistParam() :
+ org(), cur(), step( 1 ), bitDepth( 0 ), useMR( false ), applyWeight( false ), isBiPred( false ), wpCur( nullptr ), compID( MAX_NUM_COMPONENT ), maximumDistortionForEarlyExit( std::numeric_limits<Distortion>::max() ), subShift( 0 )
+
+ { }
};
/// RD cost computation class
@@ -116,7 +119,6 @@ private:
int m_iCostScale;
bool m_useQtbt;
-
public:
RdCost();
virtual ~RdCost();
@@ -169,6 +171,7 @@ public:
void setCostScale ( int iCostScale ) { m_iCostScale = iCostScale; }
Distortion getCost ( uint32_t b ) { return Distortion( m_motionLambda * b ); }
+
#if ENABLE_SPLIT_PARALLELISM
void copyState( const RdCost& other );
#endif
diff --git a/source/Lib/CommonLib/Rom.cpp b/source/Lib/CommonLib/Rom.cpp
index dfb4ce1b5fb444f4a2cc837857543a7a5ff79aab..84c9beca97ea9ca805a2651e8f4d1b721e4a5e06 100644
--- a/source/Lib/CommonLib/Rom.cpp
+++ b/source/Lib/CommonLib/Rom.cpp
@@ -186,6 +186,7 @@ const int g_aiNonLMPosThrs[] = { 3, 1, 0 };
#endif
+
// initialize ROM variables
void initROM()
{
diff --git a/source/Lib/CommonLib/Slice.cpp b/source/Lib/CommonLib/Slice.cpp
index 7e26fc803f31578221081ed0b1f997c3e4949196..be7267d1d8605577ba9aec499c404c631344f8d8 100644
--- a/source/Lib/CommonLib/Slice.cpp
+++ b/source/Lib/CommonLib/Slice.cpp
@@ -419,7 +419,6 @@ void Slice::setRefPicList( PicList& rcListPic, bool checkNumPocTotalCurr, bool b
pcRefPic = xGetLongTermRefPic(rcListPic, m_pRPS->getPOC(i), m_pRPS->getCheckLTMSBPresent(i));
}
}
-
// ref_pic_list_init
Picture* rpsCurrList0[MAX_NUM_REF+1];
Picture* rpsCurrList1[MAX_NUM_REF+1];
@@ -432,7 +431,7 @@ void Slice::setRefPicList( PicList& rcListPic, bool checkNumPocTotalCurr, bool b
// - Otherwise, when the current picture contains a P or B slice, the value of NumPocTotalCurr shall not be equal to 0.
if (getRapPicFlag())
{
- CHECK(numPicTotalCurr != 0, "Invalid state");
+ CHECK(numPicTotalCurr != 0, "Invalid state");
}
if (m_eSliceType == I_SLICE)
@@ -502,7 +501,6 @@ void Slice::setRefPicList( PicList& rcListPic, bool checkNumPocTotalCurr, bool b
m_bIsUsedAsLongTerm[REF_PIC_LIST_1][rIdx] = ( cIdx >= NumPicStCurr0 + NumPicStCurr1 );
}
}
-
// For generalized B
// note: maybe not existed case (always L0 is copied to L1 if L1 is empty)
if( bCopyL0toL1ErrorCase && isInterB() && getNumRefIdx(REF_PIC_LIST_1) == 0)
@@ -533,7 +531,7 @@ int Slice::getNumRpsCurrTempList() const
numRpsCurrTempList++;
}
}
- return numRpsCurrTempList;
+ return numRpsCurrTempList;
}
void Slice::initEqualRef()
@@ -1312,7 +1310,11 @@ int Slice::checkThatAllRefPicsAreAvailable( PicList& rcListPic, const ReferenceP
/** Function for constructing an explicit Reference Picture Set out of the available pictures in a referenced Reference Picture Set
*/
-void Slice::createExplicitReferencePictureSetFromReference( PicList& rcListPic, const ReferencePictureSet *pReferencePictureSet, bool isRAP, int pocRandomAccess, bool bUseRecoveryPoint, const bool bEfficientFieldIRAPEnabled)
+void Slice::createExplicitReferencePictureSetFromReference(PicList& rcListPic, const ReferencePictureSet *pReferencePictureSet, bool isRAP, int pocRandomAccess, bool bUseRecoveryPoint, const bool bEfficientFieldIRAPEnabled
+#if JVET_K0157
+ , bool isEncodeLtRef, bool isCompositeRefEnable
+#endif
+)
{
Picture* rpcPic;
int i, j;
@@ -1352,7 +1354,11 @@ void Slice::createExplicitReferencePictureSetFromReference( PicList& rcListPic,
}
else
{
+#if JVET_K0157
+ if (bEfficientFieldIRAPEnabled && rpcPic->getPOC() == this->getAssociatedIRAPPOC() && this->getAssociatedIRAPPOC() == this->getPOC() + (isCompositeRefEnable ? 2 : 1))
+#else
if(bEfficientFieldIRAPEnabled && rpcPic->getPOC() == this->getAssociatedIRAPPOC() && this->getAssociatedIRAPPOC() == this->getPOC()+1)
+#endif
{
irapIsInRPS = true;
}
@@ -1371,7 +1377,11 @@ void Slice::createExplicitReferencePictureSetFromReference( PicList& rcListPic,
while ( iterPic != rcListPic.end())
{
rpcPic = *(iterPic++);
+#if JVET_K0157
+ if (rpcPic->getPOC() == this->getAssociatedIRAPPOC() && this->getAssociatedIRAPPOC() == this->getPOC() + (isCompositeRefEnable ? 2 : 1))
+#else
if(rpcPic->getPOC() == this->getAssociatedIRAPPOC() && this->getAssociatedIRAPPOC() == this->getPOC()+1)
+#endif
{
pLocalRPS->setDeltaPOC(k, 1);
pLocalRPS->setUsed(k, true);
@@ -1381,6 +1391,53 @@ void Slice::createExplicitReferencePictureSetFromReference( PicList& rcListPic,
}
}
}
+#if JVET_K0157
+ if (isCompositeRefEnable && isEncodeLtRef)
+ {
+ useNewRPS = true;
+ nrOfNegativePictures = 0;
+ nrOfPositivePictures = 0;
+ for (i = 0; i<pReferencePictureSet->getNumberOfPictures(); i++)
+ {
+ j = 0;
+ k = 0;
+
+ // loop through all pictures in the reference picture buffer
+ PicList::iterator iterPic = rcListPic.begin();
+ while (iterPic != rcListPic.end())
+ {
+ j++;
+ rpcPic = *(iterPic++);
+
+ if (rpcPic->getPOC() == this->getPOC() + 1 + pReferencePictureSet->getDeltaPOC(i) && rpcPic->referenced)
+ {
+ // This picture exists as a reference picture
+ // and should be added to the explicit Reference Picture Set
+ pLocalRPS->setDeltaPOC(k, pReferencePictureSet->getDeltaPOC(i) + 1);
+ pLocalRPS->setUsed(k, pReferencePictureSet->getUsed(i) && (!isRAP));
+ if (bEfficientFieldIRAPEnabled)
+ {
+ pLocalRPS->setUsed(k, pLocalRPS->getUsed(k) && !(bUseRecoveryPoint && this->getPOC() > pocRandomAccess && this->getPOC() + pReferencePictureSet->getDeltaPOC(i) + 1 < pocRandomAccess));
+ }
+
+ if (pLocalRPS->getDeltaPOC(k) < 0)
+ {
+ nrOfNegativePictures++;
+ }
+ else
+ {
+ if (bEfficientFieldIRAPEnabled && rpcPic->getPOC() == this->getAssociatedIRAPPOC() && this->getAssociatedIRAPPOC() == this->getPOC() + 2)
+ {
+ irapIsInRPS = true;
+ }
+ nrOfPositivePictures++;
+ }
+ k++;
+ }
+ }
+ }
+ }
+#endif
pLocalRPS->setNumberOfNegativePictures(nrOfNegativePictures);
pLocalRPS->setNumberOfPositivePictures(nrOfPositivePictures);
pLocalRPS->setNumberOfPictures(nrOfNegativePictures+nrOfPositivePictures);
@@ -1641,6 +1698,9 @@ SPSNext::SPSNext( SPS& sps )
#else
#endif
, m_MTTMode ( 0 )
+#if JVET_K0157
+ , m_compositeRefEnabled ( false )
+#endif
// ADD_NEW_TOOL : (sps extension) add tool enabling flags here (with "false" as default values)
{
}
diff --git a/source/Lib/CommonLib/Slice.h b/source/Lib/CommonLib/Slice.h
index 5164daada2b1b45d73837a0b4b009635c7863015..94a7fe402e8d135c4b8865b4c8fc7809202eb273 100644
--- a/source/Lib/CommonLib/Slice.h
+++ b/source/Lib/CommonLib/Slice.h
@@ -859,6 +859,9 @@ private:
// multi type tree (QTBT + triple split)
unsigned m_MTTMode;
+#if JVET_K0157
+ bool m_compositeRefEnabled; //composite longterm reference
+#endif
// ADD_NEW_TOOL : (sps extension) add tool enabling flags and associated parameters here
public:
@@ -915,7 +918,6 @@ public:
void setUseInterEMT ( bool b ) { m_InterEMT = b; }
bool getUseInterEMT () const { return m_InterEMT; }
#endif
-
//===== additional parameters =====
// qtbt
void setCTUSize ( unsigned ctuSize ) { m_CTUSize = ctuSize; }
@@ -959,6 +961,10 @@ public:
unsigned getMTTMode () const { return m_MTTMode; }
void setMTTMode ( unsigned mode ) { m_MTTMode = mode; m_MTTEnabled = ( m_MTTMode != 0 ); }
+#if JVET_K0157
+ void setUseCompositeRef(bool b) { m_compositeRefEnabled = b; }
+ bool getUseCompositeRef() const { return m_compositeRefEnabled; }
+#endif
// ADD_NEW_TOOL : (sps extension) add access functions for tool enabling flags and associated parameters here
};
@@ -1764,9 +1770,13 @@ public:
bool isTemporalLayerSwitchingPoint( PicList& rcListPic ) const;
bool isStepwiseTemporalLayerSwitchingPointCandidate( PicList& rcListPic ) const;
int checkThatAllRefPicsAreAvailable( PicList& rcListPic, const ReferencePictureSet *pReferencePictureSet, bool printErrors, int pocRandomAccess = 0, bool bUseRecoveryPoint = false) const;
- void createExplicitReferencePictureSetFromReference( PicList& rcListPic, const ReferencePictureSet *pReferencePictureSet, bool isRAP, int pocRandomAccess, bool bUseRecoveryPoint, const bool bEfficientFieldIRAPEnabled);
+ void createExplicitReferencePictureSetFromReference(PicList& rcListPic, const ReferencePictureSet *pReferencePictureSet, bool isRAP, int pocRandomAccess, bool bUseRecoveryPoint, const bool bEfficientFieldIRAPEnabled
+#if JVET_K0157
+ , bool isEncodeLtRef, bool isCompositeRefEnable
+#endif
+ );
void setMaxNumMergeCand(uint32_t val ) { m_maxNumMergeCand = val; }
- uint32_t getMaxNumMergeCand() const { return m_maxNumMergeCand; }
+ uint32_t getMaxNumMergeCand() const { return m_maxNumMergeCand; }
void setNoOutputPriorPicsFlag( bool val ) { m_noOutputPriorPicsFlag = val; }
bool getNoOutputPriorPicsFlag() const { return m_noOutputPriorPicsFlag; }
diff --git a/source/Lib/CommonLib/TrQuant.cpp b/source/Lib/CommonLib/TrQuant.cpp
index c3d3c15ef78e2f63578faec4b4f27543bac2f696..52235829949eedd056e0b520df558d109b519393 100644
--- a/source/Lib/CommonLib/TrQuant.cpp
+++ b/source/Lib/CommonLib/TrQuant.cpp
@@ -703,6 +703,13 @@ void TrQuant::xT( const TransformUnit &tu, const ComponentID &compID, const CPel
const unsigned ucMode = getEmtMode ( tu, compID );
const unsigned ucTrIdx = getEmtTrIdx( tu, compID );
+ if( ucTrIdx != DCT2 )
+ {
+#if RExt__DECODER_DEBUG_TOOL_STATISTICS
+ CodingStatistics::IncrementStatisticTool( CodingStatisticsClassType{ STATS__TOOL_EMT, uint32_t( iWidth ), uint32_t( iHeight ), compID } );
+#endif
+ }
+
#if INTRA67_3MPM
#if HEVC_USE_4x4_DSTVII
xTrMxN_EMT(channelBitDepth, resi.buf, resi.stride, dstCoeff.buf, iWidth, iHeight, useDST, maxLog2TrDynamicRange, ucMode, ucTrIdx
diff --git a/source/Lib/CommonLib/TypeDef.h b/source/Lib/CommonLib/TypeDef.h
index 499837ae61f1eacb1a90213179ad92f3c819a48b..ed0b7a5e665abb0e9b58caa77c8856e41e57ec10 100644
--- a/source/Lib/CommonLib/TypeDef.h
+++ b/source/Lib/CommonLib/TypeDef.h
@@ -50,6 +50,9 @@
#include <assert.h>
#include <cassert>
+
+
+
#define JVET_K1000_SIMPLIFIED_EMT 1 // EMT with only DCT-2, DCT-8 and DST-7
#define JVET_K0371_ALF 1
@@ -75,6 +78,9 @@
#define REUSE_CU_RESULTS 1
#endif
+#define JVET_K0390_RATECTRL 1
+#define RATECTRL_FIX_FULLNBIT 1 //fix the QP-lambda relationship in rate control if JVET-K0154 for FULL_NBIT is enabled
+
#define JVET_K0352_MERGE_ENCOPT 1 // encoder optimization for merge
#define JVET_K0556_MAX_TT_SIZE_64 1 // Maximum TT size is set to 64x64 for P/B-slice
@@ -82,11 +88,12 @@
#define JVET_K0554 1 // when adopting, also remove the macro HM_QTBT_ONLY_QT_IMPLICIT (keep the case for value 0)
#define JVET_K0346 1 // simplifications on ATMVP
+
#define JVET_K0063_PDPC_SIMP 1 // Simplified PDPC
#define JVET_K0351_LESS_CONSTRAINT 1 // Only disallow binary split with same orientation in center partition of the ternary split and release the other constraints in K0351.
-#define JVET_K0251_QP_EXT 1 // Extending the QP parameter value range for coarse quantization
+#define JVET_K0251_QP_EXT 1 // Extending the QP parameter value range for coarse quantization
#define JVET_K_AFFINE 1
#if JVET_K_AFFINE
@@ -104,6 +111,7 @@
#define JVET_K0357_AMVR 1 // Adaptive motion vector resolution separated from JEM_TOOLS macro
+
#ifndef JVET_B0051_NON_MPM_MODE
#define JVET_B0051_NON_MPM_MODE ( 1 && JEM_TOOLS )
@@ -157,11 +165,16 @@
#ifndef ENABLE_TRACING
#define ENABLE_TRACING 0 // DISABLE by default (enable only when debugging, requires 15% run-time in decoding) -- see documentation in 'doc/DTrace for NextSoftware.pdf'
-
+#if ENABLE_TRACING
+#define K0149_BLOCK_STATISTICS 1 // enables block statistics, which can be analysed with YUView (https://github.com/IENT/YUView)
+#if K0149_BLOCK_STATISTICS
+#define BLOCK_STATS_AS_CSV 0 // statistics will be written in a comma separated value format. this is not supported by YUView
+#endif
+#endif
#endif // ! ENABLE_TRACING
#define WCG_EXT 0 // part of JEM sharp Luma qp
-#define WCG_WPSNR WCG_EXT
+#define WCG_WPSNR WCG_EXT
#if HEVC_TOOLS
#define HEVC_USE_INTRA_SMOOTHING_T32 1
@@ -180,6 +193,7 @@
#define HEVC_USE_SIGN_HIDING 1
#endif
+#define JVET_K0157 1
#define KEEP_PRED_AND_RESI_SIGNALS 0
@@ -298,7 +312,7 @@
#define SHARP_LUMA_DELTA_QP 1 ///< include non-normative LCU deltaQP and normative chromaQP change
#define ER_CHROMA_QP_WCG_PPS 1 ///< Chroma QP model for WCG used in Anchor 3.2
-#define ENABLE_QPA 0
+#define ENABLE_QPA 1 ///< Non-normative perceptual QP adaptation according to JVET-H0047 and JVET-K0206. Deactivated by default, activated using encoder arguments --PerceptQPA=1 --SliceChromaQPOffsetPeriodicity=1
@@ -1422,8 +1436,8 @@ enum AlfFilterType
struct AlfFilterShape
{
AlfFilterShape( int size )
- : filterLength( size ),
- numCoeff( size * size / 4 + 1 ),
+ : filterLength( size ),
+ numCoeff( size * size / 4 + 1 ),
filterSize( size * size / 2 + 1 )
{
if( size == 5 )
@@ -1466,7 +1480,7 @@ struct AlfFilterShape
2,
2, 2, 2,
2, 2, 2, 2, 2,
- 2, 2, 2, 1, 1
+ 2, 2, 2, 1, 1
};
golombIdx = {
diff --git a/source/Lib/CommonLib/Unit.cpp b/source/Lib/CommonLib/Unit.cpp
index 9b14f80f9b0655fa01a05983c6ecf80df72a2c5a..05eff4c40d068e0679a55b01e5c6d75815b5ed9b 100644
--- a/source/Lib/CommonLib/Unit.cpp
+++ b/source/Lib/CommonLib/Unit.cpp
@@ -274,7 +274,6 @@ CodingUnit& CodingUnit::operator=( const CodingUnit& other )
imv = other.imv;
imvNumCand = other.imvNumCand;
#endif
-
return *this;
}
diff --git a/source/Lib/CommonLib/Unit.h b/source/Lib/CommonLib/Unit.h
index 18694015491a8c7565477df51e8dc1607e0738a4..3e453c722c3e1d1257b64d7f532dd4dfb80b3724 100644
--- a/source/Lib/CommonLib/Unit.h
+++ b/source/Lib/CommonLib/Unit.h
@@ -313,7 +313,6 @@ struct CodingUnit : public UnitArea
#if JVET_K1000_SIMPLIFIED_EMT
uint8_t emtFlag;
#endif
-
// needed for fast imv mode decisions
int8_t imvNumCand;
diff --git a/source/Lib/CommonLib/UnitPartitioner.cpp b/source/Lib/CommonLib/UnitPartitioner.cpp
index 89fea3e4b6d2a41c9bbf80a244d8c26393167cbf..f12c4889f911584bba1c431475bc529d4a759f2f 100644
--- a/source/Lib/CommonLib/UnitPartitioner.cpp
+++ b/source/Lib/CommonLib/UnitPartitioner.cpp
@@ -472,11 +472,11 @@ bool QTBTPartitioner::canSplit( const PartSplit split, const CodingStructure &cs
break;
case CU_TRIH_SPLIT:
if( ( cs.sps->getSpsNext().getMTTMode() & 1 ) != 1 ) return false;
- if( area.height <= 2 * minTtSize || area.height > maxTtSize ) return false;
+ if( area.height <= 2 * minTtSize || area.height > maxTtSize || area.width > maxTtSize) return false;
break;
case CU_TRIV_SPLIT:
if( ( cs.sps->getSpsNext().getMTTMode() & 1 ) != 1 ) return false;
- if( area.width <= 2 * minTtSize || area.width > maxTtSize ) return false;
+ if( area.width <= 2 * minTtSize || area.width > maxTtSize || area.height > maxTtSize) return false;
break;
default:
break;
diff --git a/source/Lib/CommonLib/UnitTools.cpp b/source/Lib/CommonLib/UnitTools.cpp
index 9d131a11216e7456eae993006008c21eec02da1c..9086d3cf473387c333d2cd34af1b740ada730ae3 100644
--- a/source/Lib/CommonLib/UnitTools.cpp
+++ b/source/Lib/CommonLib/UnitTools.cpp
@@ -66,6 +66,30 @@ UnitArea CS::getArea( const CodingStructure &cs, const UnitArea &area, const Cha
return isDualITree( cs ) ? area.singleChan( chType ) : area;
}
+#if DMVR_JVET_LOW_LATENCY_K0217
+void CS::setRefinedMotionField(CodingStructure &cs)
+{
+ for (CodingUnit *cu : cs.cus)
+ {
+ for (auto &pu : CU::traversePUs(*cu))
+ {
+ if (pu.cs->sps->getSpsNext().getUseDMVR()
+ && pu.mergeFlag
+ && pu.mergeType == MRG_TYPE_DEFAULT_N
+ && !pu.frucMrgMode
+ && !pu.cu->LICFlag
+ && !pu.cu->affine
+ && PU::isBiPredFromDifferentDir(pu))
+ {
+ pu.mv[REF_PIC_LIST_0] += pu.mvd[REF_PIC_LIST_0];
+ pu.mv[REF_PIC_LIST_1] -= pu.mvd[REF_PIC_LIST_0];
+ pu.mvd[REF_PIC_LIST_0].setZero();
+ PU::spanMotionInfo(pu);
+ }
+ }
+ }
+}
+#endif
// CU tools
bool CU::isIntra(const CodingUnit &cu)
@@ -311,8 +335,8 @@ int PU::getIntraMPMs( const PredictionUnit &pu, unsigned* mpm, const ChannelType
CHECK(2 >= numMPMs, "Invalid number of most probable modes");
- const int offset = (int) NUM_LUMA_MODE - 5;
- const int mod = offset + 3;
+ const int offset = 61;
+ const int mod = 64;
if (leftIntraDir == aboveIntraDir)
{
@@ -619,7 +643,6 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx, co
// compute the location of the current PU
int cnt = 0;
-
const Position posLT = pu.Y().topLeft();
const Position posRT = pu.Y().topRight();
const Position posLB = pu.Y().bottomLeft();
@@ -639,7 +662,6 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx, co
// get Inter Dir
mrgCtx.interDirNeighbours[cnt] = miLeft.interDir;
-
// get Mv from Left
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField(miLeft.mv[0], miLeft.refIdx[0]);
@@ -678,7 +700,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx, co
// get Inter Dir
mrgCtx.interDirNeighbours[cnt] = miAbove.interDir;
- // get Mv from Left
+ // get Mv from Above
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField( miAbove.mv[0], miAbove.refIdx[0] );
if( slice.isInterB() )
@@ -720,7 +742,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx, co
// get Inter Dir
mrgCtx.interDirNeighbours[cnt] = miAboveRight.interDir;
- // get Mv from Left
+ // get Mv from Above-right
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField( miAboveRight.mv[0], miAboveRight.refIdx[0] );
if( slice.isInterB() )
@@ -800,7 +822,8 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx, co
bool bMrgIdxMatchATMVPCan = ( mrgCandIdx == cnt );
bool tmpLICFlag = false;
- isAvailableSubPu = cs.sps->getSpsNext().getUseATMVP() && getInterMergeSubPuMvpCand( pu, mrgCtx, tmpLICFlag, cnt );
+ isAvailableSubPu = cs.sps->getSpsNext().getUseATMVP() && getInterMergeSubPuMvpCand( pu, mrgCtx, tmpLICFlag, cnt
+ );
if( isAvailableSubPu )
{
@@ -1078,7 +1101,6 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx, co
}
-
static int xGetDistScaleFactor(const int &iCurrPOC, const int &iCurrRefPOC, const int &iColPOC, const int &iColRefPOC)
{
int iDiffPocD = iColPOC - iColRefPOC;
@@ -1124,7 +1146,6 @@ bool PU::getColocatedMVP(const PredictionUnit &pu, const RefPicList &eRefPicList
{
return false;
}
-
int iColRefIdx = mi.refIdx[eColRefPicList];
if (iColRefIdx < 0)
@@ -2165,7 +2186,6 @@ bool PU::isAffineMrgFlagCoded( const PredictionUnit &pu )
}
return getFirstAvailableAffineNeighbour( pu ) != nullptr;
}
-
void PU::getAffineMergeCand( const PredictionUnit &pu, MvField (*mvFieldNeighbours)[3], unsigned char &interDirNeighbours, int &numValidMergeCand )
{
for ( int mvNum = 0; mvNum < 3; mvNum++ )
@@ -2416,7 +2436,8 @@ void clipColBlkMv(int& mvX, int& mvY, const PredictionUnit& pu)
}
#endif
-bool PU::getInterMergeSubPuMvpCand(const PredictionUnit &pu, MergeCtx& mrgCtx, bool& LICFlag, const int count)
+bool PU::getInterMergeSubPuMvpCand(const PredictionUnit &pu, MergeCtx& mrgCtx, bool& LICFlag, const int count
+)
{
const Slice &slice = *pu.cs->slice;
#if JVET_K0346
diff --git a/source/Lib/CommonLib/UnitTools.h b/source/Lib/CommonLib/UnitTools.h
index f7640e1135dfae435d6838b6904d25c1b722c95f..2571b290bc149bd7df1f56d38a50fd8b19babbcd 100644
--- a/source/Lib/CommonLib/UnitTools.h
+++ b/source/Lib/CommonLib/UnitTools.h
@@ -49,6 +49,9 @@ namespace CS
uint64_t getEstBits ( const CodingStructure &cs );
UnitArea getArea ( const CodingStructure &cs, const UnitArea &area, const ChannelType chType );
bool isDualITree ( const CodingStructure &cs );
+#if DMVR_JVET_LOW_LATENCY_K0217
+ void setRefinedMotionField ( CodingStructure &cs );
+#endif
}
@@ -81,6 +84,7 @@ namespace CU
bool hasNonTsCodedBlock (const CodingUnit& cu);
uint32_t getNumNonZeroCoeffNonTs (const CodingUnit& cu);
+
PUTraverser traversePUs ( CodingUnit& cu);
TUTraverser traverseTUs ( CodingUnit& cu);
cPUTraverser traversePUs (const CodingUnit& cu);
@@ -126,6 +130,7 @@ namespace PU
#if JVET_K0357_AMVR
void applyImv ( PredictionUnit &pu, MergeCtx &mrgCtx, InterPrediction *interPred = NULL );
#endif
+ void getAffineMergeCand (const PredictionUnit &pu, MvField (*mvFieldNeighbours)[3], unsigned char &interDirNeighbours, int &numValidMergeCand );
#if JVET_K_AFFINE
bool isAffineMrgFlagCoded (const PredictionUnit &pu );
void getAffineMergeCand (const PredictionUnit &pu, MvField (*mvFieldNeighbours)[3], unsigned char &interDirNeighbours, int &numValidMergeCand );
@@ -133,7 +138,8 @@ namespace PU
void setAllAffineMv ( PredictionUnit &pu, Mv affLT, Mv affRT, Mv affLB, RefPicList eRefList );
#endif
#if JVET_K0346
- bool getInterMergeSubPuMvpCand(const PredictionUnit &pu, MergeCtx &mrgCtx, bool& LICFlag, const int count);
+ bool getInterMergeSubPuMvpCand(const PredictionUnit &pu, MergeCtx &mrgCtx, bool& LICFlag, const int count
+ );
bool getInterMergeSubPuRecurCand(const PredictionUnit &pu, MergeCtx &mrgCtx, const int count);
#endif
bool isBiPredFromDifferentDir (const PredictionUnit &pu);
diff --git a/source/Lib/CommonLib/dtrace.cpp b/source/Lib/CommonLib/dtrace.cpp
index 3a6c3dae5896fc452067476e74b5f6d7664b154d..66bb00825599aa8624a8973b65a3939232c5fc56 100644
--- a/source/Lib/CommonLib/dtrace.cpp
+++ b/source/Lib/CommonLib/dtrace.cpp
@@ -320,3 +320,18 @@ void CDTrace::dtrace_repeat( int k, int i_times, const char *format, /*va_list a
}
return;
}
+
+#if K0149_BLOCK_STATISTICS
+void CDTrace::dtrace_header( const char *format, /*va_list args*/... )
+{
+ if( m_trace_file )
+ {
+ va_list args;
+ va_start ( args, format );
+ vfprintf ( m_trace_file, format, args );
+ fflush( m_trace_file );
+ va_end ( args );
+ }
+ return;
+}
+#endif
diff --git a/source/Lib/CommonLib/dtrace.h b/source/Lib/CommonLib/dtrace.h
index 87140ee6aaad7f3fd1e33fecb43195059344e443..9d622f3eed7876f759e8a7adf6081113c4bdb8a4 100644
--- a/source/Lib/CommonLib/dtrace.h
+++ b/source/Lib/CommonLib/dtrace.h
@@ -47,6 +47,11 @@
#include <vector>
#include <cstdarg>
+#if K0149_BLOCK_STATISTICS
+class CodingStructure;
+struct Position;
+#endif
+
class CDTrace;
typedef std::string CType;
@@ -118,6 +123,21 @@ public:
template<bool bCount>
void dtrace ( int, const char *format, /*va_list args*/... );
void dtrace_repeat( int, int i_times, const char *format, /*va_list args*/... );
+#if K0149_BLOCK_STATISTICS
+ void dtrace_header ( const char *format, /*va_list args*/... );
+ // CTU
+ void dtrace_block_scalar( int k, const CodingStructure &cs, std::string stat_type, signed value );
+ // CU
+ void dtrace_block_scalar( int k, const CodingUnit &cu, std::string stat_type, signed value, bool isChroma = false );
+ void dtrace_block_vector( int k, const CodingUnit &cu, std::string stat_type, signed val_x, signed val_y );
+ // PU
+ void dtrace_block_scalar( int k, const PredictionUnit &pu, std::string stat_type, signed value, bool isChroma = false );
+ void dtrace_block_vector( int k, const PredictionUnit &pu, std::string stat_type, signed val_x, signed val_y );
+ void dtrace_block_affinetf( int k, const PredictionUnit &pu, std::string stat_type, signed val_x0, signed val_y0, signed val_x1, signed val_y1, signed val_x2, signed val_y2 );
+ // TU
+ void dtrace_block_scalar(int k, const TransformUnit &tu, std::string stat_type, signed value, bool isChroma = false );
+ void dtrace_block_vector(int k, const TransformUnit &tu, std::string stat_type, signed val_x, signed val_y);
+#endif
bool update ( state_type stateval );
int init( vstring channel_names );
int getLastError() { return m_error_code; }
@@ -126,6 +146,9 @@ public:
std::string getErrMessage();
int64_t getChannelCounter( int channel ) { return chanRules[channel].getCounter(); }
void decrementChannelCounter( int channel ) { chanRules[channel].decrementCounter(); }
+#if K0149_BLOCK_STATISTICS
+ bool isChannelActive( int channel ) { return chanRules[channel].active(); }
+#endif
};
diff --git a/source/Lib/CommonLib/dtrace_blockstatistics.cpp b/source/Lib/CommonLib/dtrace_blockstatistics.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..9e4552d8a5513f9b8822400100f79d572022effb
--- /dev/null
+++ b/source/Lib/CommonLib/dtrace_blockstatistics.cpp
@@ -0,0 +1,775 @@
+/* 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) 2010-2018, 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 dtrace_blockstatistics.cpp
+ * \brief DTrace block statistcis support for next software
+ */
+
+#include "dtrace_blockstatistics.h"
+#include "dtrace.h"
+#include "dtrace_next.h"
+#include "CommonLib/Unit.h"
+#include "CommonLib/Picture.h"
+#include "CommonLib/UnitTools.h"
+//#include "CommonLib/CodingStructure.h"
+
+#if K0149_BLOCK_STATISTICS
+std::string GetBlockStatisticName(BlockStatistic statistic)
+{
+ auto statisticIterator = blockstatistic2description.find(statistic);
+ // enforces that all delcared statistic enum items are also part of the map
+ assert(statisticIterator != blockstatistic2description.end() && "A block statistics declared in the enum is missing in the map for statistic description.");
+
+ return std::get<0>(statisticIterator->second);
+}
+
+std::string GetBlockStatisticTypeString(BlockStatistic statistic)
+{
+ auto statisticIterator = blockstatistic2description.find(statistic);
+ // enforces that all delcared statistic enum items are also part of the map
+ assert(statisticIterator != blockstatistic2description.end() && "A block statistics declared in the enum is missing in the map for statistic description.");
+
+ BlockStatisticType statisticType = std::get<1>(statisticIterator->second);
+ switch (statisticType) {
+ case BlockStatisticType::Flag:
+ return std::string("Flag");
+ break;
+ case BlockStatisticType::Vector:
+ return std::string("Vector");
+ break;
+ case BlockStatisticType::Integer:
+ return std::string("Integer");
+ break;
+ case BlockStatisticType::AffineTFVectors:
+ return std::string("AffineTFVectors");
+ break;
+ default:
+ assert(0);
+ break;
+ }
+ return std::string();
+}
+
+std::string GetBlockStatisticTypeSpecificInfo(BlockStatistic statistic)
+{
+ auto statisticIterator = blockstatistic2description.find(statistic);
+ // enforces that all delcared statistic enum items are also part of the map
+ assert(statisticIterator != blockstatistic2description.end() && "A block statistics declared in the enum is missing in the map for statistic description.");
+
+ return std::get<2>(statisticIterator->second);
+}
+
+void CDTrace::dtrace_block_scalar( int k, const CodingStructure &cs, std::string stat_type, signed value )
+{
+#if BLOCK_STATS_AS_CSV
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%d\n", cs.picture->poc, cs.area.lx(), cs.area.ly(), cs.area.lwidth(), cs.area.lheight(), stat_type.c_str(), value );
+#else
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s=%d\n", cs.picture->poc, cs.area.lx(), cs.area.ly(), cs.area.lwidth(), cs.area.lheight(), stat_type.c_str(), value );
+#endif
+}
+
+void CDTrace::dtrace_block_scalar( int k, const CodingUnit &cu, std::string stat_type, signed value, bool isChroma /*= false*/ )
+{
+ const CodingStructure& cs = *cu.cs;
+#if BLOCK_STATS_AS_CSV
+ if(isChroma)
+ {
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%d\n", cs.picture->poc, cu.Cb().x*2, cu.Cb().y*2, cu.Cb().width*2, cu.Cb().height*2, stat_type.c_str(), value );
+ }
+ else
+ {
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%d\n", cs.picture->poc, cu.lx(), cu.ly(), cu.lwidth(), cu.lheight(), stat_type.c_str(), value );
+ }
+#else
+ if(isChroma)
+ {
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s=%d\n", cs.picture->poc, cu.Cb().x*2, cu.Cb().y*2, cu.Cb().width*2, cu.Cb().height*2, stat_type.c_str(), value );
+ }
+ else
+ {
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s=%d\n", cs.picture->poc, cu.lx(), cu.ly(), cu.lwidth(), cu.lheight(), stat_type.c_str(), value );
+ }
+#endif
+}
+
+void CDTrace::dtrace_block_vector( int k, const CodingUnit &cu, std::string stat_type, signed val_x, signed val_y )
+{
+ const CodingStructure& cs = *cu.cs;
+#if BLOCK_STATS_AS_CSV
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%4d;%4d\n", cs.picture->poc, cu.lx(), cu.ly(), cu.lwidth(), cu.lheight(), stat_type.c_str(), val_x, val_y );
+#else
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s={%4d,%4d}\n", cs.picture->poc, cu.lx(), cu.ly(), cu.lwidth(), cu.lheight(), stat_type.c_str(), val_x, val_y );
+#endif
+}
+
+void CDTrace::dtrace_block_scalar( int k, const PredictionUnit &pu, std::string stat_type, signed value, bool isChroma /*= false*/ )
+{
+ const CodingStructure& cs = *pu.cs;
+#if BLOCK_STATS_AS_CSV
+ if(isChroma)
+ {
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%d\n", cs.picture->poc, pu.Cb().x*2, pu.Cb().y*2, pu.Cb().width*2, pu.Cb().height*2, stat_type.c_str(), value );
+ }
+ else
+ {
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%d\n", cs.picture->poc, pu.lx(), pu.ly(), pu.lwidth(), pu.lheight(), stat_type.c_str(), value );
+ }
+#else
+ if(isChroma)
+ {
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s=%d\n", cs.picture->poc, pu.Cb().x*2, pu.Cb().y*2, pu.Cb().width*2, pu.Cb().height*2, stat_type.c_str(), value );
+ }
+ else
+ {
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s=%d\n", cs.picture->poc, pu.lx(), pu.ly(), pu.lwidth(), pu.lheight(), stat_type.c_str(), value );
+ }
+#endif
+}
+
+void CDTrace::dtrace_block_vector( int k, const PredictionUnit &pu, std::string stat_type, signed val_x, signed val_y )
+{
+ const CodingStructure& cs = *pu.cs;
+#if BLOCK_STATS_AS_CSV
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%4d;%4d\n", cs.picture->poc, pu.lx(), pu.ly(), pu.lwidth(), pu.lheight(), stat_type.c_str(), val_x, val_y );
+#else
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s={%4d,%4d}\n", cs.picture->poc, pu.lx(), pu.ly(), pu.lwidth(), pu.lheight(), stat_type.c_str(), val_x, val_y );
+#endif
+}
+
+void CDTrace::dtrace_block_scalar(int k, const TransformUnit &tu, std::string stat_type, signed value, bool isChroma /*= false*/ )
+{
+ const CodingStructure& cs = *tu.cs;
+#if BLOCK_STATS_AS_CSV
+ if(isChroma)
+ {
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%d\n", cs.picture->poc, tu.Cb().x*2, tu.Cb().y*2, tu.Cb().width*2, tu.Cb().height*2, stat_type.c_str(), value );
+ }
+ else
+ {
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%d\n", cs.picture->poc, tu.lx(), tu.ly(), tu.lwidth(), tu.lheight(), stat_type.c_str(), value );
+ }
+#else
+ if(isChroma)
+ {
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s=%d\n", cs.picture->poc, tu.Cb().x*2, tu.Cb().y*2, tu.Cb().width*2, tu.Cb().height*2, stat_type.c_str(), value );
+ }
+ else
+ {
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s=%d\n", cs.picture->poc, tu.lx(), tu.ly(), tu.lwidth(), tu.lheight(), stat_type.c_str(), value );
+ }
+#endif
+}
+
+void CDTrace::dtrace_block_vector(int k, const TransformUnit &tu, std::string stat_type, signed val_x, signed val_y)
+{
+ const CodingStructure& cs = *tu.cs;
+#if BLOCK_STATS_AS_CSV
+ dtrace<false>(k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%4d;%4d\n", cs.picture->poc, pu.lx(), pu.ly(), pu.lwidth(), pu.lheight(), stat_type.c_str(), val_x, val_y);
+#else
+ dtrace<false>(k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s={%4d,%4d}\n", cs.picture->poc, tu.lx(), tu.ly(), tu.lwidth(), tu.lheight(), stat_type.c_str(), val_x, val_y);
+#endif
+}
+
+void CDTrace::dtrace_block_affinetf( int k, const PredictionUnit &pu, std::string stat_type, signed val_x0, signed val_y0, signed val_x1, signed val_y1, signed val_x2, signed val_y2 )
+{
+ const CodingStructure& cs = *pu.cs;
+#if BLOCK_STATS_AS_CSV
+ dtrace<false>( k, "BlockStat;%d;%4d;%4d;%2d;%2d;%s;%4d;%4d;%4d;%4d;%4d;%4d\n",
+ cs.picture->poc, pu.lx(), pu.ly(), pu.lwidth(), pu.lheight(), stat_type.c_str(),
+ val_x0, val_y0, val_x1, val_y1 , val_x2, val_y2 );
+#else
+ dtrace<false>( k, "BlockStat: POC %d @(%4d,%4d) [%2dx%2d] %s={%4d,%4d,%4d,%4d,%4d,%4d}\n",
+ cs.picture->poc, pu.lx(), pu.ly(), pu.lwidth(), pu.lheight(), stat_type.c_str(),
+ val_x0, val_y0, val_x1, val_y1 , val_x2, val_y2 );
+#endif
+}
+
+
+
+void writeBlockStatisticsHeader(const SPS *sps)
+{
+ static bool has_header_been_written = false;
+ if (has_header_been_written)
+ {
+ return;
+ }
+
+ // only write header when block statistics are used
+ bool write_blockstatistics = g_trace_ctx->isChannelActive( D_BLOCK_STATISTICS_ALL) || g_trace_ctx->isChannelActive( D_BLOCK_STATISTICS_CODED);
+ if(!write_blockstatistics)
+ {
+ return;
+ }
+
+ DTRACE_HEADER( g_trace_ctx, "# VTMBMS Block Statistics\n");
+ // sequence info
+ DTRACE_HEADER( g_trace_ctx, "# Sequence size: [%dx %d]\n", sps->getPicWidthInLumaSamples(), sps->getPicHeightInLumaSamples());
+ // list statistics
+ for( auto i = static_cast<int>(BlockStatistic::PredMode); i < static_cast<int>(BlockStatistic::NumBlockStatistics); i++)
+ {
+ BlockStatistic statistic = BlockStatistic(i);
+ std::string statitic_name = GetBlockStatisticName(statistic);
+ std::string statitic_type = GetBlockStatisticTypeString(statistic);
+ std::string statitic_type_specific_info = GetBlockStatisticTypeSpecificInfo(statistic);
+ DTRACE_HEADER( g_trace_ctx, "# Block Statistic Type: %s; %s; %s\n", statitic_name.c_str(), statitic_type.c_str(), statitic_type_specific_info.c_str());
+ }
+
+ has_header_been_written = true;
+}
+
+void getAndStoreBlockStatistics(const CodingStructure& cs, const UnitArea& ctuArea)
+{
+ // two differemt behaviors, depending on which information is needed
+ bool writeAll = g_trace_ctx->isChannelActive( D_BLOCK_STATISTICS_ALL);
+ bool writeCoded = g_trace_ctx->isChannelActive( D_BLOCK_STATISTICS_CODED);
+
+ CHECK(writeAll && writeCoded, "Either used D_BLOCK_STATISTICS_ALL_DATA or D_BLOCK_STATISTICS_CODED_DATA. Not both at once!")
+
+ if (writeCoded)
+ writeAllCodedData(cs, ctuArea); // this will write out important cu-based data, only if it is actually decoded and used
+ else if (writeAll)
+ writeAllData(cs, ctuArea); // this will write out all inter- or intra-prediction related data
+}
+
+void writeAllData(const CodingStructure& cs, const UnitArea& ctuArea)
+{
+ const int maxNumChannelType = cs.pcv->chrFormat != CHROMA_400 && CS::isDualITree( cs ) ? 2 : 1;
+
+ for( int ch = 0; ch < maxNumChannelType; ch++ )
+ {
+ const ChannelType chType = ChannelType( ch );
+
+ for( const CodingUnit &cu : cs.traverseCUs( CS::getArea( cs, ctuArea, chType ), chType ) )
+ {
+ if( chType == CHANNEL_TYPE_LUMA )
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::PredMode), cu.predMode);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::PartSize), cu.partSize);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::Depth), cu.depth);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::QT_Depth), cu.qtDepth);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::BT_Depth), cu.btDepth);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::MT_Depth), cu.mtDepth);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::ChromaQPAdj), cu.chromaQpAdj);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::QP), cu.qp);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::SplitSeries), (int)cu.splitSeries);
+
+ if (cs.pps->getTransquantBypassEnabledFlag())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::TransQuantBypassFlag), cu.transQuantBypass);
+ }
+
+ // skip flag
+ if (!cs.slice->isIntra())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::SkipFlag), cu.skip);
+ }
+
+ #if JVET_K1000_SIMPLIFIED_EMT && HM_EMT_NSST_AS_IN_JEM
+ if (!(!((cs.sps->getSpsNext().getUseIntraEMT() && CU::isIntra(cu)) || (cs.sps->getSpsNext().getUseInterEMT() && CU::isInter(cu))) || isChroma(cu.chType)))
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::EMTFlag), cu.emtFlag);
+ }
+ #endif
+ }
+ else if( chType == CHANNEL_TYPE_CHROMA )
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::PartSize_Chroma), cu.partSize);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::Depth_Chroma), cu.depth);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::QT_Depth_Chroma), cu.qtDepth);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::BT_Depth_Chroma), cu.btDepth);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::MT_Depth_Chroma), cu.mtDepth);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::ChromaQPAdj_Chroma), cu.chromaQpAdj);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::QP_Chroma), cu.qp);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::SplitSeries_Chroma), (int)cu.splitSeries);
+
+ if (cs.pps->getTransquantBypassEnabledFlag())
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::TransQuantBypassFlag_Chroma), cu.transQuantBypass);
+ }
+
+ #if JVET_K1000_SIMPLIFIED_EMT && HM_EMT_NSST_AS_IN_JEM
+ if (!(!((cs.sps->getSpsNext().getUseIntraEMT() && CU::isIntra(cu)) || (cs.sps->getSpsNext().getUseInterEMT() && CU::isInter(cu))) || isChroma(cu.chType)))
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::EMTFlag_Chroma), cu.emtFlag);
+ }
+ #endif
+ }
+
+
+ switch( cu.predMode )
+ {
+ case MODE_INTER:
+ {
+ for( const PredictionUnit &pu : CU::traversePUs( cu ) )
+ {
+ if (!pu.cu->skip)
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MergeFlag), pu.mergeFlag);
+ }
+ if( pu.mergeFlag )
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MergeIdx), pu.mergeIdx);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MergeType), pu.mergeType);
+ }
+#if JVET_K_AFFINE
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::AffineFlag), pu.cu->affine);
+#if JVET_K0337_AFFINE_6PARA
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::AffineType), pu.cu->affineType);
+#endif
+#endif
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::InterDir), pu.interDir);
+
+ if (pu.interDir != 2 /* PRED_L1 */)
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MVPIdxL0), pu.mvpIdx[REF_PIC_LIST_0]);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::RefIdxL0), pu.refIdx[REF_PIC_LIST_0]);
+ }
+ if (pu.interDir != 1 /* PRED_L1 */)
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MVPIdxL1), pu.mvpIdx[REF_PIC_LIST_1]);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::RefIdxL1), pu.refIdx[REF_PIC_LIST_1]);
+ }
+#if JVET_K_AFFINE
+ if (!pu.cu->affine)
+ {
+#endif
+ if (pu.interDir != 2 /* PRED_L1 */)
+ {
+ Mv mv = pu.mv[REF_PIC_LIST_0];
+ Mv mvd = pu.mvd[REF_PIC_LIST_0];
+#if JVET_K0346 || JVET_K_AFFINE
+ mv.setLowPrec();
+ mvd.setLowPrec();
+#endif
+ DTRACE_BLOCK_VECTOR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MVDL0), mvd.hor, mvd.ver);
+ DTRACE_BLOCK_VECTOR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MVL0), mv.hor, mv.ver);
+ }
+ if (pu.interDir != 1 /* PRED_L1 */)
+ {
+ Mv mv = pu.mv[REF_PIC_LIST_1];
+ Mv mvd = pu.mvd[REF_PIC_LIST_1];
+#if JVET_K0346 || JVET_K_AFFINE
+ mv.setLowPrec();
+ mvd.setLowPrec();
+#endif
+ DTRACE_BLOCK_VECTOR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MVDL1), mvd.hor, mvd.ver);
+ DTRACE_BLOCK_VECTOR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::MVL1), mv.hor, mv.ver);
+ }
+#if JVET_K_AFFINE
+ }
+ else
+ {
+ if (pu.interDir != 2 /* PRED_L1 */)
+ {
+ Mv mv[3];
+ const CMotionBuf &mb = pu.getMotionBuf();
+ mv[0] = mb.at(0, 0).mv[REF_PIC_LIST_0];
+ mv[1] = mb.at(mb.width - 1, 0).mv[REF_PIC_LIST_0];
+ mv[2] = mb.at(0, mb.height - 1).mv[REF_PIC_LIST_0];
+#if JVET_K0346 || JVET_K_AFFINE
+ // motion vectors should use low precision or they will appear to large
+ mv[0].setLowPrec();
+ mv[1].setLowPrec();
+ mv[2].setLowPrec();
+#endif
+ DTRACE_BLOCK_AFFINETF(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::AffineMVL0), mv[0].hor, mv[0].ver, mv[1].hor, mv[1].ver, mv[2].hor, mv[2].ver);
+ }
+ if (pu.interDir != 1 /* PRED_L1 */)
+ {
+ Mv mv[3];
+ const CMotionBuf &mb = pu.getMotionBuf();
+ mv[0] = mb.at(0, 0).mv[REF_PIC_LIST_1];
+ mv[1] = mb.at(mb.width - 1, 0).mv[REF_PIC_LIST_1];
+ mv[2] = mb.at(0, mb.height - 1).mv[REF_PIC_LIST_1];
+#if JVET_K0346 || JVET_K_AFFINE
+ // motion vectors should use low precision or they will appear to large
+ mv[0].setLowPrec();
+ mv[1].setLowPrec();
+ mv[2].setLowPrec();
+#endif
+ DTRACE_BLOCK_AFFINETF(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::AffineMVL1), mv[0].hor, mv[0].ver, mv[1].hor, mv[1].ver, mv[2].hor, mv[2].ver);
+ }
+ }
+#endif
+ }
+#if JVET_K0357_AMVR
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::IMVMode), cu.imv);
+#endif
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::RootCbf), cu.rootCbf);
+ }
+ break;
+ case MODE_INTRA:
+ {
+
+ if(chType == CHANNEL_TYPE_LUMA)
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::IPCM), cu.ipcm);
+ }
+ else if(chType == CHANNEL_TYPE_CHROMA)
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::IPCM_Chroma), cu.ipcm);
+ }
+
+ const uint32_t numChType = ::getNumberValidChannels( cu.chromaFormat );
+
+ for( uint32_t chType = CHANNEL_TYPE_LUMA; chType < numChType; chType++ )
+ {
+ if( cu.blocks[chType].valid() )
+ {
+ for( const PredictionUnit &pu : CU::traversePUs( cu ) )
+ {
+ if( isLuma( ChannelType( chType ) ) )
+ {
+ const uint32_t uiChFinalMode = PU::getFinalIntraMode( pu, ChannelType( chType ) );
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::Luma_IntraMode), uiChFinalMode);
+ }
+ else
+ {
+ const uint32_t uiChFinalMode = PU::getFinalIntraMode( pu, ChannelType( chType ) );
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, pu, GetBlockStatisticName(BlockStatistic::Chroma_IntraMode), uiChFinalMode);
+ #if ENABLE_CHROMA_422
+ assert(0);
+ #endif
+ }
+ }
+ }
+ }
+ }
+ break;
+ default:
+ THROW( "Invalid prediction mode" );
+ break;
+ }
+
+ for (const TransformUnit &tu : CU::traverseTUs(cu))
+ {
+ if (tu.Y().valid())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, tu, GetBlockStatisticName(BlockStatistic::Cbf_Y), tu.cbf[COMPONENT_Y]);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, tu, GetBlockStatisticName(BlockStatistic::TransformSkipFlag_Y), tu.transformSkip[COMPONENT_Y]);
+ }
+ if (!(cu.chromaFormat == CHROMA_400 || (CS::isDualITree(*cu.cs) && cu.chType == CHANNEL_TYPE_LUMA)))
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, tu, GetBlockStatisticName(BlockStatistic::Cbf_Cb), tu.cbf[COMPONENT_Cb]);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, tu, GetBlockStatisticName(BlockStatistic::Cbf_Cr), tu.cbf[COMPONENT_Cr]);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, tu, GetBlockStatisticName(BlockStatistic::TransformSkipFlag_Cb), tu.transformSkip[COMPONENT_Cb]);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_ALL, tu, GetBlockStatisticName(BlockStatistic::TransformSkipFlag_Cr), tu.transformSkip[COMPONENT_Cr]);
+ }
+ }
+ }
+ }
+}
+
+void writeAllCodedData(const CodingStructure & cs, const UnitArea & ctuArea)
+{
+ const int maxNumChannelType = cs.pcv->chrFormat != CHROMA_400 && CS::isDualITree(cs) ? 2 : 1;
+
+ for (int ch = 0; ch < maxNumChannelType; ch++)
+ {
+ const ChannelType chType = ChannelType(ch);
+ const SPS& sps = *cs.sps;
+
+ for (const CodingUnit &cu : cs.traverseCUs(CS::getArea(cs, ctuArea, chType), chType))
+ {
+ if( chType == CHANNEL_TYPE_LUMA )
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::PartSize), cu.partSize);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::Depth), cu.depth);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::QT_Depth), cu.qtDepth);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::BT_Depth), cu.btDepth);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::MT_Depth), cu.mtDepth);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::ChromaQPAdj), cu.chromaQpAdj);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::QP), cu.qp);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::SplitSeries), (int)cu.splitSeries);
+ // transquant bypass flag
+ if (cs.pps->getTransquantBypassEnabledFlag())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::TransQuantBypassFlag), cu.transQuantBypass);
+ }
+ // skip flag
+ if (!cs.slice->isIntra())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::SkipFlag), cu.skip);
+ }
+
+ // prediction mode and partitioning data
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::PredMode), cu.predMode);
+
+ if (CU::isIntra(cu) && cu.partSize == SIZE_2Nx2N)
+ {
+ if (!(!sps.getUsePCM() || cu.lumaSize().width > (1 << sps.getPCMLog2MaxSize()) || cu.lumaSize().width < (1 << sps.getPCMLog2MinSize())))
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::IPCM), cu.ipcm);
+ }
+ }
+ }
+ else if (chType == CHANNEL_TYPE_CHROMA )
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::PartSize_Chroma), cu.partSize);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::Depth_Chroma), cu.depth);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::QT_Depth_Chroma), cu.qtDepth);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::BT_Depth_Chroma), cu.btDepth);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::MT_Depth_Chroma), cu.mtDepth);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::ChromaQPAdj_Chroma), cu.chromaQpAdj);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::QP_Chroma), cu.qp);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::SplitSeries_Chroma), (int)cu.splitSeries);
+ // transquant bypass flag
+ if (cs.pps->getTransquantBypassEnabledFlag())
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::TransQuantBypassFlag_Chroma), cu.transQuantBypass);
+ }
+
+ if (CU::isIntra(cu) && cu.partSize == SIZE_2Nx2N)
+ {
+ if (!(!sps.getUsePCM() || cu.lumaSize().width > (1 << sps.getPCMLog2MaxSize()) || cu.lumaSize().width < (1 << sps.getPCMLog2MinSize())))
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::IPCM_Chroma), cu.ipcm);
+ }
+ }
+ }
+
+ for (auto &pu : CU::traversePUs(cu))
+ {
+ switch (pu.cu->predMode)
+ {
+ case MODE_INTRA:
+ {
+ if (pu.Y().valid())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::Luma_IntraMode), PU::getFinalIntraMode(pu, ChannelType(chType)));
+ }
+ if (!(pu.chromaFormat == CHROMA_400 || (CS::isDualITree(*pu.cs) && pu.chType == CHANNEL_TYPE_LUMA)))
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::Chroma_IntraMode), PU::getFinalIntraMode(pu, CHANNEL_TYPE_CHROMA));
+ }
+ break;
+ }
+ case MODE_INTER:
+ {
+ if (!pu.cu->skip)
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MergeFlag), pu.mergeFlag);
+ }
+ if (pu.mergeFlag)
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MergeType), pu.mergeType);
+ #if JVET_K_AFFINE
+ if (!(cu.cs->slice->isIntra() || !cu.cs->sps->getSpsNext().getUseAffine() || cu.partSize != SIZE_2Nx2N)
+ && !(!cu.firstPU->mergeFlag && !(cu.lumaSize().width > 8 && cu.lumaSize().height > 8))
+ && !(cu.firstPU->mergeFlag && !PU::isAffineMrgFlagCoded(*cu.firstPU)))
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::AffineFlag), pu.cu->affine);
+ if (cu.affine && !cu.firstPU->mergeFlag && cu.cs->sps->getSpsNext().getUseAffineType())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::AffineType), pu.cu->affineType);
+ }
+ }
+ #endif
+ #if JVET_K_AFFINE
+ if (!(pu.cu->affine))
+ #endif
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MergeIdx), pu.mergeIdx);
+ }
+ }
+ else
+ {
+ if (!pu.cs->slice->isInterP())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::InterDir), pu.interDir);
+ }
+ #if JVET_K_AFFINE
+ if (!(cu.cs->slice->isIntra() || !cu.cs->sps->getSpsNext().getUseAffine() || cu.partSize != SIZE_2Nx2N)
+ && !(!cu.firstPU->mergeFlag && !(cu.lumaSize().width > 8 && cu.lumaSize().height > 8))
+ && !(cu.firstPU->mergeFlag && !PU::isAffineMrgFlagCoded(*cu.firstPU)))
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::AffineFlag), pu.cu->affine);
+ if (cu.affine && !cu.firstPU->mergeFlag && cu.cs->sps->getSpsNext().getUseAffineType())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::AffineType), pu.cu->affineType);
+ }
+ }
+ #endif
+ }
+ if (pu.interDir != 2 /* PRED_L1 */)
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MVPIdxL0), pu.mvpIdx[REF_PIC_LIST_0]);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::RefIdxL0), pu.refIdx[REF_PIC_LIST_0]);
+ }
+ if (pu.interDir != 1 /* PRED_L1 */)
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MVPIdxL1), pu.mvpIdx[REF_PIC_LIST_1]);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::RefIdxL1), pu.refIdx[REF_PIC_LIST_1]);
+ }
+
+ #if JVET_K_AFFINE
+ if (!pu.cu->affine)
+ {
+ #endif
+ if (pu.interDir != 2 /* PRED_L1 */)
+ {
+ Mv mv = pu.mv[REF_PIC_LIST_0];
+ Mv mvd = pu.mvd[REF_PIC_LIST_0];
+ #if JVET_K0346 || JVET_K_AFFINE
+ mv.setLowPrec();
+ mvd.setLowPrec();
+ #endif
+ DTRACE_BLOCK_VECTOR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MVDL0), mvd.hor, mvd.ver);
+ DTRACE_BLOCK_VECTOR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MVL0), mv.hor, mv.ver);
+ }
+ if (pu.interDir != 1 /* PRED_L1 */)
+ {
+ Mv mv = pu.mv[REF_PIC_LIST_1];
+ Mv mvd = pu.mvd[REF_PIC_LIST_1];
+ #if JVET_K0346 || JVET_K_AFFINE
+ mv.setLowPrec();
+ mvd.setLowPrec();
+ #endif
+ DTRACE_BLOCK_VECTOR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MVDL0), mvd.hor, mvd.ver);
+ DTRACE_BLOCK_VECTOR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::MVL1), mv.hor, mv.ver);
+ }
+ #if JVET_K_AFFINE
+ }
+ else
+ {
+ if (pu.interDir != 2 /* PRED_L1 */)
+ {
+ Mv mv[3];
+ const CMotionBuf &mb = pu.getMotionBuf();
+ mv[0] = mb.at(0, 0).mv[REF_PIC_LIST_0];
+ mv[1] = mb.at(mb.width - 1, 0).mv[REF_PIC_LIST_0];
+ mv[2] = mb.at(0, mb.height - 1).mv[REF_PIC_LIST_0];
+ #if JVET_K0346 || JVET_K_AFFINE
+ // motion vectors should use low precision or they will appear to large
+ mv[0].setLowPrec();
+ mv[1].setLowPrec();
+ mv[2].setLowPrec();
+ #endif
+ DTRACE_BLOCK_AFFINETF(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::AffineMVL0), mv[0].hor, mv[0].ver, mv[1].hor, mv[1].ver, mv[2].hor, mv[2].ver);
+ }
+ if (pu.interDir != 1 /* PRED_L1 */)
+ {
+ Mv mv[3];
+ const CMotionBuf &mb = pu.getMotionBuf();
+ mv[0] = mb.at(0, 0).mv[REF_PIC_LIST_1];
+ mv[1] = mb.at(mb.width - 1, 0).mv[REF_PIC_LIST_1];
+ mv[2] = mb.at(0, mb.height - 1).mv[REF_PIC_LIST_1];
+ #if JVET_K0346 || JVET_K_AFFINE
+ // motion vectors should use low precision or they will appear to large
+ mv[0].setLowPrec();
+ mv[1].setLowPrec();
+ mv[2].setLowPrec();
+ #endif
+ DTRACE_BLOCK_AFFINETF(g_trace_ctx, D_BLOCK_STATISTICS_CODED, pu, GetBlockStatisticName(BlockStatistic::AffineMVL1), mv[0].hor, mv[0].ver, mv[1].hor, mv[1].ver, mv[2].hor, mv[2].ver);
+ }
+ }
+ #endif
+ #if JVET_K0357_AMVR
+ if (cu.cs->sps->getSpsNext().getUseIMV() && CU::hasSubCUNonZeroMVd(cu))
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::IMVMode), cu.imv);
+ }
+ #endif
+ break;
+ }
+ default:
+ {
+ CHECK(1, "Invalid prediction mode");
+ break;
+ }
+ }
+ } // end pu
+ if (CU::isInter(cu))
+ {
+ const PredictionUnit &pu = *cu.firstPU;
+ if (!((cu.cs->pcv->noRQT || cu.partSize == SIZE_2Nx2N) && pu.mergeFlag))
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::RootCbf), cu.rootCbf);
+ }
+ }
+ if (cu.rootCbf || CU::isIntra(cu))
+ {
+ for (const TransformUnit &tu : CU::traverseTUs(cu))
+ {
+ if (tu.Y().valid())
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, tu, GetBlockStatisticName(BlockStatistic::Cbf_Y), tu.cbf[COMPONENT_Y]);
+#if HM_EMT_NSST_AS_IN_JEM && JVET_K1000_SIMPLIFIED_EMT
+ if (!(!tu.cu->cs->pps->getUseTransformSkip() || tu.cu->transQuantBypass || !TU::hasTransformSkipFlag(*tu.cs, tu.blocks[COMPONENT_Y]) || (isLuma(COMPONENT_Y) && tu.cu->emtFlag)))
+#else
+ if (!(!tu.cu->cs->pps->getUseTransformSkip() || tu.cu->transQuantBypass || !TU::hasTransformSkipFlag(*tu.cs, tu.blocks[COMPONENT_Y])))
+#endif
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, tu, GetBlockStatisticName(BlockStatistic::TransformSkipFlag_Y), tu.transformSkip[COMPONENT_Y]);
+ }
+ }
+ if (!(cu.chromaFormat == CHROMA_400 || (CS::isDualITree(*cu.cs) && cu.chType == CHANNEL_TYPE_LUMA)))
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, tu, GetBlockStatisticName(BlockStatistic::Cbf_Cb), tu.cbf[COMPONENT_Cb]);
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, tu, GetBlockStatisticName(BlockStatistic::Cbf_Cr), tu.cbf[COMPONENT_Cr]);
+#if HM_EMT_NSST_AS_IN_JEM && JVET_K1000_SIMPLIFIED_EMT
+ if (!(!tu.cu->cs->pps->getUseTransformSkip() || tu.cu->transQuantBypass || !TU::hasTransformSkipFlag(*tu.cs, tu.blocks[COMPONENT_Cb]) || (isLuma(COMPONENT_Cb) && tu.cu->emtFlag)))
+#else
+ if (!(!tu.cu->cs->pps->getUseTransformSkip() || tu.cu->transQuantBypass || !TU::hasTransformSkipFlag(*tu.cs, tu.blocks[COMPONENT_Cb])))
+#endif
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, tu, GetBlockStatisticName(BlockStatistic::TransformSkipFlag_Cb), tu.transformSkip[COMPONENT_Cb]);
+ }
+#if HM_EMT_NSST_AS_IN_JEM && JVET_K1000_SIMPLIFIED_EMT
+ if (!(!tu.cu->cs->pps->getUseTransformSkip() || tu.cu->transQuantBypass || !TU::hasTransformSkipFlag(*tu.cs, tu.blocks[COMPONENT_Cr]) || (isLuma(COMPONENT_Cr) && tu.cu->emtFlag)))
+#else
+ if (!(!tu.cu->cs->pps->getUseTransformSkip() || tu.cu->transQuantBypass || !TU::hasTransformSkipFlag(*tu.cs, tu.blocks[COMPONENT_Cr])))
+#endif
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, tu, GetBlockStatisticName(BlockStatistic::TransformSkipFlag_Cr), tu.transformSkip[COMPONENT_Cr]);
+ }
+ }
+ }
+ }
+#if JVET_K1000_SIMPLIFIED_EMT && HM_EMT_NSST_AS_IN_JEM
+ if (!(!((cs.sps->getSpsNext().getUseIntraEMT() && CU::isIntra(cu)) || (cs.sps->getSpsNext().getUseInterEMT() && CU::isInter(cu))) || isChroma(cu.chType)))
+ {
+ if( isLuma( ChannelType( chType ) ) )
+ {
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::EMTFlag), cu.emtFlag);
+ }
+ else
+ {
+ DTRACE_BLOCK_SCALAR_CHROMA(g_trace_ctx, D_BLOCK_STATISTICS_CODED, cu, GetBlockStatisticName(BlockStatistic::EMTFlag_Chroma), cu.emtFlag);
+ }
+ }
+#endif
+ }
+ }
+}
+#endif
diff --git a/source/Lib/CommonLib/dtrace_blockstatistics.h b/source/Lib/CommonLib/dtrace_blockstatistics.h
new file mode 100644
index 0000000000000000000000000000000000000000..a81b03a27f258971585def1a1fa174e0e5197393
--- /dev/null
+++ b/source/Lib/CommonLib/dtrace_blockstatistics.h
@@ -0,0 +1,208 @@
+/* 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) 2010-2018, 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 dtrace_blockstatistics.h
+ * \brief DTrace block statistcis support for next software
+ */
+
+#ifndef _DTRACE_BLOCKSTATISTICS_H_
+#define _DTRACE_BLOCKSTATISTICS_H_
+
+#include <map>
+#include "CommonLib/CommonDef.h"
+#include "CommonLib/Unit.h"
+
+#if K0149_BLOCK_STATISTICS
+#define DTRACE_HEADER(ctx,...) ctx->dtrace_header( __VA_ARGS__ )
+#define DTRACE_BLOCK_SCALAR(ctx,channel,cs_cu_pu,stat_type,val) ctx->dtrace_block_scalar( channel, cs_cu_pu, stat_type, val )
+#define DTRACE_BLOCK_SCALAR_CHROMA(ctx,channel,cs_cu_pu,stat_type,val) ctx->dtrace_block_scalar( channel, cs_cu_pu, stat_type, val, true)
+#define DTRACE_BLOCK_VECTOR(ctx,channel,cu_pu,stat_type,v_x,v_y) ctx->dtrace_block_vector( channel, cu_pu, stat_type, v_x, v_y )
+#define DTRACE_BLOCK_AFFINETF(ctx,channel,pu,stat_type,v_x0,v_y0,v_x1,v_y1,v_x2,v_y2) ctx->dtrace_block_affinetf( channel, pu, stat_type, v_x0, v_y0, v_x1, v_y1, v_x2, v_y2 )
+
+enum class BlockStatistic {
+ // general
+ PredMode,
+ PartSize,
+ Depth,
+ QT_Depth,
+ BT_Depth,
+ MT_Depth,
+ ChromaQPAdj,
+ QP,
+ SplitSeries,
+ TransQuantBypassFlag,
+#if JVET_K1000_SIMPLIFIED_EMT
+ EMTFlag,
+#endif
+ TransformSkipFlag_Y,
+ TransformSkipFlag_Cb,
+ TransformSkipFlag_Cr,
+
+ // intra
+ IPCM,
+ Luma_IntraMode,
+ Chroma_IntraMode,
+ // inter
+ SkipFlag,
+ RootCbf,
+ Cbf_Y,
+ Cbf_Cb,
+ Cbf_Cr,
+#if JVET_K0357_AMVR
+ IMVMode,
+#endif
+ InterDir,
+ MergeFlag,
+ MergeIdx,
+ MergeType,
+ MVPIdxL0,
+ MVPIdxL1,
+ MVL0,
+ MVL1,
+ MVDL0,
+ MVDL1,
+ RefIdxL0,
+ RefIdxL1,
+#if JVET_K_AFFINE
+ AffineFlag,
+ AffineMVL0,
+ AffineMVL1,
+#if JVET_K0337_AFFINE_6PARA
+ AffineType,
+#endif
+#endif
+
+// for dual tree
+ // general
+ PartSize_Chroma,
+ Depth_Chroma,
+ QT_Depth_Chroma,
+ BT_Depth_Chroma,
+ MT_Depth_Chroma,
+ ChromaQPAdj_Chroma,
+ QP_Chroma,
+ SplitSeries_Chroma,
+ TransQuantBypassFlag_Chroma,
+
+ // intra
+ IPCM_Chroma,
+
+ NumBlockStatistics,
+};
+
+enum class BlockStatisticType {
+ Flag,
+ Vector,
+ Integer,
+ AffineTFVectors,
+};
+
+static const std::map<BlockStatistic, std::tuple<std::string, BlockStatisticType, std::string>> blockstatistic2description =
+{
+ // Statistics enum Statistics name string Statistic Type Type specific information:
+ // Value range, vector scale
+ { BlockStatistic::PredMode, std::tuple<std::string, BlockStatisticType, std::string>{"PredMode", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::MergeFlag, std::tuple<std::string, BlockStatisticType, std::string>{"MergeFlag", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::MVL0, std::tuple<std::string, BlockStatisticType, std::string>{"MVL0", BlockStatisticType::Vector, "Scale: 4"}},
+ { BlockStatistic::MVL1, std::tuple<std::string, BlockStatisticType, std::string>{"MVL1", BlockStatisticType::Vector, "Scale: 4"}},
+ { BlockStatistic::IPCM, std::tuple<std::string, BlockStatisticType, std::string>{"IPCM", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::Luma_IntraMode, std::tuple<std::string, BlockStatisticType, std::string>{"Luma_IntraMode", BlockStatisticType::Integer, "[0, " + std::to_string(NUM_INTRA_MODE) + "]"}},
+ { BlockStatistic::Chroma_IntraMode, std::tuple<std::string, BlockStatisticType, std::string>{"Chroma_IntraMode", BlockStatisticType::Integer, "[0, " + std::to_string(NUM_INTRA_MODE) + "]"}},
+ { BlockStatistic::SkipFlag, std::tuple<std::string, BlockStatisticType, std::string>{"SkipFlag", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::TransformSkipFlag_Y, std::tuple<std::string, BlockStatisticType, std::string>{"TransformSkipFlag_Y", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::TransformSkipFlag_Cb, std::tuple<std::string, BlockStatisticType, std::string>{"TransformSkipFlag_Cb", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::TransformSkipFlag_Cr, std::tuple<std::string, BlockStatisticType, std::string>{"TransformSkipFlag_Cr", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::PartSize, std::tuple<std::string, BlockStatisticType, std::string>{"PartSize", BlockStatisticType::Integer, "[0, " + std::to_string(NUMBER_OF_PART_SIZES) + "]"}},
+ { BlockStatistic::Depth, std::tuple<std::string, BlockStatisticType, std::string>{"Depth", BlockStatisticType::Integer, "[0, 7]"}},
+ { BlockStatistic::QT_Depth, std::tuple<std::string, BlockStatisticType, std::string>{"QT_Depth", BlockStatisticType::Integer, "[0, 7]"}},
+ { BlockStatistic::BT_Depth, std::tuple<std::string, BlockStatisticType, std::string>{"BT_Depth", BlockStatisticType::Integer, "[0, 7]"}},
+ { BlockStatistic::MT_Depth, std::tuple<std::string, BlockStatisticType, std::string>{"MT_Depth", BlockStatisticType::Integer, "[0, 7]"}},
+ { BlockStatistic::ChromaQPAdj, std::tuple<std::string, BlockStatisticType, std::string>{"ChromaQPAdj", BlockStatisticType::Integer, "[-10, 10]"}},
+ { BlockStatistic::QP, std::tuple<std::string, BlockStatisticType, std::string>{"QP", BlockStatisticType::Integer, "[0, 51]"}},
+ { BlockStatistic::SplitSeries, std::tuple<std::string, BlockStatisticType, std::string>{"SplitSeries", BlockStatisticType::Integer, "[0, " + std::to_string(std::numeric_limits<SplitSeries>::max()) + "]"}},
+ { BlockStatistic::RootCbf, std::tuple<std::string, BlockStatisticType, std::string>{"RootCbf", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::Cbf_Y, std::tuple<std::string, BlockStatisticType, std::string>{"Cbf_Y", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::Cbf_Cb, std::tuple<std::string, BlockStatisticType, std::string>{"Cbf_Cb", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::Cbf_Cr, std::tuple<std::string, BlockStatisticType, std::string>{"Cbf_Cr", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::TransQuantBypassFlag, std::tuple<std::string, BlockStatisticType, std::string>{"TransQuantBypassFlag", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::MergeIdx, std::tuple<std::string, BlockStatisticType, std::string>{"MergeIdx", BlockStatisticType::Integer, "[0, 7]"}},
+ { BlockStatistic::InterDir, std::tuple<std::string, BlockStatisticType, std::string>{"InterDir", BlockStatisticType::Integer, "[1, 3]"}},
+ { BlockStatistic::MergeType, std::tuple<std::string, BlockStatisticType, std::string>{"MergeType", BlockStatisticType::Integer, "[0, 2]"}},
+ { BlockStatistic::MVPIdxL0, std::tuple<std::string, BlockStatisticType, std::string>{"MVPIdxL0", BlockStatisticType::Integer, "[0, 1]"}},
+ { BlockStatistic::MVDL0, std::tuple<std::string, BlockStatisticType, std::string>{"MVDL0", BlockStatisticType::Vector, "Scale: 4"}},
+ { BlockStatistic::RefIdxL0, std::tuple<std::string, BlockStatisticType, std::string>{"RefIdxL0", BlockStatisticType::Integer, "[0, 4]"}},
+ { BlockStatistic::MVPIdxL1, std::tuple<std::string, BlockStatisticType, std::string>{"MVPIdxL1", BlockStatisticType::Integer, "[0, 1]"}},
+ { BlockStatistic::MVDL1, std::tuple<std::string, BlockStatisticType, std::string>{"MVDL1", BlockStatisticType::Vector, "Scale: 4"}},
+ { BlockStatistic::RefIdxL1, std::tuple<std::string, BlockStatisticType, std::string>{"RefIdxL1", BlockStatisticType::Integer, "[0, 4]"}},
+#if JVET_K0357_AMVR
+ { BlockStatistic::IMVMode, std::tuple<std::string, BlockStatisticType, std::string>{"IMVMode", BlockStatisticType::Integer, "[0, 2]"}},
+#endif
+#if JVET_K_AFFINE
+ { BlockStatistic::AffineFlag, std::tuple<std::string, BlockStatisticType, std::string>{"AffineFlag", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::AffineMVL0, std::tuple<std::string, BlockStatisticType, std::string>{"AffineMVL0", BlockStatisticType::AffineTFVectors, "Scale: 4"}},
+ { BlockStatistic::AffineMVL1, std::tuple<std::string, BlockStatisticType, std::string>{"AffineMVL1", BlockStatisticType::AffineTFVectors, "Scale: 4"}},
+#if JVET_K0337_AFFINE_6PARA
+ { BlockStatistic::AffineType, std::tuple<std::string, BlockStatisticType, std::string>{"AffineType", BlockStatisticType::Flag, ""} },
+#endif
+#endif
+#if JVET_K1000_SIMPLIFIED_EMT
+ { BlockStatistic::EMTFlag, std::tuple<std::string, BlockStatisticType, std::string>{"EMTFlag", BlockStatisticType::Flag, ""}},
+#endif
+
+
+ // for dual tree
+ { BlockStatistic::PartSize_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"PartSize_Chroma", BlockStatisticType::Integer, "[0, " + std::to_string(NUMBER_OF_PART_SIZES) + "]"}},
+ { BlockStatistic::Depth_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"Depth_Chroma", BlockStatisticType::Integer, "[0, 10]"}}, // todo: actual limits?
+ { BlockStatistic::QT_Depth_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"QT_Depth_Chroma", BlockStatisticType::Integer, "[0, 10]"}}, // todo: actual limits?
+ { BlockStatistic::BT_Depth_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"BT_Depth_Chroma", BlockStatisticType::Integer, "[0, 10]"}}, // todo: actual limits?
+ { BlockStatistic::MT_Depth_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"MT_Depth_Chroma", BlockStatisticType::Integer, "[0, 10]"}}, // todo: actual limits?
+ { BlockStatistic::ChromaQPAdj_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"ChromaQPAdj_Chroma", BlockStatisticType::Integer, "[-10, 10]"}}, // todo: actual limits?
+ { BlockStatistic::QP_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"QP_Chroma", BlockStatisticType::Integer, "[0, 51]"}},
+ { BlockStatistic::SplitSeries_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"SplitSeries_Chroma", BlockStatisticType::Integer, "[0, " + std::to_string(std::numeric_limits<SplitSeries>::max()) + "]"}},
+ { BlockStatistic::TransQuantBypassFlag_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"TransQuantBypassFlag_Chroma", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::IPCM_Chroma, std::tuple<std::string, BlockStatisticType, std::string>{"IPCM_Chroma", BlockStatisticType::Flag, ""}},
+
+};
+
+
+std::string GetBlockStatisticName(BlockStatistic statistic);
+std::string GetBlockStatisticTypeString(BlockStatistic statistic);
+std::string GetBlockStatisticTypeSpecificInfo(BlockStatistic statistic);
+
+void writeBlockStatisticsHeader(const SPS *sps);
+void getAndStoreBlockStatistics(const CodingStructure& cs, const UnitArea& ctuArea);
+void writeAllData(const CodingStructure& cs, const UnitArea& ctuArea);
+void writeAllCodedData(const CodingStructure& cs, const UnitArea& ctuArea);
+#endif
+
+#endif // _DTRACE_BLOCKSTATISTICS_H_
diff --git a/source/Lib/CommonLib/dtrace_next.h b/source/Lib/CommonLib/dtrace_next.h
index 89f214cc9c5d7dbeda5b15b3168f266f4e4a623a..88d6249f118683fcd18bbbee5935f68b79e943b0 100644
--- a/source/Lib/CommonLib/dtrace_next.h
+++ b/source/Lib/CommonLib/dtrace_next.h
@@ -145,8 +145,12 @@ enum DTRACE_CHANNEL
D_RDOQ_COST,
D_TMP,
D_CRC
+#if K0149_BLOCK_STATISTICS
+ ,
+ D_BLOCK_STATISTICS_ALL,
+ D_BLOCK_STATISTICS_CODED,
+#endif
};
-
#define _CNL_DEF(_s) {_s,(std::string(#_s))}
inline void tracing_uninit( CDTrace *pDtrace )
@@ -245,6 +249,11 @@ inline CDTrace* tracing_init( std::string& sTracingFile, std::string& sTracingRu
_CNL_DEF( D_RDOQ_COST ),
_CNL_DEF( D_TMP ),
_CNL_DEF( D_CRC )
+ #if K0149_BLOCK_STATISTICS
+ ,
+ _CNL_DEF( D_BLOCK_STATISTICS_ALL ),
+ _CNL_DEF( D_BLOCK_STATISTICS_CODED ),
+ #endif
};
dtrace_channels_t channels( next_channels, &next_channels[sizeof( next_channels ) / sizeof( next_channels[0] )] );
diff --git a/source/Lib/CommonLib/x86/InitX86.cpp b/source/Lib/CommonLib/x86/InitX86.cpp
index 47d109ea0e678a28b5e998c2b58363fea9dbdbb6..d94ede594c582b7bd6338a17f9203ff60b483451 100644
--- a/source/Lib/CommonLib/x86/InitX86.cpp
+++ b/source/Lib/CommonLib/x86/InitX86.cpp
@@ -169,5 +169,24 @@ void AdaptiveLoopFilter::initAdaptiveLoopFilterX86()
}
#endif
+#if ENABLE_SIMD_OPT_CPR
+void IbcHashMap::initIbcHashMapX86()
+{
+ auto vext = read_x86_extension_flags();
+ switch (vext)
+ {
+ case AVX512:
+ case AVX2:
+ case AVX:
+ case SSE42:
+ _initIbcHashMapX86<SSE42>();
+ break;
+ case SSE41:
+ default:
+ break;
+ }
+}
+#endif
+
#endif
diff --git a/source/Lib/DecoderAnalyserLib/CMakeLists.txt b/source/Lib/DecoderAnalyserLib/CMakeLists.txt
index b26321590b94b8c2eaac751d7ed8dcf980bdeedc..9b0017617cd6af649ccf69129f8205355165e636 100644
--- a/source/Lib/DecoderAnalyserLib/CMakeLists.txt
+++ b/source/Lib/DecoderAnalyserLib/CMakeLists.txt
@@ -18,7 +18,7 @@ target_compile_definitions( ${LIB_NAME} PUBLIC RExt__DECODER_DEBUG_BIT_STATISTIC
target_compile_definitions( ${LIB_NAME} PUBLIC RExt__DECODER_DEBUG_TOOL_STATISTICS=1 )
if( ENABLE_VTM )
- target_compile_definitions( ${LIB_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${LIB_NAME} PUBLIC BMS_TOOLS=0 )
endif()
if( EXTENSION_360_VIDEO )
diff --git a/source/Lib/DecoderLib/CABACReader.cpp b/source/Lib/DecoderLib/CABACReader.cpp
index 16908fa548e848a3c068886d6e637c4a0fbf8c48..91eba95ce14e8ccdce8de247ef665c4f2c5c5158 100644
--- a/source/Lib/DecoderLib/CABACReader.cpp
+++ b/source/Lib/DecoderLib/CABACReader.cpp
@@ -180,7 +180,7 @@ bool CABACReader::coding_tree_unit( CodingStructure& cs, const UnitArea& area, i
ctx += leftCTUAddr > -1 ? ( ctbAlfFlag[leftCTUAddr] ? 1 : 0 ) : 0;
ctx += aboveCTUAddr > -1 ? ( ctbAlfFlag[aboveCTUAddr] ? 1 : 0 ) : 0;
- if( alfSliceParam.chromaCtbPresentFlag && compIdx )
+ if( compIdx && alfSliceParam.chromaCtbPresentFlag )
{
ctbAlfFlag[ctuRsAddr] = 1;
}
@@ -703,7 +703,6 @@ bool CABACReader::split_cu_flag( CodingStructure& cs, Partitioner &partitioner )
bool CABACReader::coding_unit( CodingUnit &cu, Partitioner &partitioner, CUCtx& cuCtx )
{
CodingStructure& cs = *cu.cs;
-
// transquant bypass flag
if( cs.pps->getTransquantBypassEnabledFlag() )
{
@@ -802,7 +801,7 @@ void CABACReader::imv_mode( CodingUnit& cu, MergeCtx& mrgCtx )
unsigned value = 0;
unsigned ctxId = DeriveCtx::CtxIMVFlag( cu );
- value = m_BinDecoder.decodeBin( Ctx::ImvFlag( ctxId ) );
+ value = m_BinDecoder.decodeBin( Ctx::ImvFlag( ctxId ) );
DTRACE( g_trace_ctx, D_SYNTAX, "imv_mode() value=%d ctx=%d\n", value, ctxId );
if( spsNext.getImvMode() == IMV_4PEL && value )
@@ -852,7 +851,6 @@ void CABACReader::cu_pred_data( CodingUnit &cu )
intra_chroma_pred_modes( cu );
return;
}
-
MergeCtx mrgCtx;
for( auto &pu : CU::traversePUs( cu ) )
@@ -863,6 +861,7 @@ void CABACReader::cu_pred_data( CodingUnit &cu )
#if JVET_K0357_AMVR
imv_mode ( cu, mrgCtx );
#endif
+
}
diff --git a/source/Lib/DecoderLib/CMakeLists.txt b/source/Lib/DecoderLib/CMakeLists.txt
index 0ec46167283c09b10e12bbfd98ee9adabec5598a..62413e9004538890a6c4506006be8d78ea6c6959 100644
--- a/source/Lib/DecoderLib/CMakeLists.txt
+++ b/source/Lib/DecoderLib/CMakeLists.txt
@@ -17,7 +17,7 @@ add_library( ${LIB_NAME} STATIC ${SRC_FILES} ${INC_FILES} ${NATVIS_FILES} )
target_compile_definitions( ${LIB_NAME} PUBLIC )
if( ENABLE_VTM )
- target_compile_definitions( ${LIB_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${LIB_NAME} PUBLIC BMS_TOOLS=0 )
endif()
if( EXTENSION_360_VIDEO )
diff --git a/source/Lib/DecoderLib/DecCu.cpp b/source/Lib/DecoderLib/DecCu.cpp
index 370f41f93afabe3f8a81491ee1afa5cb68241f47..4c06ed241f2e349832f4efb4ac77e8089b43ad45 100644
--- a/source/Lib/DecoderLib/DecCu.cpp
+++ b/source/Lib/DecoderLib/DecCu.cpp
@@ -48,6 +48,10 @@
#if RExt__DECODER_DEBUG_TOOL_STATISTICS
#include "CommonLib/CodingStatistics.h"
#endif
+#if K0149_BLOCK_STATISTICS
+#include "CommonLib/ChromaFormat.h"
+#include "CommonLib/dtrace_blockstatistics.h"
+#endif
//! \ingroup DecoderLib
//! \{
@@ -107,6 +111,9 @@ void DecCu::decompressCtu( CodingStructure& cs, const UnitArea& ctuArea )
DTRACE_BLOCK_REC( cs.picture->getRecoBuf( currCU ), currCU, currCU.predMode );
}
}
+#if K0149_BLOCK_STATISTICS
+ getAndStoreBlockStatistics(cs, ctuArea);
+#endif
}
// ====================================================================================================================
@@ -400,8 +407,13 @@ void DecCu::xDeriveCUMV( CodingUnit &cu )
MergeCtx mrgCtx;
#if RExt__DECODER_DEBUG_TOOL_STATISTICS
+#if JVET_K_AFFINE
+ if( pu.cu->affine )
+ {
+ CodingStatistics::IncrementStatisticTool( CodingStatisticsClassType{ STATS__TOOL_AFF, pu.Y().width, pu.Y().height } );
+ }
+#endif
#endif
-
if( pu.mergeFlag )
{
{
diff --git a/source/Lib/DecoderLib/DecLib.cpp b/source/Lib/DecoderLib/DecLib.cpp
index 511967e0710bec006bc64f227b1e392b230c1f0b..590a0525cd48af9fdef09e17267eaf305c5b5e91 100644
--- a/source/Lib/DecoderLib/DecLib.cpp
+++ b/source/Lib/DecoderLib/DecLib.cpp
@@ -48,6 +48,9 @@
#include <fcntl.h>
#include "AnnexBread.h"
#include "NALread.h"
+#if K0149_BLOCK_STATISTICS
+#include "CommonLib/dtrace_blockstatistics.h"
+#endif
#if RExt__DECODER_DEBUG_TOOL_STATISTICS
#include "CommonLib/CodingStatistics.h"
@@ -512,6 +515,10 @@ void DecLib::executeLoopFilters()
// deblocking filter
m_cLoopFilter.loopFilterPic( cs );
+#if DMVR_JVET_LOW_LATENCY_K0217
+ CS::setRefinedMotionField(cs);
+#endif
+
if( cs.sps->getUseSAO() )
{
m_cSAO.SAOProcess( cs, cs.picture->getSAO() );
@@ -898,6 +905,15 @@ bool DecLib::xDecodeSlice(InputNALUnit &nalu, int &iSkipFrame, int iPOCLastDispl
}
m_apcSlicePilot->setIndependentSliceIdx(uiIndependentSliceIdx);
+#if K0149_BLOCK_STATISTICS
+ PPS *pps = m_parameterSetManager.getPPS(m_apcSlicePilot->getPPSId());
+ CHECK(pps == 0, "No PPS present");
+ SPS *sps = m_parameterSetManager.getSPS(pps->getSPSId());
+ CHECK(sps == 0, "No SPS present");
+
+ writeBlockStatisticsHeader(sps);
+#endif
+
DTRACE_UPDATE( g_trace_ctx, std::make_pair( "poc", m_apcSlicePilot->getPOC() ) );
#if HEVC_DEPENDENT_SLICES
@@ -1085,7 +1101,7 @@ bool DecLib::xDecodeSlice(InputNALUnit &nalu, int &iSkipFrame, int iPOCLastDispl
pcSlice->checkCRA(pcSlice->getRPS(), m_pocCRA, m_associatedIRAPType, m_cListPic );
// Set reference list
pcSlice->setRefPicList( m_cListPic, true, true );
-
+
if (!pcSlice->isIntra())
{
bool bLowDelay = true;
diff --git a/source/Lib/DecoderLib/DecSlice.cpp b/source/Lib/DecoderLib/DecSlice.cpp
index 88ae6e7632318e469898361eee3c930f0bcf40ac..01c9f40106e0516cc6d78e6d95956a56125e0bb3 100644
--- a/source/Lib/DecoderLib/DecSlice.cpp
+++ b/source/Lib/DecoderLib/DecSlice.cpp
@@ -225,6 +225,7 @@ void DecSlice::decompressSlice( Slice* slice, InputBitstream* bitstream )
#endif
+
isLastCtuOfSliceSegment = cabacReader.coding_tree_unit( cs, ctuArea, pic->m_prevQP, ctuRsAddr );
m_pcCuDecoder->decompressCtu( cs, ctuArea );
diff --git a/source/Lib/DecoderLib/VLCReader.cpp b/source/Lib/DecoderLib/VLCReader.cpp
index b681d5b1b2b9a968ac33c358430caf99c6719b0e..fe994754f6a0ac68f074fcbb10995ff7e83e6690 100644
--- a/source/Lib/DecoderLib/VLCReader.cpp
+++ b/source/Lib/DecoderLib/VLCReader.cpp
@@ -827,7 +827,6 @@ void HLSyntaxReader::parseSPSNext( SPSNext& spsNext, const bool usePCM )
}
#endif
#endif
-
for( int k = 0; k < SPSNext::NumReservedFlags; k++ )
{
READ_FLAG( symbol, "reserved_flag" ); if( symbol != 0 ) EXIT("Incompatible version: SPSNext reserved flag not equal to zero (bitstream was probably created with newer software version)" );
@@ -1819,7 +1818,7 @@ void HLSyntaxReader::parseSliceHeader (Slice* pcSlice, ParameterSetManager *para
pcSlice->setSubPuMvpSliceSubblkSizeEnable(uiCode);
if (pcSlice->getSubPuMvpSliceSubblkSizeEnable())
{
- READ_CODE(3, uiCode, "slice_atmvp_subblk_size_log2");
+ READ_CODE(3, uiCode, "log2_slice_sub_pu_tmvp_size_minus2");
pcSlice->setSubPuMvpSubblkLog2Size(uiCode + MIN_CU_LOG2);
}
else
diff --git a/source/Lib/EncoderLib/Analyze.h b/source/Lib/EncoderLib/Analyze.h
index 977db82526cfda2cc3ca80a2e59d0d02b1672324..2e5e4736054211ade625a795d717096a26d9026b 100644
--- a/source/Lib/EncoderLib/Analyze.h
+++ b/source/Lib/EncoderLib/Analyze.h
@@ -45,6 +45,7 @@
#include <stdio.h>
#include <memory.h>
#include <assert.h>
+#include <cinttypes>
#include "CommonLib/CommonDef.h"
#include "CommonLib/ChromaFormat.h"
#include "math.h"
@@ -84,9 +85,17 @@ public:
virtual ~Analyze() {}
Analyze() { clear(); }
- void addResult( double psnr[MAX_NUM_COMPONENT], double bits, const double MSEyuvframe[MAX_NUM_COMPONENT])
+ void addResult( double psnr[MAX_NUM_COMPONENT], double bits, const double MSEyuvframe[MAX_NUM_COMPONENT]
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+ )
{
m_dAddBits += bits;
+#if JVET_K0157
+ if (isEncodeLtRef)
+ return;
+#endif
for(uint32_t i=0; i<MAX_NUM_COMPONENT; i++)
{
m_dPSNRSum[i] += psnr[i];
@@ -173,11 +182,10 @@ public:
PSNRyuv = (MSEyuv == 0) ? 999.99 : 10.0 * log10((maxval * maxval) / MSEyuv);
}
-
#if ENABLE_QPA || WCG_WPSNR
- void printOut ( char cDelim, const ChromaFormat chFmt, const bool printMSEBasedSNR, const bool printSequenceMSE, const BitDepths &bitDepths, const bool useWPSNR = false )
+ void printOut ( char cDelim, const ChromaFormat chFmt, const bool printMSEBasedSNR, const bool printSequenceMSE, const bool printHexPsnr, const BitDepths &bitDepths, const bool useWPSNR = false )
#else
- void printOut ( char cDelim, const ChromaFormat chFmt, const bool printMSEBasedSNR, const bool printSequenceMSE, const BitDepths &bitDepths )
+ void printOut ( char cDelim, const ChromaFormat chFmt, const bool printMSEBasedSNR, const bool printSequenceMSE, const bool printHexPsnr, const BitDepths &bitDepths )
#endif
{
#if !WCG_WPSNR
@@ -226,6 +234,11 @@ public:
#endif
msg( e_msg_level, " \tTotal Frames | " "Bitrate " "Y-PSNR" );
+ if (printHexPsnr)
+ {
+ msg(e_msg_level, "xY-PSNR ");
+ }
+
if (printSequenceMSE)
{
msg( e_msg_level, " Y-MSE\n" );
@@ -244,6 +257,19 @@ public:
#endif
getPsnr(COMPONENT_Y) / (double)getNumPic() );
+ if (printHexPsnr)
+ {
+ double dPsnr;
+ uint64_t xPsnr;
+ dPsnr = getPsnr(COMPONENT_Y) / (double)getNumPic();
+
+ copy(reinterpret_cast<uint8_t *>(&dPsnr),
+ reinterpret_cast<uint8_t *>(&dPsnr) + sizeof(dPsnr),
+ reinterpret_cast<uint8_t *>(&xPsnr));
+
+ msg(e_msg_level, " %16" PRIx64 " ", xPsnr);
+ }
+
if (printSequenceMSE)
{
msg( e_msg_level, " %8.4lf\n", m_MSEyuvframe[COMPONENT_Y] / (double)getNumPic() );
@@ -267,6 +293,11 @@ public:
#endif
msg( e_msg_level, "\tTotal Frames | " "Bitrate " "Y-PSNR" );
+ if (printHexPsnr)
+ {
+ msg(e_msg_level, "xY-PSNR ");
+ }
+
if (printSequenceMSE)
{
msg( e_msg_level, " Y-MSE\n" );
@@ -285,6 +316,19 @@ public:
#endif
getPsnr(COMPONENT_Y) / (double)getNumPic() );
+ if (printHexPsnr)
+ {
+ double dPsnr;
+ uint64_t xPsnr;
+ dPsnr = getPsnr(COMPONENT_Y) / (double)getNumPic();
+
+ copy(reinterpret_cast<uint8_t *>(&dPsnr),
+ reinterpret_cast<uint8_t *>(&dPsnr) + sizeof(dPsnr),
+ reinterpret_cast<uint8_t *>(&xPsnr));
+
+ msg(e_msg_level, " %16" PRIx64 " ", xPsnr);
+ }
+
if (printSequenceMSE)
{
msg( e_msg_level, " %8.4lf\n", m_MSEyuvframe[COMPONENT_Y] / (double)getNumPic() );
@@ -313,6 +357,11 @@ public:
#endif
msg( e_msg_level, " \tTotal Frames | " "Bitrate " "Y-PSNR " "U-PSNR " "V-PSNR " "YUV-PSNR " );
+ if (printHexPsnr)
+ {
+ msg(e_msg_level, "xY-PSNR " "xU-PSNR " "xV-PSNR ");
+ }
+
if (printSequenceMSE)
{
msg( e_msg_level, " Y-MSE " "U-MSE " "V-MSE " "YUV-MSE \n" );
@@ -340,6 +389,21 @@ public:
getPsnr(COMPONENT_Cr) / (double)getNumPic(),
PSNRyuv );
+ if (printHexPsnr)
+ {
+ double dPsnr[MAX_NUM_COMPONENT];
+ uint64_t xPsnr[MAX_NUM_COMPONENT];
+ for (int i = 0; i < MAX_NUM_COMPONENT; i++)
+ {
+ dPsnr[i] = getPsnr((ComponentID)i) / (double)getNumPic();
+
+ copy(reinterpret_cast<uint8_t *>(&dPsnr[i]),
+ reinterpret_cast<uint8_t *>(&dPsnr[i]) + sizeof(dPsnr[i]),
+ reinterpret_cast<uint8_t *>(&xPsnr[i]));
+ }
+ msg(e_msg_level, " %16" PRIx64 " %16" PRIx64 " %16" PRIx64, xPsnr[COMPONENT_Y], xPsnr[COMPONENT_Cb], xPsnr[COMPONENT_Cr]);
+ }
+
if (printSequenceMSE)
{
msg( e_msg_level, " %8.4lf " "%8.4lf " "%8.4lf " "%8.4lf\n",
@@ -373,6 +437,11 @@ public:
m_ext360.printHeader(e_msg_level);
#endif
+ if (printHexPsnr)
+ {
+ msg(e_msg_level, "xY-PSNR " "xU-PSNR " "xV-PSNR ");
+ }
+
if (printSequenceMSE)
{
msg( e_msg_level, " Y-MSE " "U-MSE " "V-MSE " "YUV-MSE \n" );
@@ -400,6 +469,21 @@ public:
getPsnr(COMPONENT_Cr) / (double)getNumPic(),
PSNRyuv );
+ if (printHexPsnr)
+ {
+ double dPsnr[MAX_NUM_COMPONENT];
+ uint64_t xPsnr[MAX_NUM_COMPONENT];
+ for (int i = 0; i < MAX_NUM_COMPONENT; i++)
+ {
+ dPsnr[i] = getPsnr((ComponentID)i) / (double)getNumPic();
+
+ copy(reinterpret_cast<uint8_t *>(&dPsnr[i]),
+ reinterpret_cast<uint8_t *>(&dPsnr[i]) + sizeof(dPsnr[i]),
+ reinterpret_cast<uint8_t *>(&xPsnr[i]));
+ }
+ msg(e_msg_level, " %16" PRIx64 " %16" PRIx64 " %16" PRIx64 , xPsnr[COMPONENT_Y], xPsnr[COMPONENT_Cb], xPsnr[COMPONENT_Cr]);
+ }
+
#if EXTENSION_360_VIDEO
m_ext360.printPSNRs(getNumPic(), e_msg_level);
#endif
@@ -427,7 +511,7 @@ public:
}
- void printSummary(const ChromaFormat chFmt, const bool printSequenceMSE, const BitDepths &bitDepths, const std::string &sFilename)
+ void printSummary(const ChromaFormat chFmt, const bool printSequenceMSE, const bool printHexPsnr, const BitDepths &bitDepths, const std::string &sFilename)
{
FILE* pFile = fopen (sFilename.c_str(), "at");
diff --git a/source/Lib/EncoderLib/CABACWriter.cpp b/source/Lib/EncoderLib/CABACWriter.cpp
index d98bd514611797f6bbf6811b58f27a64d245655f..10059954eda284343993f235a96be6cbe46b8b3a 100644
--- a/source/Lib/EncoderLib/CABACWriter.cpp
+++ b/source/Lib/EncoderLib/CABACWriter.cpp
@@ -642,7 +642,6 @@ void CABACWriter::split_cu_mode_mt(const PartSplit split, const CodingStructure&
void CABACWriter::coding_unit( const CodingUnit& cu, Partitioner& partitioner, CUCtx& cuCtx )
{
CodingStructure& cs = *cu.cs;
-
// transquant bypass flag
if( cs.pps->getTransquantBypassEnabledFlag() )
{
@@ -753,6 +752,7 @@ void CABACWriter::cu_pred_data( const CodingUnit& cu )
#if JVET_K0357_AMVR
imv_mode ( cu );
#endif
+
}
@@ -1302,7 +1302,7 @@ void CABACWriter::imv_mode( const CodingUnit& cu )
}
unsigned ctxId = DeriveCtx::CtxIMVFlag( cu );
- m_BinEncoder.encodeBin( ( cu.imv > 0 ), Ctx::ImvFlag( ctxId ) );
+ m_BinEncoder.encodeBin( ( cu.imv > 0 ), Ctx::ImvFlag( ctxId ) );
DTRACE( g_trace_ctx, D_SYNTAX, "imv_mode() value=%d ctx=%d\n", (cu.imv > 0), ctxId );
if( spsNext.getImvMode() == IMV_4PEL && cu.imv > 0 )
diff --git a/source/Lib/EncoderLib/CMakeLists.txt b/source/Lib/EncoderLib/CMakeLists.txt
index 9c30ad89e76226aa20bbc92b7f6c7ff73058855e..9e75e9fb100aea05e051b09172210d93e3b66676 100644
--- a/source/Lib/EncoderLib/CMakeLists.txt
+++ b/source/Lib/EncoderLib/CMakeLists.txt
@@ -17,7 +17,7 @@ add_library( ${LIB_NAME} STATIC ${SRC_FILES} ${INC_FILES} ${NATVIS_FILES} )
target_compile_definitions( ${LIB_NAME} PUBLIC )
if( ENABLE_VTM )
- target_compile_definitions( ${LIB_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${LIB_NAME} PUBLIC BMS_TOOLS=0 )
endif()
if( EXTENSION_360_VIDEO )
diff --git a/source/Lib/EncoderLib/EncCfg.h b/source/Lib/EncoderLib/EncCfg.h
index 015b7cc95794a95eb66267544956d842c630b8b8..6996a87228967306b9a46d9dc3d4142bcba838a5 100644
--- a/source/Lib/EncoderLib/EncCfg.h
+++ b/source/Lib/EncoderLib/EncCfg.h
@@ -212,6 +212,9 @@ protected:
#if ENABLE_WPP_PARALLELISM
bool m_AltDQPCoding;
+#endif
+#if JVET_K0157
+ bool m_compositeRefEnabled; //composite reference
#endif
// ADD_NEW_TOOL : (encoder lib) add tool enabling flags and associated parameters here
@@ -567,8 +570,8 @@ public:
bool getPrintMSEBasedSequencePSNR () const { return m_printMSEBasedSequencePSNR; }
void setPrintMSEBasedSequencePSNR (bool value) { m_printMSEBasedSequencePSNR = value; }
- bool getPrintHexPsnr() const { return m_printHexPsnr; }
- void setPrintHexPsnr(bool value) { m_printHexPsnr = value; }
+ bool getPrintHexPsnr () const { return m_printHexPsnr; }
+ void setPrintHexPsnr (bool value) { m_printHexPsnr = value; }
bool getPrintFrameMSE () const { return m_printFrameMSE; }
void setPrintFrameMSE (bool value) { m_printFrameMSE = value; }
@@ -668,6 +671,11 @@ public:
+#if JVET_K0157
+ void setUseCompositeRef (bool b) { m_compositeRefEnabled = b; }
+ bool getUseCompositeRef () const { return m_compositeRefEnabled; }
+#endif
+
// ADD_NEW_TOOL : (encoder lib) add access functions here
void setMaxCUWidth ( uint32_t u ) { m_maxCUWidth = u; }
diff --git a/source/Lib/EncoderLib/EncCu.cpp b/source/Lib/EncoderLib/EncCu.cpp
index dadf0bf0252c3eef545f147d9cff7032afbf4d77..4dd1259285ddefe1ad264ddd7cb4d51eedd7031f 100644
--- a/source/Lib/EncoderLib/EncCu.cpp
+++ b/source/Lib/EncoderLib/EncCu.cpp
@@ -302,7 +302,6 @@ void EncCu::compressCtu( CodingStructure& cs, const UnitArea& area, const unsign
// init the partitioning manager
Partitioner *partitioner = PartitionerFactory::get( *cs.slice );
partitioner->initCtu( area, CH_L, *cs.slice );
-
// init current context pointer
m_CurrCtx = m_CtxBuffer.data();
@@ -322,7 +321,9 @@ void EncCu::compressCtu( CodingStructure& cs, const UnitArea& area, const unsign
const bool copyUnsplitCTUSignals = bestCS->cus.size() == 1 && KEEP_PRED_AND_RESI_SIGNALS;
cs.useSubStructure( *bestCS, partitioner->chType, CS::getArea( *bestCS, area, partitioner->chType ), copyUnsplitCTUSignals, false, false, copyUnsplitCTUSignals );
- if( !cs.pcv->ISingleTree && cs.slice->isIntra() && cs.pcv->chrFormat != CHROMA_400 )
+ if( !cs.pcv->ISingleTree &&
+ cs.slice->isIntra()
+ && cs.pcv->chrFormat != CHROMA_400 )
{
m_CABACEstimator->getCtx() = m_CurrCtx->start;
@@ -340,6 +341,12 @@ void EncCu::compressCtu( CodingStructure& cs, const UnitArea& area, const unsign
cs.useSubStructure( *bestCS, partitioner->chType, CS::getArea( *bestCS, area, partitioner->chType ), copyUnsplitCTUSignals, false, false, copyUnsplitCTUSignals );
}
+#if JVET_K0390_RATECTRL
+ if (m_pcEncCfg->getUseRateCtrl())
+ {
+ (m_pcRateCtrl->getRCPic()->getLCU(ctuRsAddr)).m_actualMSE = (double)bestCS->dist / (double)m_pcRateCtrl->getRCPic()->getLCU(ctuRsAddr).m_numberOfPixel;
+ }
+#endif
// reset context states and uninit context pointer
m_CABACEstimator->getCtx() = m_CurrCtx->start;
m_CurrCtx = 0;
@@ -555,7 +562,6 @@ void EncCu::xCompressCU( CodingStructure *&tempCS, CodingStructure *&bestCS, Par
}
#endif
-
m_modeCtrl->initCULevel( partitioner, *tempCS );
m_CurrCtx->start = m_CABACEstimator->getCtx();
@@ -704,9 +710,9 @@ void EncCu::updateLambda( Slice* slice, double dQP )
#endif
#endif
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*/ )
@@ -757,7 +763,7 @@ void EncCu::updateLambda( Slice* slice, double dQP )
dLambda *= lambdaModifier;
int qpBDoffset = slice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA);
- int iQP = max( -qpBDoffset, min( MAX_QP, (int) floor( dQP + 0.5 ) ) );
+ int iQP = Clip3(-qpBDoffset, MAX_QP, (int)floor(dQP + 0.5));
m_pcSliceEncoder->setUpLambda(slice, dLambda, iQP);
#else
@@ -1249,7 +1255,8 @@ void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestC
m_CABACEstimator->cu_transquant_bypass_flag( cu );
}
- if( !cu.cs->slice->isIntra() )
+ if( !cu.cs->slice->isIntra()
+ )
{
m_CABACEstimator->cu_skip_flag ( cu );
}
@@ -1326,7 +1333,7 @@ void EncCu::xCheckIntraPCM(CodingStructure *&tempCS, CodingStructure *&bestCS, P
cu.ipcm = true;
tempCS->addPU(tempCS->area, partitioner.chType);
-
+
tempCS->addTU( tempCS->area, partitioner.chType );
m_pcIntraSearch->IPCMSearch(*tempCS, partitioner);
@@ -1340,7 +1347,8 @@ void EncCu::xCheckIntraPCM(CodingStructure *&tempCS, CodingStructure *&bestCS, P
m_CABACEstimator->cu_transquant_bypass_flag( cu );
}
- if( !cu.cs->slice->isIntra() )
+ if( !cu.cs->slice->isIntra()
+ )
{
m_CABACEstimator->cu_skip_flag ( cu );
}
@@ -1505,6 +1513,10 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
static_vector<unsigned, MRG_MAX_NUM_CANDS> RdModeList;
bool mrgTempBufSet = false;
+#if DMVR_JVET_LOW_LATENCY_K0217
+ Mv refinedMvdL0[MRG_MAX_NUM_CANDS];
+#endif
+
for( unsigned i = 0; i < MRG_MAX_NUM_CANDS; i++ )
{
RdModeList.push_back( i );
@@ -1553,7 +1565,6 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
m_pcRdCost->setDistParam (distParam, tempCS->getOrgBuf().Y(), m_acMergeBuffer[0].Y(), sps.getBitDepth (CHANNEL_TYPE_LUMA), COMPONENT_Y, bUseHadamard);
const UnitArea localUnitArea( tempCS->area.chromaFormat, Area( 0, 0, tempCS->area.Y().width, tempCS->area.Y().height) );
-
for( uint32_t uiMergeCand = 0; uiMergeCand < mergeCtx.numValidMergeCand; uiMergeCand++ )
{
acMergeBuffer[uiMergeCand] = m_acMergeBuffer[uiMergeCand].getBuf( localUnitArea );
@@ -1570,6 +1581,9 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
{
mergeCtx.mvFieldNeighbours[2*uiMergeCand].mv = pu.mv[0];
mergeCtx.mvFieldNeighbours[2*uiMergeCand+1].mv = pu.mv[1];
+#if DMVR_JVET_LOW_LATENCY_K0217
+ refinedMvdL0[uiMergeCand] = pu.mvd[0];
+#endif
}
#if DISTORTION_TYPE_BUGFIX
@@ -1585,10 +1599,8 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
double cost = (double)uiSad + (double)uiBitsCand * sqrtLambdaForFirstPass;
updateCandList( uiMergeCand, cost, RdModeList, candCostList, uiNumMrgSATDCand );
-
CHECK( std::min( uiMergeCand + 1, uiNumMrgSATDCand ) != RdModeList.size(), "" );
}
-
// Try to limit number of candidates using SATD-costs
for( uint32_t i = 1; i < uiNumMrgSATDCand; i++ )
{
@@ -1617,7 +1629,6 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
for( uint32_t uiMrgHADIdx = 0; uiMrgHADIdx < uiNumMrgSATDCand; uiMrgHADIdx++ )
{
uint32_t uiMergeCand = RdModeList[uiMrgHADIdx];
-
if( ( (uiNoResidualPass != 0) && candHasNoResidual[uiMergeCand] )
|| ( (uiNoResidualPass == 0) && bestIsSkip ) )
{
@@ -1647,6 +1658,9 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
if( mrgTempBufSet )
{
+#if DMVR_JVET_LOW_LATENCY_K0217
+ pu.mvd[0] = refinedMvdL0[uiMergeCand];
+#endif
tempCS->getPredBuf().copyFrom( acMergeBuffer[ uiMergeCand ]);
}
else
@@ -1655,11 +1669,15 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
}
+ xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, uiNoResidualPass
#if JVET_K0357_AMVR
- xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, uiNoResidualPass, NULL, true, ( ( uiNoResidualPass == 0 ) ? &candHasNoResidual[uiMergeCand] : NULL ) );
-#else
- xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, uiNoResidualPass, ( ( uiNoResidualPass == 0 ) ? &candHasNoResidual[uiMergeCand] : NULL ) );
+ , NULL
#endif
+#if JVET_K1000_SIMPLIFIED_EMT
+ , 1
+#endif
+ , uiNoResidualPass == 0 ? &candHasNoResidual[uiMergeCand] : NULL );
+
if( m_pcEncCfg->getUseFastDecisionForMerge() && !bestIsSkip )
{
bestIsSkip = bestCS->getCU( partitioner.chType )->rootCbf == 0;
@@ -1744,7 +1762,6 @@ void EncCu::xCheckRDCostAffineMerge2Nx2N( CodingStructure *&tempCS, CodingStruct
cu.firstPU->mergeFlag = true;
cu.firstPU->mergeIdx = 0;
-
PU::getAffineMergeCand( *cu.firstPU, affineMvField, interDirNeighbours, numValidMergeCand );
if( numValidMergeCand == -1 )
{
@@ -1758,30 +1775,38 @@ void EncCu::xCheckRDCostAffineMerge2Nx2N( CodingStructure *&tempCS, CodingStruct
PU::spanMotionInfo( *cu.firstPU );
m_pcInterSearch->motionCompensation( cu );
+
+ xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 0
#if JVET_K0357_AMVR
- xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 0, NULL, true, &hasNoResidual);
-#else
- xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 0, &hasNoResidual);
+ , NULL
#endif
+#if JVET_K1000_SIMPLIFIED_EMT
+ , 1
+#endif
+ , &hasNoResidual);
if( ! (encTestMode.lossless || hasNoResidual) )
{
tempCS->initStructData( encTestMode.qp, encTestMode.lossless );
tempCS->copyStructure( *bestCS, partitioner.chType );
tempCS->getPredBuf().copyFrom( bestCS->getPredBuf() );
+
+ xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 1
#if JVET_K0357_AMVR
- xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 1, NULL, true, &hasNoResidual);
-#else
- xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 1, &hasNoResidual);
+ , NULL
+#endif
+#if JVET_K1000_SIMPLIFIED_EMT
+ , 1
#endif
+ , &hasNoResidual);
}
}
#endif
-
void EncCu::xCheckRDCostInter( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
tempCS->initStructData( encTestMode.qp, encTestMode.lossless );
+
CodingUnit &cu = tempCS->addCU( tempCS->area, partitioner.chType );
partitioner.setCUData( cu );
@@ -1798,16 +1823,24 @@ void EncCu::xCheckRDCostInter( CodingStructure *&tempCS, CodingStructure *&bestC
cu.qp = encTestMode.qp;
CU::addPUs( cu );
+
m_pcInterSearch->predInterSearch( cu, partitioner );
#if JVET_K0357_AMVR
const unsigned wIdx = gp_sizeIdxInfo->idxFrom( tempCS->area.lwidth () );
#endif
+
+
+ xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, 0
#if JVET_K0357_AMVR
- xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 0, (m_pImvTempCS ? m_pImvTempCS[wIdx][encTestMode.partSize] : NULL));
-#else
- xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, 0, NULL);
+ , m_pImvTempCS ? m_pImvTempCS[wIdx][encTestMode.partSize] : NULL
+#endif
+#if JVET_K1000_SIMPLIFIED_EMT
+ , 1
#endif
+ , 0
+ );
+
}
@@ -1843,6 +1876,7 @@ bool EncCu::xCheckRDCostInterIMV( CodingStructure *&tempCS, CodingStructure *&be
}
}
+
CodingUnit &cu = ( pcCUInfo2Reuse != nullptr ) ? *tempCS->getCU( partitioner.chType ) : tempCS->addCU( tempCS->area, partitioner.chType );
if( pcCUInfo2Reuse == nullptr )
@@ -1877,11 +1911,13 @@ bool EncCu::xCheckRDCostInterIMV( CodingStructure *&tempCS, CodingStructure *&be
cu.emtFlag = false;
#endif
+
if( pcCUInfo2Reuse != nullptr )
{
// reuse the motion info from pcCUInfo2Reuse
CU::resetMVDandMV2Int( cu, m_pcInterSearch );
+
if( !CU::hasSubCUNonZeroMVd( cu ) )
{
m_modeCtrl->useModeResult( encTestModeBase, tempCS, partitioner );
@@ -1894,26 +1930,41 @@ bool EncCu::xCheckRDCostInterIMV( CodingStructure *&tempCS, CodingStructure *&be
}
else
{
+
m_pcInterSearch->predInterSearch( cu, partitioner );
+
}
+
if( !CU::hasSubCUNonZeroMVd( cu ) )
{
m_modeCtrl->useModeResult( encTestModeBase, tempCS, partitioner );
return false;
}
- xEncodeInterResidual(tempCS, bestCS, partitioner, encTestModeBase, 0, NULL);
+ xEncodeInterResidual( tempCS, bestCS, partitioner, encTestModeBase, 0
+#if JVET_K0357_AMVR
+ , NULL
+#endif
+#if JVET_K1000_SIMPLIFIED_EMT
+ , true
+#endif
+ , 0
+ );
return true;
}
#endif
+void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode, int residualPass
#if JVET_K0357_AMVR
-void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode, int residualPass, CodingStructure* imvCS, int emtMode, bool* bestHasNonResi )
-#else
-void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode, int residualPass, bool* bestHasNonResi )
+ , CodingStructure* imvCS
+#endif
+#if JVET_K1000_SIMPLIFIED_EMT
+ , int emtMode
#endif
+ , bool* bestHasNonResi
+ )
{
if( residualPass == 1 && encTestMode.lossless )
{
@@ -1922,19 +1973,72 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
CodingUnit* cu = tempCS->getCU( partitioner.chType );
double bestCostInternal = MAX_DOUBLE;
+#if JVET_K1000_SIMPLIFIED_EMT
+ double bestCost = bestCS->cost;
+ const SPS& sps = *tempCS->sps;
+ const int maxSizeEMT = tempCS->pcv->noRQT ? EMT_INTER_MAX_CU_WITH_QTBT : EMT_INTER_MAX_CU;
+ bool swapped = false; // avoid unwanted data copy
+ bool reloadCU = false;
+ const bool considerEmtSecondPass = emtMode && sps.getSpsNext().getUseInterEMT() && partitioner.currArea().lwidth() <= maxSizeEMT && partitioner.currArea().lheight() <= maxSizeEMT;
+
+ int minEMTMode = 0;
+ int maxEMTMode = (considerEmtSecondPass?1:0);
+ if( emtMode == 2 )
+ {
+ minEMTMode = maxEMTMode = (cu->emtFlag?1:0);
+ }
+ for( int curEmtMode = minEMTMode; curEmtMode <= maxEMTMode; curEmtMode++ )
+#endif
{
+#if JVET_K1000_SIMPLIFIED_EMT
+ if( reloadCU )
+ {
+ if( bestCost == bestCS->cost ) //The first EMT pass didn't become the bestCS, so we clear the TUs generated
+ {
+ tempCS->clearTUs();
+ }
+ else if( false == swapped )
+ {
+ tempCS->initStructData( encTestMode.qp, encTestMode.lossless );
+ tempCS->copyStructure( *bestCS, partitioner.chType );
+ tempCS->getPredBuf().copyFrom( bestCS->getPredBuf() );
+ bestCost = bestCS->cost;
+ cu = tempCS->getCU( partitioner.chType );
+ swapped = true;
+ }
+ else
+ {
+ tempCS->clearTUs();
+ bestCost = bestCS->cost;
+ cu = tempCS->getCU( partitioner.chType );
+ }
+
+ //we need to restart the distortion for the new tempCS, the bit count and the cost
+ tempCS->dist = 0;
+ tempCS->fracBits = 0;
+ tempCS->cost = MAX_DOUBLE;
+ }
+
+ reloadCU = true; // enable cu reloading
+#endif
cu->skip = false;
+#if JVET_K1000_SIMPLIFIED_EMT
+ cu->emtFlag = curEmtMode;
+#endif
const bool skipResidual = residualPass == 1;
m_pcInterSearch->encodeResAndCalcRdInterCU( *tempCS, partitioner, skipResidual );
-
xEncodeDontSplit( *tempCS, partitioner );
xCheckDQP( *tempCS, partitioner );
+
+#if JVET_K1000_SIMPLIFIED_EMT
+ double emtFirstPassCost = tempCS->cost;
+#endif
#if JVET_K0357_AMVR
if( imvCS && (tempCS->cost < imvCS->cost) )
{
@@ -1944,7 +2048,6 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
}
imvCS->copyStructure( *tempCS, partitioner.chType );
}
-
#endif
if( NULL != bestHasNonResi && (bestCostInternal > tempCS->cost) )
{
@@ -1959,8 +2062,22 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
#endif
xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
+#if JVET_K1000_SIMPLIFIED_EMT
+ //now we check whether the second pass should be skipped or not
+ if( !curEmtMode && maxEMTMode )
+ {
+ const double thresholdToSkipEmtSecondPass = 1.1; // Skip checking EMT transforms
+ const bool bCond1 = !cu->firstTU->cbf[COMPONENT_Y];
- }//end emt loop
+ const bool bCond3 = emtFirstPassCost > ( bestCost * thresholdToSkipEmtSecondPass );
+
+ if( m_pcEncCfg->getFastInterEMT() && (bCond1 || bCond3 ) )
+ {
+ maxEMTMode = 0; // do not test EMT
+ }
+ }
+#endif
+ } //end emt loop
}
diff --git a/source/Lib/EncoderLib/EncCu.h b/source/Lib/EncoderLib/EncCu.h
index 560973c32561f8b35ad41df9b8726f119029e78a..101a62b582b961f3a1fc883fcfbdf683ef8117a7 100644
--- a/source/Lib/EncoderLib/EncCu.h
+++ b/source/Lib/EncoderLib/EncCu.h
@@ -121,7 +121,6 @@ private:
unsigned int m_prevPOC;
bool m_clearSubMergeStatic;
#endif
-
#if ENABLE_SPLIT_PARALLELISM || ENABLE_WPP_PARALLELISM
EncLib* m_pcEncLib;
#endif
@@ -201,13 +200,16 @@ protected:
void xCheckRDCostMerge2Nx2N ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &pm, const EncTestMode& encTestMode );
+ void xEncodeInterResidual ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode, int residualPass = 0
#if JVET_K0357_AMVR
- void xEncodeInterResidual ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode, int residualPass = 0, CodingStructure* imvCS = NULL, int emtMode = 1, bool* bestHasNonResi = NULL );
-#else
- void xEncodeInterResidual ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode, int residualPass, bool* bestHasNonResi );
+ , CodingStructure* imvCS = NULL
+#endif
+#if JVET_K1000_SIMPLIFIED_EMT
+ , int emtMode = 1
#endif
+ , bool* bestHasNonResi = NULL
+ );
#if REUSE_CU_RESULTS
-
void xReuseCachedResult ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &Partitioner );
#endif
};
diff --git a/source/Lib/EncoderLib/EncGOP.cpp b/source/Lib/EncoderLib/EncGOP.cpp
index bebe67399e6c85b36ec66fede8929284a1212f73..8c133e5668b778383b3b41ededb430b7960ca2fb 100644
--- a/source/Lib/EncoderLib/EncGOP.cpp
+++ b/source/Lib/EncoderLib/EncGOP.cpp
@@ -110,6 +110,15 @@ EncGOP::EncGOP()
#endif
m_bInitAMaxBT = true;
+#if JVET_K0157
+ m_bgPOC = -1;
+ m_picBg = NULL;
+ m_picOrig = NULL;
+ m_isEncodedLTRef = false;
+ m_isUseLTRef = false;
+ m_isPrepareLTRef = true;
+ m_lastLTRefPoc = 0;
+#endif
}
EncGOP::~EncGOP()
@@ -139,6 +148,20 @@ void EncGOP::destroy()
m_pcDeblockingTempPicYuv = NULL;
}
#endif
+#if JVET_K0157
+ if (m_picBg)
+ {
+ m_picBg->destroy();
+ delete m_picBg;
+ m_picBg = NULL;
+ }
+ if (m_picOrig)
+ {
+ m_picOrig->destroy();
+ delete m_picOrig;
+ m_picOrig = NULL;
+ }
+#endif
}
void EncGOP::init ( EncLib* pcEncLib )
@@ -159,7 +182,7 @@ void EncGOP::init ( EncLib* pcEncLib )
m_totalCoded = 0;
m_AUWriterIf = pcEncLib->getAUWriterIf();
-
+
#if WCG_EXT
pcEncLib->getRdCost()->initLumaLevelToWeightTable();
#endif
@@ -1286,7 +1309,11 @@ void trySkipOrDecodePicture( bool& decPic, bool& encPic, const EncCfg& cfg, Pict
// ====================================================================================================================
void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
std::list<PelUnitBuf*>& rcListPicYuvRecOut,
- bool isField, bool isTff, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE )
+ bool isField, bool isTff, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+)
{
// TODO: Split this function up.
@@ -1296,7 +1323,11 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
pcBitstreamRedirect = new OutputBitstream;
AccessUnit::iterator itLocationToPushSliceHeaderNALU; // used to store location where NALU containing slice header is to be inserted
- xInitGOP( iPOCLast, iNumPicRcvd, isField );
+ xInitGOP(iPOCLast, iNumPicRcvd, isField
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
m_iNumPicCoded = 0;
SEIMessages leadingSeiMessages;
@@ -1335,11 +1366,18 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
/////////////////////////////////////////////////////////////////////////////////////////////////// Initial to start encoding
int iTimeOffset;
int pocCurr;
+#if JVET_K0157
+ int multipleFactor = m_pcCfg->getUseCompositeRef() ? 2 : 1;
+#endif
if(iPOCLast == 0) //case first frame or first top field
{
pocCurr=0;
+#if JVET_K0157
+ iTimeOffset = multipleFactor;
+#else
iTimeOffset = 1;
+#endif
}
else if(iPOCLast == 1 && isField) //case first bottom field, just like the first frame, the poc computation is not right anymore, we set the right value
{
@@ -1348,11 +1386,24 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
else
{
+#if JVET_K0157
+ pocCurr = iPOCLast - iNumPicRcvd * multipleFactor + m_pcCfg->getGOPEntry(iGOPid).m_POC - ((isField && m_iGopSize>1) ? 1 : 0);
+#else
pocCurr = iPOCLast - iNumPicRcvd + m_pcCfg->getGOPEntry(iGOPid).m_POC - ((isField && m_iGopSize>1) ? 1:0);
+#endif
iTimeOffset = m_pcCfg->getGOPEntry(iGOPid).m_POC;
}
+#if JVET_K0157
+ if (m_pcCfg->getUseCompositeRef() && isEncodeLtRef)
+ {
+ pocCurr++;
+ iTimeOffset--;
+ }
+ if (pocCurr / multipleFactor >= m_pcCfg->getFramesToBeEncoded())
+#else
if(pocCurr>=m_pcCfg->getFramesToBeEncoded())
+#endif
{
if (m_pcCfg->getEfficientFieldIRAPEnabled())
{
@@ -1399,7 +1450,11 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
pcPic->allocateNewSlice();
m_pcSliceEncoder->setSliceSegmentIdx(0);
- m_pcSliceEncoder->initEncSlice ( pcPic, iPOCLast, pocCurr, iGOPid, pcSlice, isField );
+ m_pcSliceEncoder->initEncSlice(pcPic, iPOCLast, pocCurr, iGOPid, pcSlice, isField
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
DTRACE_UPDATE( g_trace_ctx, ( std::make_pair( "poc", pocCurr ) ) );
DTRACE_UPDATE( g_trace_ctx, ( std::make_pair( "final", 0 ) ) );
@@ -1425,7 +1480,6 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
{
pcSlice->setSliceType(I_SLICE);
}
-
// Set the nal unit type
pcSlice->setNalUnitType(getNalUnitType(pocCurr, m_iLastIDR, isField));
if(pcSlice->getTemporalLayerNonReferenceFlag())
@@ -1463,7 +1517,34 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
pcSlice->decodingRefreshMarking(m_pocCRA, m_bRefreshPending, rcListPic, m_pcCfg->getEfficientFieldIRAPEnabled());
+#if JVET_K0157
+ if (m_pcCfg->getUseCompositeRef() && isEncodeLtRef)
+ {
+ setUseLTRef(true);
+ setPrepareLTRef(false);
+ setNewestBgPOC(pocCurr);
+ setLastLTRefPoc(pocCurr);
+ }
+ else if (pcPic->cs->sps->getSpsNext().getUseCompositeRef() && getLastLTRefPoc() >= 0 && getEncodedLTRef()==false && !getPicBg()->getSpliceFull() && (pocCurr - getLastLTRefPoc()) > (m_pcCfg->getFrameRate() * 2))
+ {
+ setUseLTRef(false);
+ setPrepareLTRef(false);
+ setEncodedLTRef(true);
+ setNewestBgPOC(-1);
+ setLastLTRefPoc(-1);
+ }
+
+ if (pcPic->cs->sps->getSpsNext().getUseCompositeRef() && m_picBg->getSpliceFull() && getUseLTRef())
+ {
+ m_pcEncLib->selectReferencePictureSet(pcSlice, pocCurr, iGOPid, m_bgPOC);
+ }
+ else
+ {
+ m_pcEncLib->selectReferencePictureSet(pcSlice, pocCurr, iGOPid, -1);
+ }
+#else
m_pcEncLib->selectReferencePictureSet(pcSlice, pocCurr, iGOPid);
+#endif
if (!m_pcCfg->getEfficientFieldIRAPEnabled())
{
if ( pcSlice->getNalUnitType() == NAL_UNIT_CODED_SLICE_BLA_W_LP
@@ -1484,7 +1565,11 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
|| (m_pcCfg->getEfficientFieldIRAPEnabled() && isField && pcSlice->getAssociatedIRAPType() >= NAL_UNIT_CODED_SLICE_BLA_W_LP && pcSlice->getAssociatedIRAPType() <= NAL_UNIT_CODED_SLICE_CRA && pcSlice->getAssociatedIRAPPOC() == pcSlice->getPOC()+1)
)
{
- pcSlice->createExplicitReferencePictureSetFromReference(rcListPic, pcSlice->getRPS(), pcSlice->isIRAP(), m_iLastRecoveryPicPOC, m_pcCfg->getDecodingRefreshType() == 3, m_pcCfg->getEfficientFieldIRAPEnabled());
+ pcSlice->createExplicitReferencePictureSetFromReference(rcListPic, pcSlice->getRPS(), pcSlice->isIRAP(), m_iLastRecoveryPicPOC, m_pcCfg->getDecodingRefreshType() == 3, m_pcCfg->getEfficientFieldIRAPEnabled()
+#if JVET_K0157
+ , isEncodeLtRef, m_pcCfg->getUseCompositeRef()
+#endif
+ );
}
pcSlice->applyReferencePictureSet(rcListPic, pcSlice->getRPS());
@@ -1552,19 +1637,42 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
}
}
+#if JVET_K0157
+ if (pcSlice->getRPSidx() == -1)
+ arrangeLongtermPicturesInRPS(pcSlice, rcListPic);
+#else
arrangeLongtermPicturesInRPS(pcSlice, rcListPic);
+#endif
RefPicListModification* refPicListModification = pcSlice->getRefPicListModification();
refPicListModification->setRefPicListModificationFlagL0(0);
refPicListModification->setRefPicListModificationFlagL1(0);
+
+#if JVET_K0157
+ if (m_pcCfg->getUseCompositeRef() && getUseLTRef() && (pocCurr > getLastLTRefPoc()))
+ {
+ pcSlice->setNumRefIdx(REF_PIC_LIST_0, min(m_pcCfg->getGOPEntry(iGOPid).m_numRefPicsActive + 1, pcSlice->getRPS()->getNumberOfPictures()));
+ pcSlice->setNumRefIdx(REF_PIC_LIST_1, min(m_pcCfg->getGOPEntry(iGOPid).m_numRefPicsActive + 1, pcSlice->getRPS()->getNumberOfPictures()));
+ }
+ else
+ {
+ pcSlice->setNumRefIdx(REF_PIC_LIST_0, std::min(m_pcCfg->getGOPEntry(iGOPid).m_numRefPicsActive, pcSlice->getRPS()->getNumberOfPictures()));
+ pcSlice->setNumRefIdx(REF_PIC_LIST_1, std::min(m_pcCfg->getGOPEntry(iGOPid).m_numRefPicsActive, pcSlice->getRPS()->getNumberOfPictures()));
+ }
+ if (pcPic->cs->sps->getSpsNext().getUseCompositeRef() && getPrepareLTRef()) {
+ arrangeCompositeReference(pcSlice, rcListPic, pocCurr);
+ }
+#else
pcSlice->setNumRefIdx(REF_PIC_LIST_0,min(m_pcCfg->getGOPEntry(iGOPid).m_numRefPicsActive,pcSlice->getRPS()->getNumberOfPictures()));
pcSlice->setNumRefIdx(REF_PIC_LIST_1,min(m_pcCfg->getGOPEntry(iGOPid).m_numRefPicsActive,pcSlice->getRPS()->getNumberOfPictures()));
+#endif
// Set reference list
pcSlice->setRefPicList ( rcListPic );
if( m_pcCfg->getUseAMaxBT() )
{
- if( !pcSlice->isIntra() )
+ if( !pcSlice->isIntra()
+ )
{
int refLayer = pcSlice->getDepth();
if( refLayer > 9 ) refLayer = 9; // Max layer is 10
@@ -1614,7 +1722,6 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
{
pcSlice->setSliceType ( P_SLICE );
}
-
xUpdateRasInit( pcSlice );
// Do decoding refresh marking if any
@@ -1767,12 +1874,12 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
#if V0078_ADAPTIVE_LOWER_BOUND
if (estimatedCpbFullness - estimatedBits < m_pcRateCtrl->getRCPic()->getLowerBound())
{
- estimatedBits = max(200, estimatedCpbFullness - m_pcRateCtrl->getRCPic()->getLowerBound());
+ estimatedBits = std::max(200, estimatedCpbFullness - m_pcRateCtrl->getRCPic()->getLowerBound());
}
#else
if (estimatedCpbFullness - estimatedBits < (int)(m_pcRateCtrl->getCpbSize()*0.1f))
{
- estimatedBits = max(200, estimatedCpbFullness - (int)(m_pcRateCtrl->getCpbSize()*0.1f));
+ estimatedBits = std::max(200, estimatedCpbFullness - (int)(m_pcRateCtrl->getCpbSize()*0.1f));
}
#endif
@@ -1905,7 +2012,11 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
trySkipOrDecodePicture( decPic, encPic, *m_pcCfg, pcPic );
pcPic->cs->slice = pcSlice; // please keep this
- if (pcSlice->getPPS()->getSliceChromaQpFlag() && CS::isDualITree(*pcSlice->getPic()->cs))
+#if ENABLE_QPA
+ if (pcSlice->getPPS()->getSliceChromaQpFlag() && CS::isDualITree (*pcSlice->getPic()->cs) && !m_pcCfg->getUsePerceptQPA() && (m_pcCfg->getSliceChromaOffsetQpPeriodicity() == 0))
+#else
+ if (pcSlice->getPPS()->getSliceChromaQpFlag() && CS::isDualITree (*pcSlice->getPic()->cs))
+#endif
{
// overwrite chroma qp offset for dual tree
pcSlice->setSliceChromaQpDelta(COMPONENT_Cb, m_pcCfg->getChromaCbQpOffsetDualTree());
@@ -2010,6 +2121,10 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
m_pcLoopFilter->loopFilterPic( cs );
+#if DMVR_JVET_LOW_LATENCY_K0217
+ CS::setRefinedMotionField(cs);
+#endif
+
DTRACE_UPDATE( g_trace_ctx, ( std::make_pair( "final", 1 ) ) );
if( pcSlice->getSPS()->getUseSAO() )
@@ -2043,7 +2158,12 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
}
#endif
-
+#if JVET_K0157
+ if (pcPic->cs->sps->getSpsNext().getUseCompositeRef() && getPrepareLTRef())
+ {
+ updateCompositeReference(pcSlice, rcListPic, pocCurr);
+ }
+#endif
}
else // skip enc picture
{
@@ -2059,7 +2179,8 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
{
for( const CodingUnit *cu : pcPic->cs->cus )
{
- if( !pcSlice->isIntra() )
+ if( !pcSlice->isIntra()
+ )
{
m_uiBlkSize[pcSlice->getDepth()] += cu->Y().area();
m_uiNumBlk [pcSlice->getDepth()]++;
@@ -2236,7 +2357,11 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
m_pcCfg->setEncodedFlag(iGOPid, true);
double PSNR_Y;
- xCalculateAddPSNRs( isField, isTff, iGOPid, pcPic, accessUnit, rcListPic, encTime, snr_conversion, printFrameMSE, &PSNR_Y );
+ xCalculateAddPSNRs(isField, isTff, iGOPid, pcPic, accessUnit, rcListPic, encTime, snr_conversion, printFrameMSE, &PSNR_Y
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
// Only produce the Green Metadata SEI message with the last picture.
if( m_pcCfg->getSEIGreenMetadataInfoSEIEnable() && pcSlice->getPOC() == ( m_pcCfg->getFramesToBeEncoded() - 1 ) )
@@ -2300,7 +2425,10 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
pcPic->reconstructed = true;
m_bFirst = false;
m_iNumPicCoded++;
- m_totalCoded ++;
+#if JVET_K0157
+ if (!(pcPic->cs->sps->getSpsNext().getUseCompositeRef() && isEncodeLtRef))
+#endif
+ m_totalCoded ++;
/* logging: insert a newline at end of picture period */
if (m_pcCfg->getEfficientFieldIRAPEnabled())
@@ -2319,7 +2447,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
-void EncGOP::printOutSummary(uint32_t uiNumAllPicCoded, bool isField, const bool printMSEBasedSNR, const bool printSequenceMSE, const BitDepths &bitDepths)
+void EncGOP::printOutSummary(uint32_t uiNumAllPicCoded, bool isField, const bool printMSEBasedSNR, const bool printSequenceMSE, const bool printHexPsnr, const BitDepths &bitDepths)
{
#if ENABLE_QPA
const bool useWPSNR = m_pcEncLib->getUseWPSNR();
@@ -2345,25 +2473,25 @@ void EncGOP::printOutSummary(uint32_t uiNumAllPicCoded, bool isField, const bool
m_gcAnalyzeWPSNR.setFrmRate(m_pcCfg->getFrameRate()*rateMultiplier / (double)m_pcCfg->getTemporalSubsampleRatio());
}
#endif
-
+
const ChromaFormat chFmt = m_pcCfg->getChromaFormatIdc();
//-- all
msg( INFO, "\n" );
msg( DETAILS,"\nSUMMARY --------------------------------------------------------\n" );
#if ENABLE_QPA
- m_gcAnalyzeAll.printOut('a', chFmt, printMSEBasedSNR, printSequenceMSE, bitDepths, useWPSNR);
+ m_gcAnalyzeAll.printOut('a', chFmt, printMSEBasedSNR, printSequenceMSE, printHexPsnr, bitDepths, useWPSNR);
#else
- m_gcAnalyzeAll.printOut('a', chFmt, printMSEBasedSNR, printSequenceMSE, bitDepths);
+ m_gcAnalyzeAll.printOut('a', chFmt, printMSEBasedSNR, printSequenceMSE, printHexPsnr, bitDepths);
#endif
msg( DETAILS,"\n\nI Slices--------------------------------------------------------\n" );
- m_gcAnalyzeI.printOut('i', chFmt, printMSEBasedSNR, printSequenceMSE, bitDepths);
+ m_gcAnalyzeI.printOut('i', chFmt, printMSEBasedSNR, printSequenceMSE, printHexPsnr, bitDepths);
msg( DETAILS,"\n\nP Slices--------------------------------------------------------\n" );
- m_gcAnalyzeP.printOut('p', chFmt, printMSEBasedSNR, printSequenceMSE, bitDepths);
+ m_gcAnalyzeP.printOut('p', chFmt, printMSEBasedSNR, printSequenceMSE, printHexPsnr, bitDepths);
msg( DETAILS,"\n\nB Slices--------------------------------------------------------\n" );
- m_gcAnalyzeB.printOut('b', chFmt, printMSEBasedSNR, printSequenceMSE, bitDepths);
+ m_gcAnalyzeB.printOut('b', chFmt, printMSEBasedSNR, printSequenceMSE, printHexPsnr, bitDepths);
#if WCG_WPSNR
if (useLumaWPSNR)
@@ -2374,14 +2502,14 @@ void EncGOP::printOutSummary(uint32_t uiNumAllPicCoded, bool isField, const bool
#endif
if (!m_pcCfg->getSummaryOutFilename().empty())
{
- m_gcAnalyzeAll.printSummary(chFmt, printSequenceMSE, bitDepths, m_pcCfg->getSummaryOutFilename());
+ m_gcAnalyzeAll.printSummary(chFmt, printSequenceMSE, printHexPsnr, bitDepths, m_pcCfg->getSummaryOutFilename());
}
if (!m_pcCfg->getSummaryPicFilenameBase().empty())
{
- m_gcAnalyzeI.printSummary(chFmt, printSequenceMSE, bitDepths, m_pcCfg->getSummaryPicFilenameBase()+"I.txt");
- m_gcAnalyzeP.printSummary(chFmt, printSequenceMSE, bitDepths, m_pcCfg->getSummaryPicFilenameBase()+"P.txt");
- m_gcAnalyzeB.printSummary(chFmt, printSequenceMSE, bitDepths, m_pcCfg->getSummaryPicFilenameBase()+"B.txt");
+ m_gcAnalyzeI.printSummary(chFmt, printSequenceMSE, printHexPsnr, bitDepths, m_pcCfg->getSummaryPicFilenameBase()+"I.txt");
+ m_gcAnalyzeP.printSummary(chFmt, printSequenceMSE, printHexPsnr, bitDepths, m_pcCfg->getSummaryPicFilenameBase()+"P.txt");
+ m_gcAnalyzeB.printSummary(chFmt, printSequenceMSE, printHexPsnr, bitDepths, m_pcCfg->getSummaryPicFilenameBase()+"B.txt");
}
#if WCG_WPSNR
@@ -2399,13 +2527,13 @@ void EncGOP::printOutSummary(uint32_t uiNumAllPicCoded, bool isField, const bool
msg( DETAILS,"\n\nSUMMARY INTERLACED ---------------------------------------------\n" );
#if ENABLE_QPA
- m_gcAnalyzeAll_in.printOut('a', chFmt, printMSEBasedSNR, printSequenceMSE, bitDepths, useWPSNR);
+ m_gcAnalyzeAll_in.printOut('a', chFmt, printMSEBasedSNR, printSequenceMSE, printHexPsnr, bitDepths, useWPSNR);
#else
- m_gcAnalyzeAll_in.printOut('a', chFmt, printMSEBasedSNR, printSequenceMSE, bitDepths);
+ m_gcAnalyzeAll_in.printOut('a', chFmt, printMSEBasedSNR, printSequenceMSE, printHexPsnr, bitDepths);
#endif
if (!m_pcCfg->getSummaryOutFilename().empty())
{
- m_gcAnalyzeAll_in.printSummary(chFmt, printSequenceMSE, bitDepths, m_pcCfg->getSummaryOutFilename());
+ m_gcAnalyzeAll_in.printSummary(chFmt, printSequenceMSE, printHexPsnr, bitDepths, m_pcCfg->getSummaryOutFilename());
#if WCG_WPSNR
if (useLumaWPSNR)
{
@@ -2448,13 +2576,19 @@ uint64_t EncGOP::preLoopFilterPicAndCalcDist( Picture* pcPic )
// ====================================================================================================================
// Protected member functions
// ====================================================================================================================
-
-
-void EncGOP::xInitGOP( int iPOCLast, int iNumPicRcvd, bool isField )
+void EncGOP::xInitGOP( int iPOCLast, int iNumPicRcvd, bool isField
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+)
{
CHECK(!( iNumPicRcvd > 0 ), "Unspecified error");
// Exception for the first frames
+#if JVET_K0157
+ if ((isField && (iPOCLast == 0 || iPOCLast == 1)) || (!isField && (iPOCLast == 0)) || isEncodeLtRef)
+#else
if ( ( isField && (iPOCLast == 0 || iPOCLast == 1) ) || (!isField && (iPOCLast == 0)) )
+#endif
{
m_iGopSize = 1;
}
@@ -2485,7 +2619,12 @@ void EncGOP::xGetBuffer( PicList& rcListPic,
iTimeOffset--;
}
+#if JVET_K0157
+ int multipleFactor = m_pcCfg->getUseCompositeRef() ? 2 : 1;
+ for (i = 0; i < (iNumPicRcvd * multipleFactor - iTimeOffset + 1); i += multipleFactor)
+#else
for ( i = 0; i < (iNumPicRcvd - iTimeOffset + 1); i++ )
+#endif
{
iterPicYuvRec--;
}
@@ -2512,15 +2651,18 @@ void EncGOP::xGetBuffer( PicList& rcListPic,
#if ENABLE_QPA
#ifndef BETA
- #define BETA (2.0 / 3.0) // value between 0 and 1; use 0.0 for traditional PSNR
+ #define BETA 0.5 // value between 0.0 and 1; use 0.0 to obtain traditional PSNR
#endif
#define GLOBAL_AVERAGING 1 // "global" averaging of a_k across a set instead of one picture
#if FRAME_WEIGHTING
static const uint32_t DQP[16] = { 4, 12, 11, 12, 9, 12, 11, 12, 6, 12, 11, 12, 9, 12, 11, 12 };
#endif
-static inline double calcWeightedSquaredError(const CPelBuf& org, const CPelBuf& rec, double &sumAct,
- const uint32_t imageWidth, const uint32_t imageHeight, const uint32_t offsetX, const uint32_t offsetY, int blockWidth, int blockHeight)
+static inline double calcWeightedSquaredError(const CPelBuf& org, const CPelBuf& rec,
+ double &sumAct, const uint32_t bitDepth,
+ const uint32_t imageWidth, const uint32_t imageHeight,
+ const uint32_t offsetX, const uint32_t offsetY,
+ int blockWidth, int blockHeight)
{
const int O = org.stride;
const int R = rec.stride;
@@ -2534,8 +2676,8 @@ static inline double calcWeightedSquaredError(const CPelBuf& org, const CPelB
const int hAct = offsetY + (uint32_t)blockHeight < imageHeight ? blockHeight : blockHeight - 1;
const int wAct = offsetX + (uint32_t)blockWidth < imageWidth ? blockWidth : blockWidth - 1;
- uint64_t ssErr = 0; // sum of squared diffs
- uint64_t saAct = 0; // sum of abs. activity
+ uint64_t ssErr = 0; // sum of squared diffs
+ uint64_t saAct = 0; // sum of abs. activity
double msAct;
int x, y;
@@ -2544,7 +2686,7 @@ static inline double calcWeightedSquaredError(const CPelBuf& org, const CPelB
{
for (x = 0; x < blockWidth; x++)
{
- register int64_t iDiff = (int64_t)o[y*O + x] - (int64_t)r[y*R + x];
+ const int64_t iDiff = (int64_t)o[y*O + x] - (int64_t)r[y*R + x];
ssErr += uint64_t(iDiff * iDiff);
}
}
@@ -2554,13 +2696,18 @@ static inline double calcWeightedSquaredError(const CPelBuf& org, const CPelB
{
for (x = xAct; x < wAct; x++)
{
- saAct += uint64_t(abs(4 * (int64_t)o[y*O + x] - (int64_t)o[y*O + x-1] - (int64_t)o[y*O + x+1] - (int64_t)o[(y-1)*O + x] - (int64_t)o[(y+1)*O + x]));
+ const int f = 12 * (int)o[y*O + x] - 2 * ((int)o[y*O + x-1] + (int)o[y*O + x+1] + (int)o[(y-1)*O + x] + (int)o[(y+1)*O + x])
+ - (int)o[(y-1)*O + x-1] - (int)o[(y-1)*O + x+1] - (int)o[(y+1)*O + x-1] - (int)o[(y+1)*O + x+1];
+ saAct += abs(f);
}
}
// calculate weight (mean squared activity)
msAct = (double)saAct / (double(wAct - xAct) * double(hAct - yAct));
- if (msAct < 8.0) msAct = 8.0;
+
+ // lower limit, accounts for high-pass gain
+ if (msAct < double(1 << (bitDepth - 4))) msAct = double(1 << (bitDepth - 4));
+
msAct *= msAct; // because ssErr is squared
sumAct += msAct; // includes high-pass gain
@@ -2572,9 +2719,9 @@ static inline double calcWeightedSquaredError(const CPelBuf& org, const CPelB
uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1, const uint32_t rshift
#if ENABLE_QPA
- , const uint32_t chromaShift /*= 0*/
+ , const uint32_t chromaShift /*= 0*/
#endif
- )
+ )
{
uint64_t uiTotalDiff;
const Pel* pSrc0 = pic0.bufAt(0, 0);
@@ -2591,7 +2738,7 @@ uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1,
{
const uint32_t W = pic0.width; // image width
const uint32_t H = pic0.height; // image height
- const double R = double(W * H) / (1920.0 * 1080.0);
+ const double R = double(W * H) / (1920.0 * 1080.0);
const uint32_t B = Clip3<uint32_t>(0, 128 >> chromaShift, 4 * uint32_t(16.0 * sqrt(R) + 0.5)); // WPSNR block size in integer multiple of 4 (for SIMD, = 64 at full-HD)
uint32_t x, y;
@@ -2603,7 +2750,7 @@ uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1,
{
for (x = 0; x < W; x++)
{
- register int64_t iDiff = (int64_t)pSrc0[x] - (int64_t)pSrc1[x];
+ const int64_t iDiff = (int64_t)pSrc0[x] - (int64_t)pSrc1[x];
uiTotalDiff += uint64_t(iDiff * iDiff);
}
pSrc0 += pic0.stride;
@@ -2620,7 +2767,11 @@ uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1,
{
for (x = 0; x < W; x += B)
{
- wmse += calcWeightedSquaredError(pic1, pic0, sumAct, W, H, x, y, B, B);
+ wmse += calcWeightedSquaredError(pic1, pic0,
+ sumAct, BD,
+ W, H,
+ x, y,
+ B, B);
#if !GLOBAL_AVERAGING
numAct += 1.0;
#endif
@@ -2629,11 +2780,17 @@ uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1,
// integer weighted distortion
#if GLOBAL_AVERAGING
- sumAct = 1.5 * double(1 << BD);
- if ((W << chromaShift) > 2048 && (H << chromaShift) > 1280) // UHD luma
+ sumAct = 32.0 * double(1 << BD);
+
+ if ((W << chromaShift) > 2048 && (H << chromaShift) > 1280) // for UHD/4K
+ {
+ sumAct *= 0.5;
+ }
+ else if ((W << chromaShift) <= 1024 || (H << chromaShift) <= 640) // 480p
{
- sumAct /= 1.5;
+ sumAct *= 2.0;
}
+
return (wmse <= 0.0) ? 0 : uint64_t(wmse * pow(sumAct, BETA) + 0.5);
#else
return (wmse <= 0.0 || numAct <= 0.0) ? 0 : uint64_t(wmse * pow(sumAct / numAct, BETA) + 0.5);
@@ -2670,7 +2827,7 @@ uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1,
return uiTotalDiff;
}
#if WCG_WPSNR
-double EncGOP::xFindDistortionPlaneWPSNR(const CPelBuf& pic0, const CPelBuf& pic1, const uint32_t rshift, const CPelBuf& picLuma0,
+double EncGOP::xFindDistortionPlaneWPSNR(const CPelBuf& pic0, const CPelBuf& pic1, const uint32_t rshift, const CPelBuf& picLuma0,
ComponentID compID, const ChromaFormat chfmt )
{
const bool useLumaWPSNR = m_pcEncLib->getLumaLevelToDeltaQPMapping().isEnabled();
@@ -2723,9 +2880,17 @@ double EncGOP::xFindDistortionPlaneWPSNR(const CPelBuf& pic0, const CPelBuf& pic
}
#endif
-void EncGOP::xCalculateAddPSNRs( const bool isField, const bool isFieldTopFieldFirst, const int iGOPid, Picture* pcPic, const AccessUnit&accessUnit, PicList &rcListPic, const int64_t dEncTime, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE, double* PSNR_Y )
+void EncGOP::xCalculateAddPSNRs( const bool isField, const bool isFieldTopFieldFirst, const int iGOPid, Picture* pcPic, const AccessUnit&accessUnit, PicList &rcListPic, const int64_t dEncTime, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE, double* PSNR_Y
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+)
{
- xCalculateAddPSNR( pcPic, pcPic->getRecoBuf(), accessUnit, (double) dEncTime, snr_conversion, printFrameMSE, PSNR_Y );
+ xCalculateAddPSNR(pcPic, pcPic->getRecoBuf(), accessUnit, (double)dEncTime, snr_conversion, printFrameMSE, PSNR_Y
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
//In case of field coding, compute the interlaced PSNR for both fields
if(isField)
@@ -2778,19 +2943,31 @@ void EncGOP::xCalculateAddPSNRs( const bool isField, const bool isFieldTopFieldF
}
Picture* correspondingFieldPic = *(iterPic);
- if( (pcPic->topField && isFieldTopFieldFirst) || (!pcPic->topField && !isFieldTopFieldFirst))
+ if ((pcPic->topField && isFieldTopFieldFirst) || (!pcPic->topField && !isFieldTopFieldFirst))
{
- xCalculateInterlacedAddPSNR(pcPic, correspondingFieldPic, pcPic->getRecoBuf(), correspondingFieldPic->getRecoBuf(), snr_conversion, printFrameMSE, PSNR_Y );
+ xCalculateInterlacedAddPSNR(pcPic, correspondingFieldPic, pcPic->getRecoBuf(), correspondingFieldPic->getRecoBuf(), snr_conversion, printFrameMSE, PSNR_Y
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
}
else
{
- xCalculateInterlacedAddPSNR(correspondingFieldPic, pcPic, correspondingFieldPic->getRecoBuf(), pcPic->getRecoBuf(), snr_conversion, printFrameMSE, PSNR_Y );
+ xCalculateInterlacedAddPSNR(correspondingFieldPic, pcPic, correspondingFieldPic->getRecoBuf(), pcPic->getRecoBuf(), snr_conversion, printFrameMSE, PSNR_Y
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
}
}
}
}
-void EncGOP::xCalculateAddPSNR( Picture* pcPic, PelUnitBuf cPicD, const AccessUnit& accessUnit, double dEncTime, const InputColourSpaceConversion conversion, const bool printFrameMSE, double* PSNR_Y )
+void EncGOP::xCalculateAddPSNR(Picture* pcPic, PelUnitBuf cPicD, const AccessUnit& accessUnit, double dEncTime, const InputColourSpaceConversion conversion, const bool printFrameMSE, double* PSNR_Y
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+)
{
const SPS& sps = *pcPic->cs->sps;
const CPelUnitBuf& pic = cPicD;
@@ -2860,7 +3037,7 @@ void EncGOP::xCalculateAddPSNR( Picture* pcPic, PelUnitBuf cPicD, const AccessUn
#else
const uint64_t uiSSDtemp = xFindDistortionPlane(recPB, orgPB, 0);
#if WCG_WPSNR
- const double uiSSDtempWeighted = xFindDistortionPlaneWPSNR(recPB, orgPB, 0, org.get(COMPONENT_Y), compID, format);
+ const double uiSSDtempWeighted = xFindDistortionPlaneWPSNR(recPB, orgPB, 0, org.get(COMPONENT_Y), compID, format);
#endif
const uint32_t maxval = 255 << (bitDepth - 8);
#endif
@@ -2919,13 +3096,21 @@ void EncGOP::xCalculateAddPSNR( Picture* pcPic, PelUnitBuf cPicD, const AccessUn
m_vRVM_RP.push_back( uibits );
//===== add PSNR =====
- m_gcAnalyzeAll.addResult (dPSNR, (double)uibits, MSEyuvframe);
+ m_gcAnalyzeAll.addResult(dPSNR, (double)uibits, MSEyuvframe
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
#if EXTENSION_360_VIDEO
m_ext360.addResult(m_gcAnalyzeAll);
#endif
if (pcSlice->isIntra())
{
- m_gcAnalyzeI.addResult (dPSNR, (double)uibits, MSEyuvframe);
+ m_gcAnalyzeI.addResult(dPSNR, (double)uibits, MSEyuvframe
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
*PSNR_Y = dPSNR[COMPONENT_Y];
#if EXTENSION_360_VIDEO
m_ext360.addResult(m_gcAnalyzeI);
@@ -2933,7 +3118,11 @@ void EncGOP::xCalculateAddPSNR( Picture* pcPic, PelUnitBuf cPicD, const AccessUn
}
if (pcSlice->isInterP())
{
- m_gcAnalyzeP.addResult (dPSNR, (double)uibits, MSEyuvframe);
+ m_gcAnalyzeP.addResult(dPSNR, (double)uibits, MSEyuvframe
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
*PSNR_Y = dPSNR[COMPONENT_Y];
#if EXTENSION_360_VIDEO
m_ext360.addResult(m_gcAnalyzeP);
@@ -2941,7 +3130,11 @@ void EncGOP::xCalculateAddPSNR( Picture* pcPic, PelUnitBuf cPicD, const AccessUn
}
if (pcSlice->isInterB())
{
- m_gcAnalyzeB.addResult (dPSNR, (double)uibits, MSEyuvframe);
+ m_gcAnalyzeB.addResult(dPSNR, (double)uibits, MSEyuvframe
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
*PSNR_Y = dPSNR[COMPONENT_Y];
#if EXTENSION_360_VIDEO
m_ext360.addResult(m_gcAnalyzeB);
@@ -3024,7 +3217,11 @@ void EncGOP::xCalculateAddPSNR( Picture* pcPic, PelUnitBuf cPicD, const AccessUn
void EncGOP::xCalculateInterlacedAddPSNR( Picture* pcPicOrgFirstField, Picture* pcPicOrgSecondField,
PelUnitBuf cPicRecFirstField, PelUnitBuf cPicRecSecondField,
- const InputColourSpaceConversion conversion, const bool printFrameMSE, double* PSNR_Y )
+ const InputColourSpaceConversion conversion, const bool printFrameMSE, double* PSNR_Y
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+)
{
const SPS &sps = *pcPicOrgFirstField->cs->sps;
const ChromaFormat format = sps.getChromaFormatIdc();
@@ -3100,7 +3297,11 @@ void EncGOP::xCalculateInterlacedAddPSNR( Picture* pcPicOrgFirstField, Picture*
uint32_t uibits = 0; // the number of bits for the pair is not calculated here - instead the overall total is used elsewhere.
//===== add PSNR =====
- m_gcAnalyzeAll_in.addResult (dPSNR, (double)uibits, MSEyuvframe);
+ m_gcAnalyzeAll_in.addResult (dPSNR, (double)uibits, MSEyuvframe
+#if JVET_K0157
+ , isEncodeLtRef
+#endif
+ );
*PSNR_Y = dPSNR[COMPONENT_Y];
@@ -3130,13 +3331,21 @@ NalUnitType EncGOP::getNalUnitType(int pocCurr, int lastIDR, bool isField)
return NAL_UNIT_CODED_SLICE_IDR_W_RADL;
}
- if(m_pcCfg->getEfficientFieldIRAPEnabled() && isField && pocCurr == 1)
+#if JVET_K0157
+ if (m_pcCfg->getEfficientFieldIRAPEnabled() && isField && pocCurr == (m_pcCfg->getUseCompositeRef() ? 2: 1))
+#else
+ if (m_pcCfg->getEfficientFieldIRAPEnabled() && isField && pocCurr == 1)
+#endif
{
// to avoid the picture becoming an IRAP
return NAL_UNIT_CODED_SLICE_TRAIL_R;
}
- if(m_pcCfg->getDecodingRefreshType() != 3 && (pocCurr - isField) % m_pcCfg->getIntraPeriod() == 0)
+#if JVET_K0157
+ if (m_pcCfg->getDecodingRefreshType() != 3 && (pocCurr - isField) % (m_pcCfg->getIntraPeriod() * (m_pcCfg->getUseCompositeRef() ? 2 : 1)) == 0)
+#else
+ if (m_pcCfg->getDecodingRefreshType() != 3 && (pocCurr - isField) % m_pcCfg->getIntraPeriod() == 0)
+#endif
{
if (m_pcCfg->getDecodingRefreshType() == 1)
{
@@ -3355,6 +3564,223 @@ void EncGOP::arrangeLongtermPicturesInRPS(Slice *pcSlice, PicList& rcListPic)
}
}
+#if JVET_K0157
+void EncGOP::arrangeCompositeReference(Slice* pcSlice, PicList& rcListPic, int pocCurr)
+{
+ Picture* curPic = NULL;
+ PicList::iterator iterPic = rcListPic.begin();
+ const PreCalcValues *pcv = pcSlice->getPPS()->pcv;
+ m_bgPOC = pocCurr + 1;
+ if (m_picBg->getSpliceFull())
+ {
+ return;
+ }
+ while (iterPic != rcListPic.end())
+ {
+ curPic = *(iterPic++);
+ if (curPic->getPOC() == pocCurr)
+ {
+ break;
+ }
+ }
+ if (pcSlice->getSliceType() == I_SLICE)
+ {
+ return;
+ }
+
+ int width = pcv->lumaWidth;
+ int height = pcv->lumaHeight;
+ int stride = curPic->getOrigBuf().get(COMPONENT_Y).stride;
+ int cStride = curPic->getOrigBuf().get(COMPONENT_Cb).stride;
+ Pel* curLumaAddr = curPic->getOrigBuf().get(COMPONENT_Y).buf;
+ Pel* curCbAddr = curPic->getOrigBuf().get(COMPONENT_Cb).buf;
+ Pel* curCrAddr = curPic->getOrigBuf().get(COMPONENT_Cr).buf;
+ Pel* bgOrgLumaAddr = m_picOrig->getOrigBuf().get(COMPONENT_Y).buf;
+ Pel* bgOrgCbAddr = m_picOrig->getOrigBuf().get(COMPONENT_Cb).buf;
+ Pel* bgOrgCrAddr = m_picOrig->getOrigBuf().get(COMPONENT_Cr).buf;
+ int cuMaxWidth = pcv->maxCUWidth;
+ int cuMaxHeight = pcv->maxCUHeight;
+ int maxReplace = (pcv->sizeInCtus) / 2;
+ maxReplace = maxReplace < 1 ? 1 : maxReplace;
+ typedef struct tagCostStr
+ {
+ double cost;
+ int ctuIdx;
+ }CostStr;
+ CostStr* minCtuCost = new CostStr[maxReplace];
+ for (int i = 0; i < maxReplace; i++)
+ {
+ minCtuCost[i].cost = 1e10;
+ minCtuCost[i].ctuIdx = -1;
+ }
+ int bitIncrementY = pcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8;
+ int bitIncrementUV = pcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_CHROMA) - 8;
+ for (int y = 0; y < height; y += cuMaxHeight)
+ {
+ for (int x = 0; x < width; x += cuMaxWidth)
+ {
+ double lcuDist = 0.0;
+ double lcuDistCb = 0.0;
+ double lcuDistCr = 0.0;
+ int realPixelCnt = 0;
+ double lcuCost = 1e10;
+ int largeDist = 0;
+
+ for (int tmpy = 0; tmpy < cuMaxHeight; tmpy++)
+ {
+ if (y + tmpy >= height)
+ {
+ break;
+ }
+ for (int tmpx = 0; tmpx < cuMaxWidth; tmpx++)
+ {
+ if (x + tmpx >= width)
+ {
+ break;
+ }
+
+ realPixelCnt++;
+ lcuDist += abs(curLumaAddr[(y + tmpy)*stride + x + tmpx] - bgOrgLumaAddr[(y + tmpy)*stride + x + tmpx]);
+ if (abs(curLumaAddr[(y + tmpy)*stride + x + tmpx] - bgOrgLumaAddr[(y + tmpy)*stride + x + tmpx]) >(20 << bitIncrementY))
+ {
+ largeDist++;
+ }
+
+ if (tmpy % 2 == 0 && tmpx % 2 == 0)
+ {
+ lcuDistCb += abs(curCbAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2] - bgOrgCbAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2]);
+ lcuDistCr += abs(curCrAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2] - bgOrgCrAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2]);
+ }
+ }
+ }
+
+ //Test the vertical or horizontal edge for background patches candidates
+ int yInLCU = y / cuMaxHeight;
+ int xInLCU = x / cuMaxWidth;
+ int iLCUIdx = yInLCU * pcv->widthInCtus + xInLCU;
+ if ((largeDist / (double)realPixelCnt < 0.01 &&lcuDist / realPixelCnt < (3.5 * (1 << bitIncrementY)) && lcuDistCb / realPixelCnt < (0.5 * (1 << bitIncrementUV)) && lcuDistCr / realPixelCnt < (0.5 * (1 << bitIncrementUV)) && m_picBg->getSpliceIdx(iLCUIdx) == 0))
+ {
+ lcuCost = lcuDist / realPixelCnt + lcuDistCb / realPixelCnt + lcuDistCr / realPixelCnt;
+ //obtain the maxReplace smallest cost
+ //1) find the largest cost in the maxReplace candidates
+ for (int i = 0; i < maxReplace - 1; i++)
+ {
+ if (minCtuCost[i].cost > minCtuCost[i + 1].cost)
+ {
+ swap(minCtuCost[i].cost, minCtuCost[i + 1].cost);
+ swap(minCtuCost[i].ctuIdx, minCtuCost[i + 1].ctuIdx);
+ }
+ }
+ // 2) compare the current cost with the largest cost
+ if (lcuCost < minCtuCost[maxReplace - 1].cost)
+ {
+ minCtuCost[maxReplace - 1].cost = lcuCost;
+ minCtuCost[maxReplace - 1].ctuIdx = iLCUIdx;
+ }
+ }
+ }
+ }
+
+ // modify QP for background CTU
+ {
+ for (int i = 0; i < maxReplace; i++)
+ {
+ if (minCtuCost[i].ctuIdx != -1)
+ {
+ m_picBg->setSpliceIdx(minCtuCost[i].ctuIdx, pocCurr);
+ }
+ }
+ }
+ delete[]minCtuCost;
+}
+
+void EncGOP::updateCompositeReference(Slice* pcSlice, PicList& rcListPic, int pocCurr)
+{
+ Picture* curPic = NULL;
+ const PreCalcValues *pcv = pcSlice->getPPS()->pcv;
+ PicList::iterator iterPic = rcListPic.begin();
+ iterPic = rcListPic.begin();
+ while (iterPic != rcListPic.end())
+ {
+ curPic = *(iterPic++);
+ if (curPic->getPOC() == pocCurr)
+ {
+ break;
+ }
+ }
+ assert(curPic->getPOC() == pocCurr);
+
+ int width = pcv->lumaWidth;
+ int height = pcv->lumaHeight;
+ int stride = curPic->getRecoBuf().get(COMPONENT_Y).stride;
+ int cStride = curPic->getRecoBuf().get(COMPONENT_Cb).stride;
+
+ Pel* bgLumaAddr = m_picBg->getRecoBuf().get(COMPONENT_Y).buf;
+ Pel* bgCbAddr = m_picBg->getRecoBuf().get(COMPONENT_Cb).buf;
+ Pel* bgCrAddr = m_picBg->getRecoBuf().get(COMPONENT_Cr).buf;
+ Pel* curLumaAddr = curPic->getRecoBuf().get(COMPONENT_Y).buf;
+ Pel* curCbAddr = curPic->getRecoBuf().get(COMPONENT_Cb).buf;
+ Pel* curCrAddr = curPic->getRecoBuf().get(COMPONENT_Cr).buf;
+
+ int maxCuWidth = pcv->maxCUWidth;
+ int maxCuHeight = pcv->maxCUHeight;
+
+ // Update background reference
+ if (pcSlice->getSliceType() == I_SLICE)//(pocCurr == 0)
+ {
+ curPic->extendPicBorder();
+ curPic->setBorderExtension(true);
+
+ m_picBg->getRecoBuf().copyFrom(curPic->getRecoBuf());
+ m_picOrig->getOrigBuf().copyFrom(curPic->getOrigBuf());
+ }
+ else
+ {
+ //cout << "update B" << pocCurr << endl;
+ for (int y = 0; y < height; y += maxCuHeight)
+ {
+ for (int x = 0; x < width; x += maxCuWidth)
+ {
+ if (m_picBg->getSpliceIdx((y / maxCuHeight)*pcv->widthInCtus + x / maxCuWidth) == pocCurr)
+ {
+ for (int tmpy = 0; tmpy < maxCuHeight; tmpy++)
+ {
+ if (y + tmpy >= height)
+ {
+ break;
+ }
+ for (int tmpx = 0; tmpx < maxCuWidth; tmpx++)
+ {
+ if (x + tmpx >= width)
+ {
+ break;
+ }
+ bgLumaAddr[(y + tmpy)*stride + x + tmpx] = curLumaAddr[(y + tmpy)*stride + x + tmpx];
+ if (tmpy % 2 == 0 && tmpx % 2 == 0)
+ {
+ bgCbAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2] = curCbAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2];
+ bgCrAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2] = curCrAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2];
+ }
+ }
+ }
+ }
+ }
+ }
+ m_picBg->setBorderExtension(false);
+ m_picBg->extendPicBorder();
+ m_picBg->setBorderExtension(true);
+
+ curPic->extendPicBorder();
+ curPic->setBorderExtension(true);
+ m_picOrig->getOrigBuf().copyFrom(curPic->getOrigBuf());
+
+ m_picBg->setBorderExtension(false);
+ m_picBg->extendPicBorder();
+ m_picBg->setBorderExtension(true);
+ }
+}
+#endif
+
void EncGOP::applyDeblockingFilterMetric( Picture* pcPic, uint32_t uiNumSlices )
{
PelBuf cPelBuf = pcPic->getRecoBuf().get( COMPONENT_Y );
diff --git a/source/Lib/EncoderLib/EncGOP.h b/source/Lib/EncoderLib/EncGOP.h
index e7aa27d50d8fcaf15fb0e4de1b90d76ad165c62b..84c60e73bbc0a541bb385ddaa2ffd822fe42b803 100644
--- a/source/Lib/EncoderLib/EncGOP.h
+++ b/source/Lib/EncoderLib/EncGOP.h
@@ -131,6 +131,15 @@ private:
SEIWriter m_seiWriter;
+#if JVET_K0157
+ Picture * m_picBg;
+ Picture * m_picOrig;
+ int m_bgPOC;
+ bool m_isEncodedLTRef;
+ bool m_isPrepareLTRef;
+ bool m_isUseLTRef;
+ int m_lastLTRefPoc;
+#endif
//--Adaptive Loop filter
EncSampleAdaptiveOffset* m_pcSAO;
#if JVET_K0371_ALF
@@ -173,21 +182,45 @@ public:
void init ( EncLib* pcEncLib );
void compressGOP ( int iPOCLast, int iNumPicRcvd, PicList& rcListPic, std::list<PelUnitBuf*>& rcListPicYuvRec,
- bool isField, bool isTff, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE );
+ bool isField, bool isTff, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+ );
void xAttachSliceDataToNalUnit (OutputNALUnit& rNalu, OutputBitstream* pcBitstreamRedirect);
int getGOPSize() { return m_iGopSize; }
PicList* getListPic() { return m_pcListPic; }
+#if JVET_K0157
+ void setPicBg(Picture* tmpPicBg) { m_picBg = tmpPicBg; }
+ Picture* getPicBg() const { return m_picBg; }
+ void setPicOrig(Picture* tmpPicBg) { m_picOrig = tmpPicBg; }
+ Picture* getPicOrig() { return m_picOrig; }
+ void setNewestBgPOC(int poc) { m_bgPOC = poc; }
+ int getNewestBgPOC() const { return m_bgPOC; }
+ void setEncodedLTRef(bool isEncodedLTRef) { m_isEncodedLTRef = isEncodedLTRef; }
+ bool getEncodedLTRef() { return m_isEncodedLTRef; }
+ void setUseLTRef(bool isUseLTRef) { m_isUseLTRef = isUseLTRef; }
+ bool getUseLTRef() { return m_isUseLTRef; }
+ void setPrepareLTRef(bool isPrepareLTRef) { m_isPrepareLTRef = isPrepareLTRef; }
+ bool getPrepareLTRef() { return m_isPrepareLTRef; }
+ void setLastLTRefPoc(int iLastLTRefPoc) { m_lastLTRefPoc = iLastLTRefPoc; }
+ int getLastLTRefPoc() const { return m_lastLTRefPoc; }
- void printOutSummary ( uint32_t uiNumAllPicCoded, bool isField, const bool printMSEBasedSNR, const bool printSequenceMSE, const BitDepths &bitDepths );
+#endif
+ void printOutSummary ( uint32_t uiNumAllPicCoded, bool isField, const bool printMSEBasedSNR, const bool printSequenceMSE, const bool printHexPsnr, const BitDepths &bitDepths );
#if W0038_DB_OPT
uint64_t preLoopFilterPicAndCalcDist( Picture* pcPic );
#endif
EncSlice* getSliceEncoder() { return m_pcSliceEncoder; }
NalUnitType getNalUnitType( int pocCurr, int lastIdr, bool isField );
void arrangeLongtermPicturesInRPS(Slice *, PicList& );
+#if JVET_K0157
+ void arrangeCompositeReference(Slice* pcSlice, PicList& rcListPic, int pocCurr);
+ void updateCompositeReference(Slice* pcSlice, PicList& rcListPic, int pocCurr);
+#endif
#if EXTENSION_360_VIDEO
Analyze& getAnalyzeAllData() { return m_gcAnalyzeAll; }
@@ -201,15 +234,31 @@ protected:
protected:
- void xInitGOP ( int iPOCLast, int iNumPicRcvd, bool isField );
+ void xInitGOP ( int iPOCLast, int iNumPicRcvd, bool isField
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+ );
void xGetBuffer ( PicList& rcListPic, std::list<PelUnitBuf*>& rcListPicYuvRecOut,
int iNumPicRcvd, int iTimeOffset, Picture*& rpcPic, int pocCurr, bool isField );
- void xCalculateAddPSNRs ( const bool isField, const bool isFieldTopFieldFirst, const int iGOPid, Picture* pcPic, const AccessUnit&accessUnit, PicList &rcListPic, int64_t dEncTime, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE, double* PSNR_Y );
- void xCalculateAddPSNR ( Picture* pcPic, PelUnitBuf cPicD, const AccessUnit&, double dEncTime, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE, double* PSNR_Y );
+ void xCalculateAddPSNRs(const bool isField, const bool isFieldTopFieldFirst, const int iGOPid, Picture* pcPic, const AccessUnit&accessUnit, PicList &rcListPic, int64_t dEncTime, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE, double* PSNR_Y
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+ );
+ void xCalculateAddPSNR(Picture* pcPic, PelUnitBuf cPicD, const AccessUnit&, double dEncTime, const InputColourSpaceConversion snr_conversion, const bool printFrameMSE, double* PSNR_Y
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+ );
void xCalculateInterlacedAddPSNR( Picture* pcPicOrgFirstField, Picture* pcPicOrgSecondField,
PelUnitBuf cPicRecFirstField, PelUnitBuf cPicRecSecondField,
- const InputColourSpaceConversion snr_conversion, const bool printFrameMSE, double* PSNR_Y );
+ const InputColourSpaceConversion snr_conversion, const bool printFrameMSE, double* PSNR_Y
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+ );
uint64_t xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1, const uint32_t rshift
#if ENABLE_QPA
diff --git a/source/Lib/EncoderLib/EncLib.cpp b/source/Lib/EncoderLib/EncLib.cpp
index 9612ec3650d013e0083c3b68f591cdfba97b2047..8429fde15c8b81878a62b94ae85474446d06e14b 100644
--- a/source/Lib/EncoderLib/EncLib.cpp
+++ b/source/Lib/EncoderLib/EncLib.cpp
@@ -87,7 +87,9 @@ void EncLib::create ()
-
+#if JVET_K0157
+ m_iPOCLast = m_compositeRefEnabled ? -2 : -1;
+#endif
// create processing unit classes
m_cGOPEncoder. create( );
m_cSliceEncoder. create( getSourceWidth(), getSourceHeight(), m_chromaFormatIDC, m_maxCUWidth, m_maxCUHeight, m_maxTotalCUDepth );
@@ -140,8 +142,13 @@ void EncLib::create ()
if ( m_RCEnableRateControl )
{
+#if RATECTRL_FIX_FULLNBIT
+ m_cRateCtrl.init(m_framesToBeEncoded, m_RCTargetBitrate, (int)((double)m_iFrameRate / m_temporalSubsampleRatio + 0.5), m_iGOPSize, m_iSourceWidth, m_iSourceHeight,
+ m_maxCUWidth, m_maxCUHeight, getBitDepth(CHANNEL_TYPE_LUMA), m_RCKeepHierarchicalBit, m_RCUseLCUSeparateModel, m_GOPList);
+#else
m_cRateCtrl.init( m_framesToBeEncoded, m_RCTargetBitrate, (int)( (double)m_iFrameRate/m_temporalSubsampleRatio + 0.5), m_iGOPSize, m_iSourceWidth, m_iSourceHeight,
m_maxCUWidth, m_maxCUHeight,m_RCKeepHierarchicalBit, m_RCUseLCUSeparateModel, m_GOPList );
+#endif
}
}
@@ -219,6 +226,12 @@ void EncLib::init( bool isFieldCoding, AUWriterIf* auWriterIf )
omp_set_nested( true );
#endif
+#if JVET_K0157
+ if (sps0.getSpsNext().getUseCompositeRef())
+ {
+ sps0.setLongTermRefsPresent(true);
+ }
+#endif
#if U0132_TARGET_BITS_SATURATION
if (m_RCCpbSaturationEnabled)
@@ -248,6 +261,14 @@ void EncLib::init( bool isFieldCoding, AUWriterIf* auWriterIf )
xInitPPS(pps1, sps0);
}
#endif
+#if JVET_K0157
+ if (sps0.getSpsNext().getUseCompositeRef())
+ {
+ PPS &pps2 = *(m_ppsMap.allocatePS(2));
+ xInitPPS(pps2, sps0);
+ xInitPPSforLT(pps2);
+ }
+#endif
// initialize processing unit classes
m_cGOPEncoder. init( this );
@@ -356,6 +377,22 @@ void EncLib::init( bool isFieldCoding, AUWriterIf* auWriterIf )
#if ENABLE_WPP_PARALLELISM
m_entropyCodingSyncContextStateVec.resize( pps0.pcv->heightInCtus );
#endif
+#if JVET_K0157
+ if (sps0.getSpsNext().getUseCompositeRef())
+ {
+ Picture *picBg = new Picture;
+ picBg->create(sps0.getChromaFormatIdc(), Size(sps0.getPicWidthInLumaSamples(), sps0.getPicHeightInLumaSamples()), sps0.getMaxCUWidth(), sps0.getMaxCUWidth() + 16, false);
+ picBg->getRecoBuf().fill(0);
+ picBg->finalInit(sps0, pps0);
+ picBg->allocateNewSlice();
+ picBg->createSpliceIdx(pps0.pcv->sizeInCtus);
+ m_cGOPEncoder.setPicBg(picBg);
+ Picture *picOrig = new Picture;
+ picOrig->create(sps0.getChromaFormatIdc(), Size(sps0.getPicWidthInLumaSamples(), sps0.getPicHeightInLumaSamples()), sps0.getMaxCUWidth(), sps0.getMaxCUWidth() + 16, false);
+ picOrig->getOrigBuf().fill(0);
+ m_cGOPEncoder.setPicOrig(picOrig);
+ }
+#endif
}
#if HEVC_USE_SCALING_LISTS
@@ -441,6 +478,15 @@ void EncLib::xInitScalingLists(SPS &sps, PPS &pps)
}
#endif
+#if JVET_K0157
+void EncLib::xInitPPSforLT(PPS& pps)
+{
+ pps.setOutputFlagPresentFlag(true);
+ pps.setDeblockingFilterControlPresentFlag(true);
+ pps.setPPSDeblockingFilterDisabledFlag(true);
+}
+#endif
+
// ====================================================================================================================
// Public member functions
// ====================================================================================================================
@@ -483,6 +529,38 @@ void EncLib::deletePicBuffer()
void EncLib::encode( bool flush, PelStorage* pcPicYuvOrg, PelStorage* cPicYuvTrueOrg, const InputColourSpaceConversion snrCSC, std::list<PelUnitBuf*>& rcListPicYuvRecOut,
int& iNumEncoded )
{
+#if JVET_K0157
+ if (m_compositeRefEnabled && m_cGOPEncoder.getPicBg()->getSpliceFull() && m_iPOCLast >= 10 && m_iNumPicRcvd == 0 && m_cGOPEncoder.getEncodedLTRef() == false)
+ {
+ Picture* picCurr = NULL;
+ xGetNewPicBuffer(rcListPicYuvRecOut, picCurr, 2);
+ const PPS *pps = m_ppsMap.getPS(2);
+ const SPS *sps = m_spsMap.getPS(pps->getSPSId());
+
+ picCurr->M_BUFS(0, PIC_ORIGINAL).copyFrom(m_cGOPEncoder.getPicBg()->getRecoBuf());
+ picCurr->finalInit(*sps, *pps);
+ picCurr->poc = m_iPOCLast - 1;
+ m_iPOCLast -= 2;
+ if (getUseAdaptiveQP())
+ {
+ AQpPreanalyzer::preanalyze(picCurr);
+ }
+ if (m_RCEnableRateControl)
+ {
+ m_cRateCtrl.initRCGOP(m_iNumPicRcvd);
+ }
+ m_cGOPEncoder.compressGOP(m_iPOCLast, m_iNumPicRcvd, m_cListPic, rcListPicYuvRecOut,
+ false, false, snrCSC, m_printFrameMSE, true);
+ m_cGOPEncoder.setEncodedLTRef(true);
+ if (m_RCEnableRateControl)
+ {
+ m_cRateCtrl.destroyRCGOP();
+ }
+
+ iNumEncoded = 0;
+ m_iNumPicRcvd = 0;
+ }
+#endif
//PROF_ACCUM_AND_START_NEW_SET( getProfilerPic(), P_GOP_LEVEL );
if (pcPicYuvOrg != NULL)
{
@@ -493,7 +571,11 @@ void EncLib::encode( bool flush, PelStorage* pcPicYuvOrg, PelStorage* cPicYuvTru
int ppsID=-1; // Use default PPS ID
if (getWCGChromaQPControl().isEnabled())
{
+#if JVET_K0157
+ ppsID = getdQPs()[m_iPOCLast / (m_compositeRefEnabled ? 2 : 1) + 1];
+#else
ppsID=getdQPs()[ m_iPOCLast+1 ];
+#endif
ppsID+=(getSwitchPOC() != -1 && (m_iPOCLast+1 >= getSwitchPOC())?1:0);
}
xGetNewPicBuffer( rcListPicYuvRecOut,
@@ -533,8 +615,12 @@ void EncLib::encode( bool flush, PelStorage* pcPicYuvOrg, PelStorage* cPicYuvTru
}
// compress GOP
- m_cGOPEncoder.compressGOP( m_iPOCLast, m_iNumPicRcvd, m_cListPic, rcListPicYuvRecOut,
- false, false, snrCSC, m_printFrameMSE );
+ m_cGOPEncoder.compressGOP(m_iPOCLast, m_iNumPicRcvd, m_cListPic, rcListPicYuvRecOut,
+ false, false, snrCSC, m_printFrameMSE
+#if JVET_K0157
+ , false
+#endif
+ );
if ( m_RCEnableRateControl )
{
@@ -622,8 +708,11 @@ void EncLib::encode( bool flush, PelStorage* pcPicYuvOrg, PelStorage* pcPicYuvTr
if ( m_iNumPicRcvd && ((flush&&fieldNum==1) || (m_iPOCLast/2)==0 || m_iNumPicRcvd==m_iGOPSize ) )
{
// compress GOP
- m_cGOPEncoder.compressGOP( m_iPOCLast, m_iNumPicRcvd, m_cListPic, rcListPicYuvRecOut,
- true, isTff, snrCSC, m_printFrameMSE );
+ m_cGOPEncoder.compressGOP(m_iPOCLast, m_iNumPicRcvd, m_cListPic, rcListPicYuvRecOut, true, isTff, snrCSC, m_printFrameMSE
+#if JVET_K0157
+ , false
+#endif
+ );
iNumEncoded += m_iNumPicRcvd;
m_uiNumAllPicCoded += m_iNumPicRcvd;
@@ -711,8 +800,11 @@ void EncLib::xGetNewPicBuffer ( std::list<PelUnitBuf*>& rcListPicYuvRecOut, Pict
rpcPic->reconstructed = false;
rpcPic->referenced = true;
-
+#if JVET_K0157
+ m_iPOCLast += (m_compositeRefEnabled ? 2 : 1);
+#else
m_iPOCLast++;
+#endif
m_iNumPicRcvd++;
}
@@ -821,6 +913,9 @@ void EncLib::xInitSPS(SPS &sps)
sps.getSpsNext().setUseIntraEMT ( m_IntraEMT );
sps.getSpsNext().setUseInterEMT ( m_InterEMT );
#endif
+#if JVET_K0157
+ sps.getSpsNext().setUseCompositeRef ( m_compositeRefEnabled );
+#endif
// ADD_NEW_TOOL : (encoder lib) set tool enabling flags and associated parameters here
@@ -861,7 +956,7 @@ void EncLib::xInitSPS(SPS &sps)
sps.setMaxTLayers( m_maxTempLayer );
sps.setTemporalIdNestingFlag( ( m_maxTempLayer == 1 ) ? true : false );
- for (int i = 0; i < min(sps.getMaxTLayers(),(uint32_t) MAX_TLAYER); i++ )
+ for (int i = 0; i < std::min(sps.getMaxTLayers(), (uint32_t) MAX_TLAYER); i++ )
{
sps.setMaxDecPicBuffering(m_maxDecPicBuffering[i], i);
sps.setNumReorderPics(m_numReorderPics[i], i);
@@ -1151,7 +1246,7 @@ void EncLib::xInitPPS(PPS &pps, const SPS &sps)
if (getUsePerceptQPA() && !bUseDQP)
{
CHECK( m_iMaxCuDQPDepth != 0, "max. delta-QP depth must be zero!" );
- bUseDQP = true;
+ bUseDQP = (getBaseQP() < 38) && (getSourceWidth() > 512 || getSourceHeight() > 320);
}
#endif
@@ -1245,9 +1340,15 @@ void EncLib::xInitPPS(PPS &pps, const SPS &sps)
}
}
}
+ #if ENABLE_QPA
+ if ((getUsePerceptQPA() || getSliceChromaOffsetQpPeriodicity() > 0) && (getChromaFormatIdc() != CHROMA_400))
+ {
+ bChromaDeltaQPEnabled = true;
+ }
+ #endif
pps.setSliceChromaQpFlag(bChromaDeltaQPEnabled);
#endif
- if (!pps.getSliceChromaQpFlag() && sps.getSpsNext().getUseDualITree())
+ if (!pps.getSliceChromaQpFlag() && sps.getSpsNext().getUseDualITree() && (getChromaFormatIdc() != CHROMA_400))
{
pps.setSliceChromaQpFlag(m_chromaCbQpOffsetDualTree != 0 || m_chromaCrQpOffsetDualTree != 0);
}
@@ -1323,7 +1424,7 @@ void EncLib::xInitPPS(PPS &pps, const SPS &sps)
}
}
CHECK(!(bestPos <= 15), "Unspecified error");
- pps.setNumRefIdxL0DefaultActive(bestPos);
+ pps.setNumRefIdxL0DefaultActive(bestPos);
pps.setNumRefIdxL1DefaultActive(bestPos);
pps.setTransquantBypassEnabledFlag(getTransquantBypassEnabledFlag());
pps.setUseTransformSkip( m_useTransformSkip );
@@ -1495,8 +1596,20 @@ void EncLib::xInitRPS(SPS &sps, bool isFieldCoding)
// This is a function that
// determines what Reference Picture Set to use
// for a specific slice (with POC = POCCurr)
-void EncLib::selectReferencePictureSet(Slice* slice, int POCCurr, int GOPid )
+void EncLib::selectReferencePictureSet(Slice* slice, int POCCurr, int GOPid
+#if JVET_K0157
+ , int ltPoc
+#endif
+)
{
+#if JVET_K0157
+ bool isEncodeLtRef = (POCCurr == ltPoc);
+ if (m_compositeRefEnabled && isEncodeLtRef)
+ {
+ POCCurr++;
+ }
+ int rIdx = GOPid;
+#endif
slice->setRPSidx(GOPid);
for(int extraNum=m_iGOPSize; extraNum<m_extraRPSs+m_iGOPSize; extraNum++)
@@ -1511,6 +1624,9 @@ void EncLib::selectReferencePictureSet(Slice* slice, int POCCurr, int GOPid )
if(POCIndex == m_GOPList[extraNum].m_POC)
{
slice->setRPSidx(extraNum);
+#if JVET_K0157
+ rIdx = extraNum;
+#endif
}
}
else
@@ -1518,6 +1634,9 @@ void EncLib::selectReferencePictureSet(Slice* slice, int POCCurr, int GOPid )
if(POCCurr==m_GOPList[extraNum].m_POC)
{
slice->setRPSidx(extraNum);
+#if JVET_K0157
+ rIdx = extraNum;
+#endif
}
}
}
@@ -1525,9 +1644,77 @@ void EncLib::selectReferencePictureSet(Slice* slice, int POCCurr, int GOPid )
if(POCCurr == 1 && slice->getPic()->fieldPic)
{
slice->setRPSidx(m_iGOPSize+m_extraRPSs);
+#if JVET_K0157
+ rIdx = m_iGOPSize + m_extraRPSs;
+#endif
}
+#if JVET_K0157
+ ReferencePictureSet *rps = const_cast<ReferencePictureSet *>(slice->getSPS()->getRPSList()->getReferencePictureSet(slice->getRPSidx()));
+ if (m_compositeRefEnabled && ltPoc != -1 && !isEncodeLtRef)
+ {
+ if (ltPoc != -1 && rps->getNumberOfLongtermPictures() != 1 && !isEncodeLtRef)
+ {
+ int idx = rps->getNumberOfPictures();
+ int maxPicOrderCntLSB = 1 << slice->getSPS()->getBitsForPOC();
+ int ltPocLsb = ltPoc % maxPicOrderCntLSB;
+
+ rps->setNumberOfPictures(rps->getNumberOfPictures() + 1);
+ rps->setNumberOfLongtermPictures(1);
+ rps->setPOC(idx, ltPoc);
+ rps->setPocLSBLT(idx, ltPocLsb);
+ rps->setDeltaPOC(idx, -POCCurr + ltPoc);
+ rps->setUsed(idx, true);
+ }
+ }
+ else if (m_compositeRefEnabled && isEncodeLtRef)
+ {
+ ReferencePictureSet* localRPS = slice->getLocalRPS();
+ (*localRPS) = ReferencePictureSet();
+ int refPics = rps->getNumberOfPictures();
+ localRPS->setNumberOfPictures(rps->getNumberOfPictures());
+ for (int i = 0; i < refPics; i++)
+ {
+ localRPS->setDeltaPOC(i, rps->getDeltaPOC(i) + 1);
+ localRPS->setUsed(i, rps->getUsed(i));
+ }
+ localRPS->setNumberOfNegativePictures(rps->getNumberOfNegativePictures());
+ localRPS->setNumberOfPositivePictures(rps->getNumberOfPositivePictures());
+ localRPS->setInterRPSPrediction(true);
+ int deltaRPS = 1;
+ int newIdc = 0;
+ for (int i = 0; i < refPics; i++)
+ {
+ int deltaPOC = ((i != refPics) ? rps->getDeltaPOC(i) : 0); // check if the reference abs POC is >= 0
+ int refIdc = 0;
+ for (int j = 0; j < localRPS->getNumberOfPictures(); j++) // loop through the pictures in the new RPS
+ {
+ if ((deltaPOC + deltaRPS) == localRPS->getDeltaPOC(j))
+ {
+ if (localRPS->getUsed(j))
+ {
+ refIdc = 1;
+ }
+ else
+ {
+ refIdc = 2;
+ }
+ }
+ }
+ localRPS->setRefIdc(i, refIdc);
+ newIdc++;
+ }
+ localRPS->setNumRefIdc(newIdc + 1);
+ localRPS->setRefIdc(newIdc, 0);
+ localRPS->setDeltaRPS(deltaRPS);
+ localRPS->setDeltaRIdxMinus1(slice->getSPS()->getRPSList()->getNumberOfReferencePictureSets() - 1 - rIdx);
+ slice->setRPS(localRPS);
+ slice->setRPSidx(-1);
+ return;
+ }
+#else
const ReferencePictureSet *rps = (slice->getSPS()->getRPSList()->getReferencePictureSet(slice->getRPSidx()));
+#endif
slice->setRPS(rps);
}
@@ -1678,7 +1865,11 @@ int EncCfg::getQPForPicture(const uint32_t gopIndex, const Slice *pSlice) const
const int* pdQPs = getdQPs();
if ( pdQPs )
{
+#if JVET_K0157
+ qp += pdQPs[pSlice->getPOC() / (m_compositeRefEnabled ? 2 : 1)];
+#else
qp += pdQPs[ pSlice->getPOC() ];
+#endif
}
#endif
diff --git a/source/Lib/EncoderLib/EncLib.h b/source/Lib/EncoderLib/EncLib.h
index 7fbdd7aec8d55f30710f93ea2eac8991fefd86de..74d78d10a3909ce88fdf86126d6744ef29404a82 100644
--- a/source/Lib/EncoderLib/EncLib.h
+++ b/source/Lib/EncoderLib/EncLib.h
@@ -149,6 +149,9 @@ protected:
void xInitPPS (PPS &pps, const SPS &sps); ///< initialize PPS from encoder options
#if HEVC_USE_SCALING_LISTS
void xInitScalingLists (SPS &sps, PPS &pps); ///< initialize scaling lists
+#endif
+#if JVET_K0157
+ void xInitPPSforLT(PPS& pps);
#endif
void xInitHrdParameters(SPS &sps); ///< initialize HRD parameters
@@ -209,7 +212,12 @@ public:
#endif
RateCtrl* getRateCtrl () { return &m_cRateCtrl; }
- void selectReferencePictureSet(Slice* slice, int POCCurr, int GOPid );
+
+ void selectReferencePictureSet(Slice* slice, int POCCurr, int GOPid
+#if JVET_K0157
+ , int ltPoc
+#endif
+ );
int getReferencePictureSetIdxForSOP(int POCCurr, int GOPid );
bool PPSNeedsWriting(int ppsId);
@@ -240,7 +248,7 @@ public:
int& iNumEncoded, bool isTff );
- void printSummary(bool isField) { m_cGOPEncoder.printOutSummary (m_uiNumAllPicCoded, isField, m_printMSEBasedSequencePSNR, m_printSequenceMSE, m_spsMap.getFirstPS()->getBitDepths()); }
+ void printSummary(bool isField) { m_cGOPEncoder.printOutSummary (m_uiNumAllPicCoded, isField, m_printMSEBasedSequencePSNR, m_printSequenceMSE, m_printHexPsnr, m_spsMap.getFirstPS()->getBitDepths()); }
};
diff --git a/source/Lib/EncoderLib/EncModeCtrl.cpp b/source/Lib/EncoderLib/EncModeCtrl.cpp
index 6823598fa38ad57b5b996b6f8d056ef24430b368..dd8cd9bd160d0811c5c7abd42b5a20ee9e1d2c15 100644
--- a/source/Lib/EncoderLib/EncModeCtrl.cpp
+++ b/source/Lib/EncoderLib/EncModeCtrl.cpp
@@ -743,7 +743,8 @@ bool BestEncInfoCache::isValid( const CodingStructure& cs, const Partitioner& pa
BestEncodingInfo& encInfo = *m_bestEncInfo[idx1][idx2][idx3][idx4];
- if( cs.picture->poc != encInfo.poc || CS::getArea( cs, cs.area, partitioner.chType ) != encInfo.cu || !isTheSameNbHood( encInfo.cu, partitioner ) )
+ if( cs.picture->poc != encInfo.poc || CS::getArea( cs, cs.area, partitioner.chType ) != encInfo.cu || !isTheSameNbHood( encInfo.cu, partitioner )
+ )
{
return false;
}
@@ -942,7 +943,15 @@ void EncModeCtrlMTnoRQT::initCTUEncoding( const Slice &slice )
if( m_pcEncCfg->getUseE0023FastEnc() )
{
- m_skipThreshold = ( ( slice.getMinPictureDistance() <= PICTURE_DISTANCE_TH ) ? FAST_SKIP_DEPTH : SKIP_DEPTH );
+#if JVET_K0157
+ if (m_pcEncCfg->getUseCompositeRef())
+ m_skipThreshold = ( ( slice.getMinPictureDistance() <= PICTURE_DISTANCE_TH * 2 ) ? FAST_SKIP_DEPTH : SKIP_DEPTH );
+ else
+ m_skipThreshold = ((slice.getMinPictureDistance() <= PICTURE_DISTANCE_TH) ? FAST_SKIP_DEPTH : SKIP_DEPTH);
+
+#else
+ m_skipThreshold = ((slice.getMinPictureDistance() <= PICTURE_DISTANCE_TH) ? FAST_SKIP_DEPTH : SKIP_DEPTH);
+#endif
}
else
{
@@ -1057,7 +1066,6 @@ void EncModeCtrlMTnoRQT::initCULevel( Partitioner &partitioner, const CodingStru
#endif
xGetMinMaxQP( minQP, maxQP, cs, partitioner, baseQP, *cs.sps, *cs.pps, true );
-
// Add coding modes here
// NOTE: Working back to front, as a stack, which is more efficient with the container
// NOTE: First added modes will be processed at the end.
@@ -1425,11 +1433,11 @@ bool EncModeCtrlMTnoRQT::tryMode( const EncTestMode& encTestmode, const CodingSt
{
return false;
}
-
if( lastTestMode().type != ETM_INTRA && cuECtx.bestCS && cuECtx.bestCU && interHadActive( cuECtx ) )
{
// Get SATD threshold from best Inter-CU
- if( !cs.slice->isIntra() && m_pcEncCfg->getUsePbIntraFast() )
+ if( !cs.slice->isIntra() && m_pcEncCfg->getUsePbIntraFast()
+ )
{
CodingUnit* bestCU = cuECtx.bestCU;
if( bestCU && CU::isInter( *bestCU ) )
diff --git a/source/Lib/EncoderLib/EncModeCtrl.h b/source/Lib/EncoderLib/EncModeCtrl.h
index d8caa932771b8cb34f4c63b0f5780a15889ae3c8..12fd159c5f840d80c6ceae961b95d7b776c32409 100644
--- a/source/Lib/EncoderLib/EncModeCtrl.h
+++ b/source/Lib/EncoderLib/EncModeCtrl.h
@@ -385,6 +385,8 @@ struct CodedCUInfo
bool validMv[NUM_REF_PIC_LIST_01][MAX_STORED_CU_INFO_REFS];
Mv saveMv [NUM_REF_PIC_LIST_01][MAX_STORED_CU_INFO_REFS];
+
+
#if ENABLE_SPLIT_PARALLELISM
uint64_t
@@ -431,6 +433,7 @@ public:
bool getMv ( const UnitArea& area, const RefPicList refPicList, const int iRefIdx, Mv& rMv ) const;
void setMv ( const UnitArea& area, const RefPicList refPicList, const int iRefIdx, const Mv& rMv );
+
};
#if REUSE_CU_RESULTS
diff --git a/source/Lib/EncoderLib/EncSlice.cpp b/source/Lib/EncoderLib/EncSlice.cpp
index 4c7541aa94352b937ea0772e01e0585067e5f29c..a11f93c01dad8f484cb893a85e4ec2a3ec201cf2 100644
--- a/source/Lib/EncoderLib/EncSlice.cpp
+++ b/source/Lib/EncoderLib/EncSlice.cpp
@@ -40,6 +40,9 @@
#include "EncLib.h"
#include "CommonLib/UnitTools.h"
#include "CommonLib/Picture.h"
+#if K0149_BLOCK_STATISTICS
+#include "CommonLib/dtrace_blockstatistics.h"
+#endif
#if ENABLE_WPP_PARALLELISM
#include <mutex>
@@ -57,6 +60,9 @@ extern recursive_mutex g_cache_mutex;
EncSlice::EncSlice()
: m_encCABACTableIdx(I_SLICE)
+#if ENABLE_QPA
+ , m_adaptedLumaQP(-1)
+#endif
{
}
@@ -140,7 +146,120 @@ EncSlice::setUpLambda( Slice* slice, const double dLambda, int iQP)
slice->setLambdas( dLambdas );
}
+#if ENABLE_QPA
+
+static inline int apprI3Log2 (const double d) // rounded 3*log2(d)
+{
+ return d < 1.5e-13 ? -128 : int (floor (3.0 * log (d) / log (2.0) + 0.5));
+}
+
+static void filterAndCalculateAverageEnergies (const Pel* pSrc, const int iSrcStride,
+ double &hpEner, const int iHeight, const int iWidth,
+ const uint32_t uBitDepth /* luma bit-depth (4-16) */)
+{
+ uint64_t saAct = 0;
+
+ // skip first row as there may be a black border frame
+ pSrc += iSrcStride;
+ // center rows
+ for (int y = 1; y < iHeight - 1; y++)
+ {
+ // skip column as there may be a black border frame
+
+ for (int x = 1; x < iWidth - 1; x++) // and columns
+ {
+ const int f = 12 * (int)pSrc[x ] - 2 * ((int)pSrc[x-1] + (int)pSrc[x+1] + (int)pSrc[x -iSrcStride] + (int)pSrc[x +iSrcStride])
+ - (int)pSrc[x-1-iSrcStride] - (int)pSrc[x+1-iSrcStride] - (int)pSrc[x-1+iSrcStride] - (int)pSrc[x+1+iSrcStride];
+ saAct += abs (f);
+ }
+ // skip column as there may be a black border frame
+ pSrc += iSrcStride;
+ }
+ // skip last row as there may be a black border frame
+
+ hpEner = double(saAct) / double((iWidth - 2) * (iHeight - 2));
+
+ // lower limit, compensate for highpass amplification
+ if (hpEner < double(1 << (uBitDepth - 4))) hpEner = double(1 << (uBitDepth - 4));
+}
+
+#ifndef GLOBAL_AVERAGING
+ #define GLOBAL_AVERAGING 1 // "global" averaging of a_k across a set instead of one picture
+#endif
+
+#if GLOBAL_AVERAGING
+static double getAveragePictureEnergy (const CPelBuf picOrig, const uint32_t uBitDepth)
+{
+ double hpEnerPic = 5.65625 * double(1 << (uBitDepth >> 1)); // square-root of a_pic value
+
+ if (picOrig.width > 2048 && picOrig.height > 1280) // for UHD/4K
+ {
+ hpEnerPic *= (4.0 / 5.65625);
+ }
+ else if (picOrig.width <= 1024 || picOrig.height <= 640) // 480p
+ {
+ hpEnerPic *= (8.0 / 5.65625);
+ }
+
+ return hpEnerPic;
+}
+#endif
+
+static int applyQPAdaptationChroma (Picture* const pcPic, Slice* const pcSlice, EncCfg* const pcEncCfg, const int sliceQP)
+{
+ double hpEner[MAX_NUM_COMPONENT] = {0.0, 0.0, 0.0};
+ int optSliceChromaQpOffset[2] = {0, 0};
+ int savedLumaQP = -1;
+
+ for (uint32_t comp = 0; comp < getNumberValidComponents (pcPic->chromaFormat); comp++)
+ {
+ const ComponentID compID = (ComponentID)comp;
+ const CPelBuf picOrig = pcPic->getOrigBuf (pcPic->block (compID));
+
+ filterAndCalculateAverageEnergies (picOrig.buf, picOrig.stride, hpEner[comp], picOrig.height, picOrig.width,
+ pcSlice->getSPS()->getBitDepth (toChannelType (compID)) - (isChroma (compID) ? 1 : 0));
+ if (isChroma (compID))
+ {
+ const int adaptChromaQPOffset = 2.0 * hpEner[comp] <= hpEner[0] ? 0 : apprI3Log2 (2.0 * hpEner[comp] / hpEner[0]);
+ #if GLOBAL_AVERAGING
+ int averageAdaptedLumaQP = Clip3 (0, MAX_QP, sliceQP + apprI3Log2 (hpEner[0] / getAveragePictureEnergy (pcPic->getOrigBuf().Y(), pcSlice->getSPS()->getBitDepth (CH_L))));
+ #else
+ int averageAdaptedLumaQP = Clip3 (0, MAX_QP, sliceQP); // mean slice QP
+ #endif
+ #if SHARP_LUMA_DELTA_QP
+
+ // change mean picture QP index based on picture's average luma value (Sharp)
+ if (pcEncCfg->getLumaLevelToDeltaQPMapping().mode == LUMALVL_TO_DQP_NUM_MODES)
+ {
+ const CPelBuf picLuma = pcPic->getOrigBuf().Y();
+ uint64_t uAvgLuma = 0;
+
+ for (SizeType y = 0; y < picLuma.height; y++)
+ {
+ for (SizeType x = 0; x < picLuma.width; x++)
+ {
+ uAvgLuma += (uint64_t)picLuma.at (x, y);
+ }
+ }
+ uAvgLuma = (uAvgLuma + (picLuma.area() >> 1)) / picLuma.area();
+
+ averageAdaptedLumaQP = Clip3 (0, MAX_QP, averageAdaptedLumaQP + 1 - int((3 * uAvgLuma * uAvgLuma) >> uint64_t (2 * pcSlice->getSPS()->getBitDepth (CH_L) - 1)));
+ }
+ #endif
+ const int lumaChromaMappingDQP = averageAdaptedLumaQP - getScaledChromaQP (averageAdaptedLumaQP, pcEncCfg->getChromaFormatIdc());
+
+ optSliceChromaQpOffset[comp-1] = std::min (3 + lumaChromaMappingDQP, adaptChromaQPOffset + lumaChromaMappingDQP);
+ if (savedLumaQP < 0) savedLumaQP = averageAdaptedLumaQP; // save it for later
+ }
+ }
+
+ pcEncCfg->setSliceChromaOffsetQpIntraOrPeriodic (pcEncCfg->getSliceChromaOffsetQpPeriodicity(), optSliceChromaQpOffset);
+
+ return savedLumaQP;
+}
+
+#endif // ENABLE_QPA
/**
- non-referenced frame marking
@@ -156,8 +275,11 @@ EncSlice::setUpLambda( Slice* slice, const double dLambda, int iQP)
\param rpcSlice slice header class
\param isField true for field coding
*/
-
-void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCurr, const int iGOPid, Slice*& rpcSlice, const bool isField )
+void EncSlice::initEncSlice(Picture* pcPic, const int pocLast, const int pocCurr, const int iGOPid, Slice*& rpcSlice, const bool isField
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+)
{
double dQP;
double dLambda;
@@ -166,7 +288,19 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
rpcSlice->setSliceBits(0);
rpcSlice->setPic( pcPic );
rpcSlice->initSlice();
+#if JVET_K0157
+ int multipleFactor = pcPic->cs->sps->getSpsNext().getUseCompositeRef() ? 2 : 1;
+ if (pcPic->cs->sps->getSpsNext().getUseCompositeRef() && isEncodeLtRef)
+ {
+ rpcSlice->setPicOutputFlag(false);
+ }
+ else
+ {
+ rpcSlice->setPicOutputFlag(true);
+ }
+#else
rpcSlice->setPicOutputFlag( true );
+#endif
rpcSlice->setPOC( pocCurr );
#if JVET_K0072
rpcSlice->setDepQuantEnabledFlag( m_pcCfg->getDepQuantEnabledFlag() );
@@ -190,7 +324,11 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
}
else
{
+#if JVET_K0157
+ poc = poc % (m_pcCfg->getGOPSize() * multipleFactor);
+#else
poc = poc % m_pcCfg->getGOPSize();
+#endif
}
if ( poc == 0 )
@@ -199,11 +337,19 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
}
else
{
+#if JVET_K0157
+ int step = m_pcCfg->getGOPSize() * multipleFactor;
+#else
int step = m_pcCfg->getGOPSize();
+#endif
depth = 0;
for( int i=step>>1; i>=1; i>>=1 )
{
+#if JVET_K0157
+ for (int j = i; j<(m_pcCfg->getGOPSize() * multipleFactor); j += step)
+#else
for ( int j=i; j<m_pcCfg->getGOPSize(); j+=step )
+#endif
{
if ( j == poc )
{
@@ -233,11 +379,19 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
{
if(m_pcCfg->getDecodingRefreshType() == 3)
{
+#if JVET_K0157
+ eSliceType = (pocLast == 0 || pocCurr % (m_pcCfg->getIntraPeriod() * multipleFactor) == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
+#else
eSliceType = (pocLast == 0 || pocCurr % m_pcCfg->getIntraPeriod() == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
+#endif
}
else
{
+#if JVET_K0157
+ eSliceType = (pocLast == 0 || (pocCurr - (isField ? 1 : 0)) % (m_pcCfg->getIntraPeriod() * multipleFactor) == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
+#else
eSliceType = (pocLast == 0 || (pocCurr - (isField ? 1 : 0)) % m_pcCfg->getIntraPeriod() == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
+#endif
}
}
@@ -399,7 +553,7 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
dLambda *= lambdaModifier;
#endif
- iQP = max( -rpcSlice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA), min( MAX_QP, (int) floor( dQP + 0.5 ) ) );
+ iQP = Clip3( -rpcSlice->getSPS()->getQpBDOffset( CHANNEL_TYPE_LUMA ), MAX_QP, (int) floor( dQP + 0.5 ) );
#endif
m_vdRdPicLambda[iDQpIdx] = dLambda;
@@ -418,11 +572,20 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
#endif
#if W0038_CQP_ADJ
+ #if ENABLE_QPA
+ m_adaptedLumaQP = -1;
+
+ if ((m_pcCfg->getUsePerceptQPA() || m_pcCfg->getSliceChromaOffsetQpPeriodicity() > 0) && !m_pcCfg->getUseRateCtrl() && rpcSlice->getPPS()->getSliceChromaQpFlag() &&
+ (rpcSlice->isIntra() || (m_pcCfg->getSliceChromaOffsetQpPeriodicity() > 0 && (rpcSlice->getPOC() % m_pcCfg->getSliceChromaOffsetQpPeriodicity()) == 0)))
+ {
+ m_adaptedLumaQP = applyQPAdaptationChroma (pcPic, rpcSlice, m_pcCfg, iQP);
+ }
+ #endif
if(rpcSlice->getPPS()->getSliceChromaQpFlag())
{
- const bool bUseIntraOrPeriodicOffset = rpcSlice->getSliceType()==I_SLICE || (m_pcCfg->getSliceChromaOffsetQpPeriodicity()!=0 && (rpcSlice->getPOC()%m_pcCfg->getSliceChromaOffsetQpPeriodicity())==0);
- int cbQP = bUseIntraOrPeriodicOffset? m_pcCfg->getSliceChromaOffsetQpIntraOrPeriodic(false) : m_pcCfg->getGOPEntry(iGOPid).m_CbQPoffset;
- int crQP = bUseIntraOrPeriodicOffset? m_pcCfg->getSliceChromaOffsetQpIntraOrPeriodic(true) : m_pcCfg->getGOPEntry(iGOPid).m_CrQPoffset;
+ const bool bUseIntraOrPeriodicOffset = rpcSlice->isIntra() || (m_pcCfg->getSliceChromaOffsetQpPeriodicity() > 0 && (rpcSlice->getPOC() % m_pcCfg->getSliceChromaOffsetQpPeriodicity()) == 0);
+ int cbQP = bUseIntraOrPeriodicOffset ? m_pcCfg->getSliceChromaOffsetQpIntraOrPeriodic(false) : m_pcCfg->getGOPEntry(iGOPid).m_CbQPoffset;
+ int crQP = bUseIntraOrPeriodicOffset ? m_pcCfg->getSliceChromaOffsetQpIntraOrPeriodic(true) : m_pcCfg->getGOPEntry(iGOPid).m_CrQPoffset;
cbQP = Clip3( -12, 12, cbQP + rpcSlice->getPPS()->getQpOffset(COMPONENT_Cb) ) - rpcSlice->getPPS()->getQpOffset(COMPONENT_Cb);
crQP = Clip3( -12, 12, crQP + rpcSlice->getPPS()->getQpOffset(COMPONENT_Cr) ) - rpcSlice->getPPS()->getQpOffset(COMPONENT_Cr);
@@ -453,7 +616,7 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
#endif
setUpLambda(rpcSlice, dLambda, iQP);
-
+
#if WCG_EXT
// cost = Distortion + Lambda*R,
// when QP is adjusted by luma, distortion is changed, so we have to adjust lambda to match the distortion, then the cost function becomes
@@ -471,11 +634,19 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
{
if(m_pcCfg->getDecodingRefreshType() == 3)
{
+#if JVET_K0157
+ eSliceType = (pocLast == 0 || (pocCurr) % (m_pcCfg->getIntraPeriod() * multipleFactor) == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
+#else
eSliceType = (pocLast == 0 || (pocCurr) % m_pcCfg->getIntraPeriod() == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
+#endif
}
else
{
+#if JVET_K0157
+ eSliceType = (pocLast == 0 || (pocCurr - (isField ? 1 : 0)) % (m_pcCfg->getIntraPeriod() * multipleFactor) == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
+#else
eSliceType = (pocLast == 0 || (pocCurr - (isField ? 1 : 0)) % m_pcCfg->getIntraPeriod() == 0 || m_pcGOPEncoder->getGOPSize() == 0) ? I_SLICE : eSliceType;
+#endif
}
}
@@ -485,7 +656,7 @@ void EncSlice::initEncSlice( Picture* pcPic, const int pocLast, const int pocCur
if (m_pcCfg->getUseRecalculateQPAccordingToLambda())
{
dQP = xGetQPValueAccordingToLambda( dLambda );
- iQP = max( -rpcSlice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA), min( MAX_QP, (int) floor( dQP + 0.5 ) ) );
+ iQP = Clip3( -rpcSlice->getSPS()->getQpBDOffset( CHANNEL_TYPE_LUMA ), MAX_QP, (int) floor( dQP + 0.5 ) );
}
rpcSlice->setSliceQp ( iQP );
@@ -653,7 +824,7 @@ double EncSlice::calculateLambda( const Slice* slice,
dLambda *= lambdaModifier;
#endif
- iQP = max( -slice->getSPS()->getQpBDOffset(CHANNEL_TYPE_LUMA), min( MAX_QP, (int) floor( dQP + 0.5 ) ) );
+ iQP = Clip3( -slice->getSPS()->getQpBDOffset( CHANNEL_TYPE_LUMA ), MAX_QP, (int) floor( dQP + 0.5 ) );
#if JVET_K0072
if( m_pcCfg->getDepQuantEnabledFlag() )
@@ -677,68 +848,9 @@ void EncSlice::resetQP( Picture* pic, int sliceQP, double lambda )
}
#if ENABLE_QPA
-static inline int apprI2Log2 (const double d)
-{
- return d < 6.0e-20 ? -128 : int(floor(2.0 * log(d) / log(2.0) + 0.5));
-}
-
-#ifndef HLM_L1_NORM
- #define HLM_L1_NORM
-#endif
-
-static int filterAndCalculateAverageEnergies (const Pel* pSrc, const int iSrcStride,
- double &hpEner, const int iHeight, const int iWidth,
- const int iPOC = 0)
-{
- int iHpValue;
- uint32_t uHpERow, uHpEner = 0;
-
- // skip first row as there may be a black border frame
- pSrc += iSrcStride;
- // center rows
- for (int y = 1; y < iHeight - 1; y++)
- {
- uHpERow = 0;
- // skip column as there may be a black border frame
-
- for (int x = 1; x < iWidth - 1; x++) // and columns
- {
- iHpValue = 4 * (int)pSrc[x] - (int)pSrc[x-1] - (int)pSrc[x+1] - (int)pSrc[x-iSrcStride] - (int)pSrc[x+iSrcStride];
-#ifdef HLM_L1_NORM
- uHpERow += abs (iHpValue);
-#else
- uHpERow += iHpValue * iHpValue;
-#endif
- }
- // skip column as there may be a black border frame
-#ifdef HLM_L1_NORM
- uHpEner += uHpERow;
-#else
- uHpEner += (uHpERow + 64) >> 7; // avoids overflows
-#endif
- pSrc += iSrcStride;
- }
- // skip last row as there may be a black border frame
-
- hpEner = double(uHpEner) / double((iWidth - 2) * (iHeight - 2));
-#ifdef HLM_L1_NORM
- hpEner *= hpEner;
-#endif
- // lower limit, compensate for highpass amplification
- if (hpEner < 64.0) hpEner = 64.0;
-
- if (iPOC <= 0) return 0;
- return 1; // OK
-}
-
-#ifdef HLM_L1_NORM
- #undef HLM_L1_NORM
-#endif
-
-#if ENABLE_QPA
-static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, const PreCalcValues& pcv,
+static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, const PreCalcValues& pcv,
const uint32_t startAddr, const uint32_t boundingAddr, const bool useSharpLumaDQP,
- const int gopSize, const double hpEnerAvg, const double hpEnerMax)
+ const double hpEnerAvg, const double hpEnerMax, const bool useFrameWiseQPA, const int previouslyAdaptedLumaQP = -1)
{
const int iBitDepth = pcSlice->getSPS()->getBitDepth (CHANNEL_TYPE_LUMA);
const int iQPIndex = pcSlice->getSliceQp(); // initial QP index for current slice, used in following loops
@@ -746,20 +858,28 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
const TileMap& tileMap = *pcPic->tileMap;
#endif
bool sliceQPModified = false;
- double hpEnerPic = 1.0 / (1.5 * double(1 << iBitDepth)); // speedup: multiply instead of divide in loops below
+#if GLOBAL_AVERAGING
+ const double hpEnerPic = 1.0 / getAveragePictureEnergy (pcPic->getOrigBuf().Y(), iBitDepth); // inverse, speed
+#else
+ const double hpEnerPic = 1.0 / hpEnerAvg; // speedup: multiply instead of divide in loop below; 1.0 for tuning
+#endif
- if (pcv.lumaWidth > 2048 && pcv.lumaHeight > 1280) // for UHD/4K
+ if (useFrameWiseQPA || (iQPIndex >= MAX_QP))
{
- hpEnerPic *= 1.5;
- }
+ int iQPFixed;
- if ((pcPic->getPOC() & 1) && (iQPIndex >= MAX_QP))
- {
- int iQPFixed = Clip3 (0, MAX_QP, iQPIndex + ((apprI2Log2 (hpEnerAvg * hpEnerPic) + apprI2Log2 (hpEnerMax * hpEnerPic) + 1) >> 1)); // adapted slice QP = (mean(QP) + max(QP)) / 2
+ if (useFrameWiseQPA)
+ {
+ iQPFixed = (previouslyAdaptedLumaQP < 0) ? Clip3 (0, MAX_QP, iQPIndex + apprI3Log2 (hpEnerAvg * hpEnerPic)) : previouslyAdaptedLumaQP; // average-activity slice QP
+ }
+ else
+ {
+ iQPFixed = Clip3 (0, MAX_QP, iQPIndex + ((apprI3Log2 (hpEnerAvg * hpEnerPic) + apprI3Log2 (hpEnerMax * hpEnerPic) + 1) >> 1)); // adapted slice QP = (mean(QP) + max(QP)) / 2
+ }
#if SHARP_LUMA_DELTA_QP
// change new fixed QP based on average CTU luma value (Sharp)
- if (useSharpLumaDQP)
+ if (useSharpLumaDQP && (iQPIndex < MAX_QP) && (previouslyAdaptedLumaQP < 0))
{
uint64_t uAvgLuma = 0;
@@ -779,9 +899,9 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
}
#endif
- if (iQPFixed < iQPIndex) iQPFixed = iQPIndex;
+ if (iQPIndex >= MAX_QP) iQPFixed = MAX_QP;
else
- if (iQPFixed > iQPIndex)
+ if (iQPFixed != iQPIndex)
{
const double* oldLambdas = pcSlice->getLambdas();
const double corrFactor = pow (2.0, double(iQPFixed - iQPIndex) / 3.0);
@@ -816,12 +936,12 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
const uint32_t ctuRsAddr = ctuTsAddr;
#endif
- int iQPAdapt = Clip3 (0, MAX_QP, iQPIndex + apprI2Log2 (pcPic->m_uEnerHpCtu[ctuRsAddr] * hpEnerPic));
+ int iQPAdapt = Clip3 (0, MAX_QP, iQPIndex + apprI3Log2 (pcPic->m_uEnerHpCtu[ctuRsAddr] * hpEnerPic));
#if SHARP_LUMA_DELTA_QP
- if ((pcv.widthInCtus > 1) && (gopSize > 1)) // try to enforce CTU SNR greater than zero dB
+ if (pcv.widthInCtus > 1) // try to enforce CTU SNR greater than zero dB
#else
- if ((!pcSlice->isIntra()) && (gopSize > 1)) // try to enforce CTU SNR greater than zero dB
+ if (!pcSlice->isIntra()) // try to enforce CTU SNR greater than zero dB
#endif
{
const Pel dcOffset = pcPic->m_iOffsetCtu[ctuRsAddr];
@@ -832,11 +952,11 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
{
const uint64_t uAvgLuma = (uint64_t)dcOffset;
- iQPAdapt = max (0, iQPAdapt + 1 - int((3 * uAvgLuma * uAvgLuma) >> uint64_t(2 * iBitDepth - 1)));
+ iQPAdapt = std::max (0, iQPAdapt + 1 - int((3 * uAvgLuma * uAvgLuma) >> uint64_t(2 * iBitDepth - 1)));
}
#endif
- const uint32_t uRefScale = g_invQuantScales[iQPAdapt % 6] << ((iQPAdapt / 6) + iBitDepth - (pcSlice->isIntra() ? 4 : 3));
+ const uint32_t uRefScale = g_invQuantScales[iQPAdapt % 6] << ((iQPAdapt / 6) + iBitDepth - 4);
const CompArea subArea = clipArea (CompArea (COMPONENT_Y, pcPic->chromaFormat, Area ((ctuRsAddr % pcv.widthInCtus) * pcv.maxCUWidth, (ctuRsAddr / pcv.widthInCtus) * pcv.maxCUHeight, pcv.maxCUWidth, pcv.maxCUHeight)), pcPic->Y());
const Pel* pSrc = pcPic->getOrigBuf (subArea).buf;
const SizeType iSrcStride = pcPic->getOrigBuf (subArea).stride;
@@ -866,10 +986,10 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
// reduce QP index if CTU would be fully quantized to zero
if (uAbsDCless < uRefScale)
{
- const int limit = min (0, ((iQPIndex + 4) >> 3) - 6);
- const int redVal = max (limit, apprI2Log2 ((double)uAbsDCless / (double)uRefScale));
+ const int limit = std::min (0, ((iQPIndex + 4) >> 3) - 6);
+ const int redVal = std::max (limit, apprI3Log2 ((double)uAbsDCless / (double)uRefScale));
- iQPAdapt = max (0, iQPAdapt + redVal);
+ iQPAdapt = std::max (0, iQPAdapt + redVal);
}
#if SHARP_LUMA_DELTA_QP
@@ -879,7 +999,7 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
pcPic->m_iOffsetCtu[ctuRsAddr] = (Pel)iQPAdapt; // adapted QPs
- if ((pcv.widthInCtus > 1) && (gopSize > 1)) // try to reduce local bitrate peaks via minimum smoothing
+ if (pcv.widthInCtus > 1) // try to reduce local bitrate peaks via minimum smoothing of the adapted QPs
{
iQPAdapt = ctuRsAddr % pcv.widthInCtus; // horizontal offset
if (iQPAdapt == 0)
@@ -888,11 +1008,11 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
}
else // iQPAdapt >= 1
{
- iQPAdapt = (iQPAdapt > 1) ? min (pcPic->m_iOffsetCtu[ctuRsAddr - 2], pcPic->m_iOffsetCtu[ctuRsAddr]) : pcPic->m_iOffsetCtu[ctuRsAddr];
+ iQPAdapt = (iQPAdapt > 1) ? std::min (pcPic->m_iOffsetCtu[ctuRsAddr - 2], pcPic->m_iOffsetCtu[ctuRsAddr]) : pcPic->m_iOffsetCtu[ctuRsAddr];
}
if (ctuRsAddr > pcv.widthInCtus)
{
- iQPAdapt = min (iQPAdapt, (int)pcPic->m_iOffsetCtu[ctuRsAddr - 1 - pcv.widthInCtus]); // min(L, T)
+ iQPAdapt = std::min (iQPAdapt, (int)pcPic->m_iOffsetCtu[ctuRsAddr - 1 - pcv.widthInCtus]);
}
if ((ctuRsAddr > 0) && (pcPic->m_iOffsetCtu[ctuRsAddr - 1] < (Pel)iQPAdapt))
{
@@ -900,7 +1020,7 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
}
if ((ctuTsAddr == boundingAddr - 1) && (ctuRsAddr > pcv.widthInCtus)) // last CTU in the given slice
{
- iQPAdapt = min (pcPic->m_iOffsetCtu[ctuRsAddr - 1], pcPic->m_iOffsetCtu[ctuRsAddr - pcv.widthInCtus]);
+ iQPAdapt = std::min (pcPic->m_iOffsetCtu[ctuRsAddr - 1], pcPic->m_iOffsetCtu[ctuRsAddr - pcv.widthInCtus]);
if (pcPic->m_iOffsetCtu[ctuRsAddr] < (Pel)iQPAdapt)
{
pcPic->m_iOffsetCtu[ctuRsAddr] = (Pel)iQPAdapt;
@@ -914,8 +1034,6 @@ static bool applyQPAdaptation (Picture* const pcPic, Slice* const pcSlice, co
}
#endif // ENABLE_QPA
-#endif // ENABLE_QPA || ENABLE_PRIVATE
-
// ====================================================================================================================
// Public member functions
// ====================================================================================================================
@@ -1132,7 +1250,8 @@ void EncSlice::compressSlice( Picture* pcPic, const bool bCompressEntireSlice, c
#if JVET_K0346
if (pcSlice->getSPS()->getSpsNext().getUseSubPuMvp())
{
- if (!pcSlice->isIntra())
+ if (!pcSlice->isIntra()
+ )
{
if (pcSlice->getPOC() > m_pcCuEncoder->getPrevPOC() && m_pcCuEncoder->getClearSubMergeStatic())
{
@@ -1176,7 +1295,15 @@ void EncSlice::compressSlice( Picture* pcPic, const bool bCompressEntireSlice, c
else
{
m_pcCuEncoder->setPrevPOC(pcSlice->getPOC());
- m_pcCuEncoder->setClearSubMergeStatic(true);
+ if (m_pcCfg->getGOPSize() != m_pcCfg->getIntraPeriod())
+ {
+ m_pcCuEncoder->setClearSubMergeStatic(true);
+ }
+ else
+ {
+ m_pcCuEncoder->clearSubMergeStatics();
+ m_pcCuEncoder->setClearSubMergeStatic(false);
+ }
}
}
#endif
@@ -1269,22 +1396,18 @@ void EncSlice::compressSlice( Picture* pcPic, const bool bCompressEntireSlice, c
}
#if ENABLE_QPA
- #if ENABLE_QPA
double hpEnerMax = 1.0;
double hpEnerPic = 0.0;
- #endif
int iSrcOffset;
- #if ENABLE_QPA
- if (m_pcCfg->getUsePerceptQPA() && pcSlice->getPPS()->getUseDQP() && !m_pcCfg->getUseRateCtrl())
- #endif
+ if (m_pcCfg->getUsePerceptQPA() && !m_pcCfg->getUseRateCtrl())
{
for (uint32_t ctuTsAddr = startCtuTsAddr; ctuTsAddr < boundingCtuTsAddr; ctuTsAddr++)
{
#if HEVC_TILES_WPP
- const uint32_t ctuRsAddr = tileMap.getCtuTsToRsAddrMap (ctuTsAddr);
+ const uint32_t ctuRsAddr = tileMap.getCtuTsToRsAddrMap (ctuTsAddr);
#else
- const uint32_t ctuRsAddr = ctuTsAddr;
+ const uint32_t ctuRsAddr = ctuTsAddr;
#endif
const Position pos ((ctuRsAddr % widthInCtus) * pcv.maxCUWidth, (ctuRsAddr / widthInCtus) * pcv.maxCUHeight);
const CompArea subArea = clipArea (CompArea (COMPONENT_Y, pcPic->chromaFormat, Area (pos.x, pos.y, pcv.maxCUWidth, pcv.maxCUHeight)), pcPic->Y());
@@ -1315,24 +1438,23 @@ void EncSlice::compressSlice( Picture* pcPic, const bool bCompressEntireSlice, c
iSrcOffset = (iSrcOffset + (x >> 1)) / x; // slow division
filterAndCalculateAverageEnergies (pcPic->getOrigBuf (fltArea).buf, iSrcStride,
- hpEner, iFltHeight, iFltWidth, pcPic->getPOC());
+ hpEner, iFltHeight, iFltWidth,
+ pcSlice->getSPS()->getBitDepth (CHANNEL_TYPE_LUMA));
- #if ENABLE_QPA
if (hpEner > hpEnerMax) hpEnerMax = hpEner;
hpEnerPic += hpEner;
pcPic->m_uEnerHpCtu[ctuRsAddr] = hpEner;
pcPic->m_iOffsetCtu[ctuRsAddr] = (Pel)iSrcOffset;
- #endif
} // end iteration over all CTUs in current slice
}
- #if ENABLE_QPA
- if (m_pcCfg->getUsePerceptQPA() && pcSlice->getPPS()->getUseDQP() && !m_pcCfg->getUseRateCtrl() && (boundingCtuTsAddr > startCtuTsAddr))
+ if (m_pcCfg->getUsePerceptQPA() && !m_pcCfg->getUseRateCtrl() && (boundingCtuTsAddr > startCtuTsAddr))
{
const double hpEnerAvg = hpEnerPic / double(boundingCtuTsAddr - startCtuTsAddr);
- if (applyQPAdaptation (pcPic, pcSlice, pcv, startCtuTsAddr, boundingCtuTsAddr, m_pcCfg->getLumaLevelToDeltaQPMapping().mode == LUMALVL_TO_DQP_NUM_MODES, m_pcCfg->getGOPSize(), hpEnerAvg, hpEnerMax))
+ if (applyQPAdaptation (pcPic, pcSlice, pcv, startCtuTsAddr, boundingCtuTsAddr, m_pcCfg->getLumaLevelToDeltaQPMapping().mode == LUMALVL_TO_DQP_NUM_MODES,
+ hpEnerAvg, hpEnerMax, (m_pcCfg->getBaseQP() >= 38) || (m_pcCfg->getSourceWidth() <= 512 && m_pcCfg->getSourceHeight() <= 320), m_adaptedLumaQP))
{
m_CABACEstimator->initCtxModels (*pcSlice);
#if ENABLE_SPLIT_PARALLELISM || ENABLE_WPP_PARALLELISM
@@ -1356,9 +1478,7 @@ void EncSlice::compressSlice( Picture* pcPic, const bool bCompressEntireSlice, c
}
}
}
- #endif // ENABLE_QPA
-
-#endif // ENABLE_QPA || ENABLE_PRIVATE
+#endif // ENABLE_QPA
cs.pcv = pcSlice->getPPS()->pcv;
cs.fracBits = 0;
@@ -1387,6 +1507,11 @@ void EncSlice::compressSlice( Picture* pcPic, const bool bCompressEntireSlice, c
}
}
else
+#endif
+#if K0149_BLOCK_STATISTICS
+ const SPS *sps = pcSlice->getSPS();
+ CHECK(sps == 0, "No SPS present");
+ writeBlockStatisticsHeader(sps);
#endif
encodeCtus( pcPic, bCompressEntireSlice, bFastDeltaQP, startCtuTsAddr, boundingCtuTsAddr, m_pcLib );
@@ -1568,12 +1693,17 @@ void EncSlice::encodeCtus( Picture* pcPic, const bool bCompressEntireSlice, cons
#endif
+
#if ENABLE_WPP_PARALLELISM
pEncLib->getCuEncoder( dataId )->compressCtu( cs, ctuArea, ctuRsAddr, prevQP, currQP );
#else
m_pcCuEncoder->compressCtu( cs, ctuArea, ctuRsAddr, prevQP, currQP );
#endif
+#if K0149_BLOCK_STATISTICS
+ getAndStoreBlockStatistics(cs, ctuArea);
+#endif
+
pCABACWriter->resetBits();
pCABACWriter->coding_tree_unit( cs, ctuArea, prevQP, ctuRsAddr, true );
const int numberOfWrittenBits = int( pCABACWriter->getEstFracBits() >> SCALE_BITS );
@@ -1629,11 +1759,13 @@ void EncSlice::encodeCtus( Picture* pcPic, const bool bCompressEntireSlice, cons
#if !ENABLE_WPP_PARALLELISM
int actualBits = int(cs.fracBits >> SCALE_BITS);
+ actualBits -= (int)m_uiPicTotalBits;
#endif
if ( pCfg->getUseRateCtrl() )
{
#if ENABLE_WPP_PARALLELISM
int actualBits = int( cs.fracBits >> SCALE_BITS );
+ actualBits -= (int)m_uiPicTotalBits;
#endif
int actualQP = g_RCInvalidQPValue;
double actualLambda = pRdCost->getLambda();
@@ -1957,7 +2089,7 @@ void EncSlice::calculateBoundingCtuTsAddrForSlice(uint32_t &startCtuTSAddrSlice,
{
// Adjust for wavefronts (no tiles).
// WPP: if a slice does not start at the beginning of a CTB row, it must end within the same CTB row
- boundingCtuTSAddrSlice = min(boundingCtuTSAddrSlice, startCtuTSAddrSlice - (startCtuTSAddrSlice % pcPic->cs->pcv->widthInCtus) + (pcPic->cs->pcv->widthInCtus));
+ boundingCtuTSAddrSlice = std::min(boundingCtuTSAddrSlice, startCtuTSAddrSlice - (startCtuTSAddrSlice % pcPic->cs->pcv->widthInCtus) + (pcPic->cs->pcv->widthInCtus));
}
#endif
}
@@ -2011,7 +2143,7 @@ void EncSlice::xDetermineStartAndBoundingCtuTsAddr ( uint32_t& startCtuTsAddr,
pcSlice->setSliceSegmentCurStartCtuTsAddr(startCtuTsAddrSliceSegment);
// Make a joint decision based on reconstruction and dependent slice bounds
- startCtuTsAddr = max(startCtuTsAddrSlice , startCtuTsAddrSliceSegment );
+ startCtuTsAddr = std::max(startCtuTsAddrSlice, startCtuTsAddrSliceSegment);
boundingCtuTsAddr = boundingCtuTsAddrSliceSegment;
#else
startCtuTsAddr = startCtuTsAddrSlice;
diff --git a/source/Lib/EncoderLib/EncSlice.h b/source/Lib/EncoderLib/EncSlice.h
index 2c6eabb26d25d5362aba4080de659a3827d7a679..836bc4d7caccc05776643776ac7e4dbc2792c25b 100644
--- a/source/Lib/EncoderLib/EncSlice.h
+++ b/source/Lib/EncoderLib/EncSlice.h
@@ -106,7 +106,7 @@ public:
int getGopId() const { return m_gopID; }
double calculateLambda( const Slice* slice, const int GOPid, const int depth, const double refQP, const double dQP, int &iQP );
void setUpLambda( Slice* slice, const double dLambda, int iQP );
-
+
private:
#endif
#if HEVC_TILES_WPP
@@ -117,6 +117,10 @@ private:
public:
+#if ENABLE_QPA
+ int m_adaptedLumaQP;
+
+#endif
EncSlice();
virtual ~EncSlice();
@@ -126,7 +130,12 @@ public:
/// preparation of slice encoding (reference marking, QP and lambda)
void initEncSlice ( Picture* pcPic, const int pocLast, const int pocCurr,
- const int iGOPid, Slice*& rpcSlice, const bool isField );
+ const int iGOPid, Slice*& rpcSlice, const bool isField
+#if JVET_K0157
+ , bool isEncodeLtRef
+#endif
+ );
+
void resetQP ( Picture* pic, int sliceQP, double lambda );
// compress and encode slice
@@ -150,7 +159,6 @@ public:
void setSliceSegmentIdx (uint32_t i) { m_uiSliceSegmentIdx = i; }
SliceType getEncCABACTableIdx() const { return m_encCABACTableIdx; }
-
private:
double xGetQPValueAccordingToLambda ( double lambda );
};
diff --git a/source/Lib/EncoderLib/InterSearch.cpp b/source/Lib/EncoderLib/InterSearch.cpp
index 1b790b62a1f9f528b4a2c4a1bb97d17a70d9a5a9..11e5085dd69ffa6556f82e62c0f9cd5a82522737 100644
--- a/source/Lib/EncoderLib/InterSearch.cpp
+++ b/source/Lib/EncoderLib/InterSearch.cpp
@@ -749,9 +749,11 @@ void InterSearch::xMergeEstimation( PredictionUnit& pu, PelUnitBuf& origBuf, int
uiMergeIdx = uiMergeCand;
}
}
+
}
+
//! search of the best candidate for inter prediction
void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
{
@@ -796,6 +798,7 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
int bestBiPMvpL1 = 0;
Distortion biPDistTemp = std::numeric_limits<Distortion>::max();
+
MergeCtx mergeCtx;
// Loop over Prediction Units
@@ -874,7 +877,6 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
for ( int iRefList = 0; iRefList < iNumPredDir; iRefList++ )
{
RefPicList eRefPicList = ( iRefList ? REF_PIC_LIST_1 : REF_PIC_LIST_0 );
-
for ( int iRefIdxTemp = 0; iRefIdxTemp < cs.slice->getNumRefIdx(eRefPicList); iRefIdxTemp++ )
{
uiBitsTemp = uiMbBits[iRefList];
@@ -1093,7 +1095,6 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
iRefStart = 0;
iRefEnd = cs.slice->getNumRefIdx(eRefPicList)-1;
-
for ( int iRefIdxTemp = iRefStart; iRefIdxTemp <= iRefEnd; iRefIdxTemp++ )
{
uiBitsTemp = uiMbBits[2] + uiMotBits[1-iRefList];
@@ -1187,6 +1188,7 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
uiBits [1] = bitsValidList1;
uiCost [1] = costValidList1;
+
#if JVET_K_AFFINE
uiLastModeTemp = uiLastMode;
#endif
@@ -1232,6 +1234,7 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
uiMEBits = uiBits[1];
}
+
if ( cu.partSize != SIZE_2Nx2N )
{
uint32_t uiMRGIndex = 0;
@@ -1308,9 +1311,9 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
int refIdx4Para[2] = { -1, -1 };
#if JVET_K0220_ENC_CTRL
- xPredAffineInterSearch( pu, origBuf, puIdx, uiLastModeTemp, uiAffineCost, cMvHevcTemp, acMvAffine4Para, refIdx4Para );
+ xPredAffineInterSearch(pu, origBuf, puIdx, uiLastModeTemp, uiAffineCost, cMvHevcTemp, acMvAffine4Para, refIdx4Para);
#else
- xPredAffineInterSearch( pu, origBuf, puIdx, uiLastModeTemp, uiAffineCost, cMvHevcTemp, bFastSkipBi, acMvAffine4Para, refIdx4Para );
+ xPredAffineInterSearch(pu, origBuf, puIdx, uiLastModeTemp, uiAffineCost, cMvHevcTemp, bFastSkipBi, acMvAffine4Para, refIdx4Para);
#endif
if ( cu.slice->getSPS()->getSpsNext().getUseAffineType() )
{
@@ -1347,9 +1350,9 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
Distortion uiAffine6Cost = std::numeric_limits<Distortion>::max();
cu.affineType = AFFINEMODEL_6PARAM;
#if JVET_K0220_ENC_CTRL
- xPredAffineInterSearch( pu, origBuf, puIdx, uiLastModeTemp, uiAffine6Cost, cMvHevcTemp, acMvAffine4Para, refIdx4Para );
+ xPredAffineInterSearch(pu, origBuf, puIdx, uiLastModeTemp, uiAffine6Cost, cMvHevcTemp, acMvAffine4Para, refIdx4Para);
#else
- xPredAffineInterSearch( pu, origBuf, puIdx, uiLastModeTemp, uiAffine6Cost, cMvHevcTemp, bFastSkipBi, acMvAffine4Para, refIdx4Para );
+ xPredAffineInterSearch(pu, origBuf, puIdx, uiLastModeTemp, uiAffine6Cost, cMvHevcTemp, bFastSkipBi, acMvAffine4Para, refIdx4Para);
#endif
// reset to 4 parameter affine inter mode
@@ -1414,6 +1417,8 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
}
}
#endif
+
+
m_maxCompIDToPred = MAX_NUM_COMPONENT;
{
@@ -1662,6 +1667,7 @@ Distortion InterSearch::xGetAffineTemplateCost( PredictionUnit& pu, PelUnitBuf&
void InterSearch::xMotionEstimation(PredictionUnit& pu, PelUnitBuf& origBuf, RefPicList eRefPicList, Mv& rcMvPred, int iRefIdxPred, Mv& rcMv, int& riMVPIdx, uint32_t& ruiBits, Distortion& ruiCost, const AMVPInfo& amvpInfo, bool bBi)
{
+
Mv cMvHalf, cMvQter;
CHECK(eRefPicList >= MAX_NUM_REF_LIST_ADAPT_SR || iRefIdxPred>=int(MAX_IDX_ADAPT_SR), "Invalid reference picture list");
@@ -1678,8 +1684,8 @@ void InterSearch::xMotionEstimation(PredictionUnit& pu, PelUnitBuf& origBuf, Ref
// NOTE: Other buf contains predicted signal from another direction
PelUnitBuf otherBuf = m_tmpPredStorage[1 - (int)eRefPicList].getBuf( UnitAreaRelative(*pu.cu, pu ));
origBufTmp.copyFrom(origBuf);
- origBufTmp.removeHighFreq(otherBuf, m_pcEncCfg->getClipForBiPredMeEnabled(), pu.cu->slice->clpRngs() );
-
+ origBufTmp.removeHighFreq( otherBuf, m_pcEncCfg->getClipForBiPredMeEnabled(), pu.cu->slice->clpRngs()
+ );
pBuf = &origBufTmp;
fWeight = 0.5;
@@ -1701,6 +1707,17 @@ void InterSearch::xMotionEstimation(PredictionUnit& pu, PelUnitBuf& origBuf, Ref
#if JVET_K0357_AMVR
cStruct.imvShift = pu.cu->imv << 1;
#endif
+#if JVET_K0157
+ cStruct.inCtuSearch = false;
+ cStruct.zeroMV = false;
+ {
+ if (pu.cs->sps->getSpsNext().getUseCompositeRef() && pu.cs->slice->getRefPic(eRefPicList, iRefIdxPred)->longTerm)
+ {
+ cStruct.inCtuSearch = true;
+ }
+ }
+#endif
+
auto blkCache = dynamic_cast<CacheBlkInfoCtrl*>( m_modeCtrl );
bool bQTBTMV = false;
@@ -1730,7 +1747,11 @@ void InterSearch::xMotionEstimation(PredictionUnit& pu, PelUnitBuf& origBuf, Ref
{
if( !bQTBTMV )
{
- xSetSearchRange( pu, ( bBi ? rcMv : rcMvPred ), iSrchRng, cStruct.searchRange );
+ xSetSearchRange(pu, (bBi ? rcMv : rcMvPred), iSrchRng, cStruct.searchRange
+#if JVET_K0157
+ , cStruct
+#endif
+ );
}
cStruct.subShiftMode = m_pcEncCfg->getFastInterSearchMode() == FASTINTERSEARCH_MODE1 || m_pcEncCfg->getFastInterSearchMode() == FASTINTERSEARCH_MODE3 ? 2 : 0;
xPatternSearch( cStruct, rcMv, ruiCost);
@@ -1799,7 +1820,11 @@ void InterSearch::xMotionEstimation(PredictionUnit& pu, PelUnitBuf& origBuf, Ref
void InterSearch::xSetSearchRange ( const PredictionUnit& pu,
const Mv& cMvPred,
const int iSrchRng,
- SearchRange& sr )
+ SearchRange& sr
+#if JVET_K0157
+ , IntTZSearchStruct& cStruct
+#endif
+)
{
#if JVET_K0346 || JVET_K_AFFINE
const int iMvShift = cMvPred.highPrec ? 4 : 2;
@@ -1827,6 +1852,35 @@ void InterSearch::xSetSearchRange ( const PredictionUnit& pu,
sr.top = mvTL.ver;
sr.right = mvBR.hor;
sr.bottom = mvBR.ver;
+
+#if JVET_K0157
+ if (pu.cs->sps->getSpsNext().getUseCompositeRef() && cStruct.inCtuSearch)
+ {
+ Position posRB = pu.Y().bottomRight();
+ Position posTL = pu.Y().topLeft();
+ const PreCalcValues *pcv = pu.cs->pcv;
+ Position posRBinCTU(posRB.x & pcv->maxCUWidthMask, posRB.y & pcv->maxCUHeightMask);
+ Position posLTinCTU = Position(posTL.x & pcv->maxCUWidthMask, posTL.y & pcv->maxCUHeightMask).offset(-4, -4);
+ if (sr.left < -posLTinCTU.x)
+ sr.left = -posLTinCTU.x;
+ if (sr.top < -posLTinCTU.y)
+ sr.top = -posLTinCTU.y;
+ if (sr.right >((int)pcv->maxCUWidth - 4 - posRBinCTU.x))
+ sr.right = (int)pcv->maxCUWidth - 4 - posRBinCTU.x;
+ if (sr.bottom >((int)pcv->maxCUHeight - 4 - posRBinCTU.y))
+ sr.bottom = (int)pcv->maxCUHeight - 4 - posRBinCTU.y;
+ if (posLTinCTU.x == -4 || posLTinCTU.y == -4)
+ {
+ sr.left = sr.right = sr.bottom = sr.top = 0;
+ cStruct.zeroMV = 1;
+ }
+ if (posRBinCTU.x == pcv->maxCUWidthMask || posRBinCTU.y == pcv->maxCUHeightMask)
+ {
+ sr.left = sr.right = sr.bottom = sr.top = 0;
+ cStruct.zeroMV = 1;
+ }
+ }
+#endif
}
@@ -1998,7 +2052,11 @@ void InterSearch::xTZSearch( const PredictionUnit& pu,
// set search range
Mv currBestMv(cStruct.iBestX, cStruct.iBestY );
currBestMv <<= 2;
- xSetSearchRange( pu, currBestMv, m_iSearchRange>>(bFastSettings?1:0), sr );
+ xSetSearchRange(pu, currBestMv, m_iSearchRange >> (bFastSettings ? 1 : 0), sr
+#if JVET_K0157
+ , cStruct
+#endif
+ );
}
// start search
@@ -2258,7 +2316,11 @@ void InterSearch::xTZSearchSelective( const PredictionUnit& pu,
// set search range
Mv currBestMv(cStruct.iBestX, cStruct.iBestY );
currBestMv <<= 2;
- xSetSearchRange( pu, currBestMv, m_iSearchRange, sr );
+ xSetSearchRange( pu, currBestMv, m_iSearchRange, sr
+#if JVET_K0157
+ , cStruct
+#endif
+ );
}
// Initial search
@@ -2457,7 +2519,11 @@ void InterSearch::xPatternSearchFracDIF(
#if JVET_K0357_AMVR
+#if JVET_K0157
+ if (cStruct.imvShift || (pu.cs->sps->getSpsNext().getUseCompositeRef() && cStruct.zeroMV))
+#else
if( cStruct.imvShift )
+#endif
{
m_pcRdCost->setDistParam( m_cDistParam, *cStruct.pcPatternKey, cStruct.piRefY + iOffset, cStruct.iRefStride, m_lumaClpRng.bd, COMPONENT_Y, 0, 1, m_pcEncCfg->getUseHADME() && !bIsLosslessCoded );
ruiCost = m_cDistParam.distFunc( m_cDistParam );
@@ -2581,11 +2647,11 @@ void InterSearch::xPredAffineInterSearch( PredictionUnit& pu,
pu.cu->affine = true;
pu.mergeFlag = false;
+
// Uni-directional prediction
for ( int iRefList = 0; iRefList < iNumPredDir; iRefList++ )
{
RefPicList eRefPicList = ( iRefList ? REF_PIC_LIST_1 : REF_PIC_LIST_0 );
-
for ( int iRefIdxTemp = 0; iRefIdxTemp < slice.getNumRefIdx(eRefPicList); iRefIdxTemp++ )
{
// Get RefIdx bits
@@ -2715,7 +2781,6 @@ void InterSearch::xPredAffineInterSearch( PredictionUnit& pu,
{
xAffineMotionEstimation( pu, origBuf, eRefPicList, cMvPred[iRefList][iRefIdxTemp], iRefIdxTemp, cMvTemp[iRefList][iRefIdxTemp], uiBitsTemp, uiCostTemp );
}
-
// Set best AMVP Index
xCopyAffineAMVPInfo( affiAMVPInfoTemp[eRefPicList], aacAffineAMVPInfo[iRefList][iRefIdxTemp] );
xCheckBestAffineMVP( pu, affiAMVPInfoTemp[eRefPicList], eRefPicList, cMvTemp[iRefList][iRefIdxTemp], cMvPred[iRefList][iRefIdxTemp], aaiMvpIdx[iRefList][iRefIdxTemp], uiBitsTemp, uiCostTemp );
@@ -2866,7 +2931,6 @@ void InterSearch::xPredAffineInterSearch( PredictionUnit& pu,
iRefStart = 0;
iRefEnd = slice.getNumRefIdx(eRefPicList) - 1;
-
for ( int iRefIdxTemp = iRefStart; iRefIdxTemp <= iRefEnd; iRefIdxTemp++ )
{
#if JVET_K0185_AFFINE_6PARA_ENC // reuse refidx of 4-para
@@ -2955,6 +3019,7 @@ void InterSearch::xPredAffineInterSearch( PredictionUnit& pu,
uiBits[1] = bitsValidList1;
uiCost[1] = costValidList1;
+
// Affine ME result set
if ( uiCostBi <= uiCost[0] && uiCostBi <= uiCost[1] ) // Bi
{
@@ -3184,7 +3249,7 @@ void InterSearch::xCheckBestAffineMVP( PredictionUnit &pu, AffineAMVPInfo &affin
if ( iVerIdx != 0 )
{
#if JVET_K0185_AFFINE_6PARA_ENC
- secondPred = (iVerIdx == 1 ? affineAMVPInfo.mvCandRT[iMVPIdx] : affineAMVPInfo.mvCandLT[iMVPIdx]) + (acMv[0] - affineAMVPInfo.mvCandLT[iMVPIdx]);
+ secondPred = (iVerIdx == 1 ? affineAMVPInfo.mvCandRT[iMVPIdx] : affineAMVPInfo.mvCandLB[iMVPIdx]) + (acMv[0] - affineAMVPInfo.mvCandLT[iMVPIdx]);
#else
secondPred = affineAMVPInfo.mvCandRT[iMVPIdx] + (acMv[0] - affineAMVPInfo.mvCandLT[iMVPIdx]);
#endif
@@ -3228,6 +3293,7 @@ void InterSearch::xAffineMotionEstimation( PredictionUnit& pu,
Distortion& ruiCost,
bool bBi )
{
+
const int width = pu.Y().width;
const int height = pu.Y().height;
@@ -3245,7 +3311,8 @@ void InterSearch::xAffineMotionEstimation( PredictionUnit& pu,
// NOTE: Other buf contains predicted signal from another direction
PelUnitBuf otherBuf = m_tmpPredStorage[1 - (int)eRefPicList].getBuf( UnitAreaRelative( *pu.cu, pu ) );
origBufTmp.copyFrom(origBuf);
- origBufTmp.removeHighFreq(otherBuf, m_pcEncCfg->getClipForBiPredMeEnabled(), pu.cu->slice->clpRngs());
+ origBufTmp.removeHighFreq(otherBuf, m_pcEncCfg->getClipForBiPredMeEnabled(), pu.cu->slice->clpRngs()
+ );
pBuf = &origBufTmp;
fWeight = 0.5;
@@ -4184,7 +4251,8 @@ void InterSearch::xEncodeInterResidualQT(CodingStructure &cs, Partitioner &parti
#endif
}
-void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &partitioner, Distortion *puiZeroDist /*= NULL*/)
+void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &partitioner, Distortion *puiZeroDist /*= NULL*/
+)
{
const UnitArea& currArea = partitioner.currArea();
const SPS &sps = *cs.sps;
@@ -4253,10 +4321,8 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
saveCS.picture = cs.picture;
saveCS.area.repositionTo(currArea);
saveCS.clearTUs();
-
TransformUnit &bestTU = saveCS.addTU( currArea, partitioner.chType );
-
for( uint32_t c = 0; c < numTBlocks; c++ )
{
const ComponentID compID = ComponentID(c);
@@ -4541,7 +4607,6 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
for (uint32_t ch = 0; ch < numValidComp; ch++)
{
const ComponentID compID = ComponentID(ch);
-
if (tu.blocks[compID].valid())
{
if( cs.pps->getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() && isChroma(compID) && uiAbsSum[COMPONENT_Y] )
@@ -4588,7 +4653,8 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
do
{
- xEstimateInterResidualQT(*csSplit, partitioner, bCheckFull ? nullptr : puiZeroDist);
+ xEstimateInterResidualQT(*csSplit, partitioner, bCheckFull ? nullptr : puiZeroDist
+ );
csSplit->cost = m_pcRdCost->calcRdCost( csSplit->fracBits, csSplit->dist );
#if JVET_K1000_SIMPLIFIED_EMT
@@ -4656,6 +4722,9 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
{
cs.useSubStructure( *csSplit, partitioner.chType, currArea, false, false, false, true );
cs.cost = csSplit->cost;
+#if JVET_K1000_SIMPLIFIED_EMT
+ isSplit = true;
+#endif
}
}
@@ -4680,7 +4749,8 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
#endif
}
-void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual)
+void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual
+)
{
CodingUnit &cu = *cs.getCU( partitioner.chType );
@@ -4701,13 +4771,11 @@ void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &pa
// add an empty TU
cs.addTU(cs.area, partitioner.chType);
-
Distortion distortion = 0;
for (int comp = 0; comp < numValidComponents; comp++)
{
const ComponentID compID = ComponentID(comp);
-
CPelBuf reco = cs.getRecoBuf (compID);
CPelBuf org = cs.getOrgBuf (compID);
#if WCG_EXT
@@ -4747,7 +4815,6 @@ void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &pa
// Residual coding.
cs.getResiBuf().copyFrom (cs.getOrgBuf());
cs.getResiBuf().subtract (cs.getPredBuf());
-
Distortion zeroDistortion = 0;
const TempCtx ctxStart( m_CtxCache, m_CABACEstimator->getCtx() );
@@ -4755,7 +4822,6 @@ void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &pa
cs.getOrgResiBuf().copyFrom(cs.getResiBuf());
xEstimateInterResidualQT(cs, partitioner, &zeroDistortion);
-
TransformUnit &firstTU = *cs.getTU( partitioner.chType );
cu.rootCbf = false;
@@ -4777,7 +4843,7 @@ void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &pa
const int numValidTBlocks = ::getNumberValidTBlocks( *cs.pcv );
for (uint32_t i = 0; i < numValidTBlocks; i++)
{
- cu.rootCbf |= TU::getCbf( firstTU, ComponentID( i ) );
+ cu.rootCbf |= TU::getCbfAtDepth(firstTU, ComponentID(i), 0);
}
// -------------------------------------------------------
@@ -4819,7 +4885,6 @@ void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &pa
for (int comp = 0; comp < numValidComponents; comp++)
{
const ComponentID compID = ComponentID(comp);
-
CPelBuf reco = cs.getRecoBuf (compID);
CPelBuf org = cs.getOrgBuf (compID);
@@ -4874,7 +4939,6 @@ uint64_t InterSearch::xGetSymbolFracBitsInter(CodingStructure &cs, Partitioner &
{
m_CABACEstimator->cu_transquant_bypass_flag( cu );
}
-
m_CABACEstimator->cu_skip_flag( cu );
m_CABACEstimator->pred_mode ( cu );
m_CABACEstimator->cu_pred_data( cu );
@@ -4888,4 +4952,3 @@ uint64_t InterSearch::xGetSymbolFracBitsInter(CodingStructure &cs, Partitioner &
return fracBits;
}
-
diff --git a/source/Lib/EncoderLib/InterSearch.h b/source/Lib/EncoderLib/InterSearch.h
index 6cf72e338edbd4629b594809270d17edb8db4a16..cbdaaf21a92f2d7af4927affbf219613b26833ff 100644
--- a/source/Lib/EncoderLib/InterSearch.h
+++ b/source/Lib/EncoderLib/InterSearch.h
@@ -53,7 +53,6 @@
#if JVET_K0367_AFFINE_FIX_POINT
#include "CommonLib/AffineGradientSearch.h"
#endif
-
//! \ingroup EncoderLib
//! \{
@@ -64,7 +63,6 @@
static const uint32_t MAX_NUM_REF_LIST_ADAPT_SR = 2;
static const uint32_t MAX_IDX_ADAPT_SR = 33;
static const uint32_t NUM_MV_PREDICTORS = 3;
-
class EncModeCtrl;
/// encoder search class
@@ -95,7 +93,6 @@ private:
ClpRng m_lumaClpRng;
-
protected:
// interface to option
EncCfg* m_pcEncCfg;
@@ -124,7 +121,6 @@ protected:
bool m_isInitialized;
-
public:
InterSearch();
virtual ~InterSearch();
@@ -145,7 +141,6 @@ public:
void destroy ();
void setTempBuffers (CodingStructure ****pSlitCS, CodingStructure ****pFullCS, CodingStructure **pSaveCS );
-
#if ENABLE_SPLIT_PARALLELISM
void copyState ( const InterSearch& other );
#endif
@@ -178,6 +173,10 @@ protected:
int subShiftMode;
#if JVET_K0357_AMVR
unsigned imvShift;
+#endif
+#if JVET_K0157
+ bool inCtuSearch;
+ bool zeroMV;
#endif
} IntTZSearchStruct;
@@ -199,7 +198,6 @@ public:
/// set ME search range
void setAdaptiveSearchRange ( int iDir, int iRefIdx, int iSearchRange) { CHECK(iDir >= MAX_NUM_REF_LIST_ADAPT_SR || iRefIdx>=int(MAX_IDX_ADAPT_SR), "Invalid index"); m_aaiAdaptSR[iDir][iRefIdx] = iSearchRange; }
-
protected:
// -------------------------------------------------------------------------------------------------------------------
@@ -291,6 +289,9 @@ protected:
const Mv& cMvPred,
const int iSrchRng,
SearchRange& sr
+#if JVET_K0157
+ , IntTZSearchStruct & cStruct
+#endif
);
void xPatternSearchFast ( const PredictionUnit& pu,
@@ -368,7 +369,10 @@ protected:
void xCopyAffineAMVPInfo ( AffineAMVPInfo& src, AffineAMVPInfo& dst );
void xCheckBestAffineMVP ( PredictionUnit &pu, AffineAMVPInfo &affineAMVPInfo, RefPicList eRefPicList, Mv acMv[3], Mv acMvPred[3], int& riMVPIdx, uint32_t& ruiBits, Distortion& ruiCost );
+
#endif
+
+
void xExtDIFUpSamplingH ( CPelBuf* pcPattern );
void xExtDIFUpSamplingQ ( CPelBuf* pcPatternKey, Mv halfPelRef );
@@ -378,12 +382,13 @@ protected:
void setWpScalingDistParam ( int iRefIdx, RefPicList eRefPicListCur, Slice *slice );
-
public:
- void encodeResAndCalcRdInterCU (CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual);
+ void encodeResAndCalcRdInterCU (CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual
+ );
void xEncodeInterResidualQT (CodingStructure &cs, Partitioner &partitioner, const ComponentID &compID);
- void xEstimateInterResidualQT (CodingStructure &cs, Partitioner &partitioner, Distortion *puiZeroDist = NULL);
+ void xEstimateInterResidualQT (CodingStructure &cs, Partitioner &partitioner, Distortion *puiZeroDist = NULL
+ );
uint64_t xGetSymbolFracBitsInter (CodingStructure &cs, Partitioner &partitioner);
};// END CLASS DEFINITION EncSearch
diff --git a/source/Lib/EncoderLib/IntraSearch.cpp b/source/Lib/EncoderLib/IntraSearch.cpp
index b0696fe49b51d4034356e9f7f0a18fc0c46bff17..eb8519e79dbe930bd7cc2d13f507cce82b8e21ce 100644
--- a/source/Lib/EncoderLib/IntraSearch.cpp
+++ b/source/Lib/EncoderLib/IntraSearch.cpp
@@ -915,7 +915,8 @@ void IntraSearch::xEncIntraHeader(CodingStructure &cs, Partitioner &partitioner,
// CU header
if( isFirst )
{
- if( !cs.slice->isIntra() )
+ if( !cs.slice->isIntra()
+ )
{
if( cs.pps->getTransquantBypassEnabledFlag() )
{
diff --git a/source/Lib/EncoderLib/RateCtrl.cpp b/source/Lib/EncoderLib/RateCtrl.cpp
index 6c265b9324123e14fa98c0d1d16761cd5100e890..4b4fffb03c458a4b6c7ba89e41b63c7c3cf1b066 100644
--- a/source/Lib/EncoderLib/RateCtrl.cpp
+++ b/source/Lib/EncoderLib/RateCtrl.cpp
@@ -39,6 +39,10 @@
#include <cmath>
+#if JVET_K0390_RATECTRL
+#define LAMBDA_PREC 1000000
+#endif
+
using namespace std;
//sequence level
@@ -66,6 +70,9 @@ EncRCSeq::EncRCSeq()
m_useLCUSeparateModel = false;
m_adaptiveBit = 0;
m_lastLambda = 0.0;
+#if RATECTRL_FIX_FULLNBIT
+ m_bitDepth = 0;
+#endif
}
EncRCSeq::~EncRCSeq()
@@ -138,6 +145,9 @@ void EncRCSeq::create( int totalFrames, int targetBitrate, int frameRate, int GO
{
m_picPara[i].m_alpha = 0.0;
m_picPara[i].m_beta = 0.0;
+#if JVET_K0390_RATECTRL
+ m_picPara[i].m_validPix = -1;
+#endif
}
if ( m_useLCUSeparateModel )
@@ -150,6 +160,9 @@ void EncRCSeq::create( int totalFrames, int targetBitrate, int frameRate, int GO
{
m_LCUPara[i][j].m_alpha = 0.0;
m_LCUPara[i][j].m_beta = 0.0;
+#if JVET_K0390_RATECTRL
+ m_LCUPara[i][j].m_validPix = -1;
+#endif
}
}
}
@@ -217,13 +230,47 @@ void EncRCSeq::initPicPara( TRCParameter* picPara )
{
if (i>0)
{
+#if RATECTRL_FIX_FULLNBIT
+#if DISTORTION_LAMBDA_BUGFIX
+ int bitdepth_luma_scale =
+ 2
+ * (m_bitDepth - 8
+ - DISTORTION_PRECISION_ADJUSTMENT(m_bitDepth));
+#else
+#if FULL_NBIT
+ int bitdepth_luma_scale = 2 * (m_bitDepth - 8);
+#else
+ int bitdepth_luma_scale = 0;
+#endif
+#endif
+ m_picPara[i].m_alpha = 3.2003 * pow(2.0, bitdepth_luma_scale);
+ m_picPara[i].m_beta = -1.367;
+#else
m_picPara[i].m_alpha = 3.2003;
m_picPara[i].m_beta = -1.367;
+#endif
}
else
{
+#if RATECTRL_FIX_FULLNBIT
+#if DISTORTION_LAMBDA_BUGFIX
+ int bitdepth_luma_scale =
+ 2
+ * (m_bitDepth - 8
+ - DISTORTION_PRECISION_ADJUSTMENT(m_bitDepth));
+#else
+#if FULL_NBIT
+ int bitdepth_luma_scale = 2 * (m_bitDepth - 8);
+#else
+ int bitdepth_luma_scale = 0;
+#endif
+#endif
+ m_picPara[i].m_alpha = pow(2.0, bitdepth_luma_scale) * ALPHA;
+ m_picPara[i].m_beta = BETA2;
+#else
m_picPara[i].m_alpha = ALPHA;
m_picPara[i].m_beta = BETA2;
+#endif
}
}
}
@@ -276,7 +323,11 @@ void EncRCSeq::setAllBitRatio( double basicLambda, double* equaCoeffA, double* e
int* bitsRatio = new int[m_GOPSize];
for ( int i=0; i<m_GOPSize; i++ )
{
+#if JVET_K0390_RATECTRL
+ bitsRatio[i] = (int)(equaCoeffA[i] * pow(basicLambda, equaCoeffB[i]) * (double)getPicPara(getGOPID2Level(i)).m_validPix);
+#else
bitsRatio[i] = (int)( equaCoeffA[i] * pow( basicLambda, equaCoeffB[i] ) * m_numberOfPixel );
+#endif
}
initBitsRatio( bitsRatio );
delete[] bitsRatio;
@@ -353,9 +404,75 @@ void EncRCGOP::create( EncRCSeq* encRCSeq, int numPic )
lambdaRatio[7] = 12.3;
}
}
+#if JVET_K0390_RATECTRL
+ else if (encRCSeq->getAdaptiveBits() == 3) // for GOP size = 16, random access case
+ {
+ {
+#if RATECTRL_FIX_FULLNBIT
+#if DISTORTION_LAMBDA_BUGFIX
+ int bitdepth_luma_scale =
+ 2
+ * (encRCSeq->getbitDepth() - 8
+ - DISTORTION_PRECISION_ADJUSTMENT(encRCSeq->getbitDepth()));
+#else
+#if FULL_NBIT
+ int bitdepth_luma_scale = 2 * (encRCSeq->getbitDepth() - 8);
+#else
+ int bitdepth_luma_scale = 0;
+#endif
+#endif
+
+ double hierarQp = 4.2005 * log(encRCSeq->getLastLambda() / pow(2.0, bitdepth_luma_scale)) + 13.7122; // the qp of POC16
+ double qpLev2 = (hierarQp + 0.0) + 0.2016 * (hierarQp + 0.0) - 4.8848;
+ double qpLev3 = (hierarQp + 3.0) + 0.22286 * (hierarQp + 3.0) - 5.7476;
+ double qpLev4 = (hierarQp + 4.0) + 0.2333 * (hierarQp + 4.0) - 5.9;
+ double qpLev5 = (hierarQp + 5.0) + 0.3 * (hierarQp + 5.0) - 7.1444;
+
+ double lambdaLev1 = exp((hierarQp - 13.7122) / 4.2005) *pow(2.0, bitdepth_luma_scale);
+ double lambdaLev2 = exp((qpLev2 - 13.7122) / 4.2005) * pow(2.0, bitdepth_luma_scale);
+ double lambdaLev3 = exp((qpLev3 - 13.7122) / 4.2005) * pow(2.0, bitdepth_luma_scale);
+ double lambdaLev4 = exp((qpLev4 - 13.7122) / 4.2005) * pow(2.0, bitdepth_luma_scale);
+ double lambdaLev5 = exp((qpLev5 - 13.7122) / 4.2005) * pow(2.0, bitdepth_luma_scale);
+#else
+ double hierarQp = 4.2005 * log(encRCSeq->getLastLambda()) + 13.7122; // the qp of POC16
+ double qpLev2 = (hierarQp + 0.0) + 0.2016 * (hierarQp + 0.0) - 4.8848;
+ double qpLev3 = (hierarQp + 3.0) + 0.22286 * (hierarQp + 3.0) - 5.7476;
+ double qpLev4 = (hierarQp + 4.0) + 0.2333 * (hierarQp + 4.0) - 5.9;
+ double qpLev5 = (hierarQp + 5.0) + 0.3 * (hierarQp + 5.0) - 7.1444;
+
+ double lambdaLev1 = exp((hierarQp - 13.7122) / 4.2005);
+ double lambdaLev2 = exp((qpLev2 - 13.7122) / 4.2005);
+ double lambdaLev3 = exp((qpLev3 - 13.7122) / 4.2005);
+ double lambdaLev4 = exp((qpLev4 - 13.7122) / 4.2005);
+ double lambdaLev5 = exp((qpLev5 - 13.7122) / 4.2005);
+#endif
+
+ lambdaRatio[0] = 1.0;
+ lambdaRatio[1] = lambdaLev2 / lambdaLev1;
+ lambdaRatio[2] = lambdaLev3 / lambdaLev1;
+ lambdaRatio[3] = lambdaLev4 / lambdaLev1;
+ lambdaRatio[4] = lambdaLev5 / lambdaLev1;
+ lambdaRatio[5] = lambdaLev5 / lambdaLev1;
+ lambdaRatio[6] = lambdaLev4 / lambdaLev1;
+ lambdaRatio[7] = lambdaLev5 / lambdaLev1;
+ lambdaRatio[8] = lambdaLev5 / lambdaLev1;
+ lambdaRatio[9] = lambdaLev3 / lambdaLev1;
+ lambdaRatio[10] = lambdaLev4 / lambdaLev1;
+ lambdaRatio[11] = lambdaLev5 / lambdaLev1;
+ lambdaRatio[12] = lambdaLev5 / lambdaLev1;
+ lambdaRatio[13] = lambdaLev4 / lambdaLev1;
+ lambdaRatio[14] = lambdaLev5 / lambdaLev1;
+ lambdaRatio[15] = lambdaLev5 / lambdaLev1;
+ }
+ }
+#endif
xCalEquaCoeff( encRCSeq, lambdaRatio, equaCoeffA, equaCoeffB, encRCSeq->getGOPSize() );
+#if JVET_K0390_RATECTRL
+ basicLambda = xSolveEqua(encRCSeq, targetBpp, equaCoeffA, equaCoeffB, encRCSeq->getGOPSize());
+#else
basicLambda = xSolveEqua( targetBpp, equaCoeffA, equaCoeffB, encRCSeq->getGOPSize() );
+#endif
encRCSeq->setAllBitRatio( basicLambda, equaCoeffA, equaCoeffB );
delete []lambdaRatio;
@@ -396,7 +513,11 @@ void EncRCGOP::xCalEquaCoeff( EncRCSeq* encRCSeq, double* lambdaRatio, double* e
}
}
+#if JVET_K0390_RATECTRL
+double EncRCGOP::xSolveEqua(EncRCSeq* encRCSeq, double targetBpp, double* equaCoeffA, double* equaCoeffB, int GOPSize)
+#else
double EncRCGOP::xSolveEqua( double targetBpp, double* equaCoeffA, double* equaCoeffB, int GOPSize )
+#endif
{
double solution = 100.0;
double minNumber = 0.1;
@@ -406,7 +527,13 @@ double EncRCGOP::xSolveEqua( double targetBpp, double* equaCoeffA, double* equaC
double fx = 0.0;
for ( int j=0; j<GOPSize; j++ )
{
+#if JVET_K0390_RATECTRL
+ double tmpBpp = equaCoeffA[j] * pow(solution, equaCoeffB[j]);
+ double actualBpp = tmpBpp * (double)encRCSeq->getPicPara(encRCSeq->getGOPID2Level(j)).m_validPix / (double)encRCSeq->getNumPixel();
+ fx += actualBpp;
+#else
fx += equaCoeffA[j] * pow( solution, equaCoeffB[j] );
+#endif
}
if ( fabs( fx - targetBpp ) < 0.000001 )
@@ -484,6 +611,10 @@ EncRCPic::EncRCPic()
m_picActualBits = 0;
m_picQP = 0;
m_picLambda = 0.0;
+#if JVET_K0390_RATECTRL
+ m_picMSE = 0.0;
+ m_validPixelsInPic = 0;
+#endif
}
EncRCPic::~EncRCPic()
@@ -641,6 +772,10 @@ void EncRCPic::create( EncRCSeq* encRCSeq, EncRCGOP* encRCGOP, int frameLevel, l
{
LCUIdx = j*picWidthInLCU + i;
m_LCUs[LCUIdx].m_actualBits = 0;
+#if JVET_K0390_RATECTRL
+ m_LCUs[LCUIdx].m_actualSSE = 0.0;
+ m_LCUs[LCUIdx].m_actualMSE = 0.0;
+#endif
m_LCUs[LCUIdx].m_QP = 0;
m_LCUs[LCUIdx].m_lambda = 0.0;
m_LCUs[LCUIdx].m_targetBits = 0;
@@ -654,6 +789,10 @@ void EncRCPic::create( EncRCSeq* encRCSeq, EncRCGOP* encRCGOP, int frameLevel, l
m_picActualBits = 0;
m_picQP = 0;
m_picLambda = 0.0;
+#if JVET_K0390_RATECTRL
+ m_validPixelsInPic = 0;
+ m_picMSE = 0.0;
+#endif
}
void EncRCPic::destroy()
@@ -673,6 +812,19 @@ double EncRCPic::estimatePicLambda( list<EncRCPic*>& listPreviousPictures, Slice
double alpha = m_encRCSeq->getPicPara( m_frameLevel ).m_alpha;
double beta = m_encRCSeq->getPicPara( m_frameLevel ).m_beta;
double bpp = (double)m_targetBits/(double)m_numberOfPixel;
+
+#if JVET_K0390_RATECTRL
+ int lastPicValPix = 0;
+ if (listPreviousPictures.size() > 0)
+ {
+ lastPicValPix = m_encRCSeq->getPicPara(m_frameLevel).m_validPix;
+ }
+ if (lastPicValPix > 0)
+ {
+ bpp = (double)m_targetBits / (double)lastPicValPix;
+ }
+#endif
+
double estLambda;
if (eSliceType == I_SLICE)
{
@@ -727,6 +879,10 @@ double EncRCPic::estimatePicLambda( list<EncRCPic*>& listPreviousPictures, Slice
estLambda = 0.1;
}
+#if JVET_K0390_RATECTRL
+ //Avoid different results in different platforms. The problem is caused by the different results of pow() in different platforms.
+ estLambda = double(int64_t(estLambda * (double)LAMBDA_PREC + 0.5)) / (double)LAMBDA_PREC;
+#endif
m_estPicLambda = estLambda;
double totalWeight = 0.0;
@@ -764,7 +920,24 @@ double EncRCPic::estimatePicLambda( list<EncRCPic*>& listPreviousPictures, Slice
int EncRCPic::estimatePicQP( double lambda, list<EncRCPic*>& listPreviousPictures )
{
+#if RATECTRL_FIX_FULLNBIT
+#if DISTORTION_LAMBDA_BUGFIX
+ int bitdepth_luma_scale =
+ 2
+ * (m_encRCSeq->getbitDepth() - 8
+ - DISTORTION_PRECISION_ADJUSTMENT(m_encRCSeq->getbitDepth()));
+#else
+#if FULL_NBIT
+ int bitdepth_luma_scale = 2 * (m_encRCSeq->getbitDepth() - 8);
+#else
+ int bitdepth_luma_scale = 0;
+#endif
+#endif
+
+ int QP = int(4.2005 * log(lambda / pow(2.0, bitdepth_luma_scale)) + 13.7122 + 0.5);
+#else
int QP = int( 4.2005 * log( lambda ) + 13.7122 + 0.5 );
+#endif
int lastLevelQP = g_RCInvalidQPValue;
int lastPicQP = g_RCInvalidQPValue;
@@ -895,13 +1068,34 @@ double EncRCPic::getLCUEstLambda( double bpp )
estLambda = 0.1;
}
+#if JVET_K0390_RATECTRL
+ //Avoid different results in different platforms. The problem is caused by the different results of pow() in different platforms.
+ estLambda = double(int64_t(estLambda * (double)LAMBDA_PREC + 0.5)) / (double)LAMBDA_PREC;
+#endif
return estLambda;
}
int EncRCPic::getLCUEstQP( double lambda, int clipPicQP )
{
int LCUIdx = getLCUCoded();
+#if RATECTRL_FIX_FULLNBIT
+#if DISTORTION_LAMBDA_BUGFIX
+ int bitdepth_luma_scale =
+ 2
+ * (m_encRCSeq->getbitDepth() - 8
+ - DISTORTION_PRECISION_ADJUSTMENT(m_encRCSeq->getbitDepth()));
+#else
+#if FULL_NBIT
+ int bitdepth_luma_scale = 2 * (m_encRCSeq->getbitDepth() - 8);
+#else
+ int bitdepth_luma_scale = 0;
+#endif
+#endif
+
+ int estQP = int(4.2005 * log(lambda / pow(2.0, bitdepth_luma_scale)) + 13.7122 + 0.5);
+#else
int estQP = int( 4.2005 * log( lambda ) + 13.7122 + 0.5 );
+#endif
//for Lambda clip, LCU level clip
int clipNeighbourQP = g_RCInvalidQPValue;
@@ -929,6 +1123,9 @@ void EncRCPic::updateAfterCTU( int LCUIdx, int bits, int QP, double lambda, bool
m_LCUs[LCUIdx].m_actualBits = bits;
m_LCUs[LCUIdx].m_QP = QP;
m_LCUs[LCUIdx].m_lambda = lambda;
+#if JVET_K0390_RATECTRL
+ m_LCUs[LCUIdx].m_actualSSE = m_LCUs[LCUIdx].m_actualMSE * m_LCUs[LCUIdx].m_numberOfPixel;
+#endif
m_LCULeft--;
m_bitsLeft -= bits;
@@ -964,7 +1161,34 @@ void EncRCPic::updateAfterCTU( int LCUIdx, int bits, int QP, double lambda, bool
TRCParameter rcPara;
rcPara.m_alpha = alpha;
rcPara.m_beta = beta;
+#if JVET_K0390_RATECTRL
+ if (QP == g_RCInvalidQPValue && m_encRCSeq->getAdaptiveBits() == 1)
+ {
+ rcPara.m_validPix = 0;
+ }
+ else
+ {
+ rcPara.m_validPix = LCUTotalPixels;
+ }
+
+ double MSE = m_LCUs[LCUIdx].m_actualMSE;
+ double updatedK = bpp * inputLambda / MSE;
+ double updatedC = MSE / pow(bpp, -updatedK);
+ rcPara.m_alpha = updatedC * updatedK;
+ rcPara.m_beta = -updatedK - 1.0;
+
+ if (bpp > 0 && updatedK > 0.0001)
+ {
+ m_encRCSeq->setLCUPara(m_frameLevel, LCUIdx, rcPara);
+ }
+ else
+ {
+ rcPara.m_alpha = Clip3(0.0001, g_RCAlphaMaxValue, rcPara.m_alpha);
+ m_encRCSeq->setLCUPara(m_frameLevel, LCUIdx, rcPara);
+ }
+#else
m_encRCSeq->setLCUPara( m_frameLevel, LCUIdx, rcPara );
+#endif
return;
}
@@ -981,7 +1205,34 @@ void EncRCPic::updateAfterCTU( int LCUIdx, int bits, int QP, double lambda, bool
TRCParameter rcPara;
rcPara.m_alpha = alpha;
rcPara.m_beta = beta;
+#if JVET_K0390_RATECTRL
+ if (QP == g_RCInvalidQPValue && m_encRCSeq->getAdaptiveBits() == 1)
+ {
+ rcPara.m_validPix = 0;
+ }
+ else
+ {
+ rcPara.m_validPix = LCUTotalPixels;
+ }
+
+ double MSE = m_LCUs[LCUIdx].m_actualMSE;
+ double updatedK = bpp * inputLambda / MSE;
+ double updatedC = MSE / pow(bpp, -updatedK);
+ rcPara.m_alpha = updatedC * updatedK;
+ rcPara.m_beta = -updatedK - 1.0;
+
+ if (bpp > 0 && updatedK > 0.0001)
+ {
+ m_encRCSeq->setLCUPara(m_frameLevel, LCUIdx, rcPara);
+ }
+ else
+ {
+ rcPara.m_alpha = Clip3(0.0001, g_RCAlphaMaxValue, rcPara.m_alpha);
+ m_encRCSeq->setLCUPara(m_frameLevel, LCUIdx, rcPara);
+ }
+#else
m_encRCSeq->setLCUPara( m_frameLevel, LCUIdx, rcPara );
+#endif
}
@@ -1017,16 +1268,41 @@ double EncRCPic::calAverageLambda()
double totalLambdas = 0.0;
int numTotalLCUs = 0;
+#if JVET_K0390_RATECTRL
+ double totalSSE = 0.0;
+ int totalPixels = 0;
+#endif
int i;
for ( i=0; i<m_numberOfLCU; i++ )
{
if ( m_LCUs[i].m_lambda > 0.01 )
{
+#if JVET_K0390_RATECTRL
+ if (m_LCUs[i].m_QP > 0 || m_encRCSeq->getAdaptiveBits() != 1)
+ {
+ m_validPixelsInPic += m_LCUs[i].m_numberOfPixel;
+
+ totalLambdas += log(m_LCUs[i].m_lambda);
+ numTotalLCUs++;
+ }
+#else
totalLambdas += log( m_LCUs[i].m_lambda );
numTotalLCUs++;
+#endif
+
+#if JVET_K0390_RATECTRL
+ if (m_LCUs[i].m_QP > 0 || m_encRCSeq->getAdaptiveBits() != 1)
+ {
+ totalSSE += m_LCUs[i].m_actualSSE;
+ totalPixels += m_LCUs[i].m_numberOfPixel;
+ }
+#endif
}
}
+#if JVET_K0390_RATECTRL
+ setPicMSE(totalPixels > 0 ? totalSSE / (double)totalPixels : 1.0); //1.0 is useless in the following process, just to make sure the divisor not be 0
+#endif
double avgLambda;
if( numTotalLCUs == 0 )
{
@@ -1065,7 +1341,11 @@ void EncRCPic::updateAfterPicture( int actualHeaderBits, int actualTotalBits, do
{
// update parameters
double picActualBits = ( double )m_picActualBits;
+#if JVET_K0390_RATECTRL
+ double picActualBpp = picActualBits / (double)m_validPixelsInPic;
+#else
double picActualBpp = picActualBits/(double)m_numberOfPixel;
+#endif
double calLambda = alpha * pow( picActualBpp, beta );
double inputLambda = m_picLambda;
@@ -1080,7 +1360,26 @@ void EncRCPic::updateAfterPicture( int actualHeaderBits, int actualTotalBits, do
TRCParameter rcPara;
rcPara.m_alpha = alpha;
rcPara.m_beta = beta;
+#if JVET_K0390_RATECTRL
+ double avgMSE = getPicMSE();
+ double updatedK = picActualBpp * averageLambda / avgMSE;
+ double updatedC = avgMSE / pow(picActualBpp, -updatedK);
+
+ if (m_frameLevel > 0) //only use for level > 0
+ {
+ rcPara.m_alpha = updatedC * updatedK;
+ rcPara.m_beta = -updatedK - 1.0;
+ }
+
+ rcPara.m_validPix = m_validPixelsInPic;
+
+ if (m_validPixelsInPic > 0)
+ {
+ m_encRCSeq->setPicPara(m_frameLevel, rcPara);
+ }
+#else
m_encRCSeq->setPicPara( m_frameLevel, rcPara );
+#endif
return;
}
@@ -1099,8 +1398,27 @@ void EncRCPic::updateAfterPicture( int actualHeaderBits, int actualTotalBits, do
TRCParameter rcPara;
rcPara.m_alpha = alpha;
rcPara.m_beta = beta;
+#if JVET_K0390_RATECTRL
+ double picActualBpp = (double)m_picActualBits / (double)m_validPixelsInPic;
+
+ double avgMSE = getPicMSE();
+ double updatedK = picActualBpp * averageLambda / avgMSE;
+ double updatedC = avgMSE / pow(picActualBpp, -updatedK);
+ if (m_frameLevel > 0) //only use for level > 0
+ {
+ rcPara.m_alpha = updatedC * updatedK;
+ rcPara.m_beta = -updatedK - 1.0;
+ }
+
+ rcPara.m_validPix = m_validPixelsInPic;
+ if (m_validPixelsInPic > 0)
+ {
+ m_encRCSeq->setPicPara(m_frameLevel, rcPara);
+ }
+#else
m_encRCSeq->setPicPara( m_frameLevel, rcPara );
+#endif
if ( m_frameLevel == 1 )
{
@@ -1188,12 +1506,38 @@ double EncRCPic::getLCUEstLambdaAndQP(double bpp, int clipPicQP, int *estQP)
minQP = max(clipNeighbourQP - 1, minQP);
}
+#if RATECTRL_FIX_FULLNBIT
+#if DISTORTION_LAMBDA_BUGFIX
+ int bitdepth_luma_scale =
+ 2
+ * (m_encRCSeq->getbitDepth() - 8
+ - DISTORTION_PRECISION_ADJUSTMENT(m_encRCSeq->getbitDepth()));
+#else
+#if FULL_NBIT
+ int bitdepth_luma_scale = 2 * (m_encRCSeq->getbitDepth() - 8);
+#else
+ int bitdepth_luma_scale = 0;
+#endif
+#endif
+
+ double maxLambda = exp(((double)(maxQP + 0.49) - 13.7122) / 4.2005) * pow(2.0, bitdepth_luma_scale);
+ double minLambda = exp(((double)(minQP - 0.49) - 13.7122) / 4.2005) * pow(2.0, bitdepth_luma_scale);
+#else
double maxLambda=exp(((double)(maxQP+0.49)-13.7122)/4.2005);
double minLambda=exp(((double)(minQP-0.49)-13.7122)/4.2005);
+#endif
estLambda = Clip3(minLambda, maxLambda, estLambda);
+#if JVET_K0390_RATECTRL
+ //Avoid different results in different platforms. The problem is caused by the different results of pow() in different platforms.
+ estLambda = double(int64_t(estLambda * (double)LAMBDA_PREC + 0.5)) / (double)LAMBDA_PREC;
+#endif
+#if RATECTRL_FIX_FULLNBIT
+ *estQP = int(4.2005 * log(estLambda / pow(2.0, bitdepth_luma_scale)) + 13.7122 + 0.5);
+#else
*estQP = int( 4.2005 * log(estLambda) + 13.7122 + 0.5 );
+#endif
*estQP = Clip3(minQP, maxQP, *estQP);
return estLambda;
@@ -1231,7 +1575,11 @@ void RateCtrl::destroy()
}
}
+#if RATECTRL_FIX_FULLNBIT
+void RateCtrl::init(int totalFrames, int targetBitrate, int frameRate, int GOPSize, int picWidth, int picHeight, int LCUWidth, int LCUHeight, int bitDepth, int keepHierBits, bool useLCUSeparateModel, GOPEntry GOPList[MAX_GOP])
+#else
void RateCtrl::init( int totalFrames, int targetBitrate, int frameRate, int GOPSize, int picWidth, int picHeight, int LCUWidth, int LCUHeight, int keepHierBits, bool useLCUSeparateModel, GOPEntry GOPList[MAX_GOP] )
+#endif
{
destroy();
@@ -1251,7 +1599,11 @@ void RateCtrl::init( int totalFrames, int targetBitrate, int frameRate, int GOPS
{
numberOfLevel = int( log((double)GOPSize)/log(2.0) + 0.5 ) + 1;
}
+#if JVET_K0390_RATECTRL
+ if (!isLowdelay && (GOPSize == 16 || GOPSize == 8))
+#else
if ( !isLowdelay && GOPSize == 8 )
+#endif
{
numberOfLevel = int( log((double)GOPSize)/log(2.0) + 0.5 ) + 1;
}
@@ -1361,6 +1713,92 @@ void RateCtrl::init( int totalFrames, int targetBitrate, int frameRate, int GOPS
adaptiveBit = 2;
}
}
+#if JVET_K0390_RATECTRL
+ else if (GOPSize == 16 && !isLowdelay)
+ {
+ if (bpp > 0.2)
+ {
+ bitsRatio[0] = 10;
+ bitsRatio[1] = 8;
+ bitsRatio[2] = 4;
+ bitsRatio[3] = 2;
+ bitsRatio[4] = 1;
+ bitsRatio[5] = 1;
+ bitsRatio[6] = 2;
+ bitsRatio[7] = 1;
+ bitsRatio[8] = 1;
+ bitsRatio[9] = 4;
+ bitsRatio[10] = 2;
+ bitsRatio[11] = 1;
+ bitsRatio[12] = 1;
+ bitsRatio[13] = 2;
+ bitsRatio[14] = 1;
+ bitsRatio[15] = 1;
+ }
+ else if (bpp > 0.1)
+ {
+ bitsRatio[0] = 15;
+ bitsRatio[1] = 9;
+ bitsRatio[2] = 4;
+ bitsRatio[3] = 2;
+ bitsRatio[4] = 1;
+ bitsRatio[5] = 1;
+ bitsRatio[6] = 2;
+ bitsRatio[7] = 1;
+ bitsRatio[8] = 1;
+ bitsRatio[9] = 4;
+ bitsRatio[10] = 2;
+ bitsRatio[11] = 1;
+ bitsRatio[12] = 1;
+ bitsRatio[13] = 2;
+ bitsRatio[14] = 1;
+ bitsRatio[15] = 1;
+ }
+ else if (bpp > 0.05)
+ {
+ bitsRatio[0] = 40;
+ bitsRatio[1] = 17;
+ bitsRatio[2] = 7;
+ bitsRatio[3] = 2;
+ bitsRatio[4] = 1;
+ bitsRatio[5] = 1;
+ bitsRatio[6] = 2;
+ bitsRatio[7] = 1;
+ bitsRatio[8] = 1;
+ bitsRatio[9] = 7;
+ bitsRatio[10] = 2;
+ bitsRatio[11] = 1;
+ bitsRatio[12] = 1;
+ bitsRatio[13] = 2;
+ bitsRatio[14] = 1;
+ bitsRatio[15] = 1;
+ }
+ else
+ {
+ bitsRatio[0] = 40;
+ bitsRatio[1] = 15;
+ bitsRatio[2] = 6;
+ bitsRatio[3] = 3;
+ bitsRatio[4] = 1;
+ bitsRatio[5] = 1;
+ bitsRatio[6] = 3;
+ bitsRatio[7] = 1;
+ bitsRatio[8] = 1;
+ bitsRatio[9] = 6;
+ bitsRatio[10] = 3;
+ bitsRatio[11] = 1;
+ bitsRatio[12] = 1;
+ bitsRatio[13] = 3;
+ bitsRatio[14] = 1;
+ bitsRatio[15] = 1;
+ }
+
+ if (keepHierBits == 2)
+ {
+ adaptiveBit = 3;
+ }
+ }
+#endif
else
{
msg( WARNING, "\n hierarchical bit allocation is not support for the specified coding structure currently.\n" );
@@ -1397,6 +1835,27 @@ void RateCtrl::init( int totalFrames, int targetBitrate, int frameRate, int GOPS
GOPID2Level[6] = 4;
GOPID2Level[7] = 4;
}
+#if JVET_K0390_RATECTRL
+ else if (GOPSize == 16 && !isLowdelay)
+ {
+ GOPID2Level[0] = 1;
+ GOPID2Level[1] = 2;
+ GOPID2Level[2] = 3;
+ GOPID2Level[3] = 4;
+ GOPID2Level[4] = 5;
+ GOPID2Level[5] = 5;
+ GOPID2Level[6] = 4;
+ GOPID2Level[7] = 5;
+ GOPID2Level[8] = 5;
+ GOPID2Level[9] = 3;
+ GOPID2Level[10] = 4;
+ GOPID2Level[11] = 5;
+ GOPID2Level[12] = 5;
+ GOPID2Level[13] = 4;
+ GOPID2Level[14] = 5;
+ GOPID2Level[15] = 5;
+ }
+#endif
}
if ( !isLowdelay && GOPSize == 8 )
@@ -1410,11 +1869,35 @@ void RateCtrl::init( int totalFrames, int targetBitrate, int frameRate, int GOPS
GOPID2Level[6] = 4;
GOPID2Level[7] = 4;
}
+#if JVET_K0390_RATECTRL
+ else if (GOPSize == 16 && !isLowdelay)
+ {
+ GOPID2Level[0] = 1;
+ GOPID2Level[1] = 2;
+ GOPID2Level[2] = 3;
+ GOPID2Level[3] = 4;
+ GOPID2Level[4] = 5;
+ GOPID2Level[5] = 5;
+ GOPID2Level[6] = 4;
+ GOPID2Level[7] = 5;
+ GOPID2Level[8] = 5;
+ GOPID2Level[9] = 3;
+ GOPID2Level[10] = 4;
+ GOPID2Level[11] = 5;
+ GOPID2Level[12] = 5;
+ GOPID2Level[13] = 4;
+ GOPID2Level[14] = 5;
+ GOPID2Level[15] = 5;
+ }
+#endif
m_encRCSeq = new EncRCSeq;
m_encRCSeq->create( totalFrames, targetBitrate, frameRate, GOPSize, picWidth, picHeight, LCUWidth, LCUHeight, numberOfLevel, useLCUSeparateModel, adaptiveBit );
m_encRCSeq->initBitsRatio( bitsRatio );
m_encRCSeq->initGOPID2Level( GOPID2Level );
+#if RATECTRL_FIX_FULLNBIT
+ m_encRCSeq->setBitDepth(bitDepth);
+#endif
m_encRCSeq->initPicPara();
if ( useLCUSeparateModel )
{
diff --git a/source/Lib/EncoderLib/RateCtrl.h b/source/Lib/EncoderLib/RateCtrl.h
index 8a6a58c13ebb19ed10b3dfa5449a795772327535..52b44fca853c41ba99692f051005668743e3d5f4 100644
--- a/source/Lib/EncoderLib/RateCtrl.h
+++ b/source/Lib/EncoderLib/RateCtrl.h
@@ -84,12 +84,19 @@ struct TRCLCU
int m_numberOfPixel;
double m_costIntra;
int m_targetBitsLeft;
+#if JVET_K0390_RATECTRL
+ double m_actualSSE;
+ double m_actualMSE;
+#endif
};
struct TRCParameter
{
double m_alpha;
double m_beta;
+#if JVET_K0390_RATECTRL
+ int m_validPix;
+#endif
};
class EncRCSeq
@@ -147,6 +154,10 @@ public:
int getAdaptiveBits() { return m_adaptiveBit; }
double getLastLambda() { return m_lastLambda; }
void setLastLambda( double lamdba ) { m_lastLambda = lamdba; }
+#if RATECTRL_FIX_FULLNBIT
+ void setBitDepth(int bitDepth) { m_bitDepth = bitDepth; }
+ int getbitDepth() { return m_bitDepth; }
+#endif
private:
int m_totalFrames;
@@ -177,6 +188,9 @@ private:
int m_adaptiveBit;
double m_lastLambda;
+#if RATECTRL_FIX_FULLNBIT
+ int m_bitDepth;
+#endif
};
class EncRCGOP
@@ -193,7 +207,11 @@ public:
private:
int xEstGOPTargetBits( EncRCSeq* encRCSeq, int GOPSize );
void xCalEquaCoeff( EncRCSeq* encRCSeq, double* lambdaRatio, double* equaCoeffA, double* equaCoeffB, int GOPSize );
+#if JVET_K0390_RATECTRL
+ double xSolveEqua(EncRCSeq* encRCSeq, double targetBpp, double* equaCoeffA, double* equaCoeffB, int GOPSize);
+#else
double xSolveEqua( double targetBpp, double* equaCoeffA, double* equaCoeffB, int GOPSize );
+#endif
public:
EncRCSeq* getEncRCSeq() { return m_encRCSeq; }
@@ -282,6 +300,10 @@ public:
void setPicEstQP( int QP ) { m_estPicQP = QP; }
double getPicEstLambda() { return m_estPicLambda; }
void setPicEstLambda( double lambda ) { m_picLambda = lambda; }
+#if JVET_K0390_RATECTRL
+ double getPicMSE() { return m_picMSE; }
+ void setPicMSE(double avgMSE) { m_picMSE = avgMSE; }
+#endif
private:
EncRCSeq* m_encRCSeq;
@@ -309,6 +331,10 @@ private:
int m_picActualBits; // the whole picture, including header
int m_picQP; // in integer form
double m_picLambda;
+#if JVET_K0390_RATECTRL
+ double m_picMSE;
+ int m_validPixelsInPic;
+#endif
};
class RateCtrl
@@ -318,7 +344,11 @@ public:
~RateCtrl();
public:
+#if RATECTRL_FIX_FULLNBIT
+ void init(int totalFrames, int targetBitrate, int frameRate, int GOPSize, int picWidth, int picHeight, int LCUWidth, int LCUHeight, int bitDepth, int keepHierBits, bool useLCUSeparateModel, GOPEntry GOPList[MAX_GOP]);
+#else
void init( int totalFrames, int targetBitrate, int frameRate, int GOPSize, int picWidth, int picHeight, int LCUWidth, int LCUHeight, int keepHierBits, bool useLCUSeparateModel, GOPEntry GOPList[MAX_GOP] );
+#endif
void destroy();
void initRCPic( int frameLevel );
void initRCGOP( int numberOfPictures );
diff --git a/source/Lib/Utilities/CMakeLists.txt b/source/Lib/Utilities/CMakeLists.txt
index b8976132ec48f5a9b5d7f8ad0e256ec43888bb54..0b0464411f660fffc5e6a3ab431d0c0331724388 100644
--- a/source/Lib/Utilities/CMakeLists.txt
+++ b/source/Lib/Utilities/CMakeLists.txt
@@ -17,7 +17,7 @@ add_library( ${LIB_NAME} STATIC ${SRC_FILES} ${INC_FILES} ${NATVIS_FILES} )
target_compile_definitions( ${LIB_NAME} PUBLIC )
if( ENABLE_VTM )
- target_compile_definitions( ${LIB_NAME} PUBLIC JEM_TOOLS=0 )
+ target_compile_definitions( ${LIB_NAME} PUBLIC BMS_TOOLS=0 )
endif()
if( EXTENSION_360_VIDEO )
diff --git a/source/Lib/Utilities/VideoIOYuv.cpp b/source/Lib/Utilities/VideoIOYuv.cpp
index 9494596cc472e587a13d9e6c64835454ef427d1a..5f26d44d5aee4afee42ca848573ef2a0233c3900 100644
--- a/source/Lib/Utilities/VideoIOYuv.cpp
+++ b/source/Lib/Utilities/VideoIOYuv.cpp
@@ -413,24 +413,25 @@ static bool readPlane(Pel* dst,
* @param fileBitDepth component bit depth in file
* @return true for success, false in case of error
*/
-static bool writePlane(ostream& fd, const Pel* src, bool is16bit,
+static bool writePlane(ostream& fd, const Pel* src,
+ const bool is16bit,
const uint32_t stride_src,
uint32_t width444, uint32_t height444,
const ComponentID compID,
const ChromaFormat srcFormat,
const ChromaFormat fileFormat,
- const int fileBitDepth)
+ const uint32_t fileBitDepth,
+ const uint32_t packedYUVOutputMode = 0)
{
const uint32_t csx_file =getComponentScaleX(compID, fileFormat);
const uint32_t csy_file =getComponentScaleY(compID, fileFormat);
const uint32_t csx_src =getComponentScaleX(compID, srcFormat);
const uint32_t csy_src =getComponentScaleY(compID, srcFormat);
-/* const uint32_t stride_src = stride444>>csx_src;*/
-
- const uint32_t stride_file = (width444 * (is16bit ? 2 : 1)) >> csx_file;
- const uint32_t width_file = width444 >>csx_file;
- const uint32_t height_file = height444>>csy_file;
+ const uint32_t width_file = width444 >> csx_file;
+ const uint32_t height_file = height444 >> csy_file;
+ const bool writePYUV = (packedYUVOutputMode > 0) && (fileBitDepth == 10 || fileBitDepth == 12) && ((width_file & (1 + (fileBitDepth & 3))) == 0);
+ const uint32_t stride_file = writePYUV ? (width444 * fileBitDepth) >> (csx_file + 3) : (width444 * (is16bit ? 2 : 1)) >> csx_file;
std::vector<uint8_t> bufVec(stride_file);
uint8_t *buf=&(bufVec[0]);
@@ -438,15 +439,106 @@ static bool writePlane(ostream& fd, const Pel* src, bool is16bit,
const Pel *pSrcBuf = src;
const int srcbuf_stride = stride_src;
+ if (writePYUV)
+ {
+ const uint32_t mask_y_file = (1 << csy_file) - 1;
+ const uint32_t mask_y_src = (1 << csy_src ) - 1;
+ const uint32_t widthS_file = width_file >> (fileBitDepth == 12 ? 1 : 2);
+
+ for (uint32_t y444 = 0; y444 < height444; y444++)
+ {
+ if ((y444 & mask_y_file) == 0) // write a new line to file
+ {
+ if (csx_file < csx_src)
+ {
+ // eg file is 444, source is 422.
+ const uint32_t sx = csx_src - csx_file;
+
+ if (fileBitDepth == 10) // write 4 values into 5 bytes
+ {
+ for (uint32_t x = 0; x < widthS_file; x++)
+ {
+ const uint32_t src0 = pSrcBuf[(4*x ) >> sx];
+ const uint32_t src1 = pSrcBuf[(4*x+1) >> sx];
+ const uint32_t src2 = pSrcBuf[(4*x+2) >> sx];
+ const uint32_t src3 = pSrcBuf[(4*x+3) >> sx];
+
+ buf[5*x ] = ((src0 ) & 0xff); // src0:76543210
+ buf[5*x+1] = ((src1 << 2) & 0xfc) + ((src0 >> 8) & 0x03);
+ buf[5*x+2] = ((src2 << 4) & 0xf0) + ((src1 >> 6) & 0x0f);
+ buf[5*x+3] = ((src3 << 6) & 0xc0) + ((src2 >> 4) & 0x3f);
+ buf[5*x+4] = ((src3 >> 2) & 0xff); // src3:98765432
+ }
+ }
+ else if (fileBitDepth == 12) //...2 values into 3 bytes
+ {
+ for (uint32_t x = 0; x < widthS_file; x++)
+ {
+ const uint32_t src0 = pSrcBuf[(2*x ) >> sx];
+ const uint32_t src1 = pSrcBuf[(2*x+1) >> sx];
+
+ buf[3*x ] = ((src0 ) & 0xff); // src0:76543210
+ buf[3*x+1] = ((src1 << 4) & 0xf0) + ((src0 >> 8) & 0x0f);
+ buf[3*x+2] = ((src1 >> 4) & 0xff); // src1:BA987654
+ }
+ }
+ }
+ else
+ {
+ // eg file is 422, source is 444.
+ const uint32_t sx = csx_file - csx_src;
+
+ if (fileBitDepth == 10) // write 4 values into 5 bytes
+ {
+ for (uint32_t x = 0; x < widthS_file; x++)
+ {
+ const uint32_t src0 = pSrcBuf[(4*x ) << sx];
+ const uint32_t src1 = pSrcBuf[(4*x+1) << sx];
+ const uint32_t src2 = pSrcBuf[(4*x+2) << sx];
+ const uint32_t src3 = pSrcBuf[(4*x+3) << sx];
+
+ buf[5*x ] = ((src0 ) & 0xff); // src0:76543210
+ buf[5*x+1] = ((src1 << 2) & 0xfc) + ((src0 >> 8) & 0x03);
+ buf[5*x+2] = ((src2 << 4) & 0xf0) + ((src1 >> 6) & 0x0f);
+ buf[5*x+3] = ((src3 << 6) & 0xc0) + ((src2 >> 4) & 0x3f);
+ buf[5*x+4] = ((src3 >> 2) & 0xff); // src3:98765432
+ }
+ }
+ else if (fileBitDepth == 12) //...2 values into 3 bytes
+ {
+ for (uint32_t x = 0; x < widthS_file; x++)
+ {
+ const uint32_t src0 = pSrcBuf[(2*x ) << sx];
+ const uint32_t src1 = pSrcBuf[(2*x+1) << sx];
+
+ buf[3*x ] = ((src0 ) & 0xff); // src0:76543210
+ buf[3*x+1] = ((src1 << 4) & 0xf0) + ((src0 >> 8) & 0x0f);
+ buf[3*x+2] = ((src1 >> 4) & 0xff); // src1:BA987654
+ }
+ }
+ }
+ fd.write (reinterpret_cast<const char*>(buf), stride_file);
+ if (fd.eof() || fd.fail())
+ {
+ return false;
+ }
+ }
+ if ((y444 & mask_y_src) == 0)
+ {
+ pSrcBuf += srcbuf_stride;
+ }
+ }
+ }
+ else // !writePYUV
if (compID!=COMPONENT_Y && (fileFormat==CHROMA_400 || srcFormat==CHROMA_400))
{
if (fileFormat!=CHROMA_400)
{
- const uint32_t value = 1u << (fileBitDepth - 1);
+ const uint32_t value = 1 << (fileBitDepth - 1);
- for(uint32_t y=0; y< height_file; y++)
+ for (uint32_t y = 0; y < height_file; y++)
{
if (!is16bit)
{
@@ -461,7 +553,7 @@ static bool writePlane(ostream& fd, const Pel* src, bool is16bit,
uint16_t val(value);
for (uint32_t x = 0; x < width_file; x++)
{
- buf[2*x+0]= (val>>0) & 0xff;
+ buf[2*x ]= (val>>0) & 0xff;
buf[2*x+1]= (val>>8) & 0xff;
}
}
@@ -541,35 +633,41 @@ static bool writePlane(ostream& fd, const Pel* src, bool is16bit,
return true;
}
-static bool writeField(ostream& fd, const Pel* top, const Pel* bottom, bool is16bit,
+static bool writeField(ostream& fd, const Pel* top, const Pel* bottom,
+ const bool is16bit,
const uint32_t stride_src,
uint32_t width444, uint32_t height444,
const ComponentID compID,
const ChromaFormat srcFormat,
const ChromaFormat fileFormat,
- const uint32_t fileBitDepth, const bool isTff)
+ const uint32_t fileBitDepth, const bool isTff,
+ const uint32_t packedYUVOutputMode = 0)
{
const uint32_t csx_file =getComponentScaleX(compID, fileFormat);
const uint32_t csy_file =getComponentScaleY(compID, fileFormat);
const uint32_t csx_src =getComponentScaleX(compID, srcFormat);
const uint32_t csy_src =getComponentScaleY(compID, srcFormat);
- /*const uint32_t stride_src = stride444>>csx_src;*/
-
- const uint32_t stride_file = (width444 * (is16bit ? 2 : 1)) >> csx_file;
- const uint32_t width_file = width444 >>csx_file;
- const uint32_t height_file = height444>>csy_file;
+ const uint32_t width_file = width444 >> csx_file;
+ const uint32_t height_file = height444 >> csy_file;
+ const bool writePYUV = (packedYUVOutputMode > 0) && (fileBitDepth == 10 || fileBitDepth == 12) && ((width_file & (1 + (fileBitDepth & 3))) == 0);
+ const uint32_t stride_file = writePYUV ? (width444 * fileBitDepth) >> (csx_file + 3) : (width444 * (is16bit ? 2 : 1)) >> csx_file;
std::vector<uint8_t> bufVec(stride_file * 2);
uint8_t *buf=&(bufVec[0]);
+ if (writePYUV)
+ {
+ // TODO
+ }
+ else // !writePYUV
if (compID!=COMPONENT_Y && (fileFormat==CHROMA_400 || srcFormat==CHROMA_400))
{
if (fileFormat!=CHROMA_400)
{
- const uint32_t value=1<<(fileBitDepth-1);
+ const uint32_t value = 1 << (fileBitDepth - 1);
- for(uint32_t y=0; y< height_file; y++)
+ for (uint32_t y = 0; y < height_file; y++)
{
for (uint32_t field = 0; field < 2; field++)
{
@@ -588,7 +686,7 @@ static bool writeField(ostream& fd, const Pel* top, const Pel* bottom, bool is16
uint16_t val(value);
for (uint32_t x = 0; x < width_file; x++)
{
- fieldBuffer[2*x+0]= (val>>0) & 0xff;
+ fieldBuffer[2*x ]= (val>>0) & 0xff;
fieldBuffer[2*x+1]= (val>>8) & 0xff;
}
}
@@ -753,7 +851,7 @@ bool VideoIOYuv::read ( PelUnitBuf& pic, PelUnitBuf& picOrg, const InputColourSp
scalePlane( picOrg.get(compID), m_bitdepthShift[chType], minval, maxval);
}
}
-
+
#if EXTENSION_360_VIDEO
if (pic.chromaFormat != NUM_CHROMA_FORMAT)
ColourSpaceConvert(picOrg, pic, ipcsc, true);
@@ -777,7 +875,9 @@ bool VideoIOYuv::read ( PelUnitBuf& pic, PelUnitBuf& picOrg, const InputColourSp
* @return true for success, false in case of error
*/
bool VideoIOYuv::write( const CPelUnitBuf& pic,
- const InputColourSpaceConversion ipCSC, int confLeft, int confRight, int confTop, int confBottom, ChromaFormat format, const bool bClipToRec709 )
+ const InputColourSpaceConversion ipCSC,
+ const bool bPackedYUVOutputMode,
+ int confLeft, int confRight, int confTop, int confBottom, ChromaFormat format, const bool bClipToRec709 )
{
PelStorage interm;
@@ -844,12 +944,13 @@ bool VideoIOYuv::write( const CPelUnitBuf& pic,
{
const ComponentID compID = ComponentID(comp);
const ChannelType ch = toChannelType(compID);
- const uint32_t csx = ::getComponentScaleX(compID, format);
- const uint32_t csy = ::getComponentScaleY(compID, format);
+ const uint32_t csx = ::getComponentScaleX(compID, format);
+ const uint32_t csy = ::getComponentScaleY(compID, format);
const CPelBuf area = picO.get(compID);
const int planeOffset = (confLeft >> csx) + (confTop >> csy) * area.stride;
if (!writePlane (m_cHandle, area.bufAt (0, 0) + planeOffset, is16bit, area.stride,
- width444, height444, compID, picO.chromaFormat, format, m_fileBitdepth[ch]))
+ width444, height444, compID, picO.chromaFormat, format, m_fileBitdepth[ch],
+ bPackedYUVOutputMode ? 1 : 0))
{
retval = false;
}
@@ -858,7 +959,10 @@ bool VideoIOYuv::write( const CPelUnitBuf& pic,
return retval;
}
-bool VideoIOYuv::write( const CPelUnitBuf& picTop, const CPelUnitBuf& picBottom, const InputColourSpaceConversion ipCSC, int confLeft, int confRight, int confTop, int confBottom, ChromaFormat format, const bool isTff, const bool bClipToRec709 )
+bool VideoIOYuv::write( const CPelUnitBuf& picTop, const CPelUnitBuf& picBottom,
+ const InputColourSpaceConversion ipCSC,
+ const bool bPackedYUVOutputMode,
+ int confLeft, int confRight, int confTop, int confBottom, ChromaFormat format, const bool isTff, const bool bClipToRec709 )
{
PelStorage intermTop;
PelStorage intermBottom;
@@ -934,8 +1038,8 @@ bool VideoIOYuv::write( const CPelUnitBuf& picTop, const CPelUnitBuf& picBottom,
const CPelBuf areaTop = picTopO. get( compID );
const CPelBuf areaBottom = picBottomO.get( compID );
const CPelBuf areaTopY = picTopO.Y();
- const uint32_t width444 = areaTopY.width - (confLeft + confRight);
- const uint32_t height444 = areaTopY.height - (confTop + confBottom);
+ const uint32_t width444 = areaTopY.width - (confLeft + confRight);
+ const uint32_t height444 = areaTopY.height - (confTop + confBottom);
CHECK(areaTop.width == areaBottom.width , "Incompatible formats");
CHECK(areaTop.height == areaBottom.height, "Incompatible formats");
@@ -950,12 +1054,13 @@ bool VideoIOYuv::write( const CPelUnitBuf& picTop, const CPelUnitBuf& picBottom,
const uint32_t csy = ::getComponentScaleY(compID, dstChrFormat );
const int planeOffset = (confLeft>>csx) + ( confTop>>csy) * areaTop.stride; //offset is for entire frame - round up for top field and down for bottom field
- if (! writeField(m_cHandle,
+ if (!writeField (m_cHandle,
(areaTop. bufAt(0,0) + planeOffset),
(areaBottom.bufAt(0,0) + planeOffset),
is16bit,
areaTop.stride,
- width444, height444, compID, dstChrFormat, format, m_fileBitdepth[ch], isTff))
+ width444, height444, compID, dstChrFormat, format, m_fileBitdepth[ch], isTff,
+ bPackedYUVOutputMode ? 1 : 0))
{
retval=false;
}
diff --git a/source/Lib/Utilities/VideoIOYuv.h b/source/Lib/Utilities/VideoIOYuv.h
index 044d274c9ea2d02d26df263a4e2d8dca80c620c9..bf72925bde43753d9de8028bdd1b6586f4003e66 100644
--- a/source/Lib/Utilities/VideoIOYuv.h
+++ b/source/Lib/Utilities/VideoIOYuv.h
@@ -78,10 +78,16 @@ public:
// If fileFormat=NUM_CHROMA_FORMAT, use the format defined by pPicYuv
bool write( const CPelUnitBuf& pic,
- const InputColourSpaceConversion ipCSC, int confLeft = 0, int confRight = 0, int confTop = 0, int confBottom = 0, ChromaFormat format = NUM_CHROMA_FORMAT, const bool bClipToRec709 = false ); ///< write one YUV frame with padding parameter
+ const InputColourSpaceConversion ipCSC,
+ const bool bPackedYUVOutputMode,
+ int confLeft = 0, int confRight = 0, int confTop = 0, int confBottom = 0, ChromaFormat format = NUM_CHROMA_FORMAT, const bool bClipToRec709 = false ); ///< write one YUV frame with padding parameter
// If fileFormat=NUM_CHROMA_FORMAT, use the format defined by pPicYuvTop and pPicYuvBottom
- bool write( const CPelUnitBuf& picTop, const CPelUnitBuf& picBot, const InputColourSpaceConversion ipCSC, int confLeft=0, int confRight=0, int confTop=0, int confBottom=0, ChromaFormat fileFormat=NUM_CHROMA_FORMAT, const bool isTff=false, const bool bClipToRec709=false);
+ bool write( const CPelUnitBuf& picTop, const CPelUnitBuf& picBot,
+ const InputColourSpaceConversion ipCSC,
+ const bool bPackedYUVOutputMode,
+ int confLeft = 0, int confRight = 0, int confTop = 0, int confBottom = 0, ChromaFormat format = NUM_CHROMA_FORMAT, const bool isTff = false, const bool bClipToRec709 = false );
+
static void ColourSpaceConvert(const CPelUnitBuf &src, PelUnitBuf &dest, const InputColourSpaceConversion conversion, bool bIsForwards);
bool isEof (); ///< check for end-of-file