diff --git a/cfg/encoder_intra_vtm.cfg b/cfg/encoder_intra_vtm.cfg
index 5de7eca29a7a0e22936e16c0e7123cbacbd2830c..814bf20d9368cd7c9a6675e9840f2ec1ae4c2e13 100644
--- a/cfg/encoder_intra_vtm.cfg
+++ b/cfg/encoder_intra_vtm.cfg
@@ -48,9 +48,9 @@ InternalBitDepth : 10 # codec operating bit-depth
#=========== Coding Tools =================
SAO : 1 # Sample adaptive offset (0: OFF, 1: ON)
-AMP : 1 # Asymmetric motion partitions (0: OFF, 1: ON)
TransformSkip : 1 # Transform skipping (0: OFF, 1: ON)
TransformSkipFast : 1 # Fast Transform skipping (0: OFF, 1: ON)
+TransformSkipLog2MaxSize : 5
SAOLcuBoundary : 0 # SAOLcuBoundary using non-deblocked pixels (0: OFF, 1: ON)
#============ Slices ================
@@ -77,11 +77,11 @@ PCMFilterDisableFlag : 0 # 0: Enable loop filterin
TransquantBypassEnable : 0 # Value of PPS flag.
CUTransquantBypassFlagForce: 0 # Force transquant bypass mode, when transquant_bypass_enable_flag is enabled
-#============ JEM settings ======================
+#============ VTM settings ======================
LoopFilterTcOffset_div2 : 0
SEIDecodedPictureHash : 0
-CbQpOffset : 0
-CrQpOffset : 0
+CbQpOffset : 1
+CrQpOffset : 1
TemporalSubsampleRatio : 8
#============ NEXT ====================
@@ -101,8 +101,9 @@ MaxBTDepthISliceL : 3
MaxBTDepthISliceC : 3
MTT : 1
-EMT : 1
-EMTFast : 1
+MTS : 1
+MTSIntraMaxCand : 3
+MTSInterMaxCand : 4
Affine : 1
SubPuMvp : 1
MaxNumMergeCand : 6
@@ -110,7 +111,7 @@ LMChroma : 1 # use CCLM only
DepQuant : 1
IMV : 2
ALF : 1
-CPR : 0 # turned off in CTC
+IBC : 0 # turned off in CTC
# Fast tools
PBIntraFast : 1
diff --git a/cfg/encoder_lowdelay_P_vtm.cfg b/cfg/encoder_lowdelay_P_vtm.cfg
index 16ee4e63f0bb3871d8c170bcda6e269cdb2b6270..132d9db5c55967284a1d2ee5497da6c44ee39ed5 100644
--- a/cfg/encoder_lowdelay_P_vtm.cfg
+++ b/cfg/encoder_lowdelay_P_vtm.cfg
@@ -56,9 +56,9 @@ InternalBitDepth : 10 # codec operating bit-depth
#=========== Coding Tools =================
SAO : 1 # Sample adaptive offset (0: OFF, 1: ON)
-AMP : 1 # Asymmetric motion partitions (0: OFF, 1: ON)
TransformSkip : 1 # Transform skipping (0: OFF, 1: ON)
TransformSkipFast : 1 # Fast Transform skipping (0: OFF, 1: ON)
+TransformSkipLog2MaxSize : 5
SAOLcuBoundary : 0 # SAOLcuBoundary using non-deblocked pixels (0: OFF, 1: ON)
#============ Slices ================
@@ -94,11 +94,11 @@ RCLCUSeparateModel : 1 # Rate control: use LCU l
InitialQP : 0 # Rate control: initial QP
RCForceIntraQP : 0 # Rate control: force intra QP to be equal to initial QP
-#============ JEM settings ======================
+#============ VTM settings ======================
LoopFilterTcOffset_div2 : 0
SEIDecodedPictureHash : 0
-CbQpOffset : 0
-CrQpOffset : 0
+CbQpOffset : 1
+CrQpOffset : 1
#============ NEXT ====================
@@ -117,8 +117,9 @@ MaxBTDepthISliceL : 3
MaxBTDepthISliceC : 3
MTT : 1
-EMT : 1
-EMTFast : 1
+MTS : 1
+MTSIntraMaxCand : 3
+MTSInterMaxCand : 4
Affine : 1
SubPuMvp : 1
MaxNumMergeCand : 6
@@ -127,7 +128,7 @@ DepQuant : 1
IMV : 2
ALF : 1
MHIntra : 1
-CPR : 0 # turned off in CTC
+IBC : 0 # turned off in CTC
# Fast tools
PBIntraFast : 1
diff --git a/cfg/encoder_lowdelay_vtm.cfg b/cfg/encoder_lowdelay_vtm.cfg
index 935e259733e48722779be43f555e7d16b5584c91..7d8350f9257c2d87be841e0083792ada75bc11b3 100644
--- a/cfg/encoder_lowdelay_vtm.cfg
+++ b/cfg/encoder_lowdelay_vtm.cfg
@@ -56,9 +56,9 @@ InternalBitDepth : 10 # codec operating bit-depth
#=========== Coding Tools =================
SAO : 1 # Sample adaptive offset (0: OFF, 1: ON)
-AMP : 1 # Asymmetric motion partitions (0: OFF, 1: ON)
TransformSkip : 1 # Transform skipping (0: OFF, 1: ON)
TransformSkipFast : 1 # Fast Transform skipping (0: OFF, 1: ON)
+TransformSkipLog2MaxSize : 5
SAOLcuBoundary : 0 # SAOLcuBoundary using non-deblocked pixels (0: OFF, 1: ON)
#============ Slices ================
@@ -94,11 +94,11 @@ RCLCUSeparateModel : 1 # Rate control: use LCU l
InitialQP : 0 # Rate control: initial QP
RCForceIntraQP : 0 # Rate control: force intra QP to be equal to initial QP
-#============ JEM settings ======================
+#============ VTM settings ======================
LoopFilterTcOffset_div2 : 0
SEIDecodedPictureHash : 0
-CbQpOffset : 0
-CrQpOffset : 0
+CbQpOffset : 1
+CrQpOffset : 1
#============ NEXT ====================
@@ -117,8 +117,9 @@ MaxBTDepthISliceL : 3
MaxBTDepthISliceC : 3
MTT : 1
-EMT : 1
-EMTFast : 1
+MTS : 1
+MTSIntraMaxCand : 3
+MTSInterMaxCand : 4
Affine : 1
SubPuMvp : 1
MaxNumMergeCand : 6
@@ -130,7 +131,7 @@ GBi : 1
GBiFast : 1
MHIntra : 1
Triangle : 1
-CPR : 0 # turned off in CTC
+IBC : 0 # turned off in CTC
# Fast tools
PBIntraFast : 1
diff --git a/cfg/encoder_randomaccess_vtm.cfg b/cfg/encoder_randomaccess_vtm.cfg
index 2e938d715f9fb5c2b2b7b93fa69b3d3732944a48..f3404e3f2b958461236f98e56e0387d315808658 100644
--- a/cfg/encoder_randomaccess_vtm.cfg
+++ b/cfg/encoder_randomaccess_vtm.cfg
@@ -70,9 +70,9 @@ InternalBitDepth : 10 # codec operating bit-depth
#=========== Coding Tools =================
SAO : 1 # Sample adaptive offset (0: OFF, 1: ON)
-AMP : 1 # Asymmetric motion partitions (0: OFF, 1: ON)
TransformSkip : 1 # Transform skipping (0: OFF, 1: ON)
TransformSkipFast : 1 # Fast Transform skipping (0: OFF, 1: ON)
+TransformSkipLog2MaxSize : 5
SAOLcuBoundary : 0 # SAOLcuBoundary using non-deblocked pixels (0: OFF, 1: ON)
#============ Slices ================
@@ -108,11 +108,11 @@ RCLCUSeparateModel : 1 # Rate control: use LCU l
InitialQP : 0 # Rate control: initial QP
RCForceIntraQP : 0 # Rate control: force intra QP to be equal to initial QP
-#============ JEM settings ======================
+#============ VTM settings ======================
LoopFilterTcOffset_div2 : 0
SEIDecodedPictureHash : 0
-CbQpOffset : 0
-CrQpOffset : 0
+CbQpOffset : 1
+CrQpOffset : 1
#============ NEXT ====================
@@ -131,8 +131,9 @@ MaxBTDepthISliceL : 3
MaxBTDepthISliceC : 3
MTT : 1
-EMT : 1
-EMTFast : 1
+MTS : 1
+MTSIntraMaxCand : 3
+MTSInterMaxCand : 4
Affine : 1
SubPuMvp : 1
MaxNumMergeCand : 6
@@ -145,7 +146,7 @@ GBiFast : 1
BIO : 1
MHIntra : 1
Triangle : 1
-CPR : 0 # turned off in CTC
+IBC : 0 # turned off in CTC
# Fast tools
PBIntraFast : 1
diff --git a/source/App/EncoderApp/EncApp.cpp b/source/App/EncoderApp/EncApp.cpp
index 4e47a75dfb014ebec6739b5edf17f62e7ff91491..6afd8db7333f1376da1346246b9cd70f6ae5befa 100644
--- a/source/App/EncoderApp/EncApp.cpp
+++ b/source/App/EncoderApp/EncApp.cpp
@@ -141,7 +141,6 @@ void EncApp::xInitLibCfg()
m_cEncLib.setAccessUnitDelimiter ( m_AccessUnitDelimiter );
m_cEncLib.setMaxTempLayer ( m_maxTempLayer );
- m_cEncLib.setUseAMP( m_enableAMP );
//===== Slice ========
@@ -224,13 +223,23 @@ void EncApp::xInitLibCfg()
m_cEncLib.setDisableMotionCompression ( m_DisableMotionCompression );
m_cEncLib.setMTTMode ( m_MTT );
m_cEncLib.setUseLMChroma ( m_LMChroma );
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ m_cEncLib.setCclmCollocatedChromaFlag ( m_cclmCollocatedChromaFlag );
+#endif
#if ENABLE_WPP_PARALLELISM
m_cEncLib.setUseAltDQPCoding ( m_AltDQPCoding );
#endif
+#if JVET_M0464_UNI_MTS
+ m_cEncLib.setIntraMTS ( m_MTS & 1 );
+ m_cEncLib.setIntraMTSMaxCand ( m_MTSIntraMaxCand );
+ m_cEncLib.setInterMTS ( ( m_MTS >> 1 ) & 1 );
+ m_cEncLib.setInterMTSMaxCand ( m_MTSInterMaxCand );
+#else
m_cEncLib.setIntraEMT ( m_EMT & 1 );
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
m_cEncLib.setUseCompositeRef ( m_compositeRefEnabled );
m_cEncLib.setUseGBi ( m_GBi );
m_cEncLib.setUseGBiFast ( m_GBiFast );
@@ -249,13 +258,13 @@ void EncApp::xInitLibCfg()
m_cEncLib.setUseMHIntra ( m_MHIntra );
m_cEncLib.setUseTriangle ( m_Triangle );
- m_cEncLib.setCPRMode ( m_CPRMode );
- m_cEncLib.setCPRLocalSearchRangeX ( m_CPRLocalSearchRangeX );
- m_cEncLib.setCPRLocalSearchRangeY ( m_CPRLocalSearchRangeY );
- m_cEncLib.setCPRHashSearch ( m_CPRHashSearch );
- m_cEncLib.setCPRHashSearchMaxCand ( m_CPRHashSearchMaxCand );
- m_cEncLib.setCPRHashSearchRange4SmallBlk ( m_CPRHashSearchRange4SmallBlk );
- m_cEncLib.setCPRFastMethod ( m_CPRFastMethod );
+ m_cEncLib.setIBCMode ( m_IBCMode );
+ m_cEncLib.setIBCLocalSearchRangeX ( m_IBCLocalSearchRangeX );
+ m_cEncLib.setIBCLocalSearchRangeY ( m_IBCLocalSearchRangeY );
+ m_cEncLib.setIBCHashSearch ( m_IBCHashSearch );
+ m_cEncLib.setIBCHashSearchMaxCand ( m_IBCHashSearchMaxCand );
+ m_cEncLib.setIBCHashSearchRange4SmallBlk ( m_IBCHashSearchRange4SmallBlk );
+ m_cEncLib.setIBCFastMethod ( m_IBCFastMethod );
m_cEncLib.setUseWrapAround ( m_wrapAround );
m_cEncLib.setWrapAroundOffset ( m_wrapAroundOffset );
diff --git a/source/App/EncoderApp/EncAppCfg.cpp b/source/App/EncoderApp/EncAppCfg.cpp
index d5fff470e0c3445ffd2155c40bb63ef4b479b522..10241f5caae994d3e4c68d692541232a9ccedee5 100644
--- a/source/App/EncoderApp/EncAppCfg.cpp
+++ b/source/App/EncoderApp/EncAppCfg.cpp
@@ -828,6 +828,20 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
("LMChroma", m_LMChroma, 1, " LMChroma prediction "
"\t0: Disable LMChroma\n"
"\t1: Enable LMChroma\n")
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ ("CclmCollocatedChroma", m_cclmCollocatedChromaFlag, false, "Specifies the location of the top-left downsampled luma sample in cross-component linear model intra prediction relative to the top-left luma sample\n"
+ "\t0: horizontally co-sited, vertically shifted by 0.5 units of luma samples\n"
+ "\t1: collocated\n")
+#endif
+#if JVET_M0464_UNI_MTS
+ ("MTS", m_MTS, 0, "Multiple Transform Set (MTS)\n"
+ "\t0: Disable MTS\n"
+ "\t1: Enable only Intra MTS\n"
+ "\t2: Enable only Inter MTS\n"
+ "\t3: Enable both Intra & Inter MTS\n")
+ ("MTSIntraMaxCand", m_MTSIntraMaxCand, 3, "Number of additional candidates to test in encoder search for MTS in intra slices\n")
+ ("MTSInterMaxCand", m_MTSInterMaxCand, 4, "Number of additional candidates to test in encoder search for MTS in inter slices\n")
+#else
("EMT,-emt", m_EMT, 0, "Enhanced Multiple Transform (EMT)\n"
"\t0: Disable EMT\n"
"\t1: Enable only Intra EMT\n"
@@ -838,6 +852,7 @@ 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
("CompositeLTReference", m_compositeRefEnabled, false, "Enable Composite Long Term Reference Frame")
("GBi", m_GBi, false, "Enable Generalized Bi-prediction(GBi)")
("GBiFast", m_GBiFast, false, "Fast methods for Generalized Bi-prediction(GBi)\n")
@@ -850,13 +865,13 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
("MHIntra", m_MHIntra, false, "Enable MHIntra mode")
("Triangle", m_Triangle, false, "Enable triangular shape motion vector prediction (0:off, 1:on)")
- ( "CPR", m_CPRMode, 0u, "CPRMode (0x1:enabled, 0x0:disabled) [default: disabled]")
- ( "CPRLocalSearchRangeX", m_CPRLocalSearchRangeX, 128u, "Search range of CPR local search in x direction")
- ( "CPRLocalSearchRangeY", m_CPRLocalSearchRangeY, 128u, "Search range of CPR local search in y direction")
- ( "CPRHashSearch", m_CPRHashSearch, 1u, "Hash based CPR search")
- ( "CPRHashSearchMaxCand", m_CPRHashSearchMaxCand, 256u, "Max candidates for hash based CPR search")
- ( "CPRHashSearchRange4SmallBlk", m_CPRHashSearchRange4SmallBlk, 256u, "Small block search range in based CPR search")
- ( "CPRFastMethod", m_CPRFastMethod, 6u, "Fast methods for CPR")
+ ( "IBC", m_IBCMode, 0u, "IBCMode (0x1:enabled, 0x0:disabled) [default: disabled]")
+ ( "IBCLocalSearchRangeX", m_IBCLocalSearchRangeX, 128u, "Search range of IBC local search in x direction")
+ ( "IBCLocalSearchRangeY", m_IBCLocalSearchRangeY, 128u, "Search range of IBC local search in y direction")
+ ( "IBCHashSearch", m_IBCHashSearch, 1u, "Hash based IBC search")
+ ( "IBCHashSearchMaxCand", m_IBCHashSearchMaxCand, 256u, "Max candidates for hash based IBC search")
+ ( "IBCHashSearchRange4SmallBlk", m_IBCHashSearchRange4SmallBlk, 256u, "Small block search range in based IBC search")
+ ( "IBCFastMethod", m_IBCFastMethod, 6u, "Fast methods for IBC")
("WrapAround", m_wrapAround, false, "Enable horizontal wrap-around motion compensation for inter prediction (0:off, 1:on) [default: off]")
("WrapAroundOffset", m_wrapAroundOffset, 0u, "Offset in luma samples used for computing the horizontal wrap-around position")
@@ -943,8 +958,8 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
("CbQpOffset,-cbqpofs", m_cbQpOffset, 0, "Chroma Cb QP Offset")
("CrQpOffset,-crqpofs", m_crQpOffset, 0, "Chroma Cr QP Offset")
- ("CbQpOffsetDualTree", m_cbQpOffsetDualTree, 1, "Chroma Cb QP Offset for dual tree")
- ("CrQpOffsetDualTree", m_crQpOffsetDualTree, 1, "Chroma Cr QP Offset for dual tree")
+ ("CbQpOffsetDualTree", m_cbQpOffsetDualTree, 0, "Chroma Cb QP Offset for dual tree")
+ ("CrQpOffsetDualTree", m_crQpOffsetDualTree, 0, "Chroma Cr QP Offset for dual tree")
#if ER_CHROMA_QP_WCG_PPS
("WCGPPSEnable", m_wcgChromaQpControl.enabled, false, "1: Enable the WCG PPS chroma modulation scheme. 0 (default) disabled")
("WCGPPSCbQpScale", m_wcgChromaQpControl.chromaCbQpScale, 1.0, "WCG PPS Chroma Cb QP Scale")
@@ -983,14 +998,18 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
("DeblockingFilterMetric", m_DeblockingFilterMetric, false)
#endif
// Coding tools
- ("AMP", m_enableAMP, true, "Enable asymmetric motion partitions")
("CrossComponentPrediction", m_crossComponentPredictionEnabledFlag, false, "Enable the use of cross-component prediction (not valid in V1 profiles)")
("ReconBasedCrossCPredictionEstimate", m_reconBasedCrossCPredictionEstimate, false, "When determining the alpha value for cross-component prediction, use the decoded residual rather than the pre-transform encoder-side residual")
("SaoLumaOffsetBitShift", saoOffsetBitShift[CHANNEL_TYPE_LUMA], 0, "Specify the luma SAO bit-shift. If negative, automatically calculate a suitable value based upon bit depth and initial QP")
("SaoChromaOffsetBitShift", saoOffsetBitShift[CHANNEL_TYPE_CHROMA], 0, "Specify the chroma SAO bit-shift. If negative, automatically calculate a suitable value based upon bit depth and initial QP")
("TransformSkip", m_useTransformSkip, false, "Intra transform skipping")
+#if JVET_M0464_UNI_MTS
+ ("TransformSkipFast", m_useTransformSkipFast, false, "Fast encoder search for transform skipping, winner takes it all mode.")
+ ("TransformSkipLog2MaxSize", m_log2MaxTransformSkipBlockSize, 5U, "Specify transform-skip maximum size. Minimum 2, Maximum 5. (not valid in V1 profiles)")
+#else
("TransformSkipFast", m_useTransformSkipFast, false, "Fast intra transform skipping")
("TransformSkipLog2MaxSize", m_log2MaxTransformSkipBlockSize, 2U, "Specify transform-skip maximum size. Minimum 2. (not valid in V1 profiles)")
+#endif
("ImplicitResidualDPCM", m_rdpcmEnabledFlag[RDPCM_SIGNAL_IMPLICIT], false, "Enable implicitly signalled residual DPCM for intra (also known as sample-adaptive intra predict) (not valid in V1 profiles)")
("ExplicitResidualDPCM", m_rdpcmEnabledFlag[RDPCM_SIGNAL_EXPLICIT], false, "Enable explicitly signalled residual DPCM for inter (not valid in V1 profiles)")
("ResidualRotation", m_transformSkipRotationEnabledFlag, false, "Enable rotation of transform-skipped and transquant-bypassed TUs through 180 degrees prior to entropy coding (not valid in V1 profiles)")
@@ -1922,15 +1941,19 @@ bool EncAppCfg::xCheckParameter()
#endif
xConfirmPara( m_LMChroma, "LMChroma only allowed with NEXT profile" );
xConfirmPara( m_LargeCTU, "Large CTU is only allowed with NEXT profile" );
- xConfirmPara( m_SubPuMvpMode != 0, "Sub-PU motion vector prediction is only allowed with NEXT profile" );
- xConfirmPara( m_BIO, "BIO only allowed with NEXT profile" );
xConfirmPara( m_DisableMotionCompression, "Disable motion data compression only allowed with NEXT profile" );
xConfirmPara( m_MTT, "Multi type tree is only allowed with NEXT profile" );
xConfirmPara( m_ImvMode, "IMV is only allowed with NEXT profile" );
- xConfirmPara(m_CPRMode, "CPR Mode only allowed with NEXT profile");
+ xConfirmPara(m_IBCMode, "IBC Mode only allowed with NEXT profile");
xConfirmPara( m_useFastLCTU, "Fast large CTU can only be applied when encoding with NEXT profile" );
+#if JVET_M0464_UNI_MTS
+ xConfirmPara( m_MTS, "MTS only allowed with NEXT profile" );
+ xConfirmPara( m_MTSIntraMaxCand, "MTS only allowed with NEXT profile" );
+ xConfirmPara( m_MTSInterMaxCand, "MTS only allowed with NEXT profile" );
+#else
xConfirmPara( m_EMT, "EMT only allowed with NEXT profile" );
xConfirmPara( m_FastEMT, "EMT only allowed with NEXT profile" );
+#endif
xConfirmPara( m_compositeRefEnabled, "Composite Reference Frame is only allowed with NEXT profile" );
xConfirmPara( m_GBi, "GBi is only allowed with NEXT profile" );
xConfirmPara( m_GBiFast, "GBiFast is only allowed with NEXT profile" );
@@ -2059,7 +2082,11 @@ bool EncAppCfg::xCheckParameter()
xConfirmPara(m_profile == Profile::MAINSTILLPICTURE && m_framesToBeEncoded > 1, "Number of frames to be encoded must be 1 when main still picture profile is used.");
xConfirmPara(m_crossComponentPredictionEnabledFlag==true, "CrossComponentPrediction must not be used for non main-RExt profiles.");
+#if JVET_M0464_UNI_MTS
+ xConfirmPara(m_log2MaxTransformSkipBlockSize>=6, "Transform Skip Log2 Max Size must be less or equal to 5.");
+#else
xConfirmPara(m_log2MaxTransformSkipBlockSize!=2, "Transform Skip Log2 Max Size must be 2 for V1 profiles.");
+#endif
xConfirmPara(m_transformSkipRotationEnabledFlag==true, "UseResidualRotation must not be enabled for non main-RExt profiles.");
xConfirmPara(m_transformSkipContextEnabledFlag==true, "UseSingleSignificanceMapContext must not be enabled for non main-RExt profiles.");
xConfirmPara(m_rdpcmEnabledFlag[RDPCM_SIGNAL_IMPLICIT]==true, "ImplicitResidualDPCM must not be enabled for non main-RExt profiles.");
@@ -2148,6 +2175,7 @@ bool EncAppCfg::xCheckParameter()
xConfirmPara (m_log2MaxTransformSkipBlockSize < 2, "Transform Skip Log2 Max Size must be at least 2 (4x4)");
+#if !JVET_M0464_UNI_MTS
if (m_log2MaxTransformSkipBlockSize!=2 && m_useTransformSkipFast)
{
msg( WARNING, "***************************************************************************\n");
@@ -2156,6 +2184,7 @@ bool EncAppCfg::xCheckParameter()
msg( WARNING, "** It may be better to disable transform skip fast mode **\n");
msg( WARNING, "***************************************************************************\n");
}
+#endif
xConfirmPara( m_quadtreeTULog2MaxSize * m_tuLog2MaxSize >= 0, "Setting of TULog2MaxSize and QuadtreeTULog2MaxSize is mutually exclusive - use only one of the parameters" );
@@ -2268,8 +2297,14 @@ bool EncAppCfg::xCheckParameter()
m_maxNumAffineMergeCand = m_SubPuMvpMode;
}
+#if JVET_M0464_UNI_MTS
+ xConfirmPara( m_MTS < 0 || m_MTS > 3, "MTS must be greater than 0 smaller than 4" );
+ xConfirmPara( m_MTSIntraMaxCand < 0 || m_MTSIntraMaxCand > 5, "m_MTSIntraMaxCand must be greater than 0 and smaller than 6" );
+ xConfirmPara( m_MTSInterMaxCand < 0 || m_MTSInterMaxCand > 5, "m_MTSInterMaxCand must be greater than 0 and smaller than 6" );
+#else
xConfirmPara( m_EMT < 0 || m_EMT >3, "EMT must be 0, 1, 2 or 3" );
xConfirmPara( m_FastEMT < 0 || m_FastEMT >3, "FEMT must be 0, 1, 2 or 3" );
+#endif
if( m_usePCM)
{
for (uint32_t channelType = 0; channelType < MAX_NUM_CHANNEL_TYPE; channelType++)
@@ -3101,8 +3136,18 @@ void EncAppCfg::xPrintParameter()
msg( VERBOSE, "AltDQPCoding:%d ", m_AltDQPCoding );
#endif
msg( VERBOSE, "LMChroma:%d ", m_LMChroma );
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ if( m_LMChroma && m_chromaFormatIDC == CHROMA_420 )
+ {
+ msg( VERBOSE, "CclmCollocatedChroma:%d ", m_cclmCollocatedChromaFlag );
+ }
+#endif
+#if JVET_M0464_UNI_MTS
+ msg( VERBOSE, "MTS: %1d(intra) %1d(inter) ", m_MTS & 1, ( m_MTS >> 1 ) & 1 );
+#else
msg( VERBOSE, "EMT: %1d(intra) %1d(inter) ", m_EMT & 1, ( m_EMT >> 1 ) & 1 );
- msg(VERBOSE, "CompositeLTReference:%d ", m_compositeRefEnabled);
+#endif
+ msg( VERBOSE, "CompositeLTReference:%d ", m_compositeRefEnabled);
msg( VERBOSE, "GBi:%d ", m_GBi );
msg( VERBOSE, "GBiFast:%d ", m_GBiFast );
#if LUMA_ADAPTIVE_DEBLOCKING_FILTER_QP_OFFSET
@@ -3111,7 +3156,7 @@ void EncAppCfg::xPrintParameter()
msg(VERBOSE, "MHIntra:%d ", m_MHIntra);
msg( VERBOSE, "Triangle:%d ", m_Triangle );
}
- msg(VERBOSE, "CPR:%d ", m_CPRMode);
+ msg(VERBOSE, "IBC:%d ", m_IBCMode);
msg( VERBOSE, "WrapAround:%d ", m_wrapAround);
if( m_wrapAround )
{
@@ -3127,7 +3172,11 @@ void EncAppCfg::xPrintParameter()
msg( VERBOSE, "FastMrg:%d ", m_useFastMrg );
msg( VERBOSE, "PBIntraFast:%d ", m_usePbIntraFast );
if( m_ImvMode == 2 ) msg( VERBOSE, "IMV4PelFast:%d ", m_Imv4PelFast );
+#if JVET_M0464_UNI_MTS
+ if( m_MTS ) msg( VERBOSE, "MTSMaxCand: %1d(intra) %1d(inter) ", m_MTSIntraMaxCand, m_MTSInterMaxCand );
+#else
if( m_EMT ) msg( VERBOSE, "EMTFast: %1d(intra) %1d(inter) ", ( m_FastEMT & m_EMT & 1 ), ( m_FastEMT >> 1 ) & ( m_EMT >> 1 ) & 1 );
+#endif
msg( VERBOSE, "AMaxBT:%d ", m_useAMaxBT );
msg( VERBOSE, "E0023FastEnc:%d ", m_e0023FastEnc );
msg( VERBOSE, "ContentBasedFastQtbt:%d ", m_contentBasedFastQtbt );
diff --git a/source/App/EncoderApp/EncAppCfg.h b/source/App/EncoderApp/EncAppCfg.h
index 94bedb0c472ea8232629567dcc2dd166985bae70..8b67d948decc874a2639cc402a27ee0d900cf67b 100644
--- a/source/App/EncoderApp/EncAppCfg.h
+++ b/source/App/EncoderApp/EncAppCfg.h
@@ -148,7 +148,6 @@ protected:
bool m_transformSkipRotationEnabledFlag; ///< control flag for transform-skip/transquant-bypass residual rotation
bool m_transformSkipContextEnabledFlag; ///< control flag for transform-skip/transquant-bypass single significance map context
bool m_rdpcmEnabledFlag[NUMBER_OF_RDPCM_SIGNALLING_MODES];///< control flags for residual DPCM
- bool m_enableAMP;
bool m_persistentRiceAdaptationEnabledFlag; ///< control flag for Golomb-Rice parameter adaptation over each slice
bool m_cabacBypassAlignmentEnabledFlag;
@@ -214,8 +213,17 @@ protected:
bool m_AltDQPCoding;
#endif
int m_LMChroma;
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ bool m_cclmCollocatedChromaFlag;
+#endif
+#if JVET_M0464_UNI_MTS
+ int m_MTS; ///< XZ: Multiple Transform Set
+ int m_MTSIntraMaxCand; ///< XZ: Number of additional candidates to test
+ int m_MTSInterMaxCand; ///< XZ: Number of additional candidates to test
+#else
int m_EMT; ///< XZ: Enhanced Multiple Transform
int m_FastEMT; ///< XZ: Fast Methods of Enhanced Multiple Transform
+#endif
bool m_compositeRefEnabled;
bool m_GBi;
@@ -231,13 +239,13 @@ protected:
bool m_Triangle;
- unsigned m_CPRMode;
- unsigned m_CPRLocalSearchRangeX;
- unsigned m_CPRLocalSearchRangeY;
- unsigned m_CPRHashSearch;
- unsigned m_CPRHashSearchMaxCand;
- unsigned m_CPRHashSearchRange4SmallBlk;
- unsigned m_CPRFastMethod;
+ unsigned m_IBCMode;
+ unsigned m_IBCLocalSearchRangeX;
+ unsigned m_IBCLocalSearchRangeY;
+ unsigned m_IBCHashSearch;
+ unsigned m_IBCHashSearchMaxCand;
+ unsigned m_IBCHashSearchRange4SmallBlk;
+ unsigned m_IBCFastMethod;
bool m_wrapAround;
unsigned m_wrapAroundOffset;
diff --git a/source/Lib/CommonLib/BinaryDecisionTree.cpp b/source/Lib/CommonLib/BinaryDecisionTree.cpp
index 2fa1f97cc03e0b766e1b0895b02f4550ac3c7ddb..db6492688bd70ea46f6b41b511cf7d3e4597f400 100644
--- a/source/Lib/CommonLib/BinaryDecisionTree.cpp
+++ b/source/Lib/CommonLib/BinaryDecisionTree.cpp
@@ -38,6 +38,8 @@
#include "BinaryDecisionTree.h"
#include "CommonDef.h"
+#if !REMOVE_BIN_DECISION_TREE
+
#include <algorithm>
struct DecisionTreeBuilder
@@ -183,3 +185,5 @@ void DecisionTree::reduce( unsigned offset /*= 0*/, int depth /*= -1 */ )
isAvail[offset] = avail;
}
+
+#endif
diff --git a/source/Lib/CommonLib/BinaryDecisionTree.h b/source/Lib/CommonLib/BinaryDecisionTree.h
index 271be4529e69cf35bf6e71f9d0bc1da5c5f39811..7ab0d6558cd9ece144e9545591a441480d742214 100644
--- a/source/Lib/CommonLib/BinaryDecisionTree.h
+++ b/source/Lib/CommonLib/BinaryDecisionTree.h
@@ -40,6 +40,8 @@
#include "CommonDef.h"
+#if !REMOVE_BIN_DECISION_TREE
+
#define MAX_DEPTH_DECISION_TREE 5
#define MAX_NODES_DECISION_TREE ( 2 * ( 1 << MAX_DEPTH_DECISION_TREE ) )
@@ -92,3 +94,5 @@ DecisionTreeBuilder* decision( unsigned id, DecisionTreeBuilder* sub0, DecisionT
DecisionTreeBuilder* decision( unsigned id, unsigned id0, DecisionTreeBuilder* sub1 );
#endif
+
+#endif
diff --git a/source/Lib/CommonLib/CodingStatistics.h b/source/Lib/CommonLib/CodingStatistics.h
index 7754f0a40d7240e120c459399fe19fec7d850b4e..748138fdf7a5339d69a21cab167088f17ac5e40c 100644
--- a/source/Lib/CommonLib/CodingStatistics.h
+++ b/source/Lib/CommonLib/CodingStatistics.h
@@ -108,6 +108,9 @@ enum CodingStatisticsType
STATS__TOOL_TOTAL_FRAME,// This is a special case and is not included in the report.
STATS__TOOL_AFF,
STATS__TOOL_EMT,
+#if JVET_M0444_SMVD
+ STATS__CABAC_BITS__SYMMVD_FLAG,
+#endif
STATS__TOOL_TOTAL,
STATS__NUM_STATS
};
@@ -184,6 +187,9 @@ static inline const char* getName(CodingStatisticsType name)
"CABAC_BITS__TRIANGLE_FLAG",
"CABAC_BITS__TRIANGLE_INDEX",
"CABAC_BITS__MULTI_REF_LINE",
+#if JVET_M0444_SMVD
+ "CABAC_BITS__SYMMVD_FLAG",
+#endif
"TOOL_FRAME",
"TOOL_AFFINE",
"TOOL_EMT",
diff --git a/source/Lib/CommonLib/CodingStructure.cpp b/source/Lib/CommonLib/CodingStructure.cpp
index 9f3226fc7d193fff292d3274b9469aaaface4146..8d5013fcc154824f7131c04f8dc76f23b3576441 100644
--- a/source/Lib/CommonLib/CodingStructure.cpp
+++ b/source/Lib/CommonLib/CodingStructure.cpp
@@ -1299,7 +1299,7 @@ const TransformUnit* CodingStructure::getTURestricted( const Position &pos, cons
}
}
-CprLumaCoverage CodingStructure::getCprLumaCoverage(const CompArea& chromaArea) const
+IbcLumaCoverage CodingStructure::getIbcLumaCoverage(const CompArea& chromaArea) const
{
CHECK(chType != CHANNEL_TYPE_CHROMA, "Error");
@@ -1307,7 +1307,7 @@ CprLumaCoverage CodingStructure::getCprLumaCoverage(const CompArea& chromaArea)
CompArea lumaArea = CompArea(COMPONENT_Y, chromaArea.chromaFormat, chromaArea.lumaPos(), recalcSize(chromaArea.chromaFormat, CHANNEL_TYPE_CHROMA, CHANNEL_TYPE_LUMA, chromaArea.size()));
lumaArea = clipArea(lumaArea, picture->block(COMPONENT_Y));
const unsigned int fullArea = lumaArea.area();
- unsigned int cprArea = 0;
+ unsigned int ibcArea = 0;
for (SizeType y = 0; y < lumaArea.height; y += MIN_PU_SIZE)
{
for (SizeType x = 0; x < lumaArea.width; x += MIN_PU_SIZE)
@@ -1315,19 +1315,19 @@ CprLumaCoverage CodingStructure::getCprLumaCoverage(const CompArea& chromaArea)
Position pos = lumaArea.offset(x, y);
if (picture->cs->getMotionInfo(pos).isInter) // need to change if inter slice allows dualtree
{
- cprArea += unitAreaSubBlock;
+ ibcArea += unitAreaSubBlock;
}
}
}
- CprLumaCoverage coverage = CPR_LUMA_COVERAGE_FULL;
- if (cprArea == 0)
+ IbcLumaCoverage coverage = IBC_LUMA_COVERAGE_FULL;
+ if (ibcArea == 0)
{
- coverage = CPR_LUMA_COVERAGE_NONE;
+ coverage = IBC_LUMA_COVERAGE_NONE;
}
- else if (cprArea < fullArea)
+ else if (ibcArea < fullArea)
{
- coverage = CPR_LUMA_COVERAGE_PARTIAL;
+ coverage = IBC_LUMA_COVERAGE_PARTIAL;
}
return coverage;
diff --git a/source/Lib/CommonLib/CodingStructure.h b/source/Lib/CommonLib/CodingStructure.h
index 451e1f13db3b12674aed680282b2cc4c1fe5a08b..3727cc1fd33e8af22c8d7a44362b79ba00a14524 100644
--- a/source/Lib/CommonLib/CodingStructure.h
+++ b/source/Lib/CommonLib/CodingStructure.h
@@ -58,12 +58,12 @@ enum PictureType
PIC_ORG_RESI,
NUM_PIC_TYPES
};
-enum CprLumaCoverage
+enum IbcLumaCoverage
{
- CPR_LUMA_COVERAGE_FULL = 0,
- CPR_LUMA_COVERAGE_PARTIAL,
- CPR_LUMA_COVERAGE_NONE,
- NUM_CPR_LUMA_COVERAGE,
+ IBC_LUMA_COVERAGE_FULL = 0,
+ IBC_LUMA_COVERAGE_PARTIAL,
+ IBC_LUMA_COVERAGE_NONE,
+ NUM_IBC_LUMA_COVERAGE,
};
extern XUCache g_globalUnitCache;
@@ -89,6 +89,10 @@ public:
int prevQP[MAX_NUM_CHANNEL_TYPE];
int currQP[MAX_NUM_CHANNEL_TYPE];
int chromaQpAdj;
+#if JVET_M0170_MRG_SHARELIST
+ Position sharedBndPos;
+ Size sharedBndSize;
+#endif
bool isLossless;
const SPS *sps;
const PPS *pps;
@@ -154,7 +158,7 @@ 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;
- CprLumaCoverage getCprLumaCoverage(const CompArea& chromaArea) const;
+ IbcLumaCoverage getIbcLumaCoverage(const CompArea& chromaArea) const;
// ---------------------------------------------------------------------------
// encoding search utilities
// ---------------------------------------------------------------------------
diff --git a/source/Lib/CommonLib/CommonDef.h b/source/Lib/CommonLib/CommonDef.h
index 986b0cd35278cd112addd6da1da2a08e8e8aa241..27d76437a551bb84a7d284e129c8d27423cb3d9c 100644
--- a/source/Lib/CommonLib/CommonDef.h
+++ b/source/Lib/CommonLib/CommonDef.h
@@ -211,8 +211,14 @@ static const int DM_CHROMA_IDX = NUM_INTRA_MODE; ///< chro
static const uint8_t INTER_MODE_IDX = 255; ///< index for inter modes
+#if JVET_M0464_UNI_MTS
+static const uint32_t NUM_TRAFO_MODES_MTS = 6; ///< Max Intra CU size applying EMT, supported values: 8, 16, 32, 64, 128
+static const uint32_t MTS_INTRA_MAX_CU_SIZE = 32; ///< Max Intra CU size applying EMT, supported values: 8, 16, 32, 64, 128
+static const uint32_t MTS_INTER_MAX_CU_SIZE = 32; ///< Max Inter CU size applying EMT, supported values: 8, 16, 32, 64, 128
+#else
static const uint32_t EMT_INTRA_MAX_CU_WITH_QTBT = 32; ///< Max Intra CU size applying EMT, supported values: 8, 16, 32, 64, 128
static const uint32_t EMT_INTER_MAX_CU_WITH_QTBT = 32; ///< Max Inter CU size applying EMT, supported values: 8, 16, 32, 64, 128
+#endif
static const int NUM_MOST_PROBABLE_MODES = 6;
static const int LM_SYMBOL_NUM = (1 + NUM_LMC_MODE);
@@ -321,7 +327,9 @@ static const int MAX_ENCODER_DEBLOCKING_QUALITY_LAYERS = 8 ;
static const uint32_t LUMA_LEVEL_TO_DQP_LUT_MAXSIZE = 1024; ///< max LUT size for QP offset based on luma
#endif
+#if !JVET_M0464_UNI_MTS
static const int NUM_EMT_CU_FLAG_CTX = 6; ///< number of context models for EMT CU-level flag
+#endif
//QTBT high level parameters
//for I slice luma CTB configuration para.
@@ -388,12 +396,12 @@ static const int TRIANGLE_MAX_NUM_CANDS = 40;
static const int TRIANGLE_MAX_NUM_SATD_CANDS = 3;
static const int TRIANGLE_MIN_SIZE = 8 * 8;
-static const int CPR_MAX_CAND_SIZE = 16; // max block size for cpr search
-static const int CPR_NUM_CANDIDATES = 64; ///< Maximum number of candidates to store/test
+static const int IBC_MAX_CAND_SIZE = 16; // max block size for ibc search
+static const int IBC_NUM_CANDIDATES = 64; ///< Maximum number of candidates to store/test
static const int CHROMA_REFINEMENT_CANDIDATES = 8; /// 8 candidates BV to choose from
-static const int CPR_FAST_METHOD_NOINTRA_CPRCBF0 = 0x01;
-static const int CPR_FAST_METHOD_BUFFERBV = 0X02;
-static const int CPR_FAST_METHOD_ADAPTIVE_SEARCHRANGE = 0X04;
+static const int IBC_FAST_METHOD_NOINTRA_IBCCBF0 = 0x01;
+static const int IBC_FAST_METHOD_BUFFERBV = 0X02;
+static const int IBC_FAST_METHOD_ADAPTIVE_SEARCHRANGE = 0X04;
// ====================================================================================================================
// Macro functions
@@ -536,6 +544,45 @@ template <typename ValueType> inline ValueType rightShift (const ValueType
template <typename ValueType> inline ValueType leftShift_round (const ValueType value, const int shift) { return (shift >= 0) ? ( value << shift) : ((value + (ValueType(1) << (-shift - 1))) >> -shift); }
template <typename ValueType> inline ValueType rightShift_round(const ValueType value, const int shift) { return (shift >= 0) ? ((value + (ValueType(1) << (shift - 1))) >> shift) : ( value << -shift); }
+static inline int floorLog2(uint32_t x)
+{
+ if (x == 0)
+ {
+ return -1;
+ }
+#ifdef __GNUC__
+ return 31 - __builtin_clz(x);
+#else
+ int result = 0;
+ if (x & 0xffff0000)
+ {
+ x >>= 16;
+ result += 16;
+ }
+ if (x & 0xff00)
+ {
+ x >>= 8;
+ result += 8;
+ }
+ if (x & 0xf0)
+ {
+ x >>= 4;
+ result += 4;
+ }
+ if (x & 0xc)
+ {
+ x >>= 2;
+ result += 2;
+ }
+ if (x & 0x2)
+ {
+ x >>= 1;
+ result += 1;
+ }
+ return result;
+#endif
+}
+
//CASE-BREAK for breakpoints
#if defined ( _MSC_VER ) && defined ( _DEBUG )
#define _CASE(_x) if(_x)
diff --git a/source/Lib/CommonLib/ContextModelling.cpp b/source/Lib/CommonLib/ContextModelling.cpp
index 57dd1636b913daa109bea8939a3d05175f594746..d52ea51e00383a3719d16ecc733dd3f968afda60 100644
--- a/source/Lib/CommonLib/ContextModelling.cpp
+++ b/source/Lib/CommonLib/ContextModelling.cpp
@@ -81,7 +81,11 @@ CoeffCodingContext::CoeffCodingContext(const TransformUnit& tu, ComponentID comp
, m_lastOffsetY (0)
, m_lastShiftX (0)
, m_lastShiftY (0)
+#if JVET_M0464_UNI_MTS
+ , m_TrafoBypass (tu.cs->sps->getSpsRangeExtension().getTransformSkipContextEnabledFlag() && (tu.cu->transQuantBypass || tu.mtsIdx==1))
+#else
, m_TrafoBypass (tu.cs->sps->getSpsRangeExtension().getTransformSkipContextEnabledFlag() && (tu.cu->transQuantBypass || tu.transformSkip[m_compID]))
+#endif
, m_scanPosLast (-1)
, m_subSetId (-1)
, m_subSetPos (-1)
@@ -96,7 +100,9 @@ CoeffCodingContext::CoeffCodingContext(const TransformUnit& tu, ComponentID comp
, m_parFlagCtxSet ( Ctx::ParFlag[m_chType] )
, m_gtxFlagCtxSet { Ctx::GtxFlag[m_chType], Ctx::GtxFlag[m_chType+2] }
, m_sigCoeffGroupFlag ()
+#if !JVET_M0464_UNI_MTS
, m_emtNumSigCoeff (0)
+#endif
{
// LOGTODO
unsigned log2sizeX = m_log2BlockWidth;
@@ -150,6 +156,116 @@ void CoeffCodingContext::initSubblock( int SubsetId, bool sigGroupFlag )
+
+#if JVET_M0421_SPLIT_SIG
+void DeriveCtx::CtxSplit( const CodingStructure& cs, Partitioner& partitioner, unsigned& ctxSpl, unsigned& ctxQt, unsigned& ctxHv, unsigned& ctxHorBt, unsigned& ctxVerBt, bool* _canSplit /*= nullptr */ )
+{
+ const Position pos = partitioner.currArea().blocks[partitioner.chType];
+ const unsigned curSliceIdx = cs.slice->getIndependentSliceIdx();
+#if HEVC_TILES_WPP
+ const unsigned curTileIdx = cs.picture->tileMap->getTileIdxMap( partitioner.currArea().lumaPos() );
+#endif
+
+ // get left depth
+#if HEVC_TILES_WPP
+ const CodingUnit* cuLeft = cs.getCURestricted( pos.offset( -1, 0 ), curSliceIdx, curTileIdx, partitioner.chType );
+#else
+ const CodingUnit* cuLeft = cs.getCURestricted( pos.offset( -1, 0 ), curSliceIdx, partitioner.chType );
+#endif
+
+ // get above depth
+#if HEVC_TILES_WPP
+ const CodingUnit* cuAbove = cs.getCURestricted( pos.offset( 0, -1 ), curSliceIdx, curTileIdx, partitioner.chType );
+#else
+ const CodingUnit* cuAbove = cs.getCURestricted( pos.offset( 0, -1 ), curSliceIdx, partitioner.chType );
+#endif
+
+ bool canSplit[6];
+
+ if( _canSplit == nullptr )
+ {
+ partitioner.canSplit( cs, canSplit[0], canSplit[1], canSplit[2], canSplit[3], canSplit[4], canSplit[5] );
+ }
+ else
+ {
+ memcpy( canSplit, _canSplit, 6 * sizeof( bool ) );
+ }
+
+ ///////////////////////
+ // CTX do split (0-8)
+ ///////////////////////
+ const unsigned widthCurr = partitioner.currArea().blocks[partitioner.chType].width;
+ const unsigned heightCurr = partitioner.currArea().blocks[partitioner.chType].height;
+
+ ctxSpl = 0;
+
+ if( cuLeft )
+ {
+ const unsigned heightLeft = cuLeft->blocks[partitioner.chType].height;
+ ctxSpl += ( heightLeft < heightCurr ? 1 : 0 );
+ }
+ if( cuAbove )
+ {
+ const unsigned widthAbove = cuAbove->blocks[partitioner.chType].width;
+ ctxSpl += ( widthAbove < widthCurr ? 1 : 0 );
+ }
+
+ unsigned numSplit = 0;
+ if( canSplit[1] ) numSplit += 2;
+ if( canSplit[2] ) numSplit += 1;
+ if( canSplit[3] ) numSplit += 1;
+ if( canSplit[4] ) numSplit += 1;
+ if( canSplit[5] ) numSplit += 1;
+
+ if( numSplit > 0 ) numSplit--;
+
+ ctxSpl += 3 * ( numSplit >> 1 );
+
+ //////////////////////////
+ // CTX is qt split (0-5)
+ //////////////////////////
+ ctxQt = ( cuLeft && cuLeft->qtDepth > partitioner.currQtDepth ) ? 1 : 0;
+ ctxQt += ( cuAbove && cuAbove->qtDepth > partitioner.currQtDepth ) ? 1 : 0;
+ ctxQt += partitioner.currQtDepth < 2 ? 0 : 3;
+
+ ////////////////////////////
+ // CTX is ver split (0-4)
+ ////////////////////////////
+ ctxHv = 0;
+
+ const unsigned numHor = ( canSplit[2] ? 1 : 0 ) + ( canSplit[4] ? 1 : 0 );
+ const unsigned numVer = ( canSplit[3] ? 1 : 0 ) + ( canSplit[5] ? 1 : 0 );
+
+ if( numVer == numHor )
+ {
+ const Area& area = partitioner.currArea().blocks[partitioner.chType];
+
+ const unsigned wAbove = cuAbove ? cuAbove->blocks[partitioner.chType].width : 1;
+ const unsigned hLeft = cuLeft ? cuLeft ->blocks[partitioner.chType].height : 1;
+
+ const unsigned depAbove = area.width / wAbove;
+ const unsigned depLeft = area.height / hLeft;
+
+ if( depAbove == depLeft || !cuLeft || !cuAbove ) ctxHv = 0;
+ else if( depAbove < depLeft ) ctxHv = 1;
+ else ctxHv = 2;
+ }
+ else if( numVer < numHor )
+ {
+ ctxHv = 3;
+ }
+ else
+ {
+ ctxHv = 4;
+ }
+
+ //////////////////////////
+ // CTX is h/v bt (0-3)
+ //////////////////////////
+ ctxHorBt = ( partitioner.currBtDepth >= 2 ? 1 : 0 );
+ ctxVerBt = ( partitioner.currBtDepth >= 2 ? 3 : 2 );
+}
+#else
unsigned DeriveCtx::CtxCUsplit( const CodingStructure& cs, Partitioner& partitioner )
{
auto adPartitioner = dynamic_cast<AdaptiveDepthPartitioner*>( &partitioner );
@@ -186,6 +302,7 @@ unsigned DeriveCtx::CtxCUsplit( const CodingStructure& cs, Partitioner& partitio
return ctxId;
}
+#endif
unsigned DeriveCtx::CtxQtCbf( const ComponentID compID, const unsigned trDepth, const bool prevCbCbf )
{
@@ -258,6 +375,7 @@ unsigned DeriveCtx::CtxIMVFlag( const CodingUnit& cu )
return ctxId;
}
+#if !JVET_M0421_SPLIT_SIG
unsigned DeriveCtx::CtxBTsplit(const CodingStructure& cs, Partitioner& partitioner)
{
const Position pos = partitioner.currArea().blocks[partitioner.chType];
@@ -306,6 +424,7 @@ unsigned DeriveCtx::CtxBTsplit(const CodingStructure& cs, Partitioner& partition
return ctx;
}
+#endif
unsigned DeriveCtx::CtxTriangleFlag( const CodingUnit& cu )
{
const CodingStructure *cs = cu.cs;
@@ -320,6 +439,17 @@ unsigned DeriveCtx::CtxTriangleFlag( const CodingUnit& cu )
return ctxId;
}
+#if JVET_M0502_PRED_MODE_CTX
+unsigned DeriveCtx::CtxPredModeFlag( const CodingUnit& cu )
+{
+ const CodingUnit *cuLeft = cu.cs->getCURestricted(cu.lumaPos().offset(-1, 0), cu, CH_L);
+ const CodingUnit *cuAbove = cu.cs->getCURestricted(cu.lumaPos().offset(0, -1), cu, CH_L);
+
+ unsigned ctxId = ((cuAbove && cuAbove->predMode == MODE_INTRA) || (cuLeft && cuLeft->predMode == MODE_INTRA)) ? 1 : 0;
+
+ return ctxId;
+}
+#endif
void MergeCtx::setMergeInfo( PredictionUnit& pu, int candIdx )
{
@@ -342,12 +472,15 @@ void MergeCtx::setMergeInfo( PredictionUnit& pu, int candIdx )
pu.mvpNum [REF_PIC_LIST_1] = NOT_VALID;
if (interDirNeighbours[candIdx] == 1 && pu.cs->slice->getRefPic(REF_PIC_LIST_0, mvFieldNeighbours[candIdx << 1].refIdx)->getPOC() == pu.cs->slice->getPOC())
{
- pu.cu->cpr = true;
+ pu.cu->ibc = true;
pu.bv = pu.mv[REF_PIC_LIST_0];
pu.bv.changePrecision(MV_PRECISION_INTERNAL, MV_PRECISION_INT); // used for only integer resolution
}
pu.cu->GBiIdx = ( interDirNeighbours[candIdx] == 3 ) ? GBiIdx[candIdx] : GBI_DEFAULT;
+#if JVET_M0068_M0171_MMVD_CLEANUP
+ PU::restrictBiPredMergeCandsOne(pu);
+#endif
}
void MergeCtx::setMmvdMergeCandiInfo(PredictionUnit& pu, int candIdx)
{
@@ -378,47 +511,78 @@ void MergeCtx::setMmvdMergeCandiInfo(PredictionUnit& pu, int candIdx)
const int poc0 = slice.getRefPOC(REF_PIC_LIST_0, refList0);
const int poc1 = slice.getRefPOC(REF_PIC_LIST_1, refList1);
const int currPoc = slice.getPOC();
+#if !JVET_M0068_M0171_MMVD_CLEANUP
int refSign = 1;
if ((poc0 - currPoc) * (currPoc - poc1) > 0)
{
refSign = -1;
}
+#endif
if (fPosPosition == 0)
{
tempMv[0] = Mv(offset, 0);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
tempMv[1] = Mv(offset * refSign, 0);
+#endif
}
else if (fPosPosition == 1)
{
tempMv[0] = Mv(-offset, 0);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
tempMv[1] = Mv(-offset * refSign, 0);
+#endif
}
else if (fPosPosition == 2)
{
tempMv[0] = Mv(0, offset);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
tempMv[1] = Mv(0, offset * refSign);
+#endif
}
else
{
tempMv[0] = Mv(0, -offset);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
tempMv[1] = Mv(0, -offset * refSign);
+#endif
}
+#if JVET_M0068_M0171_MMVD_CLEANUP
+ if ((poc0 - currPoc) == (poc1 - currPoc))
+ {
+ tempMv[1] = tempMv[0];
+ }
+ else if (abs(poc1 - currPoc) > abs(poc0 - currPoc))
+#else
if (abs(poc1 - currPoc) > abs(poc0 - currPoc))
+#endif
{
const int scale = PU::getDistScaleFactor(currPoc, poc0, currPoc, poc1);
+#if JVET_M0068_M0171_MMVD_CLEANUP
+ tempMv[1] = tempMv[0];
+ tempMv[0] = tempMv[1].scaleMv(scale);
+#else
if (scale != 4096)
{
tempMv[0] = tempMv[0].scaleMv(scale);
}
+#endif
}
+#if JVET_M0068_M0171_MMVD_CLEANUP
+ else
+#else
else if (abs(poc1 - currPoc) < abs(poc0 - currPoc))
+#endif
{
const int scale = PU::getDistScaleFactor(currPoc, poc1, currPoc, poc0);
+#if JVET_M0068_M0171_MMVD_CLEANUP
+ tempMv[1] = tempMv[0].scaleMv(scale);
+#else
if (scale != 4096)
{
tempMv[1] = tempMv[1].scaleMv(scale);
}
+#endif
}
pu.interDir = 3;
@@ -489,4 +653,8 @@ void MergeCtx::setMmvdMergeCandiInfo(PredictionUnit& pu, int candIdx)
pu.mvpNum[REF_PIC_LIST_1] = NOT_VALID;
pu.cu->GBiIdx = (interDirNeighbours[fPosBaseIdx] == 3) ? GBiIdx[fPosBaseIdx] : GBI_DEFAULT;
+
+#if JVET_M0068_M0171_MMVD_CLEANUP
+ PU::restrictBiPredMergeCandsOne(pu);
+#endif
}
diff --git a/source/Lib/CommonLib/ContextModelling.h b/source/Lib/CommonLib/ContextModelling.h
index 76fba82a0d454ebbfd5cd5c9538fdc2e8efb7b6c..4f2c0d90327606deffffdf26773083927762a763 100644
--- a/source/Lib/CommonLib/ContextModelling.h
+++ b/source/Lib/CommonLib/ContextModelling.h
@@ -107,7 +107,11 @@ public:
const int diag = posX + posY;
int numPos = 0;
int sumAbs = 0;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+#define UPDATE(x) {int a=abs(x);sumAbs+=std::min(4+(a&1),a);numPos+=!!a;}
+#else
#define UPDATE(x) {int a=abs(x);sumAbs+=std::min(2+(a&1),a);numPos+=!!a;}
+#endif
if( posX < m_width-1 )
{
UPDATE( pData[1] );
@@ -183,8 +187,10 @@ public:
return std::min(sum, 31);
}
+#if !JVET_M0464_UNI_MTS
unsigned emtNumSigCoeff() const { return m_emtNumSigCoeff; }
void setEmtNumSigCoeff( unsigned val ) { m_emtNumSigCoeff = val; }
+#endif
private:
// constant
@@ -235,7 +241,9 @@ private:
CtxSet m_parFlagCtxSet;
CtxSet m_gtxFlagCtxSet[2];
std::bitset<MLS_GRP_NUM> m_sigCoeffGroupFlag;
+#if !JVET_M0464_UNI_MTS
unsigned m_emtNumSigCoeff;
+#endif
};
@@ -294,14 +302,21 @@ public:
namespace DeriveCtx
{
+#if JVET_M0421_SPLIT_SIG
+void CtxSplit ( const CodingStructure& cs, Partitioner& partitioner, unsigned& ctxSpl, unsigned& ctxQt, unsigned& ctxHv, unsigned& ctxHorBt, unsigned& ctxVerBt, bool* canSplit = nullptr );
+#else
unsigned CtxCUsplit ( const CodingStructure& cs, Partitioner& partitioner );
unsigned CtxBTsplit ( const CodingStructure& cs, Partitioner& partitioner );
+#endif
unsigned CtxQtCbf ( const ComponentID compID, const unsigned trDepth, const bool prevCbCbf );
unsigned CtxInterDir ( const PredictionUnit& pu );
unsigned CtxSkipFlag ( const CodingUnit& cu );
unsigned CtxIMVFlag ( const CodingUnit& cu );
unsigned CtxAffineFlag( const CodingUnit& cu );
unsigned CtxTriangleFlag( const CodingUnit& cu );
+#if JVET_M0502_PRED_MODE_CTX
+unsigned CtxPredModeFlag( const CodingUnit& cu );
+#endif
}
#endif // __CONTEXTMODELLING__
diff --git a/source/Lib/CommonLib/Contexts.cpp b/source/Lib/CommonLib/Contexts.cpp
index 84a6947a12bbad73f44d4f0065dc535cd29fb93e..cbc516d354ffa006e10826ec03b36f8bb78fb434 100644
--- a/source/Lib/CommonLib/Contexts.cpp
+++ b/source/Lib/CommonLib/Contexts.cpp
@@ -349,6 +349,15 @@ std::vector<std::vector<uint8_t>> ContextSetCfg::sm_InitTables( NUMBER_OF_SLICE_
// clang-format off
const CtxSet ContextSetCfg::SplitFlag = ContextSetCfg::addCtxSet
({
+#if JVET_M0421_SPLIT_SIG
+ // |-------- do split ctx -------------------|
+ { 93, 124, 141, 123, 125, 141, 139, 126, 157, },
+ { 108, 139, 156, 138, 140, 141, 139, 141, 143, },
+ { 153, 154, 172, 153, 140, 156, 154, 127, 159, },
+#if JVET_M0453_CABAC_ENGINE
+ { DWS, DWS, DWS, DWS, DWS, DWS, DWS, DWS, DWS, },
+#endif
+#else
#if JVET_M0453_CABAC_ENGINE
{ 107, 110, 127, 106, 123, 140,},
{ 138, 140, 142, 106, 123, 125,},
@@ -359,8 +368,40 @@ const CtxSet ContextSetCfg::SplitFlag = ContextSetCfg::addCtxSet
{ 138, 111, 143, 107, 138, 140, },
{ 138, 141, 158, 151, 124, 126, },
#endif
+#endif
+});
+
+#if JVET_M0421_SPLIT_SIG
+const CtxSet ContextSetCfg::SplitQtFlag = ContextSetCfg::addCtxSet
+({
+ { 153, 126, 142, 137, 109, 155, },
+ { 153, 126, 157, 122, 138, 140, },
+ { 153, 125, 127, 137, 153, 155, },
+#if JVET_M0453_CABAC_ENGINE
+ { DWS, DWS, DWS, DWS, DWS, DWS, },
+#endif
+});
+
+const CtxSet ContextSetCfg::SplitHvFlag = ContextSetCfg::addCtxSet
+({
+ { 154, 168, 155, 153, 155, },
+ { 154, 168, 170, 153, 170, },
+ { 154, 153, 140, 153, 154, },
+#if JVET_M0453_CABAC_ENGINE
+ { DWS, DWS, DWS, DWS, DWS, },
+#endif
});
+const CtxSet ContextSetCfg::Split12Flag = ContextSetCfg::addCtxSet
+({
+ { 140, 154, 140, 154, },
+ { 155, 169, 140, 154, },
+ { 155, 154, 155, 154, },
+#if JVET_M0453_CABAC_ENGINE
+ { DWS, DWS, DWS, DWS, },
+#endif
+});
+#else
const CtxSet ContextSetCfg::BTSplitFlag = ContextSetCfg::addCtxSet
({
// |-------- 1st bin, 9 ctx for luma + 3 ctx for chroma------| |--2nd bin--| |3rd bin|
@@ -375,6 +416,7 @@ const CtxSet ContextSetCfg::BTSplitFlag = ContextSetCfg::addCtxSet
{ 139, 141, 157, 139, 155, 142, 153, 125, 141, 154, 154, 154, 154, 154, 154, 140, },
#endif
});
+#endif
const CtxSet ContextSetCfg::SkipFlag = ContextSetCfg::addCtxSet
({
@@ -477,15 +519,28 @@ const CtxSet ContextSetCfg::PartSize = ContextSetCfg::addCtxSet
const CtxSet ContextSetCfg::PredMode = ContextSetCfg::addCtxSet
({
#if JVET_M0453_CABAC_ENGINE
+#if JVET_M0502_PRED_MODE_CTX
+ { 193, 193, },
+ { 151, 151, },
+ { CNU, CNU, },
+ { 1, 1, },
+#else
{ 193,},
{ 151,},
{ CNU,},
{ 1,},
+#endif
+#else
+#if JVET_M0502_PRED_MODE_CTX
+ { 178, 178, },
+ { 194, 194, },
+ { CNU, CNU, },
#else
{ 178, },
{ 194, },
{ CNU, },
#endif
+#endif
});
const CtxSet ContextSetCfg::MultiRefLineIdx = ContextSetCfg::addCtxSet
@@ -1010,6 +1065,18 @@ const CtxSet ContextSetCfg::MVPIdx = ContextSetCfg::addCtxSet
#endif
});
+#if JVET_M0444_SMVD
+const CtxSet ContextSetCfg::SmvdFlag = ContextSetCfg::addCtxSet
+( {
+ { 154, },
+ { 110, },
+ { CNU, },
+#if JVET_M0453_CABAC_ENGINE
+ { DWS, }
+#endif
+} );
+#endif
+
const CtxSet ContextSetCfg::SaoMergeFlag = ContextSetCfg::addCtxSet
({
#if JVET_M0453_CABAC_ENGINE
@@ -1038,6 +1105,7 @@ const CtxSet ContextSetCfg::SaoTypeIdx = ContextSetCfg::addCtxSet
#endif
});
+#if !JVET_M0464_UNI_MTS
const CtxSet ContextSetCfg::TransformSkipFlag = ContextSetCfg::addCtxSet
({
#if JVET_M0453_CABAC_ENGINE
@@ -1051,6 +1119,7 @@ const CtxSet ContextSetCfg::TransformSkipFlag = ContextSetCfg::addCtxSet
{ 109, 42, },
#endif
});
+#endif
const CtxSet ContextSetCfg::TransquantBypassFlag = ContextSetCfg::addCtxSet
({
@@ -1082,6 +1151,21 @@ const CtxSet ContextSetCfg::RdpcmDir = ContextSetCfg::addCtxSet
#endif
});
+#if JVET_M0464_UNI_MTS
+const CtxSet ContextSetCfg::MTSIndex = ContextSetCfg::addCtxSet
+({
+#if JVET_M0453_CABAC_ENGINE
+ { CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, },
+ { CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, },
+ { CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, },
+ { DWS, DWS, DWS, DWS, DWS, DWS, DWS, DWS, DWS, DWS, DWS, },
+#else
+ { CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, },
+ { CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, },
+ { CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, CNU, },
+#endif
+});
+#else
const CtxSet ContextSetCfg::EMTTuIndex = ContextSetCfg::addCtxSet
({
#if JVET_M0453_CABAC_ENGINE
@@ -1109,6 +1193,7 @@ const CtxSet ContextSetCfg::EMTCuFlag = ContextSetCfg::addCtxSet
{ CNU, CNU, 140, 155, 155, CNU, },
#endif
});
+#endif
const CtxSet ContextSetCfg::CrossCompPred = ContextSetCfg::addCtxSet
({
diff --git a/source/Lib/CommonLib/Contexts.h b/source/Lib/CommonLib/Contexts.h
index dae5f60a09d173e5164df8e19676f7e36dc99559..fcc59f0152a14eddf03460f23822adff43834592 100644
--- a/source/Lib/CommonLib/Contexts.h
+++ b/source/Lib/CommonLib/Contexts.h
@@ -245,7 +245,13 @@ class ContextSetCfg
public:
// context sets: specify offset and size
static const CtxSet SplitFlag;
+#if JVET_M0421_SPLIT_SIG
+ static const CtxSet SplitQtFlag;
+ static const CtxSet SplitHvFlag;
+ static const CtxSet Split12Flag;
+#else
static const CtxSet BTSplitFlag;
+#endif
static const CtxSet SkipFlag;
static const CtxSet MergeFlag;
static const CtxSet MergeIdx;
@@ -276,12 +282,18 @@ public:
static const CtxSet MVPIdx;
static const CtxSet SaoMergeFlag;
static const CtxSet SaoTypeIdx;
+#if JVET_M0464_UNI_MTS
+ static const CtxSet MTSIndex;
+#else
static const CtxSet TransformSkipFlag;
+#endif
static const CtxSet TransquantBypassFlag;
static const CtxSet RdpcmFlag;
static const CtxSet RdpcmDir;
+#if !JVET_M0464_UNI_MTS
static const CtxSet EMTTuIndex;
static const CtxSet EMTCuFlag;
+#endif
static const CtxSet CrossCompPred;
static const CtxSet ChromaQpAdjFlag;
static const CtxSet ChromaQpAdjIdc;
@@ -292,6 +304,9 @@ public:
static const CtxSet MHIntraPredMode;
static const CtxSet TriangleFlag;
static const CtxSet TriangleIdx;
+#if JVET_M0444_SMVD
+ static const CtxSet SmvdFlag;
+#endif
static const unsigned NumberOfContexts;
// combined sets for less complex copying
diff --git a/source/Lib/CommonLib/DepQuant.cpp b/source/Lib/CommonLib/DepQuant.cpp
index 267f168f0f060c558bb5b4df476be8b218aa5aa3..820920082c726b577d39675070e5048683ce6888 100644
--- a/source/Lib/CommonLib/DepQuant.cpp
+++ b/source/Lib/CommonLib/DepQuant.cpp
@@ -642,7 +642,11 @@ namespace DQIntern
const int channelBitDepth = sps.getBitDepth( chType );
const int maxLog2TrDynamicRange = sps.getMaxLog2TrDynamicRange( chType );
const int nomTransformShift = getTransformShift( channelBitDepth, area.size(), maxLog2TrDynamicRange );
+#if JVET_M0464_UNI_MTS
+ const bool clipTransformShift = ( tu.mtsIdx==1 && sps.getSpsRangeExtension().getExtendedPrecisionProcessingFlag() );
+#else
const bool clipTransformShift = ( tu.transformSkip[ compID ] && sps.getSpsRangeExtension().getExtendedPrecisionProcessingFlag() );
+#endif
const int transformShift = ( clipTransformShift ? std::max<int>( 0, nomTransformShift ) : nomTransformShift );
// quant parameters
@@ -726,7 +730,11 @@ namespace DQIntern
const TCoeff minTCoeff = -( 1 << maxLog2TrDynamicRange );
const TCoeff maxTCoeff = ( 1 << maxLog2TrDynamicRange ) - 1;
const int nomTransformShift = getTransformShift( channelBitDepth, area.size(), maxLog2TrDynamicRange );
+#if JVET_M0464_UNI_MTS
+ const bool clipTransformShift = ( tu.mtsIdx==1 && sps.getSpsRangeExtension().getExtendedPrecisionProcessingFlag() );
+#else
const bool clipTransformShift = ( tu.transformSkip[ compID ] && sps.getSpsRangeExtension().getExtendedPrecisionProcessingFlag() );
+#endif
const int transformShift = ( clipTransformShift ? std::max<int>( 0, nomTransformShift ) : nomTransformShift );
#if HM_QTBT_AS_IN_JEM_QUANT
Intermediate_Int shift = IQUANT_SHIFT + 1 - qpPer - transformShift + ( TU::needsBlockSizeTrafoScale( area ) ? ADJ_DEQUANT_SHIFT : 0 );
@@ -855,7 +863,11 @@ namespace DQIntern
{
m_rdCost = std::numeric_limits<int64_t>::max()>>1;
m_numSigSbb = 0;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ m_remRegBins = 4; // just large enough for last scan pos
+#else
m_remRegBins = 3; // just large enough for last scan pos
+#endif
m_refSbbCtxId = -1;
m_sigFracBits = m_sigFracBitsArray[ 0 ];
m_coeffFracBits = m_gtxFracBitsArray[ 0 ];
@@ -869,7 +881,11 @@ namespace DQIntern
int64_t rdCostA = m_rdCost + pqDataA.deltaDist;
int64_t rdCostB = m_rdCost + pqDataB.deltaDist;
int64_t rdCostZ = m_rdCost;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ if( m_remRegBins >= 4 )
+#else
if( m_remRegBins >= 3 )
+#endif
{
if( pqDataA.absLevel < 4 )
rdCostA += m_coeffFracBits.bits[pqDataA.absLevel];
@@ -1008,14 +1024,22 @@ namespace DQIntern
m_sbbFracBits = prvState->m_sbbFracBits;
m_remRegBins = prvState->m_remRegBins - 1;
m_goRicePar = prvState->m_goRicePar;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ if( m_remRegBins >= 4 )
+#else
if( m_remRegBins >= 3 )
+#endif
{
TCoeff rem = (decision.absLevel - 4) >> 1;
if( m_goRicePar < 3 && rem > (3<<m_goRicePar)-1 )
{
m_goRicePar++;
}
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ m_remRegBins -= (decision.absLevel < 2 ? decision.absLevel : 3);
+#else
m_remRegBins -= std::min<TCoeff>( decision.absLevel, 2 );
+#endif
}
::memcpy( m_absLevelsAndCtxInit, prvState->m_absLevelsAndCtxInit, 48*sizeof(uint8_t) );
}
@@ -1025,11 +1049,19 @@ namespace DQIntern
m_refSbbCtxId = -1;
if ( scanInfo.sbbSize == 4 )
{
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ m_remRegBins = MAX_NUM_REG_BINS_2x2SUBBLOCK - (decision.absLevel < 2 ? decision.absLevel : 3);
+#else
m_remRegBins = MAX_NUM_REG_BINS_2x2SUBBLOCK - MAX_NUM_GT2_BINS_2x2SUBBLOCK - std::min<TCoeff>( decision.absLevel, 2 );
+#endif
}
else
{
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ m_remRegBins = MAX_NUM_REG_BINS_4x4SUBBLOCK - (decision.absLevel < 2 ? decision.absLevel : 3);
+#else
m_remRegBins = MAX_NUM_REG_BINS_4x4SUBBLOCK - MAX_NUM_GT2_BINS_4x4SUBBLOCK - std::min<TCoeff>( decision.absLevel, 2 );
+#endif
}
m_goRicePar = ( ((decision.absLevel - 4) >> 1) > (3<<0)-1 ? 1 : 0 );
::memset( m_absLevelsAndCtxInit, 0, 48*sizeof(uint8_t) );
@@ -1038,12 +1070,20 @@ namespace DQIntern
uint8_t* levels = reinterpret_cast<uint8_t*>(m_absLevelsAndCtxInit);
levels[ scanInfo.insidePos ] = (uint8_t)std::min<TCoeff>( 255, decision.absLevel );
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ if (m_remRegBins >= 4)
+#else
if (m_remRegBins >= 3)
+#endif
{
TCoeff tinit = m_absLevelsAndCtxInit[8 + scanInfo.nextInsidePos];
TCoeff sumAbs1 = (tinit >> 3) & 31;
TCoeff sumNum = tinit & 7;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+#define UPDATE(k) {TCoeff t=levels[scanInfo.nextNbInfoSbb.inPos[k]]; sumAbs1+=std::min<TCoeff>(4+(t&1),t); sumNum+=!!t; }
+#else
#define UPDATE(k) {TCoeff t=levels[scanInfo.nextNbInfoSbb.inPos[k]]; sumAbs1+=std::min<TCoeff>(2+(t&1),t); sumNum+=!!t; }
+#endif
if (numIPos == 1)
{
UPDATE(0);
@@ -1174,11 +1214,19 @@ namespace DQIntern
currState.m_numSigSbb = 0;
if (scanInfo.sbbSize == 4)
{
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ currState.m_remRegBins = MAX_NUM_REG_BINS_2x2SUBBLOCK;
+#else
currState.m_remRegBins = MAX_NUM_REG_BINS_2x2SUBBLOCK - MAX_NUM_GT2_BINS_2x2SUBBLOCK;
+#endif
}
else
{
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ currState.m_remRegBins = MAX_NUM_REG_BINS_4x4SUBBLOCK;
+#else
currState.m_remRegBins = MAX_NUM_REG_BINS_4x4SUBBLOCK - MAX_NUM_GT2_BINS_4x4SUBBLOCK;
+#endif
}
currState.m_goRicePar = 0;
currState.m_refSbbCtxId = currState.m_stateId;
@@ -1193,7 +1241,11 @@ namespace DQIntern
if( nbOut->num )
{
TCoeff sumAbs = 0, sumAbs1 = 0, sumNum = 0;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+#define UPDATE(k) {TCoeff t=absLevels[nbOut->outPos[k]]; sumAbs+=t; sumAbs1+=std::min<TCoeff>(4+(t&1),t); sumNum+=!!t; }
+#else
#define UPDATE(k) {TCoeff t=absLevels[nbOut->outPos[k]]; sumAbs+=t; sumAbs1+=std::min<TCoeff>(2+(t&1),t); sumNum+=!!t; }
+#endif
UPDATE(0);
if( nbOut->num > 1 )
{
diff --git a/source/Lib/CommonLib/CprHashMap.cpp b/source/Lib/CommonLib/IbcHashMap.cpp
similarity index 94%
rename from source/Lib/CommonLib/CprHashMap.cpp
rename to source/Lib/CommonLib/IbcHashMap.cpp
index 0ed0ac828fe838e947e7ddf40541bfdffbcaf6e5..9d876292deeedcab45d80f3057b14018b8ba17a4 100644
--- a/source/Lib/CommonLib/CprHashMap.cpp
+++ b/source/Lib/CommonLib/IbcHashMap.cpp
@@ -31,46 +31,46 @@
* THE POSSIBILITY OF SUCH DAMAGE.
*/
-/** \file CprHashMap.cpp
- \brief CPR hash map encoder class
+/** \file IbcHashMap.cpp
+ \brief IBC hash map encoder class
*/
#include "CommonLib/dtrace_codingstruct.h"
#include "CommonLib/Picture.h"
#include "CommonLib/UnitTools.h"
-#include "CprHashMap.h"
+#include "IbcHashMap.h"
using namespace std;
-//! \ingroup CprHashMap
+//! \ingroup IbcHashMap
//! \{
// ====================================================================================================================
// Constructor / destructor / create / destroy
// ====================================================================================================================
-CprHashMap::CprHashMap()
+IbcHashMap::IbcHashMap()
{
m_picWidth = 0;
m_picHeight = 0;
m_pos2Hash = NULL;
m_computeCrc32c = xxComputeCrc32c16bit;
-#if ENABLE_SIMD_OPT_CPR
+#if ENABLE_SIMD_OPT_IBC
#ifdef TARGET_SIMD_X86
- initCprHashMapX86();
+ initIbcHashMapX86();
#endif
#endif
}
-CprHashMap::~CprHashMap()
+IbcHashMap::~IbcHashMap()
{
destroy();
}
-void CprHashMap::init(const int picWidth, const int picHeight)
+void IbcHashMap::init(const int picWidth, const int picHeight)
{
if (picWidth != m_picWidth || picHeight != m_picHeight)
{
@@ -87,7 +87,7 @@ void CprHashMap::init(const int picWidth, const int picHeight)
}
}
-void CprHashMap::destroy()
+void IbcHashMap::destroy()
{
if (m_pos2Hash != NULL)
{
@@ -168,7 +168,7 @@ static const uint32_t crc32Table[256] = {
0xBE2DA0A5L, 0x4C4623A6L, 0x5F16D052L, 0xAD7D5351L
};
-uint32_t CprHashMap::xxComputeCrc32c16bit(uint32_t crc, const Pel pel)
+uint32_t IbcHashMap::xxComputeCrc32c16bit(uint32_t crc, const Pel pel)
{
const void *buf = &pel;
const uint8_t *p = (const uint8_t *)buf;
@@ -184,7 +184,7 @@ uint32_t CprHashMap::xxComputeCrc32c16bit(uint32_t crc, const Pel pel)
// CRC calculation in C code
////////////////////////////////////////////////////////
-unsigned int CprHashMap::xxCalcBlockHash(const Pel* pel, const int stride, const int width, const int height, unsigned int crc)
+unsigned int IbcHashMap::xxCalcBlockHash(const Pel* pel, const int stride, const int width, const int height, unsigned int crc)
{
for (int y = 0; y < height; y++)
{
@@ -198,7 +198,7 @@ unsigned int CprHashMap::xxCalcBlockHash(const Pel* pel, const int stride, const
}
template<ChromaFormat chromaFormat>
-void CprHashMap::xxBuildPicHashMap(const PelUnitBuf& pic)
+void IbcHashMap::xxBuildPicHashMap(const PelUnitBuf& pic)
{
const int chromaScalingX = getChannelTypeScaleX(CHANNEL_TYPE_CHROMA, chromaFormat);
const int chromaScalingY = getChannelTypeScaleY(CHANNEL_TYPE_CHROMA, chromaFormat);
@@ -243,7 +243,7 @@ void CprHashMap::xxBuildPicHashMap(const PelUnitBuf& pic)
}
}
-void CprHashMap::rebuildPicHashMap(const PelUnitBuf& pic)
+void IbcHashMap::rebuildPicHashMap(const PelUnitBuf& pic)
{
m_hash2Pos.clear();
@@ -267,7 +267,7 @@ void CprHashMap::rebuildPicHashMap(const PelUnitBuf& pic)
}
}
-bool CprHashMap::cprHashMatch(const Area& lumaArea, std::vector<Position>& cand, const CodingStructure& cs, const int maxCand, const int searchRange4SmallBlk)
+bool IbcHashMap::ibcHashMatch(const Area& lumaArea, std::vector<Position>& cand, const CodingStructure& cs, const int maxCand, const int searchRange4SmallBlk)
{
cand.clear();
@@ -332,7 +332,7 @@ bool CprHashMap::cprHashMatch(const Area& lumaArea, std::vector<Position>& cand,
return cand.size() > 0;
}
-int CprHashMap::getHashHitRatio(const Area& lumaArea)
+int IbcHashMap::getHashHitRatio(const Area& lumaArea)
{
int maxX = std::min((int)(lumaArea.x + lumaArea.width), m_picWidth);
int maxY = std::min((int)(lumaArea.y + lumaArea.height), m_picHeight);
diff --git a/source/Lib/CommonLib/CprHashMap.h b/source/Lib/CommonLib/IbcHashMap.h
similarity index 90%
rename from source/Lib/CommonLib/CprHashMap.h
rename to source/Lib/CommonLib/IbcHashMap.h
index 6ac4069b579a061702f130f3cdd8fa9031a7c16e..e343aab6e042abdd15ff0227a3ba38d921f24fdd 100644
--- a/source/Lib/CommonLib/CprHashMap.h
+++ b/source/Lib/CommonLib/IbcHashMap.h
@@ -31,12 +31,12 @@
* THE POSSIBILITY OF SUCH DAMAGE.
*/
-/** \file CprHashMap.h
- \brief CPR hash map encoder class (header)
+/** \file IbcHashMap.h
+ \brief IBC hash map encoder class (header)
*/
-#ifndef __CPRHASHMAP__
-#define __CPRHASHMAP__
+#ifndef __IBCHASHMAP__
+#define __IBCHASHMAP__
// Include files
#include "CommonLib/CommonDef.h"
@@ -55,7 +55,7 @@
// Class definition
// ====================================================================================================================
-class CprHashMap
+class IbcHashMap
{
private:
int m_picWidth;
@@ -73,23 +73,23 @@ private:
public:
uint32_t (*m_computeCrc32c) (uint32_t crc, const Pel pel);
- CprHashMap();
- virtual ~CprHashMap();
+ IbcHashMap();
+ virtual ~IbcHashMap();
void init(const int picWidth, const int picHeight);
void destroy();
void rebuildPicHashMap(const PelUnitBuf& pic);
- bool cprHashMatch(const Area& lumaArea, std::vector<Position>& cand, const CodingStructure& cs, const int maxCand, const int searchRange4SmallBlk);
+ bool ibcHashMatch(const Area& lumaArea, std::vector<Position>& cand, const CodingStructure& cs, const int maxCand, const int searchRange4SmallBlk);
int getHashHitRatio(const Area& lumaArea);
#ifdef TARGET_SIMD_X86
- void initCprHashMapX86();
+ void initIbcHashMapX86();
template <X86_VEXT vext>
- void _initCprHashMapX86();
+ void _initIbcHashMapX86();
#endif
};
//! \}
-#endif // __CPRHASHMAP__
+#endif // __IBCHASHMAP__
diff --git a/source/Lib/CommonLib/InterPrediction.cpp b/source/Lib/CommonLib/InterPrediction.cpp
index 923ebb2aa285311d32686dbfac9d5761c7668ed1..205d74394e0cf0adc52f44834c3280f42728888d 100644
--- a/source/Lib/CommonLib/InterPrediction.cpp
+++ b/source/Lib/CommonLib/InterPrediction.cpp
@@ -374,10 +374,10 @@ void InterPrediction::xPredInterUni(const PredictionUnit& pu, const RefPicList&
int iRefIdx = pu.refIdx[eRefPicList];
Mv mv[3];
- bool isCPR = false;
+ bool isIBC = false;
if (pu.cs->slice->getRefPic(eRefPicList, iRefIdx)->getPOC() == pu.cs->slice->getPOC())
{
- isCPR = true;
+ isIBC = true;
}
if( pu.cu->affine )
{
@@ -413,7 +413,7 @@ void InterPrediction::xPredInterUni(const PredictionUnit& pu, const RefPicList&
{
xPredInterBlk( compID, pu, pu.cu->slice->getRefPic( eRefPicList, iRefIdx ), mv[0], pcYuvPred, bi, pu.cu->slice->clpRng( compID )
, bioApplied
- , isCPR
+ , isIBC
);
}
@@ -426,7 +426,7 @@ void InterPrediction::xPredInterBi(PredictionUnit& pu, PelUnitBuf &pcYuvPred)
const Slice &slice = *pu.cs->slice;
bool bioApplied = false;
- if (pu.cs->sps->getSpsNext().getUseBIO())
+ if (pu.cs->sps->getBDOFEnabledFlag())
{
if (pu.cu->affine || m_subPuMC)
{
@@ -446,6 +446,13 @@ void InterPrediction::xPredInterBi(PredictionUnit& pu, PelUnitBuf &pcYuvPred)
}
}
+#if JVET_M0444_SMVD
+ if (bioApplied && pu.cu->smvdMode)
+ {
+ bioApplied = false;
+ }
+#endif
+
if (pu.cu->cs->sps->getSpsNext().getUseGBi() && bioApplied && pu.cu->GBiIdx != GBI_DEFAULT)
{
bioApplied = false;
@@ -517,7 +524,7 @@ void InterPrediction::xPredInterBi(PredictionUnit& pu, PelUnitBuf &pcYuvPred)
void InterPrediction::xPredInterBlk ( const ComponentID& compID, const PredictionUnit& pu, const Picture* refPic, const Mv& _mv, PelUnitBuf& dstPic, const bool& bi, const ClpRng& clpRng
, const bool& bioApplied
- , bool isCPR
+ , bool isIBC
)
{
JVET_J0090_SET_REF_PICTURE( refPic, compID );
@@ -529,7 +536,7 @@ void InterPrediction::xPredInterBlk ( const ComponentID& compID, const Predictio
int xFrac = _mv.hor & ((1 << shiftHor) - 1);
int yFrac = _mv.ver & ((1 << shiftVer) - 1);
- if (isCPR)
+ if (isIBC)
{
xFrac = yFrac = 0;
JVET_J0090_SET_CACHE_ENABLE( false );
@@ -582,9 +589,42 @@ void InterPrediction::xPredInterBlk ( const ComponentID& compID, const Predictio
JVET_J0090_SET_CACHE_ENABLE( true );
if (bioApplied && compID == COMPONENT_Y)
{
+#if JVET_M0487_INT_EXTEND
+ const int shift = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
+ const Pel* refPel = refBuf.buf - refBuf.stride - 1;
+ Pel* dstPel = m_filteredBlockTmp[2 + m_iRefListIdx][compID] + dstBuf.stride + 1;
+ for (int w = 0; w < (width - 2 * BIO_EXTEND_SIZE); w++)
+ {
+ Pel val = leftShift_round(refPel[w], shift);
+ dstPel[w] = val - (Pel)IF_INTERNAL_OFFS;
+ }
+
+ refPel = refBuf.buf - 1;
+ dstPel = m_filteredBlockTmp[2 + m_iRefListIdx][compID] + 2 * dstBuf.stride + 1;
+ for (int h = 0; h < (height - 2 * BIO_EXTEND_SIZE - 2); h++)
+ {
+ Pel val = leftShift_round(refPel[0], shift);
+ dstPel[0] = val - (Pel)IF_INTERNAL_OFFS;
+
+ val = leftShift_round(refPel[width - 3], shift);
+ dstPel[width - 3] = val - (Pel)IF_INTERNAL_OFFS;
+
+ refPel += refBuf.stride;
+ dstPel += dstBuf.stride;
+ }
+
+ refPel = refBuf.buf + (height - 2 * BIO_EXTEND_SIZE - 2)*refBuf.stride - 1;
+ dstPel = m_filteredBlockTmp[2 + m_iRefListIdx][compID] + (height - 2 * BIO_EXTEND_SIZE)*dstBuf.stride + 1;
+ for (int w = 0; w < (width - 2 * BIO_EXTEND_SIZE); w++)
+ {
+ Pel val = leftShift_round(refPel[w], shift);
+ dstPel[w] = val - (Pel)IF_INTERNAL_OFFS;
+ }
+#else
refBuf.buf = refBuf.buf - refBuf.stride - 1;
dstBuf.buf = m_filteredBlockTmp[2 + m_iRefListIdx][compID] + dstBuf.stride + 1;
bioSampleExtendBilinearFilter(refBuf.buf, refBuf.stride, dstBuf.buf, dstBuf.stride, width - 2, height - 2, 1, xFrac, yFrac, rndRes, chFmt, clpRng);
+#endif
// restore data
width = backupWidth;
@@ -686,7 +726,7 @@ void InterPrediction::xPredAffineBlk( const ComponentID& compID, const Predictio
roundAffineMv(iMvScaleTmpHor, iMvScaleTmpVer, shift);
// clip and scale
- if (sps.getUseWrapAround())
+ if (sps.getWrapAroundEnabledFlag())
{
m_storedMv[h / AFFINE_MIN_BLOCK_SIZE * MVBUFFER_SIZE + w / AFFINE_MIN_BLOCK_SIZE].set(iMvScaleTmpHor, iMvScaleTmpVer);
Mv tmpMv(iMvScaleTmpHor, iMvScaleTmpVer);
@@ -710,7 +750,7 @@ void InterPrediction::xPredAffineBlk( const ComponentID& compID, const Predictio
m_storedMv[((h << iScaleY) / AFFINE_MIN_BLOCK_SIZE + 1)* MVBUFFER_SIZE + ((w << iScaleX) / AFFINE_MIN_BLOCK_SIZE + 1)] +
Mv(2, 2));
curMv.set(curMv.getHor() >> 2, curMv.getVer() >> 2);
- if (sps.getUseWrapAround())
+ if (sps.getWrapAroundEnabledFlag())
{
clipMv(curMv, Position(pu.Y().x + (w << iScaleX), pu.Y().y + (h << iScaleY)), Size(blockWidth << iScaleX, blockHeight << iScaleY), sps);
}
@@ -898,6 +938,7 @@ void InterPrediction::applyBiOptFlow(const PredictionUnit &pu, const CPelUnitBuf
} // yu
}
+#if !JVET_M0487_INT_EXTEND
void InterPrediction::bioSampleExtendBilinearFilter(Pel const* src, int srcStride, Pel *dst, int dstStride, int width, int height, int dim, int fracX, int fracY, bool isLast, const ChromaFormat fmt, const ClpRng& clpRng)
{
Pel const* pSrc = NULL;
@@ -959,6 +1000,7 @@ void InterPrediction::bioSampleExtendBilinearFilter(Pel const* src, int srcStrid
}
}
}
+#endif
bool InterPrediction::xCalcBiPredSubBlkDist(const PredictionUnit &pu, const Pel* pYuvSrc0, const int src0Stride, const Pel* pYuvSrc1, const int src1Stride, const BitDepths &clipBitDepths)
{
@@ -1070,7 +1112,7 @@ void InterPrediction::motionCompensation( PredictionUnit &pu, PelUnitBuf &predBu
, const bool luma, const bool chroma
)
{
- // dual tree handling for CPR as the only ref
+ // dual tree handling for IBC as the only ref
if (!luma || !chroma)
{
if (!luma && chroma)
@@ -1111,7 +1153,7 @@ void InterPrediction::motionCompensation( PredictionUnit &pu, PelUnitBuf &predBu
}
else
{
- if (pu.mergeType != MRG_TYPE_DEFAULT_N && pu.mergeType != MRG_TYPE_CPR)
+ if (pu.mergeType != MRG_TYPE_DEFAULT_N && pu.mergeType != MRG_TYPE_IBC)
{
xSubPuMC( pu, predBuf, eRefPicList );
}
diff --git a/source/Lib/CommonLib/InterPrediction.h b/source/Lib/CommonLib/InterPrediction.h
index 2f07955698f401aeb58e5de66cc1c30d8540229b..145431e2777417e854efa484d18fe1fba7418ad4 100644
--- a/source/Lib/CommonLib/InterPrediction.h
+++ b/source/Lib/CommonLib/InterPrediction.h
@@ -63,6 +63,9 @@ class Mv;
class InterPrediction : public WeightPrediction
{
private:
+#if JVET_M0170_MRG_SHARELIST
+ int m_shareState;
+#endif
Distortion m_bioDistThres;
Distortion m_bioSubBlkDistThres;
@@ -101,8 +104,9 @@ protected:
int rightShiftMSB(int numer, int denom);
void applyBiOptFlow(const PredictionUnit &pu, const CPelUnitBuf &yuvSrc0, const CPelUnitBuf &yuvSrc1, const int &refIdx0, const int &refIdx1, PelUnitBuf &yuvDst, const BitDepths &clipBitDepths);
bool xCalcBiPredSubBlkDist(const PredictionUnit &pu, const Pel* yuvSrc0, const int src0Stride, const Pel* yuvSrc1, const int src1Stride, const BitDepths &clipBitDepths);
+#if !JVET_M0487_INT_EXTEND
void bioSampleExtendBilinearFilter(Pel const* src, int srcStride, Pel *dst, int dstStride, int width, int height, int dim, int fracX, int fracY, bool isLast, const ChromaFormat fmt, const ClpRng& clpRng);
-
+#endif
void xPredInterUni ( const PredictionUnit& pu, const RefPicList& eRefPicList, PelUnitBuf& pcYuvPred, const bool& bi
, const bool& bioApplied
, const bool luma, const bool chroma
@@ -110,7 +114,7 @@ protected:
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
, const bool& bioApplied
- , bool isCPR
+ , bool isIBC
);
void xAddBIOAvg4 (const Pel* src0, int src0Stride, const Pel* src1, int src1Stride, Pel *dst, int dstStride, const Pel *gradX0, const Pel *gradX1, const Pel *gradY0, const Pel*gradY1, int gradStride, int width, int height, int tmpx, int tmpy, int shift, int offset, const ClpRng& clpRng);
@@ -156,7 +160,9 @@ public:
#if JVET_J0090_MEMORY_BANDWITH_MEASURE
void cacheAssign( CacheModel *cache );
#endif
-
+#if JVET_M0170_MRG_SHARELIST
+ void setShareState(int shareStateIn) {m_shareState = shareStateIn;}
+#endif
};
//! \}
diff --git a/source/Lib/CommonLib/IntraPrediction.cpp b/source/Lib/CommonLib/IntraPrediction.cpp
index 3745b4bcc71eed36c838a90c714dccd25cd290d8..e49e92eceee0a95924f7d469f72ff4934e80877e 100644
--- a/source/Lib/CommonLib/IntraPrediction.cpp
+++ b/source/Lib/CommonLib/IntraPrediction.cpp
@@ -1563,6 +1563,24 @@ void IntraPrediction::xGetLumaRecPixels(const PredictionUnit &pu, CompArea chrom
pDst[i] = ( piSrc[2 * i] * 2 + piSrc[2 * i - 1] + piSrc[2 * i + 1] + 2 ) >> 2;
}
}
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ else if( pu.cs->sps->getSpsNext().getCclmCollocatedChromaFlag() )
+ {
+ piSrc = pRecSrc0 - iRecStride2;
+
+ if( i == 0 && !bLeftAvaillable )
+ {
+ pDst[i] = ( piSrc[2 * i] * 2 + piSrc[2 * i - iRecStride] + piSrc[2 * i + iRecStride] + 2 ) >> 2;
+ }
+ else
+ {
+ pDst[i] = ( piSrc[2 * i - iRecStride]
+ + piSrc[2 * i ] * 4 + piSrc[2 * i - 1] + piSrc[2 * i + 1]
+ + piSrc[2 * i + iRecStride]
+ + 4 ) >> 3;
+ }
+ }
+#endif
else
{
piSrc = pRecSrc0 - iRecStride2;
@@ -1592,9 +1610,30 @@ void IntraPrediction::xGetLumaRecPixels(const PredictionUnit &pu, CompArea chrom
}
for (int j = 0; j < uiCHeight + addedLeftBelow; j++)
{
- pDst[0] = ( ( piSrc[1 ] * 2 + piSrc[0 ] + piSrc[2 ] )
- + ( piSrc[1 + iRecStride] * 2 + piSrc[iRecStride] + piSrc[2 + iRecStride] )
- + 4 ) >> 3;
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ if( pu.cs->sps->getSpsNext().getCclmCollocatedChromaFlag() )
+ {
+ if( j == 0 && !bAboveAvaillable )
+ {
+ pDst[0] = ( piSrc[1] * 2 + piSrc[0] + piSrc[2] + 2 ) >> 2;
+ }
+ else
+ {
+ pDst[0] = ( piSrc[1 - iRecStride]
+ + piSrc[1 ] * 4 + piSrc[0] + piSrc[2]
+ + piSrc[1 + iRecStride]
+ + 4 ) >> 3;
+ }
+ }
+ else
+ {
+#endif
+ pDst[0] = ( ( piSrc[1 ] * 2 + piSrc[0 ] + piSrc[2 ] )
+ + ( piSrc[1 + iRecStride] * 2 + piSrc[iRecStride] + piSrc[2 + iRecStride] )
+ + 4 ) >> 3;
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ }
+#endif
piSrc += iRecStride2;
pDst += iDstStride;
@@ -1607,16 +1646,48 @@ void IntraPrediction::xGetLumaRecPixels(const PredictionUnit &pu, CompArea chrom
{
for( int i = 0; i < uiCWidth; i++ )
{
- if( i == 0 && !bLeftAvaillable )
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ if( pu.cs->sps->getSpsNext().getCclmCollocatedChromaFlag() )
{
- pDst0[i] = ( pRecSrc0[2 * i] + pRecSrc0[2 * i + iRecStride] + 1 ) >> 1;
+ if( i == 0 && !bLeftAvaillable )
+ {
+ if( j == 0 && !bAboveAvaillable )
+ {
+ pDst0[i] = pRecSrc0[2 * i];
+ }
+ else
+ {
+ pDst0[i] = ( pRecSrc0[2 * i] * 2 + pRecSrc0[2 * i - iRecStride] + pRecSrc0[2 * i + iRecStride] + 2 ) >> 2;
+ }
+ }
+ else if( j == 0 && !bAboveAvaillable )
+ {
+ pDst0[i] = ( pRecSrc0[2 * i] * 2 + pRecSrc0[2 * i - 1] + pRecSrc0[2 * i + 1] + 2 ) >> 2;
+ }
+ else
+ {
+ pDst0[i] = ( pRecSrc0[2 * i - iRecStride]
+ + pRecSrc0[2 * i ] * 4 + pRecSrc0[2 * i - 1] + pRecSrc0[2 * i + 1]
+ + pRecSrc0[2 * i + iRecStride]
+ + 4 ) >> 3;
+ }
}
else
{
- pDst0[i] = ( pRecSrc0[2 * i ] * 2 + pRecSrc0[2 * i + 1 ] + pRecSrc0[2 * i - 1 ]
- + pRecSrc0[2 * i + iRecStride] * 2 + pRecSrc0[2 * i + 1 + iRecStride] + pRecSrc0[2 * i - 1 + iRecStride]
- + 4 ) >> 3;
+#endif
+ if( i == 0 && !bLeftAvaillable )
+ {
+ pDst0[i] = ( pRecSrc0[2 * i] + pRecSrc0[2 * i + iRecStride] + 1 ) >> 1;
+ }
+ else
+ {
+ pDst0[i] = ( pRecSrc0[2 * i ] * 2 + pRecSrc0[2 * i + 1 ] + pRecSrc0[2 * i - 1 ]
+ + pRecSrc0[2 * i + iRecStride] * 2 + pRecSrc0[2 * i + 1 + iRecStride] + pRecSrc0[2 * i - 1 + iRecStride]
+ + 4 ) >> 3;
+ }
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
}
+#endif
}
pDst0 += iDstStride;
@@ -1784,6 +1855,37 @@ void IntraPrediction::xGetLMParameters(const PredictionUnit &pu, const Component
if (leftAvailable || aboveAvailable)
{
+#if JVET_M0064_CCLM_SIMPLIFICATION
+ int diff = maxLuma[0] - minLuma[0];
+ if (diff > 0)
+ {
+ int diffC = maxLuma[1] - minLuma[1];
+ int x = floorLog2( diff );
+ static const uint8_t DivSigTable[1 << 4] = {
+ // 4bit significands - 8 ( MSB is omitted )
+ 0, 7, 6, 5, 5, 4, 4, 3, 3, 2, 2, 1, 1, 1, 1, 0
+ };
+ int normDiff = (diff << 4 >> x) & 15;
+ int v = DivSigTable[normDiff] | 8;
+ x += normDiff != 0;
+
+ int y = floorLog2( abs( diffC ) ) + 1;
+ int add = 1 << y >> 1;
+ a = (diffC * v + add) >> y;
+ iShift = 3 + x - y;
+ if ( iShift < 1 ) {
+ iShift = 1;
+ a = ( (a == 0)? 0: (a < 0)? -15 : 15 ); // a=Sign(a)*15
+ }
+ b = minLuma[1] - ((a * minLuma[0]) >> iShift);
+ }
+ else
+ {
+ a = 0;
+ b = minLuma[1];
+ iShift = 0;
+ }
+#else // original
a = 0;
iShift = 16;
int shift = (internalBitDepth > 8) ? internalBitDepth - 9 : 0;
@@ -1795,6 +1897,7 @@ void IntraPrediction::xGetLMParameters(const PredictionUnit &pu, const Component
a = (((maxLuma[1] - minLuma[1]) * g_aiLMDivTableHigh[diff - 1] + div + add) >> shift);
}
b = minLuma[1] - ((a * minLuma[0]) >> iShift);
+#endif
}
else
{
diff --git a/source/Lib/CommonLib/LoopFilter.cpp b/source/Lib/CommonLib/LoopFilter.cpp
index ffc117b1f64e07628663ae4b3ba1fb9d866ea34a..f8543f5bc8a63203456999609cfe618e55c3a381 100644
--- a/source/Lib/CommonLib/LoopFilter.cpp
+++ b/source/Lib/CommonLib/LoopFilter.cpp
@@ -554,7 +554,7 @@ void LoopFilter::xEdgeFilterLuma(const CodingUnit& cu, const DeblockEdgeDir edge
unsigned uiBsAbsIdx = 0, uiBs = 0;
int iOffset, iSrcStep;
- bool bPCMFilter = (sps.getUsePCM() && sps.getPCMFilterDisableFlag()) ? true : false;
+ bool bPCMFilter = (sps.getPCMEnabledFlag() && sps.getPCMFilterDisableFlag()) ? true : false;
bool bPartPNoFilter = false;
bool bPartQNoFilter = false;
int betaOffsetDiv2 = slice.getDeblockingFilterBetaOffsetDiv2();
@@ -704,7 +704,7 @@ void LoopFilter::xEdgeFilterChroma(const CodingUnit& cu, const DeblockEdgeDir ed
int iOffset, iSrcStep;
unsigned uiLoopLength;
- bool bPCMFilter = (sps.getUsePCM() && sps.getPCMFilterDisableFlag()) ? true : false;
+ bool bPCMFilter = (sps.getPCMEnabledFlag() && sps.getPCMFilterDisableFlag()) ? true : false;
bool bPartPNoFilter = false;
bool bPartQNoFilter = false;
const int tcOffsetDiv2 = slice.getDeblockingFilterTcOffsetDiv2();
diff --git a/source/Lib/CommonLib/Mv.cpp b/source/Lib/CommonLib/Mv.cpp
index 91f379414cddf1c4f579c3ccc62eb057ba94b84b..7edfa2c56903738f79ce174ae35d224d855d16b3 100644
--- a/source/Lib/CommonLib/Mv.cpp
+++ b/source/Lib/CommonLib/Mv.cpp
@@ -61,7 +61,7 @@ void clipMv( Mv& rcMv, const Position& pos,
int iVerMax = ( sps.getPicHeightInLumaSamples() + iOffset - ( int ) pos.y - 1 ) << iMvShift;
int iVerMin = ( -( int ) sps.getMaxCUHeight() - iOffset - ( int ) pos.y + 1 ) << iMvShift;
- if( sps.getUseWrapAround() )
+ if( sps.getWrapAroundEnabledFlag() )
{
int iHorMax = ( sps.getPicWidthInLumaSamples() + sps.getMaxCUWidth() - size.width + iOffset - ( int ) pos.x - 1 ) << iMvShift;
int iHorMin = ( -( int ) sps.getMaxCUWidth() - iOffset - ( int ) pos.x + 1 ) << iMvShift;
diff --git a/source/Lib/CommonLib/Mv.h b/source/Lib/CommonLib/Mv.h
index 7b2872a9e3de226d00f631e0e636ec46021bf802..56287bec0b64953893c6313ce5b7d1ae3349c595 100644
--- a/source/Lib/CommonLib/Mv.h
+++ b/source/Lib/CommonLib/Mv.h
@@ -165,8 +165,13 @@ public:
const Mv scaleMv( int iScale ) const
{
+#if JVET_M0479_18BITS_MV_CLIP
+ const int mvx = Clip3( -131072, 131071, (iScale * getHor() + 127 + (iScale * getHor() < 0)) >> 8 );
+ const int mvy = Clip3( -131072, 131071, (iScale * getVer() + 127 + (iScale * getVer() < 0)) >> 8 );
+#else
const int mvx = Clip3( -32768, 32767, (iScale * getHor() + 127 + (iScale * getHor() < 0)) >> 8 );
const int mvy = Clip3( -32768, 32767, (iScale * getVer() + 127 + (iScale * getVer() < 0)) >> 8 );
+#endif
return Mv( mvx, mvy );
}
@@ -201,6 +206,13 @@ public:
{
roundToPrecision(src, m_amvrPrecision[amvr]);
}
+
+#if JVET_M0444_SMVD
+ Mv getSymmvdMv(const Mv& curMvPred, const Mv& tarMvPred)
+ {
+ return Mv(tarMvPred.hor - hor + curMvPred.hor, tarMvPred.ver - ver + curMvPred.ver);
+ }
+#endif
};// END CLASS DEFINITION MV
namespace std
diff --git a/source/Lib/CommonLib/Picture.cpp b/source/Lib/CommonLib/Picture.cpp
index 0937fd1bf8ff1f7c45d51d9ee2d68d0995539ad1..4c96307cb2196d52f9886c1189987f91d09c69c8 100644
--- a/source/Lib/CommonLib/Picture.cpp
+++ b/source/Lib/CommonLib/Picture.cpp
@@ -1004,7 +1004,7 @@ void Picture::extendPicBorder()
Pel* pi = piTxt;
// do left and right margins
- if (cs->sps->getUseWrapAround())
+ if (cs->sps->getWrapAroundEnabledFlag())
{
int xoffset = cs->sps->getWrapAroundOffset() >> getComponentScaleX( compID, cs->area.chromaFormat );
for (int y = 0; y < p.height; y++)
diff --git a/source/Lib/CommonLib/Quant.cpp b/source/Lib/CommonLib/Quant.cpp
index bed8fd5d85d0e325b213b748d921275be5672e63..ca477139b28c97e7342b3cae0b7c14d1494355ac 100644
--- a/source/Lib/CommonLib/Quant.cpp
+++ b/source/Lib/CommonLib/Quant.cpp
@@ -301,7 +301,11 @@ void Quant::dequant(const TransformUnit &tu,
CHECK(uiWidth > m_uiMaxTrSize, "Unsupported transformation size");
// Represents scaling through forward transform
+#if JVET_M0464_UNI_MTS
+ const bool bClipTransformShiftTo0 = tu.mtsIdx!=1 && sps->getSpsRangeExtension().getExtendedPrecisionProcessingFlag();
+#else
const bool bClipTransformShiftTo0 = (tu.transformSkip[compID] != 0) && sps->getSpsRangeExtension().getExtendedPrecisionProcessingFlag();
+#endif
const int originalTransformShift = getTransformShift(channelBitDepth, area.size(), maxLog2TrDynamicRange);
const int iTransformShift = bClipTransformShiftTo0 ? std::max<int>(0, originalTransformShift) : originalTransformShift;
@@ -717,7 +721,11 @@ void Quant::quant(TransformUnit &tu, const ComponentID &compID, const CCoeffBuf
const CCoeffBuf &piCoef = pSrc;
CoeffBuf piQCoef = tu.getCoeffs(compID);
+#if JVET_M0464_UNI_MTS
+ const bool useTransformSkip = tu.mtsIdx==1;
+#else
const bool useTransformSkip = tu.transformSkip[compID];
+#endif
const int maxLog2TrDynamicRange = sps.getMaxLog2TrDynamicRange(toChannelType(compID));
{
@@ -820,7 +828,11 @@ bool Quant::xNeedRDOQ(TransformUnit &tu, const ComponentID &compID, const CCoeff
const CCoeffBuf piCoef = pSrc;
+#if JVET_M0464_UNI_MTS
+ const bool useTransformSkip = tu.mtsIdx==1;
+#else
const bool useTransformSkip = tu.transformSkip[compID];
+#endif
const int maxLog2TrDynamicRange = sps.getMaxLog2TrDynamicRange(toChannelType(compID));
#if HEVC_USE_SCALING_LISTS
diff --git a/source/Lib/CommonLib/QuantRDOQ.cpp b/source/Lib/CommonLib/QuantRDOQ.cpp
index 8204a8c7f3959988ac7bff07cb74a9c81afb980b..bcf25fe5fd2006cf085f2430ca6b3add0080e613 100644
--- a/source/Lib/CommonLib/QuantRDOQ.cpp
+++ b/source/Lib/CommonLib/QuantRDOQ.cpp
@@ -185,7 +185,11 @@ inline int QuantRDOQ::xGetICRate( const uint32_t uiAbsLevel,
const bool useLimitedPrefixLength,
const int maxLog2TrDynamicRange ) const
{
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ if( remRegBins < 4 )
+#else
if( remRegBins < 3 )
+#endif
{
int iRate = int( xGetIEPRate() ); // cost of sign bit
uint32_t symbol = ( uiAbsLevel == 0 ? goRiceZero : uiAbsLevel <= goRiceZero ? uiAbsLevel-1 : uiAbsLevel );
@@ -226,7 +230,11 @@ inline int QuantRDOQ::xGetICRate( const uint32_t uiAbsLevel,
}
int iRate = int( xGetIEPRate() ); // cost of sign bit
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ const uint32_t cthres = 4;
+#else
const uint32_t cthres = ( remGt2Bins ? 4 : 2 );
+#endif
if( uiAbsLevel >= cthres )
{
uint32_t symbol = ( uiAbsLevel - cthres ) >> 1;
@@ -266,10 +274,14 @@ inline int QuantRDOQ::xGetICRate( const uint32_t uiAbsLevel,
iRate += fracBitsGt1.intBits[1];
iRate += fracBitsPar.intBits[( uiAbsLevel - 2 ) & 1];
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ iRate += fracBitsGt2.intBits[1];
+#else
if( remGt2Bins )
{
iRate += fracBitsGt2.intBits[1];
}
+#endif
}
else if( uiAbsLevel == 1 )
{
@@ -542,7 +554,11 @@ void QuantRDOQ::quant(TransformUnit &tu, const ComponentID &compID, const CCoeff
const CCoeffBuf &piCoef = pSrc;
CoeffBuf piQCoef = tu.getCoeffs(compID);
+#if JVET_M0464_UNI_MTS
+ const bool useTransformSkip = tu.mtsIdx==1;
+#else
const bool useTransformSkip = tu.transformSkip[compID];
+#endif
bool useRDOQ = useTransformSkip ? m_useRDOQTS : m_useRDOQ;
@@ -598,7 +614,11 @@ void QuantRDOQ::xRateDistOptQuant(TransformUnit &tu, const ComponentID &compID,
// Represents scaling through forward transform
int iTransformShift = getTransformShift(channelBitDepth, rect.size(), maxLog2TrDynamicRange);
+#if JVET_M0464_UNI_MTS
+ if (tu.mtsIdx==1 && extendedPrecision)
+#else
if (tu.transformSkip[compID] && extendedPrecision)
+#endif
{
iTransformShift = std::max<int>(0, iTransformShift);
}
@@ -686,7 +706,11 @@ void QuantRDOQ::xRateDistOptQuant(TransformUnit &tu, const ComponentID &compID,
bool is2x2subblock = ( iCGSizeM1 == 3 );
int remGt2Bins = ( is2x2subblock ? MAX_NUM_GT2_BINS_2x2SUBBLOCK : MAX_NUM_GT2_BINS_4x4SUBBLOCK );
- int remRegBins = ( is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK ) - remGt2Bins;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ int remRegBins = ( is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK );
+#else
+ int remRegBins = (is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK) - remGt2Bins;
+#endif
uint32_t goRiceParam = 0;
double *pdCostCoeffGroupSig = m_pdCostCoeffGroupSig;
@@ -762,7 +786,11 @@ void QuantRDOQ::xRateDistOptQuant(TransformUnit &tu, const ComponentID &compID,
uint32_t uiGt1Ctx = cctx.greater1CtxIdAbs ( ctxOffset );
uint32_t uiGt2Ctx = cctx.greater2CtxIdAbs ( ctxOffset );
uint32_t goRiceZero = 0;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ if( remRegBins < 4 )
+#else
if( remRegBins < 3 )
+#endif
{
unsigned sumAbs = cctx.templateAbsSum( iScanPos, piDstCoeff );
goRiceParam = g_auiGoRiceParsCoeff [ sumAbs ];
@@ -786,7 +814,11 @@ void QuantRDOQ::xRateDistOptQuant(TransformUnit &tu, const ComponentID &compID,
uiLevel = xGetCodedLevel( pdCostCoeff[ iScanPos ], pdCostCoeff0[ iScanPos ], pdCostSig[ iScanPos ],
lLevelDouble, uiMaxAbsLevel, &fracBitsSig, fracBitsPar, fracBitsGt1, fracBitsGt2, remGt2Bins, remRegBins, goRiceZero, goRiceParam, iQBits, errorScale, 0, extendedPrecision, maxLog2TrDynamicRange );
#if HEVC_USE_SIGN_HIDING
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ sigRateDelta[ uiBlkPos ] = ( remRegBins < 4 ? 0 : fracBitsSig.intBits[1] - fracBitsSig.intBits[0] );
+#else
sigRateDelta[ uiBlkPos ] = ( remRegBins < 3 ? 0 : fracBitsSig.intBits[1] - fracBitsSig.intBits[0] );
+#endif
#endif
}
@@ -805,7 +837,11 @@ void QuantRDOQ::xRateDistOptQuant(TransformUnit &tu, const ComponentID &compID,
}
else // uiLevel == 0
{
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ if( remRegBins < 4 )
+#else
if( remRegBins < 3 )
+#endif
{
int rateNow = xGetICRate( uiLevel, fracBitsPar, fracBitsGt1, fracBitsGt2, remGt2Bins, remRegBins, goRiceZero, goRiceParam, extendedPrecision, maxLog2TrDynamicRange );
rateIncUp [ uiBlkPos ] = xGetICRate( uiLevel+1, fracBitsPar, fracBitsGt1, fracBitsGt2, remGt2Bins, remRegBins, goRiceZero, goRiceParam, extendedPrecision, maxLog2TrDynamicRange ) - rateNow;
@@ -825,18 +861,32 @@ void QuantRDOQ::xRateDistOptQuant(TransformUnit &tu, const ComponentID &compID,
remRegBins = ( is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK ) - remGt2Bins;
goRiceParam = 0;
}
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ else if( remRegBins >= 4 )
+#else
else if( remRegBins >= 3 )
+#endif
{
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ const uint32_t baseLevel = 4;
+#else
const uint32_t baseLevel = ( remGt2Bins ? 4 : 2 );
+#endif
if( goRiceParam < 3 && ((uiLevel-baseLevel)>>1) > (3<<goRiceParam)-1 )
{
goRiceParam++;
}
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
if( uiLevel >= 2 && remGt2Bins )
{
remGt2Bins--;
}
+#endif
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ remRegBins -= (uiLevel < 2 ? uiLevel : 3) + (iScanPos != iLastScanPos);
+#else
remRegBins -= std::min<int>( uiLevel, 2 ) + (iScanPos != iLastScanPos);
+#endif
}
}
else
diff --git a/source/Lib/CommonLib/RdCost.h b/source/Lib/CommonLib/RdCost.h
index 52422e36358f1de2a5be8d17dcd38c5d88c2f4f0..ead8fd5cb20aa52963c84a7c55dc6f1f34f1b880 100644
--- a/source/Lib/CommonLib/RdCost.h
+++ b/source/Lib/CommonLib/RdCost.h
@@ -118,7 +118,7 @@ private:
double m_motionLambda;
int m_iCostScale;
- double m_dCost; // for cpr
+ double m_dCost; // for ibc
public:
RdCost();
virtual ~RdCost();
@@ -162,7 +162,7 @@ public:
}
void setCostScale ( int iCostScale ) { m_iCostScale = iCostScale; }
Distortion getCost ( uint32_t b ) { return Distortion( m_motionLambda * b ); }
- // for cpr
+ // for ibc
void getMotionCost(int add, bool isTransquantBypass) { m_dCost = m_dLambdaMotionSAD[(isTransquantBypass && m_costMode == COST_MIXED_LOSSLESS_LOSSY_CODING) ? 1 : 0] + add; }
void setPredictors(Mv* pcMv)
diff --git a/source/Lib/CommonLib/Rom.cpp b/source/Lib/CommonLib/Rom.cpp
index 5ae6612d5944d43ab9242a1f1e84a2c4214b3bea..a485d9fadf0a04f6d7a1aa3e0a9cc7ae868502a9 100644
--- a/source/Lib/CommonLib/Rom.cpp
+++ b/source/Lib/CommonLib/Rom.cpp
@@ -180,6 +180,7 @@ public:
return rtn;
}
};
+#if !JVET_M0064_CCLM_SIMPLIFICATION
int g_aiLMDivTableLow[] = {
0, 0, 21845, 0, 13107, 43690, 18724, 0, 50972, 39321, 53620, 21845, 15123, 9362, 4369, 0, 3855,
58254, 17246, 52428, 49932, 59578, 25644, 43690, 28835, 40329, 16990, 37449, 56496, 34952, 4228, 0, 61564, 34695,
@@ -244,6 +245,7 @@ int g_aiLMDivTableHigh[] = {
134, 134, 134, 133, 133, 133, 132, 132, 132, 132, 131, 131, 131, 131, 130, 130, 130, 130,
129, 129, 129, 129, 128, 128, 128, 128,
};
+#endif
const int8_t g_GbiLog2WeightBase = 3;
const int8_t g_GbiWeightBase = (1 << g_GbiLog2WeightBase);
const int8_t g_GbiWeights[GBI_NUM] = { -2, 3, 4, 5, 10 };
@@ -619,7 +621,7 @@ const uint8_t g_chroma422IntraAngleMappingTable[NUM_INTRA_MODE] =
{ 0, 1, 2, 2, 2, 2, 2, 2, 2, 3, 4, 6, 8, 10, 12, 13, 14, 16, 18, 20, 22, 23, 24, 26, 28, 30, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 44, 44, 45, 46, 46, 46, 47, 48, 48, 48, 49, 50, 51, 52, 52, 52, 53, 54, 54, 54, 55, 56, 56, 56, 57, 58, 59, 60, DM_CHROMA_IDX };
-
+#if !REMOVE_BIN_DECISION_TREE
// ====================================================================================================================
// Decision tree templates
// ====================================================================================================================
@@ -635,6 +637,7 @@ const DecisionTreeTemplate g_mtSplitDTT = compile(
/*0*/ DTT_SPLIT_TT_VERT,
/*1*/ DTT_SPLIT_BT_VERT ) ) ) );
+#endif
// ====================================================================================================================
diff --git a/source/Lib/CommonLib/Rom.h b/source/Lib/CommonLib/Rom.h
index df42ecc8f5862ce7f546e1f25162ec795a075213..2a856a448d8bd7b4f06454e42cbf9806a8022191 100644
--- a/source/Lib/CommonLib/Rom.h
+++ b/source/Lib/CommonLib/Rom.h
@@ -113,7 +113,9 @@ extern const uint8_t g_chroma422IntraAngleMappingTable[NUM_INTRA_MODE];
extern const TMatrixCoeff g_as_DST_MAT_4 [TRANSFORM_NUMBER_OF_DIRECTIONS][4][4];
#endif
+#if !JVET_M0464_UNI_MTS
extern const uint32_t g_EmtSigNumThr;
+#endif
extern const TMatrixCoeff g_trCoreDCT2P2 [TRANSFORM_NUMBER_OF_DIRECTIONS][ 2][ 2];
extern const TMatrixCoeff g_trCoreDCT2P4 [TRANSFORM_NUMBER_OF_DIRECTIONS][ 4][ 4];
@@ -132,6 +134,7 @@ extern const TMatrixCoeff g_trCoreDST7P8 [TRANSFORM_NUMBER_OF_DIRECTIONS][ 8][
extern const TMatrixCoeff g_trCoreDST7P16 [TRANSFORM_NUMBER_OF_DIRECTIONS][ 16][ 16];
extern const TMatrixCoeff g_trCoreDST7P32 [TRANSFORM_NUMBER_OF_DIRECTIONS][ 32][ 32];
+#if !REMOVE_BIN_DECISION_TREE
// ====================================================================================================================
// Decision tree templates
// ====================================================================================================================
@@ -155,6 +158,7 @@ extern const DecisionTreeTemplate g_mtSplitDTT;
// decision tree for QTBT split
extern const DecisionTreeTemplate g_qtbtSplitDTT;
+#endif
// ====================================================================================================================
// Misc.
@@ -210,8 +214,10 @@ extern const uint32_t g_scalingListSizeX[SCALING_LIST_SIZE_NUM];
extern MsgLevel g_verbosity;
+#if !JVET_M0064_CCLM_SIMPLIFICATION
extern int g_aiLMDivTableLow[];
extern int g_aiLMDivTableHigh[];
+#endif
extern const int g_aiNonLMPosThrs[];
diff --git a/source/Lib/CommonLib/RomTr.cpp b/source/Lib/CommonLib/RomTr.cpp
index 3f0815e85b0fec545134cc9fa15920ed2cbe5191..2df3c5a131c9bf299a36c6a32995206b127253f8 100644
--- a/source/Lib/CommonLib/RomTr.cpp
+++ b/source/Lib/CommonLib/RomTr.cpp
@@ -396,13 +396,23 @@ const TMatrixCoeff g_trCoreDCT8P8[TRANSFORM_NUMBER_OF_DIRECTIONS][8][8] =
};
const TMatrixCoeff g_trCoreDCT8P16[TRANSFORM_NUMBER_OF_DIRECTIONS][16][16] =
{
+#if JVET_M0497_FAST_DST7
+ DEFINE_DCT8_P16_MATRIX(88, 88, 87, 85, 81, 77, 73, 68, 62, 55, 48, 40, 33, 25, 17, 8),
+ DEFINE_DCT8_P16_MATRIX(88, 88, 87, 85, 81, 77, 73, 68, 62, 55, 48, 40, 33, 25, 17, 8)
+#else
DEFINE_DCT8_P16_MATRIX(90, 89, 87, 83, 81, 77, 72, 66, 62, 56, 49, 41, 33, 25, 17, 9),
DEFINE_DCT8_P16_MATRIX(90, 89, 87, 83, 81, 77, 72, 66, 62, 56, 49, 41, 33, 25, 17, 9)
+#endif
};
const TMatrixCoeff g_trCoreDCT8P32[TRANSFORM_NUMBER_OF_DIRECTIONS][32][32] =
{
+#if JVET_M0497_FAST_DST7
+ DEFINE_DCT8_P32_MATRIX(90, 90, 89, 88, 87, 86, 85, 84, 82, 80, 78, 77, 74, 72, 68, 66, 63, 60, 56, 53, 50, 46, 42, 38, 34, 30, 26, 21, 17, 13, 9, 4),
+ DEFINE_DCT8_P32_MATRIX(90, 90, 89, 88, 87, 86, 85, 84, 82, 80, 78, 77, 74, 72, 68, 66, 63, 60, 56, 53, 50, 46, 42, 38, 34, 30, 26, 21, 17, 13, 9, 4)
+#else
DEFINE_DCT8_P32_MATRIX(90, 90, 89, 88, 88, 86, 85, 84, 82, 80, 78, 77, 74, 72, 68, 66, 63, 60, 56, 53, 50, 45, 42, 38, 34, 30, 26, 21, 17, 13, 9, 4),
DEFINE_DCT8_P32_MATRIX(90, 90, 89, 88, 88, 86, 85, 84, 82, 80, 78, 77, 74, 72, 68, 66, 63, 60, 56, 53, 50, 45, 42, 38, 34, 30, 26, 21, 17, 13, 9, 4)
+#endif
};
// DST-7
@@ -418,13 +428,23 @@ const TMatrixCoeff g_trCoreDST7P8[TRANSFORM_NUMBER_OF_DIRECTIONS][8][8] =
};
const TMatrixCoeff g_trCoreDST7P16[TRANSFORM_NUMBER_OF_DIRECTIONS][16][16] =
{
+#if JVET_M0497_FAST_DST7
+ DEFINE_DST7_P16_MATRIX(8, 17, 25, 33, 40, 48, 55, 62, 68, 73, 77, 81, 85, 87, 88, 88),
+ DEFINE_DST7_P16_MATRIX(8, 17, 25, 33, 40, 48, 55, 62, 68, 73, 77, 81, 85, 87, 88, 88)
+#else
DEFINE_DST7_P16_MATRIX(9, 17, 25, 33, 41, 49, 56, 62, 66, 72, 77, 81, 83, 87, 89, 90),
DEFINE_DST7_P16_MATRIX(9, 17, 25, 33, 41, 49, 56, 62, 66, 72, 77, 81, 83, 87, 89, 90)
+#endif
};
const TMatrixCoeff g_trCoreDST7P32[TRANSFORM_NUMBER_OF_DIRECTIONS][32][32] =
{
+#if JVET_M0497_FAST_DST7
+ DEFINE_DST7_P32_MATRIX(4, 9, 13, 17, 21, 26, 30, 34, 38, 42, 46, 50, 53, 56, 60, 63, 66, 68, 72, 74, 77, 78, 80, 82, 84, 85, 86, 87, 88, 89, 90, 90),
+ DEFINE_DST7_P32_MATRIX(4, 9, 13, 17, 21, 26, 30, 34, 38, 42, 46, 50, 53, 56, 60, 63, 66, 68, 72, 74, 77, 78, 80, 82, 84, 85, 86, 87, 88, 89, 90, 90)
+#else
DEFINE_DST7_P32_MATRIX(4, 9, 13, 17, 21, 26, 30, 34, 38, 42, 45, 50, 53, 56, 60, 63, 66, 68, 72, 74, 77, 78, 80, 82, 84, 85, 86, 88, 88, 89, 90, 90),
DEFINE_DST7_P32_MATRIX(4, 9, 13, 17, 21, 26, 30, 34, 38, 42, 45, 50, 53, 56, 60, 63, 66, 68, 72, 74, 77, 78, 80, 82, 84, 85, 86, 88, 88, 89, 90, 90)
+#endif
};
//--------------------------------------------------------------------------------------------------
diff --git a/source/Lib/CommonLib/SampleAdaptiveOffset.cpp b/source/Lib/CommonLib/SampleAdaptiveOffset.cpp
index 14763c7f70bc10b327679af805c93509af802360..a048e366d9a302f393798f6aeaa650b54a0f729a 100644
--- a/source/Lib/CommonLib/SampleAdaptiveOffset.cpp
+++ b/source/Lib/CommonLib/SampleAdaptiveOffset.cpp
@@ -614,7 +614,7 @@ void SampleAdaptiveOffset::SAOProcess( CodingStructure& cs, SAOBlkParam* saoBlkP
void SampleAdaptiveOffset::xPCMLFDisableProcess(CodingStructure& cs)
{
const PreCalcValues& pcv = *cs.pcv;
- const bool bPCMFilter = (cs.sps->getUsePCM() && cs.sps->getPCMFilterDisableFlag()) ? true : false;
+ const bool bPCMFilter = (cs.sps->getPCMEnabledFlag() && cs.sps->getPCMFilterDisableFlag()) ? true : false;
if( bPCMFilter || cs.pps->getTransquantBypassEnabledFlag() )
{
diff --git a/source/Lib/CommonLib/Slice.cpp b/source/Lib/CommonLib/Slice.cpp
index fa7f7a33b712149641546842dc329cd17775e176..00f16676bcbef8eece7b89a92ad3b667df5a08cc 100644
--- a/source/Lib/CommonLib/Slice.cpp
+++ b/source/Lib/CommonLib/Slice.cpp
@@ -74,6 +74,9 @@ Slice::Slice()
, m_signDataHidingEnabledFlag ( false )
#endif
, m_bCheckLDC ( false )
+#if JVET_M0444_SMVD
+, m_biDirPred ( false )
+#endif
, m_iSliceQpDelta ( 0 )
, m_iDepth ( 0 )
#if HEVC_VPS
@@ -134,6 +137,9 @@ Slice::Slice()
, m_uiMaxTTSizeIChroma ( 0 )
, m_uiMaxBTSize ( 0 )
, m_MotionCandLut (NULL)
+#if JVET_M0170_MRG_SHARELIST
+, m_MotionCandLuTsBkup (NULL)
+#endif
{
for(uint32_t i=0; i<NUM_REF_PIC_LIST_01; i++)
{
@@ -192,6 +198,12 @@ void Slice::initSlice()
m_bCheckLDC = false;
+#if JVET_M0444_SMVD
+ m_biDirPred = false;
+ m_symRefIdx[0] = -1;
+ m_symRefIdx[1] = -1;
+#endif
+
for (uint32_t component = 0; component < MAX_NUM_COMPONENT; component++)
{
m_iSliceChromaQpDelta[component] = 0;
@@ -423,7 +435,7 @@ void Slice::setRefPicList( PicList& rcListPic, bool checkNumPocTotalCurr, bool b
pcRefPic = xGetLongTermRefPic(rcListPic, m_pRPS->getPOC(i), m_pRPS->getCheckLTMSBPresent(i));
}
}
- if (getSPS()->getSpsNext().getCPRMode())
+ if (getSPS()->getSpsNext().getIBCMode())
{
RefPicSetLtCurr[NumPicLtCurr] = getPic();
//getPic()->setIsLongTerm(true);
@@ -442,7 +454,7 @@ 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())
{
- if (getSPS()->getSpsNext().getCPRMode())
+ if (getSPS()->getSpsNext().getIBCMode())
{
CHECK(numPicTotalCurr != 1, "Invalid state");
}
@@ -517,7 +529,7 @@ void Slice::setRefPicList( PicList& rcListPic, bool checkNumPocTotalCurr, bool b
m_bIsUsedAsLongTerm[REF_PIC_LIST_1][rIdx] = ( cIdx >= NumPicStCurr0 + NumPicStCurr1 );
}
}
- if (getSPS()->getSpsNext().getCPRMode())
+ if (getSPS()->getSpsNext().getIBCMode())
{
m_apcRefPicList[REF_PIC_LIST_0][m_aiNumRefIdx[REF_PIC_LIST_0] - 1] = getPic();
m_bIsUsedAsLongTerm[REF_PIC_LIST_0][m_aiNumRefIdx[REF_PIC_LIST_0] - 1] = true;
@@ -552,7 +564,7 @@ int Slice::getNumRpsCurrTempList() const
numRpsCurrTempList++;
}
}
- if (getSPS()->getSpsNext().getCPRMode())
+ if (getSPS()->getSpsNext().getIBCMode())
{
return numRpsCurrTempList + 1;
}
@@ -764,6 +776,13 @@ void Slice::copySliceInfo(Slice *pSrc, bool cpyAlmostAll)
m_bCheckLDC = pSrc->m_bCheckLDC;
m_iSliceQpDelta = pSrc->m_iSliceQpDelta;
+
+#if JVET_M0444_SMVD
+ m_biDirPred = pSrc->m_biDirPred;
+ m_symRefIdx[0] = pSrc->m_symRefIdx[0];
+ m_symRefIdx[1] = pSrc->m_symRefIdx[1];
+#endif
+
for (uint32_t component = 0; component < MAX_NUM_COMPONENT; component++)
{
m_iSliceChromaQpDelta[component] = pSrc->m_iSliceChromaQpDelta[component];
@@ -1595,6 +1614,12 @@ void Slice::initMotionLUTs()
m_MotionCandLut->currCnt = 0;
m_MotionCandLut->motionCand = nullptr;
m_MotionCandLut->motionCand = new MotionInfo[MAX_NUM_HMVP_CANDS];
+#if JVET_M0170_MRG_SHARELIST
+ m_MotionCandLuTsBkup = new LutMotionCand;
+ m_MotionCandLuTsBkup->currCnt = 0;
+ m_MotionCandLuTsBkup->motionCand = nullptr;
+ m_MotionCandLuTsBkup->motionCand = new MotionInfo[MAX_NUM_HMVP_CANDS];
+#endif
}
void Slice::destroyMotionLUTs()
{
@@ -1602,18 +1627,31 @@ void Slice::destroyMotionLUTs()
m_MotionCandLut->motionCand = nullptr;
delete m_MotionCandLut;
m_MotionCandLut = NULL;
+#if JVET_M0170_MRG_SHARELIST
+ delete[] m_MotionCandLuTsBkup->motionCand;
+ m_MotionCandLuTsBkup->motionCand = nullptr;
+ delete m_MotionCandLuTsBkup;
+ m_MotionCandLuTsBkup = NULL;
+#endif
}
void Slice::resetMotionLUTs()
{
m_MotionCandLut->currCnt = 0;
+#if JVET_M0170_MRG_SHARELIST
+ m_MotionCandLuTsBkup->currCnt = 0;
+#endif
}
MotionInfo Slice::getMotionInfoFromLUTs(int MotCandIdx) const
{
return m_MotionCandLut->motionCand[MotCandIdx];
}
-
-
+#if JVET_M0170_MRG_SHARELIST
+MotionInfo Slice::getMotionInfoFromLUTBkup(int MotCandIdx) const
+{
+ return m_MotionCandLuTsBkup->motionCand[MotCandIdx];
+}
+#endif
void Slice::addMotionInfoToLUTs(LutMotionCand* lutMC, MotionInfo newMi)
{
@@ -1733,13 +1771,19 @@ SPSNext::SPSNext( SPS& sps )
, m_NextEnabled ( false )
// disable all tool enabling flags by default
, m_LargeCTU ( false )
- , m_SubPuMvp ( false )
, m_IMV ( false )
- , m_BIO ( false )
, m_DisableMotionCompression ( false )
, m_LMChroma ( false )
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ , m_cclmCollocatedChromaFlag ( false )
+#endif
+#if JVET_M0464_UNI_MTS
+ , m_IntraMTS ( false )
+ , m_InterMTS ( false )
+#else
, m_IntraEMT ( false )
, m_InterEMT ( false )
+#endif
, m_Affine ( false )
, m_AffineType ( false )
, m_MTTEnabled ( false )
@@ -1756,11 +1800,10 @@ SPSNext::SPSNext( SPS& sps )
#endif
// default values for additional parameters
- , m_subPuMrgMode ( 0 )
, m_ImvMode ( IMV_OFF )
, m_MTTMode ( 0 )
, m_compositeRefEnabled ( false )
- , m_CPRMode ( 0 )
+ , m_IBCMode ( 0 )
// ADD_NEW_TOOL : (sps extension) add tool enabling flags here (with "false" as default values)
{
}
@@ -1809,14 +1852,16 @@ SPS::SPS()
, m_uiQuadtreeTUMaxDepthInter ( 0)
, m_uiQuadtreeTUMaxDepthIntra ( 0)
// Tool list
-, m_usePCM (false)
+, m_pcmEnabledFlag (false)
, m_pcmLog2MaxSize ( 5)
, m_uiPCMLog2MinSize ( 7)
, m_bPCMFilterDisableFlag (false)
+, m_sbtmvpEnabledFlag (false)
+, m_bdofEnabledFlag (false)
, m_uiBitsForPOC ( 8)
, m_numLongTermRefPicSPS ( 0)
, m_uiMaxTrSize ( 32)
-, m_bUseSAO (false)
+, m_saoEnabledFlag (false)
, m_bTemporalIdNestingFlag (false)
#if HEVC_USE_SCALING_LISTS
, m_scalingListEnabledFlag (false)
@@ -1827,7 +1872,7 @@ SPS::SPS()
, m_vuiParametersPresentFlag (false)
, m_vuiParameters ()
, m_spsNextExtension (*this)
-, m_useWrapAround (false)
+, m_wrapAroundEnabledFlag (false)
, m_wrapAroundOffset ( 0)
{
for(int ch=0; ch<MAX_NUM_CHANNEL_TYPE; ch++)
diff --git a/source/Lib/CommonLib/Slice.h b/source/Lib/CommonLib/Slice.h
index 60b4d5a27c78a5a582e08aa593d92b6992993f68..05b8ea8f046b66ed3bc1e72b0d4288b506ea31dd 100644
--- a/source/Lib/CommonLib/Slice.h
+++ b/source/Lib/CommonLib/Slice.h
@@ -791,6 +791,8 @@ public:
class SPS;
+
+// Deprecated: SPSNext is going to be removed! Do not add any parameters to SPSNext
class SPSNext
{
private:
@@ -799,13 +801,19 @@ private:
bool m_NextEnabled;
//===== tool enabling flags (4 bytes - NOTE: last flag must be used for new extensions) =====
bool m_LargeCTU; // 5
- bool m_SubPuMvp;
bool m_IMV; // 9
- bool m_BIO;
bool m_DisableMotionCompression; // 13
bool m_LMChroma; // 17
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ bool m_cclmCollocatedChromaFlag;
+#endif
+#if JVET_M0464_UNI_MTS
+ bool m_IntraMTS; // 18
+ bool m_InterMTS; // 19
+#else
bool m_IntraEMT; // 18
bool m_InterEMT; // 19
+#endif
bool m_Affine;
bool m_AffineType;
bool m_GBi; //
@@ -828,18 +836,13 @@ public:
private:
//===== additional parameters =====
// qtbt
- // sub-pu merging
- int m_subPuMrgMode;
//imv
ImvMode m_ImvMode;
// multi type tree (QTBT + triple split)
unsigned m_MTTMode;
- bool m_compositeRefEnabled; //composite longterm reference
-
- unsigned m_CPRMode;
-
- // ADD_NEW_TOOL : (sps extension) add tool enabling flags and associated parameters here
+ bool m_compositeRefEnabled; //composite longterm reference
+ unsigned m_IBCMode;
public:
SPSNext( SPS& sps );
@@ -852,17 +855,12 @@ public:
//===== tool enabling flags and extension bit =====
void setUseLargeCTU ( bool b ) { m_LargeCTU = b; }
bool getUseLargeCTU () const { return m_LargeCTU; }
- bool getUseSubPuMvp() const { return m_SubPuMvp; }
- void setSubPuMvpMode(int n) { m_subPuMrgMode = n; m_SubPuMvp = n != 0; }
- bool getUseATMVP() const { return (m_subPuMrgMode & 1) == 1; }
void setUseIMV ( bool b ) { m_IMV = b; }
bool getUseIMV () const { return m_IMV; }
void setUseAffine ( bool b ) { m_Affine = b; }
bool getUseAffine () const { return m_Affine; }
void setUseAffineType ( bool b ) { m_AffineType = b; }
bool getUseAffineType () const { return m_AffineType; }
- void setUseBIO(bool b) { m_BIO = b; }
- bool getUseBIO() const { return m_BIO; }
void setDisableMotCompress ( bool b ) { m_DisableMotionCompression = b; }
bool getDisableMotCompress () const { return m_DisableMotionCompression; }
bool getMTTEnabled () const { return m_MTTEnabled; }
@@ -872,10 +870,21 @@ public:
#endif
void setUseLMChroma ( bool b ) { m_LMChroma = b; }
bool getUseLMChroma () const { return m_LMChroma; }
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ void setCclmCollocatedChromaFlag( bool b ) { m_cclmCollocatedChromaFlag = b; }
+ bool getCclmCollocatedChromaFlag() const { return m_cclmCollocatedChromaFlag; }
+#endif
+#if JVET_M0464_UNI_MTS
+ void setUseIntraMTS ( bool b ) { m_IntraMTS = b; }
+ bool getUseIntraMTS () const { return m_IntraMTS; }
+ void setUseInterMTS ( bool b ) { m_InterMTS = b; }
+ bool getUseInterMTS () const { return m_InterMTS; }
+#else
void setUseIntraEMT ( bool b ) { m_IntraEMT = b; }
bool getUseIntraEMT () const { return m_IntraEMT; }
void setUseInterEMT ( bool b ) { m_InterEMT = b; }
bool getUseInterEMT () const { return m_InterEMT; }
+#endif
void setUseGBi ( bool b ) { m_GBi = b; }
bool getUseGBi () const { return m_GBi; }
#if LUMA_ADAPTIVE_DEBLOCKING_FILTER_QP_OFFSET
@@ -894,8 +903,6 @@ public:
void setImvMode(ImvMode m) { m_ImvMode = m; m_IMV = m != 0; }
ImvMode getImvMode () const { return m_ImvMode; }
-
-
// multi type tree
unsigned getMTTMode () const { return m_MTTMode; }
void setMTTMode ( unsigned mode ) { m_MTTMode = mode; m_MTTEnabled = ( m_MTTMode != 0 ); }
@@ -907,10 +914,8 @@ public:
bool getUseMHIntra () const { return m_MHIntra; }
void setUseTriangle ( bool b ) { m_Triangle = b; }
bool getUseTriangle () const { return m_Triangle; }
- void setCPRMode (unsigned CPRMode) { m_CPRMode = CPRMode; }
- unsigned getCPRMode () const { return m_CPRMode; }
- // ADD_NEW_TOOL : (sps extension) add access functions for tool enabling flags and associated parameters here
-
+ void setIBCMode (unsigned IBCMode) { m_IBCMode = IBCMode; }
+ unsigned getIBCMode () const { return m_IBCMode; }
};
@@ -973,10 +978,9 @@ private:
uint32_t m_uiQuadtreeTULog2MinSize;
uint32_t m_uiQuadtreeTUMaxDepthInter;
uint32_t m_uiQuadtreeTUMaxDepthIntra;
- bool m_usePCM;
+ bool m_pcmEnabledFlag;
uint32_t m_pcmLog2MaxSize;
uint32_t m_uiPCMLog2MinSize;
- bool m_useAMP;
// Parameter
BitDepths m_bitDepths;
@@ -984,6 +988,9 @@ private:
int m_pcmBitDepths[MAX_NUM_CHANNEL_TYPE];
bool m_bPCMFilterDisableFlag;
+ bool m_sbtmvpEnabledFlag;
+ bool m_bdofEnabledFlag;
+
uint32_t m_uiBitsForPOC;
uint32_t m_numLongTermRefPicSPS;
uint32_t m_ltRefPicPocLsbSps[MAX_NUM_LONG_TERM_REF_PICS];
@@ -991,7 +998,7 @@ private:
// Max physical transform size
uint32_t m_uiMaxTrSize;
- bool m_bUseSAO;
+ bool m_saoEnabledFlag;
bool m_bTemporalIdNestingFlag; // temporal_id_nesting_flag
@@ -1017,9 +1024,9 @@ private:
static const int m_winUnitY[NUM_CHROMA_FORMAT];
PTL m_pcPTL;
- bool m_useALF;
+ bool m_alfEnabledFlag;
- bool m_useWrapAround;
+ bool m_wrapAroundEnabledFlag;
unsigned m_wrapAroundOffset;
public:
@@ -1135,16 +1142,14 @@ public:
uint32_t getMaxCUHeight() const { return m_uiMaxCUHeight; }
void setMaxCodingDepth( uint32_t u ) { m_uiMaxCodingDepth = u; }
uint32_t getMaxCodingDepth() const { return m_uiMaxCodingDepth; }
- void setUsePCM( bool b ) { m_usePCM = b; }
- bool getUsePCM() const { return m_usePCM; }
+ void setPCMEnabledFlag( bool b ) { m_pcmEnabledFlag = b; }
+ bool getPCMEnabledFlag() const { return m_pcmEnabledFlag; }
void setPCMLog2MaxSize( uint32_t u ) { m_pcmLog2MaxSize = u; }
uint32_t getPCMLog2MaxSize() const { return m_pcmLog2MaxSize; }
void setPCMLog2MinSize( uint32_t u ) { m_uiPCMLog2MinSize = u; }
uint32_t getPCMLog2MinSize() const { return m_uiPCMLog2MinSize; }
void setBitsForPOC( uint32_t u ) { m_uiBitsForPOC = u; }
uint32_t getBitsForPOC() const { return m_uiBitsForPOC; }
- bool getUseAMP() const { return m_useAMP; }
- void setUseAMP( bool b ) { m_useAMP = b; }
void setQuadtreeTULog2MaxSize( uint32_t u ) { m_uiQuadtreeTULog2MaxSize = u; }
uint32_t getQuadtreeTULog2MaxSize() const { return m_uiQuadtreeTULog2MaxSize; }
void setQuadtreeTULog2MinSize( uint32_t u ) { m_uiQuadtreeTULog2MinSize = u; }
@@ -1176,11 +1181,20 @@ public:
int getQpBDOffset(ChannelType type) const { return m_qpBDOffset[type]; }
void setQpBDOffset(ChannelType type, int i) { m_qpBDOffset[type] = i; }
- void setUseSAO(bool bVal) { m_bUseSAO = bVal; }
- bool getUseSAO() const { return m_bUseSAO; }
+ void setSAOEnabledFlag(bool bVal) { m_saoEnabledFlag = bVal; }
+ bool getSAOEnabledFlag() const { return m_saoEnabledFlag; }
+
+ bool getALFEnabledFlag() const { return m_alfEnabledFlag; }
+ void setALFEnabledFlag( bool b ) { m_alfEnabledFlag = b; }
+
+ bool getSBTMVPEnabledFlag() const { return m_sbtmvpEnabledFlag; }
+ void setSBTMVPEnabledFlag(bool b) { m_sbtmvpEnabledFlag = b; }
+
+ void setBDOFEnabledFlag(bool b) { m_bdofEnabledFlag = b; }
+ bool getBDOFEnabledFlag() const { return m_bdofEnabledFlag; }
- uint32_t getMaxTLayers() const { return m_uiMaxTLayers; }
- void setMaxTLayers( uint32_t uiMaxTLayers ) { CHECK( uiMaxTLayers > MAX_TLAYER, "Invalid number T-layers" ); m_uiMaxTLayers = uiMaxTLayers; }
+ uint32_t getMaxTLayers() const { return m_uiMaxTLayers; }
+ void setMaxTLayers( uint32_t uiMaxTLayers ) { CHECK( uiMaxTLayers > MAX_TLAYER, "Invalid number T-layers" ); m_uiMaxTLayers = uiMaxTLayers; }
bool getTemporalIdNestingFlag() const { return m_bTemporalIdNestingFlag; }
void setTemporalIdNestingFlag( bool bValue ) { m_bTemporalIdNestingFlag = bValue; }
@@ -1220,11 +1234,8 @@ public:
const SPSNext& getSpsNext() const { return m_spsNextExtension; }
SPSNext& getSpsNext() { return m_spsNextExtension; }
- bool getUseALF() const { return m_useALF; }
- void setUseALF( bool b ) { m_useALF = b; }
-
- void setUseWrapAround(bool b) { m_useWrapAround = b; }
- bool getUseWrapAround() const { return m_useWrapAround; }
+ void setWrapAroundEnabledFlag(bool b) { m_wrapAroundEnabledFlag = b; }
+ bool getWrapAroundEnabledFlag() const { return m_wrapAroundEnabledFlag; }
void setWrapAroundOffset(unsigned offset) { m_wrapAroundOffset = offset; }
unsigned getWrapAroundOffset() const { return m_wrapAroundOffset; }
};
@@ -1561,6 +1572,11 @@ private:
#endif
bool m_bCheckLDC;
+#if JVET_M0444_SMVD
+ bool m_biDirPred;
+ int m_symRefIdx[2];
+#endif
+
// Data
int m_iSliceQpDelta;
int m_iSliceChromaQpDelta[MAX_NUM_COMPONENT];
@@ -1649,7 +1665,10 @@ private:
AlfSliceParam m_alfSliceParam;
LutMotionCand* m_MotionCandLut;
-
+#if JVET_M0170_MRG_SHARELIST
+public:
+ LutMotionCand* m_MotionCandLuTsBkup;
+#endif
public:
Slice();
virtual ~Slice();
@@ -1749,6 +1768,12 @@ public:
void setCheckLDC( bool b ) { m_bCheckLDC = b; }
void setMvdL1ZeroFlag( bool b) { m_bLMvdL1Zero = b; }
+#if JVET_M0444_SMVD
+ void setBiDirPred( bool b, int refIdx0, int refIdx1 ) { m_biDirPred = b; m_symRefIdx[0] = refIdx0; m_symRefIdx[1] = refIdx1; }
+ bool getBiDirPred() const { return m_biDirPred; }
+ int getSymRefIdx( int refList ) const { return m_symRefIdx[refList]; }
+#endif
+
bool isIntra() const { return m_eSliceType == I_SLICE; }
bool isInterB() const { return m_eSliceType == B_SLICE; }
bool isInterP() const { return m_eSliceType == P_SLICE; }
@@ -1925,6 +1950,10 @@ public:
void destroyMotionLUTs ();
void resetMotionLUTs();
int getAvailableLUTMrgNum() const { return m_MotionCandLut->currCnt; }
+#if JVET_M0170_MRG_SHARELIST
+ int getAvailableLUTBkupMrgNum() const { return m_MotionCandLuTsBkup->currCnt; }
+ MotionInfo getMotionInfoFromLUTBkup(int MotCandIdx) const;
+#endif
MotionInfo getMotionInfoFromLUTs(int MotCandIdx) const;
LutMotionCand* getMotionLUTs() { return m_MotionCandLut; }
diff --git a/source/Lib/CommonLib/TrQuant.cpp b/source/Lib/CommonLib/TrQuant.cpp
index 92276925911e26cfe517cc9f236c50e32c3fc62b..f2afd109dbc34b8db435b25fb24882642c21634b 100644
--- a/source/Lib/CommonLib/TrQuant.cpp
+++ b/source/Lib/CommonLib/TrQuant.cpp
@@ -91,7 +91,13 @@ TrQuant::TrQuant() : m_quant( nullptr )
{
// allocate temporary buffers
m_plTempCoeff = (TCoeff*) xMalloc( TCoeff, MAX_CU_SIZE * MAX_CU_SIZE );
-
+#if JVET_M0464_UNI_MTS
+ m_mtsCoeffs = new TCoeff*[ NUM_TRAFO_MODES_MTS ];
+ for( int i = 0; i < NUM_TRAFO_MODES_MTS; i++ )
+ {
+ m_mtsCoeffs[i] = (TCoeff*) xMalloc( TCoeff, MAX_CU_SIZE * MAX_CU_SIZE );
+ }
+#endif
}
TrQuant::~TrQuant()
@@ -108,6 +114,17 @@ TrQuant::~TrQuant()
xFree( m_plTempCoeff );
m_plTempCoeff = nullptr;
}
+#if JVET_M0464_UNI_MTS
+ if( m_mtsCoeffs )
+ {
+ for( int i = 0; i < NUM_TRAFO_MODES_MTS; i++ )
+ {
+ xFree( m_mtsCoeffs[i] );
+ m_mtsCoeffs[i] = nullptr;
+ }
+ m_mtsCoeffs = nullptr;
+ }
+#endif
}
#if ENABLE_SPLIT_PARALLELISM
@@ -187,8 +204,11 @@ void TrQuant::invTransformNxN( TransformUnit &tu, const ComponentID &compID, Pel
DTRACE_COEFF_BUF( D_TCOEFF, tempCoeff, tu, tu.cu->predMode, compID );
-
+#if JVET_M0464_UNI_MTS
+ if( tu.mtsIdx == 1 )
+#else
if( tu.transformSkip[compID] )
+#endif
{
xITransformSkip( tempCoeff, pResi, tu, compID );
}
@@ -207,7 +227,11 @@ void TrQuant::invRdpcmNxN(TransformUnit& tu, const ComponentID &compID, PelBuf &
{
const CompArea &area = tu.blocks[compID];
+#if JVET_M0464_UNI_MTS
+ if (CU::isRDPCMEnabled(*tu.cu) && (tu.mtsIdx==1 || tu.cu->transQuantBypass))
+#else
if (CU::isRDPCMEnabled(*tu.cu) && ((tu.transformSkip[compID] != 0) || tu.cu->transQuantBypass))
+#endif
{
const uint32_t uiWidth = area.width;
const uint32_t uiHeight = area.height;
@@ -267,19 +291,34 @@ void TrQuant::invRdpcmNxN(TransformUnit& tu, const ComponentID &compID, PelBuf &
void TrQuant::getTrTypes ( TransformUnit tu, const ComponentID compID, int &trTypeHor, int &trTypeVer )
{
+#if JVET_M0464_UNI_MTS
+ bool mtsActivated = CU::isIntra( *tu.cu ) ? tu.cs->sps->getSpsNext().getUseIntraMTS() : tu.cs->sps->getSpsNext().getUseInterMTS();
+#else
bool emtActivated = CU::isIntra( *tu.cu ) ? tu.cs->sps->getSpsNext().getUseIntraEMT() : tu.cs->sps->getSpsNext().getUseInterEMT();
+#endif
trTypeHor = DCT2;
trTypeVer = DCT2;
+#if JVET_M0464_UNI_MTS
+ if ( mtsActivated )
+#else
if ( emtActivated )
+#endif
{
if( compID == COMPONENT_Y )
{
+#if JVET_M0464_UNI_MTS
+ if ( tu.mtsIdx > 1 )
+ {
+ int indHor = ( tu.mtsIdx - 2 ) & 1;
+ int indVer = ( tu.mtsIdx - 2 ) >> 1;
+#else
if ( tu.cu->emtFlag )
{
int indHor = tu.emtIdx & 1;
int indVer = tu.emtIdx >> 1;
+#endif
trTypeHor = indHor ? DCT8 : DST7;
trTypeVer = indVer ? DCT8 : DST7;
@@ -437,7 +476,67 @@ void TrQuant::xQuant(TransformUnit &tu, const ComponentID &compID, const CCoeffB
m_quant->quant( tu, compID, pSrc, uiAbsSum, cQP, ctx );
}
+#if JVET_M0464_UNI_MTS
+void TrQuant::transformNxN(TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, std::vector<TrMode>* trModes, const int maxCand)
+{
+ CodingStructure &cs = *tu.cs;
+ const SPS &sps = *cs.sps;
+ const CompArea &rect = tu.blocks[compID];
+ const uint32_t width = rect.width;
+ const uint32_t height = rect.height;
+
+ const CPelBuf resiBuf = cs.getResiBuf(rect);
+
+ CHECK( sps.getMaxTrSize() < width, "Unsupported transformation size" );
+
+ int pos = 0;
+ std::vector<TrCost> trCosts;
+ std::vector<TrMode>::iterator it = trModes->begin();
+ const double facBB[] = { 1.2, 1.3, 1.3, 1.4, 1.5 };
+ while( it != trModes->end() )
+ {
+ tu.mtsIdx = it->first;
+ CoeffBuf tempCoeff( m_mtsCoeffs[tu.mtsIdx], rect );
+
+ if( tu.mtsIdx == 1 )
+ {
+ xTransformSkip( tu, compID, resiBuf, tempCoeff.buf );
+ }
+ else
+ {
+ xT( tu, compID, resiBuf, tempCoeff, width, height );
+ }
+
+ int sumAbs = 0;
+ for( int pos = 0; pos < width*height; pos++ )
+ {
+ sumAbs += abs( tempCoeff.buf[pos] );
+ }
+
+ trCosts.push_back( TrCost( sumAbs, pos++ ) );
+ it++;
+ }
+
+ int numTests = 0;
+ std::vector<TrCost>::iterator itC = trCosts.begin();
+ const double fac = facBB[g_aucLog2[std::max(width, height)]-2];
+ const double thr = fac * trCosts.begin()->first;
+ const double thrTS = trCosts.begin()->first;
+ while( itC != trCosts.end() )
+ {
+ const bool testTr = itC->first <= ( itC->second == 1 ? thrTS : thr ) && numTests <= maxCand;
+ trModes->at( itC->second ).second = testTr;
+ numTests += testTr;
+ itC++;
+ }
+}
+#endif
+
+#if JVET_M0464_UNI_MTS
+void TrQuant::transformNxN(TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, const Ctx &ctx, const bool loadTr)
+#else
void TrQuant::transformNxN(TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, const Ctx &ctx)
+#endif
{
CodingStructure &cs = *tu.cs;
const SPS &sps = *cs.sps;
@@ -483,11 +582,21 @@ void TrQuant::transformNxN(TransformUnit &tu, const ComponentID &compID, const Q
{
CHECK( sps.getMaxTrSize() < uiWidth, "Unsupported transformation size" );
+#if JVET_M0464_UNI_MTS
+ CoeffBuf tempCoeff( loadTr ? m_mtsCoeffs[tu.mtsIdx] : m_plTempCoeff, rect );
+#else
CoeffBuf tempCoeff( m_plTempCoeff, rect );
+#endif
DTRACE_PEL_BUF( D_RESIDUALS, resiBuf, tu, tu.cu->predMode, compID );
+#if JVET_M0464_UNI_MTS
+ if( !loadTr )
+ {
+ if( tu.mtsIdx == 1 )
+#else
if( tu.transformSkip[compID] )
+#endif
{
xTransformSkip( tu, compID, resiBuf, tempCoeff.buf );
}
@@ -495,6 +604,9 @@ void TrQuant::transformNxN(TransformUnit &tu, const ComponentID &compID, const Q
{
xT( tu, compID, resiBuf, tempCoeff, uiWidth, uiHeight );
}
+#if JVET_M0464_UNI_MTS
+ }
+#endif
DTRACE_COEFF_BUF( D_TCOEFF, tempCoeff, tu, tu.cu->predMode, compID );
@@ -567,7 +679,11 @@ void TrQuant::applyForwardRDPCM(TransformUnit &tu, const ComponentID &compID, co
void TrQuant::rdpcmNxN(TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, RDPCMMode &rdpcmMode)
{
+#if JVET_M0464_UNI_MTS
+ if (!CU::isRDPCMEnabled(*tu.cu) || (tu.mtsIdx!=1 && !tu.cu->transQuantBypass))
+#else
if (!CU::isRDPCMEnabled(*tu.cu) || (!tu.transformSkip[compID] && !tu.cu->transQuantBypass))
+#endif
{
rdpcmMode = RDPCM_OFF;
}
diff --git a/source/Lib/CommonLib/TrQuant.h b/source/Lib/CommonLib/TrQuant.h
index c663f9e76ab738e0139daf5024b520fa9843cd3f..aa58a29981b7f6a8ac09601972a487a8811cc6f5 100644
--- a/source/Lib/CommonLib/TrQuant.h
+++ b/source/Lib/CommonLib/TrQuant.h
@@ -87,7 +87,12 @@ public:
void invTransformNxN (TransformUnit &tu, const ComponentID &compID, PelBuf &pResi, const QpParam &cQPs);
+#if JVET_M0464_UNI_MTS
+ void transformNxN (TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, std::vector<TrMode>* trModes, const int maxCand);
+ void transformNxN (TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, const Ctx &ctx, const bool loadTr=false);
+#else
void transformNxN (TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, const Ctx &ctx);
+#endif
void rdpcmNxN (TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, RDPCMMode &rdpcmMode);
void applyForwardRDPCM(TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, const RDPCMMode &rdpcmMode);
@@ -122,6 +127,9 @@ protected:
private:
Quant *m_quant; //!< Quantizer
+#if JVET_M0464_UNI_MTS
+ TCoeff** m_mtsCoeffs;
+#endif
// forward Transform
diff --git a/source/Lib/CommonLib/TrQuant_EMT.cpp b/source/Lib/CommonLib/TrQuant_EMT.cpp
index e717c6c2466c8997d5d700479b2e2cc5a0d3a628..3ea8ce7306e663478e28f20dfac2a5251e1b70ba 100644
--- a/source/Lib/CommonLib/TrQuant_EMT.cpp
+++ b/source/Lib/CommonLib/TrQuant_EMT.cpp
@@ -947,23 +947,314 @@ void fastInverseDST7_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int
void fastForwardDST7_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
{
+#if JVET_M0497_FAST_DST7 && !JVET_M0497_MATRIX_MULT
+ int j, k;
+ TCoeff a[5], b[5], c[5], d[5], t;
+ TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
+
+ const TMatrixCoeff *iT = g_trCoreDST7P16[TRANSFORM_FORWARD][0];
+
+ TCoeff *pCoef = dst;
+ const int reducedLine = line - iSkipLine;
+ const int cutoff = 16 - iSkipLine2;
+
+ for (j = 0; j < reducedLine; j++)
+ {
+ for (k = 0; k < 5; k++)
+ {
+ a[k] = src[ k] + src[11 + k];
+ b[k] = src[9 - k] + src[11 + k];
+ c[k] = src[ k] - src[ 9 - k];
+ d[k] = src[ k] + src[ 9 - k] - src[11 + k];
+ }
+
+ t = iT[10] * src[10];
+
+ dst[ 1 * line] = ( iT[ 2]*d[0] + iT[ 5]*d[1] + iT[ 8]*d[2] + iT[11]*d[3] + iT[14]*d[4] + add) >> shift;
+ dst[ 4 * line] = ( iT[ 8]*d[0] + iT[14]*d[1] + iT[ 5]*d[2] - iT[ 2]*d[3] - iT[11]*d[4] + add) >> shift;
+ dst[ 7 * line] = ( iT[14]*d[0] + iT[ 2]*d[1] - iT[11]*d[2] - iT[ 5]*d[3] + iT[ 8]*d[4] + add) >> shift;
+ dst[10 * line] = ( iT[11]*d[0] - iT[ 8]*d[1] - iT[ 2]*d[2] + iT[14]*d[3] - iT[ 5]*d[4] + add) >> shift;
+ dst[13 * line] = ( iT[ 5]*d[0] - iT[11]*d[1] + iT[14]*d[2] - iT[ 8]*d[3] + iT[ 2]*d[4] + add) >> shift;
+
+ dst[5 * line] = ( iT[10] * (src[0] + src[1] - src[3] - src[4] + src[6] + src[7] - src[9] - src[10] + src[12] + src[13] - src[15]) + add) >> shift;
+
+ dst[ 0 * line] = ( iT[0]*a[0] + iT[9]*b[0] + iT[1]*a[1] + iT[8]*b[1] + iT[2]*a[2] + iT[7]*b[2] + iT[3]*a[3] + iT[6]*b[3] + iT[4]*a[4] + iT[5]*b[4] + t + add ) >> shift;
+ dst[ 2 * line] = ( iT[4]*c[0] - iT[5]*b[0] + iT[9]*c[1] - iT[0]*b[1] + iT[6]*c[2] + iT[3]*a[2] + iT[1]*c[3] + iT[8]*a[3] + iT[7]*a[4] + iT[2]*b[4] - t + add ) >> shift;
+ dst[ 3 * line] = ( iT[6]*a[0] + iT[3]*b[0] + iT[2]*c[1] + iT[7]*a[1] + iT[9]*c[2] + iT[0]*a[2] + iT[4]*c[3] - iT[5]*b[3] - iT[1]*a[4] - iT[8]*b[4] + t + add ) >> shift;
+ dst[ 6 * line] = ( iT[8]*a[0] + iT[1]*c[0] + iT[6]*c[1] - iT[3]*b[1] - iT[5]*a[2] - iT[4]*b[2] - iT[7]*c[3] - iT[2]*a[3] - iT[0]*c[4] + iT[9]*b[4] + t + add ) >> shift;
+ dst[ 8 * line] = ( iT[4]*c[0] + iT[5]*a[0] - iT[0]*c[1] + iT[9]*b[1] - iT[3]*c[2] - iT[6]*a[2] + iT[1]*c[3] - iT[8]*b[3] + iT[2]*c[4] + iT[7]*a[4] - t + add ) >> shift;
+ dst[ 9 * line] = ( iT[7]*c[0] + iT[2]*a[0] - iT[4]*a[1] - iT[5]*b[1] - iT[8]*c[2] + iT[1]*b[2] + iT[9]*a[3] + iT[0]*b[3] + iT[3]*c[4] - iT[6]*b[4] + t + add ) >> shift;
+ dst[11 * line] = ( iT[9]*a[0] + iT[0]*b[0] - iT[8]*c[1] - iT[1]*a[1] + iT[2]*c[2] - iT[7]*b[2] + iT[6]*a[3] + iT[3]*b[3] - iT[5]*c[4] - iT[4]*a[4] - t + add ) >> shift;
+ dst[12 * line] = ( iT[7]*c[0] - iT[2]*b[0] - iT[5]*c[1] - iT[4]*a[1] + iT[8]*a[2] + iT[1]*b[2] - iT[0]*a[3] - iT[9]*b[3] - iT[6]*c[4] + iT[3]*b[4] + t + add ) >> shift;
+ dst[14 * line] = ( iT[3]*a[0] + iT[6]*b[0] - iT[7]*a[1] - iT[2]*b[1] + iT[0]*c[2] + iT[9]*a[2] - iT[4]*c[3] - iT[5]*a[3] + iT[8]*c[4] + iT[1]*a[4] - t + add ) >> shift;
+ dst[15 * line] = ( iT[1]*c[0] - iT[8]*b[0] - iT[3]*c[1] + iT[6]*b[1] + iT[5]*c[2] - iT[4]*b[2] - iT[7]*c[3] + iT[2]*b[3] + iT[9]*c[4] - iT[0]*b[4] + t + add ) >> shift;
+
+ src += 16;
+ dst++;
+ }
+
+ if (iSkipLine)
+ {
+ dst = pCoef + reducedLine;
+ for (j = 0; j < cutoff; j++)
+ {
+ memset(dst, 0, sizeof(TCoeff)*iSkipLine);
+ dst += line;
+ }
+ }
+
+ if (iSkipLine2)
+ {
+ dst = pCoef + line * cutoff;
+ memset(dst, 0, sizeof(TCoeff) * line * iSkipLine2);
+ }
+#else
_fastForwardMM< 16 >( src, dst, shift, line, iSkipLine, iSkipLine2, g_trCoreDST7P16[TRANSFORM_FORWARD][0] );
+#endif
}
void fastInverseDST7_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
{
+#if JVET_M0497_FAST_DST7 && !JVET_M0497_MATRIX_MULT
+ int j, k;
+ TCoeff a[5], b[5], c[5], d[5], t;
+
+ TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
+
+ const TMatrixCoeff *iT = g_trCoreDST7P16[TRANSFORM_INVERSE][0];
+
+ const int reducedLine = line - iSkipLine;
+
+ for (j = 0; j < reducedLine; j++)
+ {
+ for (k = 0; k < 5; k++)
+ {
+ a[k] = src[ k * line] + src[(10 - k) * line];
+ b[k] = src[(11 + k) * line] + src[(10 - k) * line];
+ c[k] = src[ k * line] - src[(11 + k) * line];
+ d[k] = src[ k * line] + src[(11 + k) * line] - src[(10 - k)*line];
+ }
+
+ t = iT[10] * src[5 * line];
+
+ dst[ 2] = Clip3(outputMinimum, outputMaximum, (int)( iT[ 2]*d[0] + iT[ 8]*d[1] + iT[14]*d[2] + iT[11]*d[3] + iT[ 5]*d[4] + add ) >> shift);
+ dst[ 5] = Clip3(outputMinimum, outputMaximum, (int)( iT[ 5]*d[0] + iT[14]*d[1] + iT[ 2]*d[2] - iT[ 8]*d[3] - iT[11]*d[4] + add ) >> shift);
+ dst[ 8] = Clip3(outputMinimum, outputMaximum, (int)( iT[ 8]*d[0] + iT[ 5]*d[1] - iT[11]*d[2] - iT[ 2]*d[3] + iT[14]*d[4] + add ) >> shift);
+ dst[11] = Clip3(outputMinimum, outputMaximum, (int)( iT[11]*d[0] - iT[ 2]*d[1] - iT[ 5]*d[2] + iT[14]*d[3] - iT[ 8]*d[4] + add ) >> shift);
+ dst[14] = Clip3(outputMinimum, outputMaximum, (int)( iT[14]*d[0] - iT[11]*d[1] + iT[ 8]*d[2] - iT[ 5]*d[3] + iT[ 2]*d[4] + add ) >> shift);
+
+ dst[10] = Clip3(outputMinimum, outputMaximum, (int)( iT[10]*(src[ 0*line]-src[ 2*line]+src[ 3*line]-src[5*line]
+ +src[ 6*line]-src[ 8*line]+src[ 9*line]-src[11*line]
+ +src[12*line]-src[14*line]+src[15*line]) + add ) >> shift);
+
+ dst[ 0] = Clip3(outputMinimum, outputMaximum, (int)( iT[0]*a[0] + iT[9]*b[0] + iT[2]*a[1] + iT[7]*b[1] + iT[4]*a[2] + iT[5]*b[2] + iT[6]*a[3] + iT[3]*b[3] + iT[8]*a[4] + iT[1]*b[4] + t + add ) >> shift);
+ dst[ 1] = Clip3(outputMinimum, outputMaximum, (int)( iT[1]*c[0] - iT[8]*b[0] + iT[5]*c[1] - iT[4]*b[1] + iT[9]*c[2] - iT[0]*b[2] + iT[2]*a[3] + iT[7]*c[3] + iT[6]*a[4] + iT[3]*c[4] + t + add ) >> shift);
+ dst[ 3] = Clip3(outputMinimum, outputMaximum, (int)( iT[3]*a[0] + iT[6]*b[0] + iT[0]*c[1] + iT[9]*a[1] + iT[1]*a[2] + iT[8]*c[2] + iT[4]*c[3] - iT[5]*b[3] - iT[2]*a[4] - iT[7]*b[4] - t + add ) >> shift);
+ dst[ 4] = Clip3(outputMinimum, outputMaximum, (int)( iT[4]*c[0] - iT[5]*b[0] + iT[6]*c[1] + iT[3]*a[1] + iT[7]*a[2] + iT[2]*b[2] - iT[1]*c[3] + iT[8]*b[3] - iT[9]*c[4] - iT[0]*a[4] - t + add ) >> shift);
+ dst[ 6] = Clip3(outputMinimum, outputMaximum, (int)( iT[6]*a[0] + iT[3]*b[0] + iT[9]*c[1] + iT[0]*a[1] - iT[1]*a[2] - iT[8]*b[2] - iT[4]*c[3] - iT[5]*a[3] - iT[2]*c[4] + iT[7]*b[4] + t + add ) >> shift);
+ dst[ 7] = Clip3(outputMinimum, outputMaximum, (int)( iT[7]*c[0] - iT[2]*b[0] + iT[8]*a[1] + iT[1]*b[1] - iT[6]*c[2] + iT[3]*b[2] - iT[9]*a[3] - iT[0]*b[3] + iT[5]*c[4] - iT[4]*b[4] + t + add ) >> shift);
+ dst[ 9] = Clip3(outputMinimum, outputMaximum, (int)( iT[9]*a[0] + iT[0]*b[0] + iT[2]*c[1] - iT[7]*b[1] - iT[5]*c[2] - iT[4]*a[2] + iT[3]*a[3] + iT[6]*b[3] + iT[8]*c[4] - iT[1]*b[4] - t + add ) >> shift);
+ dst[12] = Clip3(outputMinimum, outputMaximum, (int)( iT[1]*c[0] + iT[8]*a[0] - iT[5]*a[1] - iT[4]*b[1] - iT[0]*c[2] + iT[9]*b[2] + iT[7]*c[3] - iT[2]*b[3] - iT[6]*c[4] - iT[3]*a[4] + t + add ) >> shift);
+ dst[13] = Clip3(outputMinimum, outputMaximum, (int)( iT[7]*c[0] + iT[2]*a[0] - iT[8]*c[1] + iT[1]*b[1] + iT[3]*c[2] - iT[6]*b[2] + iT[0]*a[3] + iT[9]*b[3] - iT[5]*a[4] - iT[4]*b[4] + t + add ) >> shift);
+ dst[15] = Clip3(outputMinimum, outputMaximum, (int)( iT[4]*c[0] + iT[5]*a[0] - iT[3]*c[1] - iT[6]*a[1] + iT[2]*c[2] + iT[7]*a[2] - iT[1]*c[3] - iT[8]*a[3] + iT[0]*c[4] + iT[9]*a[4] - t + add ) >> shift);
+
+ src++;
+ dst += 16;
+ }
+
+ if (iSkipLine)
+ {
+ memset(dst, 0, (iSkipLine * 16) * sizeof(TCoeff));
+ }
+#else
_fastInverseMM< 16 >( src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_trCoreDST7P16[TRANSFORM_INVERSE][0]);
+#endif
}
void fastForwardDST7_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
{
+#if JVET_M0497_FAST_DST7 && !JVET_M0497_MATRIX_MULT
+ int j, k;
+ TCoeff a[10][6];
+ TCoeff t[2];
+ TCoeff b[6];
+ TCoeff c[2];
+
+ TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
+ const TMatrixCoeff *iT = g_trCoreDST7P32[TRANSFORM_FORWARD][0];
+ TCoeff *pCoef = dst;
+ const int reducedLine = line - iSkipLine;
+ const int cutoff = 32 - iSkipLine2;
+
+ for (j = 0; j < reducedLine; j++)
+ {
+ for (k = 0; k < 6; k++)
+ {
+ a[0][k] = src[ k] - src[11 - k];
+ a[1][k] = src[ k] + src[13 + k];
+ a[2][k] = src[ k] + src[24 - k];
+ a[3][k] = src[ k] - src[26 + k];
+ a[4][k] = src[ 6 + k] + src[18 - k];
+ a[5][k] = src[ 6 + k] + src[19 + k];
+ a[6][k] = src[ 6 + k] - src[31 - k];
+ a[7][k] = src[13 + k] - src[24 - k];
+ a[8][k] = src[13 + k] + src[26 + k];
+ a[9][k] = src[19 + k] + src[31 - k];
+
+ b[k] = src[k] + src[11 - k] - src[13 + k] - src[24 - k] + src[26 + k];
+ }
+ for (k = 0; k < 2; k++)
+ {
+ c[k] = src[k] + src[3 - k] - src[5 + k] - src[8 - k] + src[10 + k] + src[13 - k] - src[15 + k] - src[18 - k] + src[20 + k] + src[23 - k] - src[25 + k] - src[28 - k] + src[30 + k];
+ }
+
+ t[0] = iT[12] * src[12] + iT[25] * src[25];
+ t[1] = iT[12] * src[25] - iT[25] * src[12];
+
+ dst[ 0 * line] = ( iT[0] * a[3][0] + iT[11] * a[6][5] + iT[13] * a[8][0] + iT[24] * a[9][5] + iT[1] * a[3][1] + iT[10] * a[6][4] + iT[14] * a[8][1] + iT[23] * a[9][4] + iT[2] * a[3][2] + iT[9] * a[6][3] + iT[15] * a[8][2] + iT[22] * a[9][3] + iT[3] * a[3][3] + iT[8] * a[6][2] + iT[16] * a[8][3] + iT[21] * a[9][2] + iT[4] * a[3][4] + iT[7] * a[6][1] + iT[17] * a[8][4] + iT[20] * a[9][1] + iT[5] * a[3][5] + iT[6] * a[6][0] + iT[18] * a[8][5] + iT[19] * a[9][0] + t[0] + add) >> shift;
+ dst[ 1 * line] = (-iT[0] * a[5][2] + iT[11] * a[0][3] + iT[13] * a[4][2] + iT[24] * a[6][2] + iT[1] * a[9][1] + iT[10] * a[8][4] + iT[14] * a[3][4] + iT[23] * a[6][1] + iT[2] * a[0][0] - iT[9] * a[5][5] + iT[15] * a[6][5] + iT[22] * a[4][5] - iT[3] * a[5][3] + iT[8] * a[0][2] + iT[16] * a[4][3] + iT[21] * a[6][3] + iT[4] * a[9][0] + iT[7] * a[8][5] + iT[17] * a[3][5] + iT[20] * a[6][0] + iT[5] * a[0][1] - iT[6] * a[5][4] + iT[18] * a[6][4] + iT[19] * a[4][4] - t[1] + add) >> shift;
+ dst[ 3 * line] = (-iT[0] * a[9][4] - iT[11] * a[5][4] + iT[13] * a[2][1] - iT[24] * a[7][1] - iT[1] * a[0][3] - iT[10] * a[1][3] + iT[14] * a[3][3] + iT[23] * a[2][3] + iT[2] * a[8][5] + iT[9] * a[9][0] + iT[15] * a[6][0] + iT[22] * a[3][5] - iT[3] * a[1][4] - iT[8] * a[0][4] + iT[16] * a[2][4] + iT[21] * a[3][4] - iT[4] * a[5][3] - iT[7] * a[9][3] - iT[17] * a[7][2] + iT[20] * a[2][2] + iT[5] * a[8][0] + iT[6] * a[1][0] - iT[18] * a[4][5] - iT[19] * a[7][0] + t[1] + add) >> shift;
+ dst[ 4 * line] = (-iT[0] * a[3][2] - iT[11] * a[2][2] + iT[13] * a[1][2] + iT[24] * a[0][2] + iT[1] * a[6][0] + iT[10] * a[3][5] + iT[14] * a[9][0] + iT[23] * a[8][5] - iT[2] * a[2][3] - iT[9] * a[3][3] + iT[15] * a[0][3] + iT[22] * a[1][3] - iT[3] * a[7][0] + iT[8] * a[2][0] - iT[16] * a[9][5] - iT[21] * a[5][5] + iT[4] * a[4][4] + iT[7] * a[6][4] + iT[17] * a[0][1] - iT[20] * a[5][4] - iT[5] * a[7][4] - iT[6] * a[4][1] + iT[18] * a[8][4] + iT[19] * a[1][4] - t[0] + add) >> shift;
+ dst[ 5 * line] = (-iT[0] * a[3][5] - iT[11] * a[6][0] - iT[13] * a[8][5] - iT[24] * a[9][0] + iT[1] * a[6][5] + iT[10] * a[3][0] + iT[14] * a[9][5] + iT[23] * a[8][0] - iT[2] * a[7][4] + iT[9] * a[2][4] - iT[15] * a[9][1] - iT[22] * a[5][1] - iT[3] * a[7][1] - iT[8] * a[4][4] + iT[16] * a[8][1] + iT[21] * a[1][1] + iT[4] * a[6][2] + iT[7] * a[4][2] - iT[17] * a[5][2] + iT[20] * a[0][3] - iT[5] * a[3][2] - iT[6] * a[2][2] + iT[18] * a[1][2] + iT[19] * a[0][2] + t[0] + add) >> shift;
+ dst[ 8 * line] = ( iT[0] * a[9][3] + iT[11] * a[8][2] + iT[13] * a[3][2] + iT[24] * a[6][3] + iT[1] * a[1][5] + iT[10] * a[0][5] - iT[14] * a[2][5] - iT[23] * a[3][5] - iT[2] * a[1][3] - iT[9] * a[8][3] + iT[15] * a[7][3] + iT[22] * a[4][2] - iT[3] * a[9][5] - iT[8] * a[5][5] + iT[16] * a[2][0] - iT[21] * a[7][0] - iT[4] * a[1][1] - iT[7] * a[0][1] + iT[17] * a[2][1] + iT[20] * a[3][1] + iT[5] * a[5][1] + iT[6] * a[9][1] + iT[18] * a[7][4] - iT[19] * a[2][4] + t[1] + add) >> shift;
+ dst[ 9 * line] = (-iT[0] * a[2][1] - iT[11] * a[3][1] + iT[13] * a[0][1] + iT[24] * a[1][1] + iT[1] * a[7][3] - iT[10] * a[2][3] + iT[14] * a[9][2] + iT[23] * a[5][2] + iT[2] * a[4][0] + iT[9] * a[7][5] - iT[15] * a[1][5] - iT[22] * a[8][5] + iT[3] * a[3][4] + iT[8] * a[2][4] - iT[16] * a[1][4] - iT[21] * a[0][4] + iT[4] * a[6][3] + iT[7] * a[3][2] + iT[17] * a[9][3] + iT[20] * a[8][2] + iT[5] * a[4][5] + iT[6] * a[6][5] + iT[18] * a[0][0] - iT[19] * a[5][5] - t[0] + add) >> shift;
+ dst[10 * line] = (-iT[0] * a[6][1] - iT[11] * a[4][1] + iT[13] * a[5][1] - iT[24] * a[0][4] + iT[1] * a[2][2] - iT[10] * a[7][2] - iT[14] * a[5][3] - iT[23] * a[9][3] + iT[2] * a[6][4] + iT[9] * a[4][4] - iT[15] * a[5][4] + iT[22] * a[0][1] - iT[3] * a[2][5] + iT[8] * a[7][5] + iT[16] * a[5][0] + iT[21] * a[9][0] - iT[4] * a[7][0] - iT[7] * a[4][5] + iT[17] * a[8][0] + iT[20] * a[1][0] + iT[5] * a[4][2] + iT[6] * a[7][3] - iT[18] * a[1][3] - iT[19] * a[8][3] + t[0] + add) >> shift;
+ dst[11 * line] = ( iT[0] * a[1][3] + iT[11] * a[0][3] - iT[13] * a[2][3] - iT[24] * a[3][3] + iT[1] * a[9][1] + iT[10] * a[5][1] - iT[14] * a[2][4] + iT[23] * a[7][4] + iT[2] * a[8][0] + iT[9] * a[9][5] + iT[15] * a[6][5] + iT[22] * a[3][0] - iT[3] * a[0][2] + iT[8] * a[5][3] - iT[16] * a[6][3] - iT[21] * a[4][3] - iT[4] * a[5][0] + iT[7] * a[0][5] + iT[17] * a[4][0] + iT[20] * a[6][0] - iT[5] * a[9][4] - iT[6] * a[5][4] + iT[18] * a[2][1] - iT[19] * a[7][1] - t[1] + add) >> shift;
+ dst[13 * line] = (-iT[0] * a[0][0] - iT[11] * a[1][0] + iT[13] * a[3][0] + iT[24] * a[2][0] - iT[1] * a[5][4] + iT[10] * a[0][1] + iT[14] * a[4][4] + iT[23] * a[6][4] + iT[2] * a[9][3] + iT[9] * a[5][3] - iT[15] * a[2][2] + iT[22] * a[7][2] - iT[3] * a[8][3] - iT[8] * a[9][2] - iT[16] * a[6][2] - iT[21] * a[3][3] + iT[4] * a[1][4] + iT[7] * a[8][4] - iT[17] * a[7][4] - iT[20] * a[4][1] - iT[5] * a[0][5] - iT[6] * a[1][5] + iT[18] * a[3][5] + iT[19] * a[2][5] + t[1] + add) >> shift;
+ dst[14 * line] = ( iT[0] * a[4][2] + iT[11] * a[7][3] - iT[13] * a[1][3] - iT[24] * a[8][3] + iT[1] * a[4][1] + iT[10] * a[6][1] + iT[14] * a[0][4] - iT[23] * a[5][1] - iT[2] * a[3][0] - iT[9] * a[2][0] + iT[15] * a[1][0] + iT[22] * a[0][0] - iT[3] * a[6][3] - iT[8] * a[4][3] + iT[16] * a[5][3] - iT[21] * a[0][2] - iT[4] * a[7][5] - iT[7] * a[4][0] + iT[17] * a[8][5] + iT[20] * a[1][5] + iT[5] * a[6][4] + iT[6] * a[3][1] + iT[18] * a[9][4] + iT[19] * a[8][1] - t[0] + add) >> shift;
+ dst[15 * line] = (-iT[0] * a[7][4] - iT[11] * a[4][1] + iT[13] * a[8][4] + iT[24] * a[1][4] + iT[1] * a[2][2] + iT[10] * a[3][2] - iT[14] * a[0][2] - iT[23] * a[1][2] + iT[2] * a[2][1] - iT[9] * a[7][1] - iT[15] * a[5][4] - iT[22] * a[9][4] - iT[3] * a[7][5] + iT[8] * a[2][5] - iT[16] * a[9][0] - iT[21] * a[5][0] - iT[4] * a[2][0] - iT[7] * a[3][0] + iT[17] * a[0][0] + iT[20] * a[1][0] - iT[5] * a[2][3] + iT[6] * a[7][3] + iT[18] * a[5][2] + iT[19] * a[9][2] + t[0] + add) >> shift;
+ dst[16 * line] = (-iT[0] * a[0][1] + iT[11] * a[5][4] - iT[13] * a[6][4] - iT[24] * a[4][4] + iT[1] * a[0][3] - iT[10] * a[5][2] + iT[14] * a[6][2] + iT[23] * a[4][2] - iT[2] * a[0][5] + iT[9] * a[5][0] - iT[15] * a[6][0] - iT[22] * a[4][0] - iT[3] * a[0][4] - iT[8] * a[1][4] + iT[16] * a[3][4] + iT[21] * a[2][4] + iT[4] * a[0][2] + iT[7] * a[1][2] - iT[17] * a[3][2] - iT[20] * a[2][2] - iT[5] * a[0][0] - iT[6] * a[1][0] + iT[18] * a[3][0] + iT[19] * a[2][0] - t[1] + add) >> shift;
+ dst[18 * line] = ( iT[0] * a[0][5] + iT[11] * a[1][5] - iT[13] * a[3][5] - iT[24] * a[2][5] - iT[1] * a[1][0] - iT[10] * a[0][0] + iT[14] * a[2][0] + iT[23] * a[3][0] - iT[2] * a[5][1] + iT[9] * a[0][4] + iT[15] * a[4][1] + iT[22] * a[6][1] - iT[3] * a[8][1] - iT[8] * a[1][1] + iT[16] * a[4][4] + iT[21] * a[7][1] - iT[4] * a[9][2] - iT[7] * a[5][2] + iT[17] * a[2][3] - iT[20] * a[7][3] - iT[5] * a[9][3] - iT[6] * a[8][2] - iT[18] * a[3][2] - iT[19] * a[6][3] + t[1] + add) >> shift;
+ dst[20 * line] = (-iT[0] * a[4][0] - iT[11] * a[6][0] - iT[13] * a[0][5] + iT[24] * a[5][0] + iT[1] * a[6][5] + iT[10] * a[4][5] - iT[14] * a[5][5] + iT[23] * a[0][0] - iT[2] * a[6][1] - iT[9] * a[3][4] - iT[15] * a[9][1] - iT[22] * a[8][4] + iT[3] * a[4][4] + iT[8] * a[7][1] - iT[16] * a[1][1] - iT[21] * a[8][1] - iT[4] * a[3][3] - iT[7] * a[2][3] + iT[17] * a[1][3] + iT[20] * a[0][3] + iT[5] * a[7][2] - iT[6] * a[2][2] + iT[18] * a[9][3] + iT[19] * a[5][3] + t[0] + add) >> shift;
+ dst[21 * line] = (-iT[0] * a[1][2] - iT[11] * a[8][2] + iT[13] * a[7][2] + iT[24] * a[4][3] - iT[1] * a[1][5] - iT[10] * a[8][5] + iT[14] * a[7][5] + iT[23] * a[4][0] - iT[2] * a[5][2] - iT[9] * a[9][2] - iT[15] * a[7][3] + iT[22] * a[2][3] - iT[3] * a[5][5] - iT[8] * a[9][5] - iT[16] * a[7][0] + iT[21] * a[2][0] - iT[4] * a[8][1] - iT[7] * a[9][4] - iT[17] * a[6][4] - iT[20] * a[3][1] - iT[5] * a[8][4] - iT[6] * a[9][1] - iT[18] * a[6][1] - iT[19] * a[3][4] - t[1] + add) >> shift;
+ dst[23 * line] = (-iT[0] * a[8][4] - iT[11] * a[9][1] - iT[13] * a[6][1] - iT[24] * a[3][4] + iT[1] * a[8][2] + iT[10] * a[1][2] - iT[14] * a[4][3] - iT[23] * a[7][2] + iT[2] * a[0][1] + iT[9] * a[1][1] - iT[15] * a[3][1] - iT[22] * a[2][1] - iT[3] * a[5][0] - iT[8] * a[9][0] - iT[16] * a[7][5] + iT[21] * a[2][5] + iT[4] * a[9][5] + iT[7] * a[8][0] + iT[17] * a[3][0] + iT[20] * a[6][5] - iT[5] * a[5][2] + iT[6] * a[0][3] + iT[18] * a[4][2] + iT[19] * a[6][2] + t[1] + add) >> shift;
+ dst[24 * line] = (-iT[0] * a[2][3] + iT[11] * a[7][3] + iT[13] * a[5][2] + iT[24] * a[9][2] + iT[1] * a[4][1] + iT[10] * a[7][4] - iT[14] * a[1][4] - iT[23] * a[8][4] - iT[2] * a[4][5] - iT[9] * a[7][0] + iT[15] * a[1][0] + iT[22] * a[8][0] + iT[3] * a[4][3] + iT[8] * a[6][3] + iT[16] * a[0][2] - iT[21] * a[5][3] - iT[4] * a[2][5] - iT[7] * a[3][5] + iT[17] * a[0][5] + iT[20] * a[1][5] + iT[5] * a[2][1] + iT[6] * a[3][1] - iT[18] * a[0][1] - iT[19] * a[1][1] - t[0] + add) >> shift;
+ dst[25 * line] = ( iT[0] * a[4][5] + iT[11] * a[6][5] + iT[13] * a[0][0] - iT[24] * a[5][5] + iT[1] * a[3][1] + iT[10] * a[2][1] - iT[14] * a[1][1] - iT[23] * a[0][1] - iT[2] * a[7][2] - iT[9] * a[4][3] + iT[15] * a[8][2] + iT[22] * a[1][2] - iT[3] * a[6][2] - iT[8] * a[3][3] - iT[16] * a[9][2] - iT[21] * a[8][3] - iT[4] * a[2][4] + iT[7] * a[7][4] + iT[17] * a[5][1] + iT[20] * a[9][1] + iT[5] * a[4][0] + iT[6] * a[6][0] + iT[18] * a[0][5] - iT[19] * a[5][0] + t[0] + add) >> shift;
+ dst[26 * line] = ( iT[0] * a[8][0] + iT[11] * a[1][0] - iT[13] * a[4][5] - iT[24] * a[7][0] + iT[1] * a[5][4] + iT[10] * a[9][4] + iT[14] * a[7][1] - iT[23] * a[2][1] - iT[2] * a[1][2] - iT[9] * a[0][2] + iT[15] * a[2][2] + iT[22] * a[3][2] - iT[3] * a[9][2] - iT[8] * a[8][3] - iT[16] * a[3][3] - iT[21] * a[6][2] + iT[4] * a[0][4] - iT[7] * a[5][1] + iT[17] * a[6][1] + iT[20] * a[4][1] + iT[5] * a[8][5] + iT[6] * a[1][5] - iT[18] * a[4][0] - iT[19] * a[7][5] - t[1] + add) >> shift;
+ dst[28 * line] = (-iT[0] * a[5][1] - iT[11] * a[9][1] - iT[13] * a[7][4] + iT[24] * a[2][4] + iT[1] * a[8][2] + iT[10] * a[9][3] + iT[14] * a[6][3] + iT[23] * a[3][2] - iT[2] * a[9][4] - iT[9] * a[8][1] - iT[15] * a[3][1] - iT[22] * a[6][4] + iT[3] * a[9][0] + iT[8] * a[5][0] - iT[16] * a[2][5] + iT[21] * a[7][5] - iT[4] * a[5][5] + iT[7] * a[0][0] + iT[17] * a[4][5] + iT[20] * a[6][5] + iT[5] * a[1][3] + iT[6] * a[0][3] - iT[18] * a[2][3] - iT[19] * a[3][3] + t[1] + add) >> shift;
+ dst[29 * line] = (-iT[0] * a[6][4] - iT[11] * a[3][1] - iT[13] * a[9][4] - iT[24] * a[8][1] + iT[1] * a[7][3] + iT[10] * a[4][2] - iT[14] * a[8][3] - iT[23] * a[1][3] + iT[2] * a[3][5] + iT[9] * a[2][5] - iT[15] * a[1][5] - iT[22] * a[0][5] - iT[3] * a[2][4] - iT[8] * a[3][4] + iT[16] * a[0][4] + iT[21] * a[1][4] - iT[4] * a[4][3] - iT[7] * a[7][2] + iT[17] * a[1][2] + iT[20] * a[8][2] + iT[5] * a[3][0] + iT[6] * a[6][5] + iT[18] * a[8][0] + iT[19] * a[9][5] - t[0] + add) >> shift;
+ dst[30 * line] = (-iT[0] * a[7][2] + iT[11] * a[2][2] - iT[13] * a[9][3] - iT[24] * a[5][3] - iT[1] * a[6][0] - iT[10] * a[4][0] + iT[14] * a[5][0] - iT[23] * a[0][5] - iT[2] * a[4][2] - iT[9] * a[6][2] - iT[15] * a[0][3] + iT[22] * a[5][2] + iT[3] * a[2][0] - iT[8] * a[7][0] - iT[16] * a[5][5] - iT[21] * a[9][5] + iT[4] * a[7][1] - iT[7] * a[2][1] + iT[17] * a[9][4] + iT[20] * a[5][4] + iT[5] * a[6][1] + iT[6] * a[4][1] - iT[18] * a[5][1] + iT[19] * a[0][4] + t[0] + add) >> shift;
+ dst[31 * line] = (-iT[0] * a[8][5] - iT[11] * a[1][5] + iT[13] * a[4][0] + iT[24] * a[7][5] + iT[1] * a[1][0] + iT[10] * a[8][0] - iT[14] * a[7][0] - iT[23] * a[4][5] + iT[2] * a[8][4] + iT[9] * a[1][4] - iT[15] * a[4][1] - iT[22] * a[7][4] - iT[3] * a[1][1] - iT[8] * a[8][1] + iT[16] * a[7][1] + iT[21] * a[4][4] - iT[4] * a[8][3] - iT[7] * a[1][3] + iT[17] * a[4][2] + iT[20] * a[7][3] + iT[5] * a[1][2] + iT[6] * a[8][2] - iT[18] * a[7][2] - iT[19] * a[4][3] - t[1] + add) >> shift;
+
+ dst[ 2 * line] = (iT[ 4]*b[0] + iT[ 9]*b[1] + iT[14]*b[2] + iT[19]*b[3] + iT[24]*b[4] + iT[29]*b[5] + add) >> shift;
+ dst[ 7 * line] = (iT[14]*b[0] + iT[29]*b[1] + iT[19]*b[2] + iT[ 4]*b[3] - iT[ 9]*b[4] - iT[24]*b[5] + add) >> shift;
+ dst[12 * line] = (iT[24]*b[0] + iT[14]*b[1] - iT[ 9]*b[2] - iT[29]*b[3] - iT[ 4]*b[4] + iT[19]*b[5] + add) >> shift;
+ dst[17 * line] = (iT[29]*b[0] - iT[ 4]*b[1] - iT[24]*b[2] + iT[ 9]*b[3] + iT[19]*b[4] - iT[14]*b[5] + add) >> shift;
+ dst[22 * line] = (iT[19]*b[0] - iT[24]*b[1] + iT[ 4]*b[2] + iT[14]*b[3] - iT[29]*b[4] + iT[ 9]*b[5] + add) >> shift;
+ dst[27 * line] = (iT[ 9]*b[0] - iT[19]*b[1] + iT[29]*b[2] - iT[24]*b[3] + iT[14]*b[4] - iT[ 4]*b[5] + add) >> shift;
+
+ dst[ 6 * line] = (iT[12]*c[0] + iT[25]*c[1] + add) >> shift;
+ dst[19 * line] = (iT[25]*c[0] - iT[12]*c[1] + add) >> shift;
+
+ src += 32;
+ dst++;
+ }
+
+ if (iSkipLine)
+ {
+ dst = pCoef + reducedLine;
+ for (j = 0; j < cutoff; j++)
+ {
+ memset(dst, 0, sizeof(TCoeff)*iSkipLine);
+ dst += line;
+ }
+ }
+
+ if (iSkipLine2)
+ {
+ dst = pCoef + line * cutoff;
+ memset(dst, 0, sizeof(TCoeff) * line * iSkipLine2);
+ }
+#else
_fastForwardMM< 32 >( src, dst, shift, line, iSkipLine, iSkipLine2, g_trCoreDST7P32[TRANSFORM_FORWARD][0] );
+#endif
}
void fastInverseDST7_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
{
+#if JVET_M0497_FAST_DST7 && !JVET_M0497_MATRIX_MULT
+ int j, k;
+ TCoeff a[10][6];
+ TCoeff t[2];
+ TCoeff b[6];
+ TCoeff c[2];
+
+ TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
+ const TMatrixCoeff *iT = g_trCoreDST7P32[TRANSFORM_INVERSE][0];
+ const int reducedLine = line - iSkipLine;
+
+ for (j = 0; j < reducedLine; j++)
+ {
+ for (k = 0; k < 6; k++)
+ {
+ a[0][k] = src[ k * line] + src[(12 - k) * line];
+ a[1][k] = src[ k * line] - src[(13 + k) * line];
+ a[2][k] = src[ k * line] + src[(25 - k) * line];
+ a[3][k] = src[ k * line] - src[(26 + k) * line];
+ a[4][k] = src[( 7 + k) * line] + src[(18 - k) * line];
+ a[5][k] = src[( 7 + k) * line] - src[(20 + k) * line];
+ a[6][k] = src[( 7 + k) * line] + src[(31 - k) * line];
+ a[7][k] = src[(13 + k) * line] + src[(25 - k) * line];
+ a[8][k] = src[(13 + k) * line] - src[(26 + k) * line];
+ a[9][k] = src[(20 + k) * line] + src[(31 - k) * line];
+
+ b[k] = src[k * line] - src[(12-k) * line] + src[(13+k) * line] - src[(25-k) * line] + src[(26+k) * line];
+ }
+ for (k = 0; k < 2; k++)
+ {
+ c[k] = src[k * line] - src[(4-k) * line] + src[(5+k) * line] - src[(9-k) * line] + src[(10+k) * line] - src[(14-k) * line] + src[(15+k)*line] - src[(19-k)*line] + src[(20+k)*line] - src[(24-k)*line] + src[(25+k)*line] - src[(29-k)*line] + src[(30+k)*line];
+ }
+
+ t[0] = iT[12] * src[6*line] + iT[25] * src[19*line];
+ t[1] = iT[25] * src[6*line] - iT[12] * src[19*line];
+
+ dst[ 0] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[1][0] - iT[11] * a[8][0] + iT[13] * a[7][0] + iT[24] * a[4][5] - iT[1] * a[8][5] + iT[10] * a[1][5] + iT[14] * a[4][0] + iT[23] * a[7][5] + iT[2] * a[1][1] - iT[9] * a[8][1] + iT[15] * a[7][1] + iT[22] * a[4][4] - iT[3] * a[8][4] + iT[8] * a[1][4] + iT[16] * a[4][1] + iT[21] * a[7][4] + iT[4] * a[1][2] - iT[7] * a[8][2] + iT[17] * a[7][2] + iT[20] * a[4][3] - iT[5] * a[8][3] + iT[6] * a[1][3] + iT[18] * a[4][2] + iT[19] * a[7][3] + t[0] + add) >> shift);
+ dst[ 1] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[4][2] - iT[11] * a[6][2] + iT[13] * a[0][3] + iT[24] * a[5][2] + iT[1] * a[2][0] + iT[10] * a[7][0] + iT[14] * a[5][5] - iT[23] * a[9][5] + iT[2] * a[7][2] + iT[9] * a[2][2] - iT[15] * a[9][3] + iT[22] * a[5][3] - iT[3] * a[6][0] - iT[8] * a[4][0] + iT[16] * a[5][0] + iT[21] * a[0][5] - iT[4] * a[4][1] - iT[7] * a[6][1] + iT[17] * a[0][4] + iT[20] * a[5][1] + iT[5] * a[2][1] + iT[6] * a[7][1] + iT[18] * a[5][4] - iT[19] * a[9][4] + t[1] + add) >> shift);
+ dst[ 2] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[2][4] - iT[11] * a[3][4] + iT[13] * a[0][4] + iT[24] * a[1][4] + iT[1] * a[4][3] + iT[10] * a[7][2] + iT[14] * a[1][2] - iT[23] * a[8][2] + iT[2] * a[3][0] - iT[9] * a[6][5] - iT[15] * a[8][0] + iT[22] * a[9][5] - iT[3] * a[6][4] + iT[8] * a[3][1] + iT[16] * a[9][4] - iT[21] * a[8][1] + iT[4] * a[7][3] + iT[7] * a[4][2] - iT[17] * a[8][3] + iT[20] * a[1][3] - iT[5] * a[3][5] - iT[6] * a[2][5] + iT[18] * a[1][5] + iT[19] * a[0][5] + t[1] + add) >> shift);
+ dst[ 3] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[5][4] + iT[11] * a[0][1] - iT[13] * a[4][4] - iT[24] * a[6][4] - iT[1] * a[1][3] - iT[10] * a[0][3] + iT[14] * a[2][3] + iT[23] * a[3][3] - iT[2] * a[0][4] - iT[9] * a[1][4] + iT[15] * a[3][4] + iT[22] * a[2][4] + iT[3] * a[0][0] + iT[8] * a[5][5] - iT[16] * a[6][5] - iT[21] * a[4][5] + iT[4] * a[5][0] - iT[7] * a[9][0] + iT[17] * a[7][5] + iT[20] * a[2][5] - iT[5] * a[8][2] + iT[6] * a[9][3] - iT[18] * a[6][3] + iT[19] * a[3][2] + t[0] + add) >> shift);
+ dst[ 5] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[1][5] + iT[11] * a[8][5] - iT[13] * a[7][5] - iT[24] * a[4][0] + iT[1] * a[5][1] + iT[10] * a[0][4] - iT[14] * a[4][1] - iT[23] * a[6][1] - iT[2] * a[8][3] + iT[9] * a[9][2] - iT[15] * a[6][2] + iT[22] * a[3][3] - iT[3] * a[0][2] - iT[8] * a[1][2] + iT[16] * a[3][2] + iT[21] * a[2][2] - iT[4] * a[9][4] + iT[7] * a[5][4] + iT[17] * a[2][1] + iT[20] * a[7][1] + iT[5] * a[1][0] - iT[6] * a[8][0] + iT[18] * a[7][0] + iT[19] * a[4][5] - t[0] + add) >> shift);
+ dst[ 6] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[7][5] - iT[11] * a[2][5] + iT[13] * a[9][0] - iT[24] * a[5][0] + iT[1] * a[3][4] - iT[10] * a[6][1] - iT[14] * a[8][4] + iT[23] * a[9][1] + iT[2] * a[4][2] + iT[9] * a[7][3] + iT[15] * a[1][3] - iT[22] * a[8][3] - iT[3] * a[2][2] - iT[8] * a[3][2] + iT[16] * a[0][2] + iT[21] * a[1][2] - iT[4] * a[6][4] - iT[7] * a[4][4] + iT[17] * a[5][4] + iT[20] * a[0][1] + iT[5] * a[7][0] + iT[6] * a[2][0] - iT[18] * a[9][5] + iT[19] * a[5][5] - t[1] + add) >> shift);
+ dst[ 7] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[6][3] - iT[11] * a[4][3] + iT[13] * a[5][3] + iT[24] * a[0][2] + iT[1] * a[7][1] + iT[10] * a[4][4] - iT[14] * a[8][1] + iT[23] * a[1][1] - iT[2] * a[7][5] - iT[9] * a[4][0] + iT[15] * a[8][5] - iT[22] * a[1][5] + iT[3] * a[7][3] + iT[8] * a[2][3] - iT[16] * a[9][2] + iT[21] * a[5][2] - iT[4] * a[6][5] + iT[7] * a[3][0] + iT[17] * a[9][5] - iT[20] * a[8][0] + iT[5] * a[6][1] - iT[6] * a[3][4] - iT[18] * a[9][1] + iT[19] * a[8][4] - t[1] + add) >> shift);
+ dst[ 8] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[1][1] - iT[11] * a[0][1] + iT[13] * a[2][1] + iT[24] * a[3][1] + iT[1] * a[1][3] - iT[10] * a[8][3] + iT[14] * a[7][3] + iT[23] * a[4][2] - iT[2] * a[9][1] + iT[9] * a[8][4] - iT[15] * a[3][4] + iT[22] * a[6][1] + iT[3] * a[5][5] + iT[8] * a[0][0] - iT[16] * a[4][5] - iT[21] * a[6][5] + iT[4] * a[0][5] + iT[7] * a[1][5] - iT[17] * a[3][5] - iT[20] * a[2][5] + iT[5] * a[5][3] - iT[6] * a[9][3] + iT[18] * a[7][2] + iT[19] * a[2][2] - t[0] + add) >> shift);
+ dst[10] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[8][3] - iT[11] * a[1][3] - iT[13] * a[4][2] - iT[24] * a[7][3] - iT[1] * a[8][0] + iT[10] * a[1][0] + iT[14] * a[4][5] + iT[23] * a[7][0] + iT[2] * a[5][3] + iT[9] * a[0][2] - iT[15] * a[4][3] - iT[22] * a[6][3] - iT[3] * a[5][0] - iT[8] * a[0][5] + iT[16] * a[4][0] + iT[21] * a[6][0] + iT[4] * a[1][4] + iT[7] * a[0][4] - iT[17] * a[2][4] - iT[20] * a[3][4] - iT[5] * a[1][1] - iT[6] * a[0][1] + iT[18] * a[2][1] + iT[19] * a[3][1] + t[0] + add) >> shift);
+ dst[11] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[7][0] + iT[11] * a[2][0] - iT[13] * a[9][5] + iT[24] * a[5][5] + iT[1] * a[2][5] + iT[10] * a[7][5] + iT[14] * a[5][0] - iT[23] * a[9][0] - iT[2] * a[2][1] - iT[9] * a[3][1] + iT[15] * a[0][1] + iT[22] * a[1][1] - iT[3] * a[7][4] - iT[8] * a[4][1] + iT[16] * a[8][4] - iT[21] * a[1][4] + iT[4] * a[3][2] - iT[7] * a[6][3] - iT[17] * a[8][2] + iT[20] * a[9][3] + iT[5] * a[4][2] + iT[6] * a[6][2] - iT[18] * a[0][3] - iT[19] * a[5][2] + t[1] + add) >> shift);
+ dst[13] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[9][5] - iT[11] * a[8][0] + iT[13] * a[3][0] - iT[24] * a[6][5] - iT[1] * a[8][5] + iT[10] * a[9][0] - iT[14] * a[6][0] + iT[23] * a[3][5] + iT[2] * a[5][4] - iT[9] * a[9][4] + iT[15] * a[7][1] + iT[22] * a[2][1] - iT[3] * a[1][4] + iT[8] * a[8][4] - iT[16] * a[7][4] - iT[21] * a[4][1] - iT[4] * a[0][2] - iT[7] * a[5][3] + iT[17] * a[6][3] + iT[20] * a[4][3] + iT[5] * a[0][3] + iT[6] * a[1][3] - iT[18] * a[3][3] - iT[19] * a[2][3] + t[0] + add) >> shift);
+ dst[15] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[9][1] + iT[11] * a[5][1] + iT[13] * a[2][4] + iT[24] * a[7][4] + iT[1] * a[9][3] - iT[10] * a[5][3] - iT[14] * a[2][2] - iT[23] * a[7][2] - iT[2] * a[9][5] + iT[9] * a[5][5] + iT[15] * a[2][0] + iT[22] * a[7][0] + iT[3] * a[9][4] - iT[8] * a[8][1] + iT[16] * a[3][1] - iT[21] * a[6][4] - iT[4] * a[9][2] + iT[7] * a[8][3] - iT[17] * a[3][3] + iT[20] * a[6][2] + iT[5] * a[9][0] - iT[6] * a[8][5] + iT[18] * a[3][5] - iT[19] * a[6][0] - t[0] + add) >> shift);
+ dst[16] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[4][4] + iT[11] * a[7][1] + iT[13] * a[1][1] - iT[24] * a[8][1] + iT[1] * a[6][2] - iT[10] * a[3][3] - iT[14] * a[9][2] + iT[23] * a[8][3] - iT[2] * a[6][1] - iT[9] * a[4][1] + iT[15] * a[5][1] + iT[22] * a[0][4] - iT[3] * a[4][5] - iT[8] * a[6][5] + iT[16] * a[0][0] + iT[21] * a[5][5] - iT[4] * a[6][0] + iT[7] * a[3][5] + iT[17] * a[9][0] - iT[20] * a[8][5] + iT[5] * a[6][3] + iT[6] * a[4][3] - iT[18] * a[5][3] - iT[19] * a[0][2] - t[1] + add) >> shift);
+ dst[17] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[7][2] - iT[11] * a[4][3] + iT[13] * a[8][2] - iT[24] * a[1][2] + iT[1] * a[7][1] + iT[10] * a[2][1] - iT[14] * a[9][4] + iT[23] * a[5][4] - iT[2] * a[3][5] + iT[9] * a[6][0] + iT[15] * a[8][5] - iT[22] * a[9][0] - iT[3] * a[2][3] - iT[8] * a[7][3] - iT[16] * a[5][2] + iT[21] * a[9][2] + iT[4] * a[4][5] + iT[7] * a[7][0] + iT[17] * a[1][0] - iT[20] * a[8][0] - iT[5] * a[2][4] - iT[6] * a[3][4] + iT[18] * a[0][4] + iT[19] * a[1][4] - t[1] + add) >> shift);
+ dst[18] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[9][0] + iT[11] * a[8][5] - iT[13] * a[3][5] + iT[24] * a[6][0] + iT[1] * a[5][1] - iT[10] * a[9][1] + iT[14] * a[7][4] + iT[23] * a[2][4] + iT[2] * a[0][3] + iT[9] * a[5][2] - iT[15] * a[6][2] - iT[22] * a[4][2] + iT[3] * a[1][2] + iT[8] * a[0][2] - iT[16] * a[2][2] - iT[21] * a[3][2] - iT[4] * a[8][1] + iT[7] * a[1][1] + iT[17] * a[4][4] + iT[20] * a[7][1] + iT[5] * a[9][5] - iT[6] * a[8][0] + iT[18] * a[3][0] - iT[19] * a[6][5] - t[0] + add) >> shift);
+ dst[20] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[8][2] - iT[11] * a[9][3] + iT[13] * a[6][3] - iT[24] * a[3][2] + iT[1] * a[0][1] + iT[10] * a[5][4] - iT[14] * a[6][4] - iT[23] * a[4][4] + iT[2] * a[1][5] + iT[9] * a[0][5] - iT[15] * a[2][5] - iT[22] * a[3][5] - iT[3] * a[9][2] + iT[8] * a[5][2] + iT[16] * a[2][3] + iT[21] * a[7][3] + iT[4] * a[5][5] - iT[7] * a[9][5] + iT[17] * a[7][0] + iT[20] * a[2][0] + iT[5] * a[0][4] + iT[6] * a[5][1] - iT[18] * a[6][1] - iT[19] * a[4][1] + t[0] + add) >> shift);
+ dst[21] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[2][1] - iT[11] * a[7][1] - iT[13] * a[5][4] + iT[24] * a[9][4] - iT[1] * a[6][2] - iT[10] * a[4][2] + iT[14] * a[5][2] + iT[23] * a[0][3] - iT[2] * a[2][4] - iT[9] * a[7][4] - iT[15] * a[5][1] + iT[22] * a[9][1] - iT[3] * a[6][5] - iT[8] * a[4][5] + iT[16] * a[5][5] + iT[21] * a[0][0] - iT[4] * a[4][0] - iT[7] * a[7][5] - iT[17] * a[1][5] + iT[20] * a[8][5] - iT[5] * a[7][2] - iT[6] * a[4][3] + iT[18] * a[8][2] - iT[19] * a[1][2] + t[1] + add) >> shift);
+ dst[22] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[6][1] - iT[11] * a[3][4] - iT[13] * a[9][1] + iT[24] * a[8][4] + iT[1] * a[4][3] + iT[10] * a[6][3] - iT[14] * a[0][2] - iT[23] * a[5][3] + iT[2] * a[7][0] + iT[9] * a[4][5] - iT[15] * a[8][0] + iT[22] * a[1][0] - iT[3] * a[3][1] + iT[8] * a[6][4] + iT[16] * a[8][1] - iT[21] * a[9][4] - iT[4] * a[2][3] - iT[7] * a[3][3] + iT[17] * a[0][3] + iT[20] * a[1][3] - iT[5] * a[7][5] - iT[6] * a[2][5] + iT[18] * a[9][0] - iT[19] * a[5][0] + t[1] + add) >> shift);
+ dst[23] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[0][3] - iT[11] * a[1][3] + iT[13] * a[3][3] + iT[24] * a[2][3] - iT[1] * a[8][0] + iT[10] * a[9][5] - iT[14] * a[6][5] + iT[23] * a[3][0] + iT[2] * a[8][2] - iT[9] * a[1][2] - iT[15] * a[4][3] - iT[22] * a[7][2] + iT[3] * a[0][5] + iT[8] * a[5][0] - iT[16] * a[6][0] - iT[21] * a[4][0] + iT[4] * a[8][4] - iT[7] * a[9][1] + iT[17] * a[6][1] - iT[20] * a[3][4] - iT[5] * a[5][4] - iT[6] * a[0][1] + iT[18] * a[4][4] + iT[19] * a[6][4] + t[0] + add) >> shift);
+ dst[26] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[3][0] - iT[11] * a[2][0] + iT[13] * a[1][0] + iT[24] * a[0][0] - iT[1] * a[2][5] - iT[10] * a[3][5] + iT[14] * a[0][5] + iT[23] * a[1][5] + iT[2] * a[4][4] + iT[9] * a[6][4] - iT[15] * a[0][1] - iT[22] * a[5][4] - iT[3] * a[4][1] - iT[8] * a[7][4] - iT[16] * a[1][4] + iT[21] * a[8][4] + iT[4] * a[2][2] + iT[7] * a[7][2] + iT[17] * a[5][3] - iT[20] * a[9][3] + iT[5] * a[3][3] - iT[6] * a[6][2] - iT[18] * a[8][3] + iT[19] * a[9][2] - t[1] + add) >> shift);
+ dst[27] = Clip3(outputMinimum, outputMaximum, (int)(-iT[0] * a[3][3] + iT[11] * a[6][2] + iT[13] * a[8][3] - iT[24] * a[9][2] - iT[1] * a[2][0] - iT[10] * a[3][0] + iT[14] * a[0][0] + iT[23] * a[1][0] - iT[2] * a[6][3] + iT[9] * a[3][2] + iT[15] * a[9][3] - iT[22] * a[8][2] - iT[3] * a[4][0] - iT[8] * a[6][0] + iT[16] * a[0][5] + iT[21] * a[5][0] - iT[4] * a[7][4] - iT[7] * a[2][4] + iT[17] * a[9][1] - iT[20] * a[5][1] - iT[5] * a[4][4] - iT[6] * a[7][1] - iT[18] * a[1][1] + iT[19] * a[8][1] - t[1] + add) >> shift);
+ dst[28] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[0][4] + iT[11] * a[5][1] - iT[13] * a[6][1] - iT[24] * a[4][1] + iT[1] * a[9][3] - iT[10] * a[8][2] + iT[14] * a[3][2] - iT[23] * a[6][3] - iT[2] * a[1][0] - iT[9] * a[0][0] + iT[15] * a[2][0] + iT[22] * a[3][0] + iT[3] * a[8][1] - iT[8] * a[9][4] + iT[16] * a[6][4] - iT[21] * a[3][1] - iT[4] * a[5][2] - iT[7] * a[0][3] + iT[17] * a[4][2] + iT[20] * a[6][2] + iT[5] * a[1][5] - iT[6] * a[8][5] + iT[18] * a[7][5] + iT[19] * a[4][0] - t[0] + add) >> shift);
+ dst[30] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[5][3] - iT[11] * a[9][3] + iT[13] * a[7][2] + iT[24] * a[2][2] + iT[1] * a[0][1] + iT[10] * a[1][1] - iT[14] * a[3][1] - iT[23] * a[2][1] + iT[2] * a[9][0] - iT[9] * a[5][0] - iT[15] * a[2][5] - iT[22] * a[7][5] - iT[3] * a[5][2] + iT[8] * a[9][2] - iT[16] * a[7][3] - iT[21] * a[2][3] - iT[4] * a[0][0] - iT[7] * a[1][0] + iT[17] * a[3][0] + iT[20] * a[2][0] - iT[5] * a[9][1] + iT[6] * a[5][1] + iT[18] * a[2][4] + iT[19] * a[7][4] + t[0] + add) >> shift);
+ dst[31] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[3][5] + iT[11] * a[2][5] - iT[13] * a[1][5] - iT[24] * a[0][5] - iT[1] * a[3][4] - iT[10] * a[2][4] + iT[14] * a[1][4] + iT[23] * a[0][4] + iT[2] * a[3][3] + iT[9] * a[2][3] - iT[15] * a[1][3] - iT[22] * a[0][3] - iT[3] * a[3][2] - iT[8] * a[2][2] + iT[16] * a[1][2] + iT[21] * a[0][2] + iT[4] * a[3][1] + iT[7] * a[2][1] - iT[17] * a[1][1] - iT[20] * a[0][1] - iT[5] * a[3][0] - iT[6] * a[2][0] + iT[18] * a[1][0] + iT[19] * a[0][0] + t[1] + add) >> shift);
+
+ dst[ 4] = Clip3(outputMinimum, outputMaximum, (int)(iT[ 4] * b[0] + iT[14] * b[1] + iT[24] * b[2] + iT[29] * b[3] + iT[19] * b[4] + iT[ 9] * b[5] + add) >> shift);
+ dst[ 9] = Clip3(outputMinimum, outputMaximum, (int)(iT[ 9] * b[0] + iT[29] * b[1] + iT[14] * b[2] - iT[ 4] * b[3] - iT[24] * b[4] - iT[19] * b[5] + add) >> shift);
+ dst[14] = Clip3(outputMinimum, outputMaximum, (int)(iT[14] * b[0] + iT[19] * b[1] - iT[ 9] * b[2] - iT[24] * b[3] + iT[ 4] * b[4] + iT[29] * b[5] + add) >> shift);
+ dst[19] = Clip3(outputMinimum, outputMaximum, (int)(iT[19] * b[0] + iT[ 4] * b[1] - iT[29] * b[2] + iT[ 9] * b[3] + iT[14] * b[4] - iT[24] * b[5] + add) >> shift);
+ dst[24] = Clip3(outputMinimum, outputMaximum, (int)(iT[24] * b[0] - iT[ 9] * b[1] - iT[ 4] * b[2] + iT[19] * b[3] - iT[29] * b[4] + iT[14] * b[5] + add) >> shift);
+ dst[29] = Clip3(outputMinimum, outputMaximum, (int)(iT[29] * b[0] - iT[24] * b[1] + iT[19] * b[2] - iT[14] * b[3] + iT[ 9] * b[4] - iT[ 4] * b[5] + add) >> shift);
+
+ dst[12] = Clip3(outputMinimum, outputMaximum, (int)(iT[12]*c[0] + iT[25]*c[1] + add) >> shift);
+ dst[25] = Clip3(outputMinimum, outputMaximum, (int)(iT[25]*c[0] - iT[12]*c[1] + add) >> shift);
+
+ src++;
+ dst += 32;
+ }
+
+ if (iSkipLine)
+ {
+ memset(dst, 0, (iSkipLine * 32) * sizeof(TCoeff));
+ }
+#else
_fastInverseMM< 32 >( src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_trCoreDST7P32[TRANSFORM_INVERSE][0] );
+#endif
}
@@ -1049,22 +1340,314 @@ void fastInverseDCT8_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int
void fastForwardDCT8_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
{
+#if JVET_M0497_FAST_DST7 && !JVET_M0497_MATRIX_MULT
+ int j, k;
+ TCoeff a[5], b[5], c[5], d[5], t;
+ TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
+
+ const TMatrixCoeff *iT = g_trCoreDST7P16[TRANSFORM_FORWARD][0];
+
+ TCoeff *pCoef = dst;
+ const int reducedLine = line - iSkipLine;
+ const int cutoff = 16 - iSkipLine2;
+
+ for (j = 0; j < reducedLine; j++)
+ {
+ for (k = 0; k < 5; k++)
+ {
+ a[k] = src[15 - k] + src[ 4 - k];
+ b[k] = src[ 6 + k] + src[ 4 - k];
+ c[k] = src[15 - k] - src[ 6 + k];
+ d[k] = src[15 - k] + src[ 6 + k] - src[ 4 - k];
+ }
+
+ t = iT[10] * src[5];
+
+ dst[ 1 * line] = ( - iT[ 2]*d[0] - iT[ 5]*d[1] - iT[ 8]*d[2] - iT[11]*d[3] - iT[14]*d[4] + add) >> shift;
+ dst[ 4 * line] = ( iT[ 8]*d[0] + iT[14]*d[1] + iT[ 5]*d[2] - iT[ 2]*d[3] - iT[11]*d[4] + add) >> shift;
+ dst[ 7 * line] = ( - iT[14]*d[0] - iT[ 2]*d[1] + iT[11]*d[2] + iT[ 5]*d[3] - iT[ 8]*d[4] + add) >> shift;
+ dst[10 * line] = ( iT[11]*d[0] - iT[ 8]*d[1] - iT[ 2]*d[2] + iT[14]*d[3] - iT[ 5]*d[4] + add) >> shift;
+ dst[13 * line] = ( - iT[ 5]*d[0] + iT[11]*d[1] - iT[14]*d[2] + iT[ 8]*d[3] - iT[ 2]*d[4] + add) >> shift;
+
+ dst[ 5 * line] = ( - iT[10] * (src[15] + src[14] - src[12] - src[11] + src[9] + src[8] - src[6] - src[5] + src[3] + src[2] - src[0]) + add) >> shift;
+
+ dst[ 0 * line] = ( iT[0]*a[0] + iT[9]*b[0] + iT[1]*a[1] + iT[8]*b[1] + iT[2]*a[2] + iT[7]*b[2] + iT[3]*a[3] + iT[6]*b[3] + iT[4]*a[4] + iT[5]*b[4] + t + add ) >> shift;
+ dst[ 2 * line] = ( iT[4]*c[0] - iT[5]*b[0] + iT[9]*c[1] - iT[0]*b[1] + iT[6]*c[2] + iT[3]*a[2] + iT[1]*c[3] + iT[8]*a[3] + iT[7]*a[4] + iT[2]*b[4] - t + add ) >> shift;
+ dst[ 3 * line] = ( - iT[6]*a[0] - iT[3]*b[0] - iT[2]*c[1] - iT[7]*a[1] - iT[9]*c[2] - iT[0]*a[2] - iT[4]*c[3] + iT[5]*b[3] + iT[1]*a[4] + iT[8]*b[4] - t + add ) >> shift;
+ dst[ 6 * line] = ( iT[8]*a[0] + iT[1]*c[0] + iT[6]*c[1] - iT[3]*b[1] - iT[5]*a[2] - iT[4]*b[2] - iT[7]*c[3] - iT[2]*a[3] - iT[0]*c[4] + iT[9]*b[4] + t + add ) >> shift;
+ dst[ 8 * line] = ( iT[4]*c[0] + iT[5]*a[0] - iT[0]*c[1] + iT[9]*b[1] - iT[3]*c[2] - iT[6]*a[2] + iT[1]*c[3] - iT[8]*b[3] + iT[2]*c[4] + iT[7]*a[4] - t + add ) >> shift;
+ dst[ 9 * line] = ( - iT[7]*c[0] - iT[2]*a[0] + iT[4]*a[1] + iT[5]*b[1] + iT[8]*c[2] - iT[1]*b[2] - iT[9]*a[3] - iT[0]*b[3] - iT[3]*c[4] + iT[6]*b[4] - t + add ) >> shift;
+ dst[11 * line] = ( - iT[9]*a[0] - iT[0]*b[0] + iT[8]*c[1] + iT[1]*a[1] - iT[2]*c[2] + iT[7]*b[2] - iT[6]*a[3] - iT[3]*b[3] + iT[5]*c[4] + iT[4]*a[4] + t + add ) >> shift;
+ dst[12 * line] = ( iT[7]*c[0] - iT[2]*b[0] - iT[5]*c[1] - iT[4]*a[1] + iT[8]*a[2] + iT[1]*b[2] - iT[0]*a[3] - iT[9]*b[3] - iT[6]*c[4] + iT[3]*b[4] + t + add ) >> shift;
+ dst[14 * line] = ( iT[3]*a[0] + iT[6]*b[0] - iT[7]*a[1] - iT[2]*b[1] + iT[0]*c[2] + iT[9]*a[2] - iT[4]*c[3] - iT[5]*a[3] + iT[8]*c[4] + iT[1]*a[4] - t + add ) >> shift;
+ dst[15 * line] = ( - iT[1]*c[0] + iT[8]*b[0] + iT[3]*c[1] - iT[6]*b[1] - iT[5]*c[2] + iT[4]*b[2] + iT[7]*c[3] - iT[2]*b[3] - iT[9]*c[4] + iT[0]*b[4] - t + add ) >> shift;
+
+ src += 16;
+ dst++;
+ }
+
+ if (iSkipLine)
+ {
+ dst = pCoef + reducedLine;
+ for (j = 0; j < cutoff; j++)
+ {
+ memset(dst, 0, sizeof(TCoeff)*iSkipLine);
+ dst += line;
+ }
+ }
+
+ if (iSkipLine2)
+ {
+ dst = pCoef + line * cutoff;
+ memset(dst, 0, sizeof(TCoeff) * line * iSkipLine2);
+ }
+#else
_fastForwardMM< 16 >( src, dst, shift, line, iSkipLine, iSkipLine2, g_trCoreDCT8P16[TRANSFORM_FORWARD][0] );
+#endif
}
void fastInverseDCT8_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
{
+#if JVET_M0497_FAST_DST7 && !JVET_M0497_MATRIX_MULT
+ int j, k;
+ TCoeff a[5], b[5], c[5], d[5], t;
+
+ TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
+
+ const TMatrixCoeff *iT = g_trCoreDST7P16[TRANSFORM_INVERSE][0];
+
+ const int reducedLine = line - iSkipLine;
+
+ for (j = 0; j < reducedLine; j++)
+ {
+ for (k = 0; k < 5; k++)
+ {
+ a[k] = src[(15 - k ) * line] + src[( 4 - k) * line];
+ b[k] = src[( 6 + k ) * line] + src[( 4 - k) * line];
+ c[k] = src[(15 - k ) * line] - src[( 6 + k) * line];
+ d[k] = src[(15 - k ) * line] + src[( 6 + k) * line] - src[(4 - k) * line];
+ }
+
+ t = iT[10] * src[5*line];
+
+ dst[ 1] = Clip3(outputMinimum, outputMaximum, (int)( - iT[ 2]*d[0] - iT[ 5]*d[1] - iT[ 8]*d[2] - iT[11]*d[3] - iT[14]*d[4] + add) >> shift);
+ dst[ 4] = Clip3(outputMinimum, outputMaximum, (int)( iT[ 8]*d[0] + iT[14]*d[1] + iT[ 5]*d[2] - iT[ 2]*d[3] - iT[11]*d[4] + add) >> shift);
+ dst[ 7] = Clip3(outputMinimum, outputMaximum, (int)( - iT[14]*d[0] - iT[ 2]*d[1] + iT[11]*d[2] + iT[ 5]*d[3] - iT[ 8]*d[4] + add) >> shift);
+ dst[10] = Clip3(outputMinimum, outputMaximum, (int)( iT[11]*d[0] - iT[ 8]*d[1] - iT[ 2]*d[2] + iT[14]*d[3] - iT[ 5]*d[4] + add) >> shift);
+ dst[13] = Clip3(outputMinimum, outputMaximum, (int)( - iT[ 5]*d[0] + iT[11]*d[1] - iT[14]*d[2] + iT[ 8]*d[3] - iT[ 2]*d[4] + add) >> shift);
+
+ dst[ 5] = Clip3(outputMinimum, outputMaximum, (int)( - iT[10] * (src[15 * line] + src[14 * line] - src[12 * line] - src[11 * line] + src[9 * line] + src[8 * line] - src[6 * line] - src[5 * line] + src[3 * line] + src[2 * line] - src[0 * line]) + add) >> shift);
+
+ dst[ 0] = Clip3(outputMinimum, outputMaximum, (int)( iT[0]*a[0] + iT[9]*b[0] + iT[1]*a[1] + iT[8]*b[1] + iT[2]*a[2] + iT[7]*b[2] + iT[3]*a[3] + iT[6]*b[3] + iT[4]*a[4] + iT[5]*b[4] + t + add ) >> shift );
+ dst[ 2] = Clip3(outputMinimum, outputMaximum, (int)( iT[4]*c[0] - iT[5]*b[0] + iT[9]*c[1] - iT[0]*b[1] + iT[6]*c[2] + iT[3]*a[2] + iT[1]*c[3] + iT[8]*a[3] + iT[7]*a[4] + iT[2]*b[4] - t + add ) >> shift );
+ dst[ 3] = Clip3(outputMinimum, outputMaximum, (int)( - iT[6]*a[0] - iT[3]*b[0] - iT[2]*c[1] - iT[7]*a[1] - iT[9]*c[2] - iT[0]*a[2] - iT[4]*c[3] + iT[5]*b[3] + iT[1]*a[4] + iT[8]*b[4] - t + add ) >> shift );
+ dst[ 6] = Clip3(outputMinimum, outputMaximum, (int)( iT[8]*a[0] + iT[1]*c[0] + iT[6]*c[1] - iT[3]*b[1] - iT[5]*a[2] - iT[4]*b[2] - iT[7]*c[3] - iT[2]*a[3] - iT[0]*c[4] + iT[9]*b[4] + t + add ) >> shift );
+ dst[ 8] = Clip3(outputMinimum, outputMaximum, (int)( iT[4]*c[0] + iT[5]*a[0] - iT[0]*c[1] + iT[9]*b[1] - iT[3]*c[2] - iT[6]*a[2] + iT[1]*c[3] - iT[8]*b[3] + iT[2]*c[4] + iT[7]*a[4] - t + add ) >> shift );
+ dst[ 9] = Clip3(outputMinimum, outputMaximum, (int)( - iT[7]*c[0] - iT[2]*a[0] + iT[4]*a[1] + iT[5]*b[1] + iT[8]*c[2] - iT[1]*b[2] - iT[9]*a[3] - iT[0]*b[3] - iT[3]*c[4] + iT[6]*b[4] - t + add ) >> shift );
+ dst[11] = Clip3(outputMinimum, outputMaximum, (int)( - iT[9]*a[0] - iT[0]*b[0] + iT[8]*c[1] + iT[1]*a[1] - iT[2]*c[2] + iT[7]*b[2] - iT[6]*a[3] - iT[3]*b[3] + iT[5]*c[4] + iT[4]*a[4] + t + add ) >> shift );
+ dst[12] = Clip3(outputMinimum, outputMaximum, (int)( iT[7]*c[0] - iT[2]*b[0] - iT[5]*c[1] - iT[4]*a[1] + iT[8]*a[2] + iT[1]*b[2] - iT[0]*a[3] - iT[9]*b[3] - iT[6]*c[4] + iT[3]*b[4] + t + add ) >> shift );
+ dst[14] = Clip3(outputMinimum, outputMaximum, (int)( iT[3]*a[0] + iT[6]*b[0] - iT[7]*a[1] - iT[2]*b[1] + iT[0]*c[2] + iT[9]*a[2] - iT[4]*c[3] - iT[5]*a[3] + iT[8]*c[4] + iT[1]*a[4] - t + add ) >> shift );
+ dst[15] = Clip3(outputMinimum, outputMaximum, (int)( - iT[1]*c[0] + iT[8]*b[0] + iT[3]*c[1] - iT[6]*b[1] - iT[5]*c[2] + iT[4]*b[2] + iT[7]*c[3] - iT[2]*b[3] - iT[9]*c[4] + iT[0]*b[4] - t + add ) >> shift );
+
+ src++;
+ dst += 16;
+ }
+
+ if (iSkipLine)
+ {
+ memset(dst, 0, (iSkipLine * 16) * sizeof(TCoeff));
+ }
+#else
_fastInverseMM< 16 >( src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_trCoreDCT8P16[TRANSFORM_INVERSE][0] );
+#endif
}
void fastForwardDCT8_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
{
+#if JVET_M0497_FAST_DST7 && !JVET_M0497_MATRIX_MULT
+ int j, k;
+ TCoeff a[10][6];
+ TCoeff t[2];
+ TCoeff b[6];
+ TCoeff c[2];
+
+ TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
+ const TMatrixCoeff *iT = g_trCoreDST7P32[TRANSFORM_FORWARD][0];
+ TCoeff *pCoef = dst;
+ const int reducedLine = line - iSkipLine;
+ const int cutoff = 32 - iSkipLine2;
+
+ for (j = 0; j < reducedLine; j++)
+ {
+ for (k = 0; k < 6; k++)
+ {
+ a[0][k] = src[31-k] - src[20+k];
+ a[1][k] = src[31-k] + src[18-k];
+ a[2][k] = src[31-k] + src[ 7+k];
+ a[3][k] = src[31-k] - src[ 5-k];
+ a[4][k] = src[25-k] + src[13+k];
+ a[5][k] = src[25-k] + src[12-k];
+ a[6][k] = src[25-k] - src[ k];
+ a[7][k] = src[18-k] - src[ 7+k];
+ a[8][k] = src[18-k] + src[ 5-k];
+ a[9][k] = src[12-k] + src[ k];
+
+ b[k] = src[31-k] + src[20+k] - src[18-k] - src[7+k] + src[5-k];
+ }
+
+ for (k = 0; k < 2; k++)
+ {
+ c[k] = src[31-k] + src[28+k] - src[26-k] - src[23+k] + src[21-k] + src[18+k] - src[16-k] - src[13+k] + src[11-k] + src[8+k] - src[6-k] - src[3+k] + src[1-k];
+ }
+
+ t[0] = iT[12] * src[19] + iT[25] * src[6];
+ t[1] = iT[12] * src[6] - iT[25] * src[19];
+
+ dst[ 0 * line] = ( iT[0] * a[3][0] + iT[11] * a[6][5] + iT[13] * a[8][0] + iT[24] * a[9][5] + iT[1] * a[3][1] + iT[10] * a[6][4] + iT[14] * a[8][1] + iT[23] * a[9][4] + iT[2] * a[3][2] + iT[9] * a[6][3] + iT[15] * a[8][2] + iT[22] * a[9][3] + iT[3] * a[3][3] + iT[8] * a[6][2] + iT[16] * a[8][3] + iT[21] * a[9][2] + iT[4] * a[3][4] + iT[7] * a[6][1] + iT[17] * a[8][4] + iT[20] * a[9][1] + iT[5] * a[3][5] + iT[6] * a[6][0] + iT[18] * a[8][5] + iT[19] * a[9][0] + t[0] + add) >> shift;
+ dst[ 1 * line] = ( iT[0] * a[5][2] - iT[11] * a[0][3] - iT[13] * a[4][2] - iT[24] * a[6][2] - iT[1] * a[9][1] - iT[10] * a[8][4] - iT[14] * a[3][4] - iT[23] * a[6][1] - iT[2] * a[0][0] + iT[9] * a[5][5] - iT[15] * a[6][5] - iT[22] * a[4][5] + iT[3] * a[5][3] - iT[8] * a[0][2] - iT[16] * a[4][3] - iT[21] * a[6][3] - iT[4] * a[9][0] - iT[7] * a[8][5] - iT[17] * a[3][5] - iT[20] * a[6][0] - iT[5] * a[0][1] + iT[6] * a[5][4] - iT[18] * a[6][4] - iT[19] * a[4][4] + t[1] + add) >> shift;
+ dst[ 3 * line] = ( iT[0] * a[9][4] + iT[11] * a[5][4] - iT[13] * a[2][1] + iT[24] * a[7][1] + iT[1] * a[0][3] + iT[10] * a[1][3] - iT[14] * a[3][3] - iT[23] * a[2][3] - iT[2] * a[8][5] - iT[9] * a[9][0] - iT[15] * a[6][0] - iT[22] * a[3][5] + iT[3] * a[1][4] + iT[8] * a[0][4] - iT[16] * a[2][4] - iT[21] * a[3][4] + iT[4] * a[5][3] + iT[7] * a[9][3] + iT[17] * a[7][2] - iT[20] * a[2][2] - iT[5] * a[8][0] - iT[6] * a[1][0] + iT[18] * a[4][5] + iT[19] * a[7][0] - t[1] + add) >> shift;
+ dst[ 4 * line] = ( - iT[0] * a[3][2] - iT[11] * a[2][2] + iT[13] * a[1][2] + iT[24] * a[0][2] + iT[1] * a[6][0] + iT[10] * a[3][5] + iT[14] * a[9][0] + iT[23] * a[8][5] - iT[2] * a[2][3] - iT[9] * a[3][3] + iT[15] * a[0][3] + iT[22] * a[1][3] - iT[3] * a[7][0] + iT[8] * a[2][0] - iT[16] * a[9][5] - iT[21] * a[5][5] + iT[4] * a[4][4] + iT[7] * a[6][4] + iT[17] * a[0][1] - iT[20] * a[5][4] - iT[5] * a[7][4] - iT[6] * a[4][1] + iT[18] * a[8][4] + iT[19] * a[1][4] - t[0] + add) >> shift;
+ dst[ 5 * line] = ( iT[0] * a[3][5] + iT[11] * a[6][0] + iT[13] * a[8][5] + iT[24] * a[9][0] - iT[1] * a[6][5] - iT[10] * a[3][0] - iT[14] * a[9][5] - iT[23] * a[8][0] + iT[2] * a[7][4] - iT[9] * a[2][4] + iT[15] * a[9][1] + iT[22] * a[5][1] + iT[3] * a[7][1] + iT[8] * a[4][4] - iT[16] * a[8][1] - iT[21] * a[1][1] - iT[4] * a[6][2] - iT[7] * a[4][2] + iT[17] * a[5][2] - iT[20] * a[0][3] + iT[5] * a[3][2] + iT[6] * a[2][2] - iT[18] * a[1][2] - iT[19] * a[0][2] - t[0] + add) >> shift;
+ dst[ 8 * line] = ( iT[0] * a[9][3] + iT[11] * a[8][2] + iT[13] * a[3][2] + iT[24] * a[6][3] + iT[1] * a[1][5] + iT[10] * a[0][5] - iT[14] * a[2][5] - iT[23] * a[3][5] - iT[2] * a[1][3] - iT[9] * a[8][3] + iT[15] * a[7][3] + iT[22] * a[4][2] - iT[3] * a[9][5] - iT[8] * a[5][5] + iT[16] * a[2][0] - iT[21] * a[7][0] - iT[4] * a[1][1] - iT[7] * a[0][1] + iT[17] * a[2][1] + iT[20] * a[3][1] + iT[5] * a[5][1] + iT[6] * a[9][1] + iT[18] * a[7][4] - iT[19] * a[2][4] + t[1] + add) >> shift;
+ dst[ 9 * line] = ( iT[0] * a[2][1] + iT[11] * a[3][1] - iT[13] * a[0][1] - iT[24] * a[1][1] - iT[1] * a[7][3] + iT[10] * a[2][3] - iT[14] * a[9][2] - iT[23] * a[5][2] - iT[2] * a[4][0] - iT[9] * a[7][5] + iT[15] * a[1][5] + iT[22] * a[8][5] - iT[3] * a[3][4] - iT[8] * a[2][4] + iT[16] * a[1][4] + iT[21] * a[0][4] - iT[4] * a[6][3] - iT[7] * a[3][2] - iT[17] * a[9][3] - iT[20] * a[8][2] - iT[5] * a[4][5] - iT[6] * a[6][5] - iT[18] * a[0][0] + iT[19] * a[5][5] + t[0] + add) >> shift;
+ dst[10 * line] = ( - iT[0] * a[6][1] - iT[11] * a[4][1] + iT[13] * a[5][1] - iT[24] * a[0][4] + iT[1] * a[2][2] - iT[10] * a[7][2] - iT[14] * a[5][3] - iT[23] * a[9][3] + iT[2] * a[6][4] + iT[9] * a[4][4] - iT[15] * a[5][4] + iT[22] * a[0][1] - iT[3] * a[2][5] + iT[8] * a[7][5] + iT[16] * a[5][0] + iT[21] * a[9][0] - iT[4] * a[7][0] - iT[7] * a[4][5] + iT[17] * a[8][0] + iT[20] * a[1][0] + iT[5] * a[4][2] + iT[6] * a[7][3] - iT[18] * a[1][3] - iT[19] * a[8][3] + t[0] + add) >> shift;
+ dst[11 * line] = ( - iT[0] * a[1][3] - iT[11] * a[0][3] + iT[13] * a[2][3] + iT[24] * a[3][3] - iT[1] * a[9][1] - iT[10] * a[5][1] + iT[14] * a[2][4] - iT[23] * a[7][4] - iT[2] * a[8][0] - iT[9] * a[9][5] - iT[15] * a[6][5] - iT[22] * a[3][0] + iT[3] * a[0][2] - iT[8] * a[5][3] + iT[16] * a[6][3] + iT[21] * a[4][3] + iT[4] * a[5][0] - iT[7] * a[0][5] - iT[17] * a[4][0] - iT[20] * a[6][0] + iT[5] * a[9][4] + iT[6] * a[5][4] - iT[18] * a[2][1] + iT[19] * a[7][1] + t[1] + add) >> shift;
+ dst[13 * line] = ( iT[0] * a[0][0] + iT[11] * a[1][0] - iT[13] * a[3][0] - iT[24] * a[2][0] + iT[1] * a[5][4] - iT[10] * a[0][1] - iT[14] * a[4][4] - iT[23] * a[6][4] - iT[2] * a[9][3] - iT[9] * a[5][3] + iT[15] * a[2][2] - iT[22] * a[7][2] + iT[3] * a[8][3] + iT[8] * a[9][2] + iT[16] * a[6][2] + iT[21] * a[3][3] - iT[4] * a[1][4] - iT[7] * a[8][4] + iT[17] * a[7][4] + iT[20] * a[4][1] + iT[5] * a[0][5] + iT[6] * a[1][5] - iT[18] * a[3][5] - iT[19] * a[2][5] - t[1] + add) >> shift;
+ dst[14 * line] = ( iT[0] * a[4][2] + iT[11] * a[7][3] - iT[13] * a[1][3] - iT[24] * a[8][3] + iT[1] * a[4][1] + iT[10] * a[6][1] + iT[14] * a[0][4] - iT[23] * a[5][1] - iT[2] * a[3][0] - iT[9] * a[2][0] + iT[15] * a[1][0] + iT[22] * a[0][0] - iT[3] * a[6][3] - iT[8] * a[4][3] + iT[16] * a[5][3] - iT[21] * a[0][2] - iT[4] * a[7][5] - iT[7] * a[4][0] + iT[17] * a[8][5] + iT[20] * a[1][5] + iT[5] * a[6][4] + iT[6] * a[3][1] + iT[18] * a[9][4] + iT[19] * a[8][1] - t[0] + add) >> shift;
+ dst[15 * line] = ( iT[0] * a[7][4] + iT[11] * a[4][1] - iT[13] * a[8][4] - iT[24] * a[1][4] - iT[1] * a[2][2] - iT[10] * a[3][2] + iT[14] * a[0][2] + iT[23] * a[1][2] - iT[2] * a[2][1] + iT[9] * a[7][1] + iT[15] * a[5][4] + iT[22] * a[9][4] + iT[3] * a[7][5] - iT[8] * a[2][5] + iT[16] * a[9][0] + iT[21] * a[5][0] + iT[4] * a[2][0] + iT[7] * a[3][0] - iT[17] * a[0][0] - iT[20] * a[1][0] + iT[5] * a[2][3] - iT[6] * a[7][3] - iT[18] * a[5][2] - iT[19] * a[9][2] - t[0] + add) >> shift;
+ dst[16 * line] = ( - iT[0] * a[0][1] + iT[11] * a[5][4] - iT[13] * a[6][4] - iT[24] * a[4][4] + iT[1] * a[0][3] - iT[10] * a[5][2] + iT[14] * a[6][2] + iT[23] * a[4][2] - iT[2] * a[0][5] + iT[9] * a[5][0] - iT[15] * a[6][0] - iT[22] * a[4][0] - iT[3] * a[0][4] - iT[8] * a[1][4] + iT[16] * a[3][4] + iT[21] * a[2][4] + iT[4] * a[0][2] + iT[7] * a[1][2] - iT[17] * a[3][2] - iT[20] * a[2][2] - iT[5] * a[0][0] - iT[6] * a[1][0] + iT[18] * a[3][0] + iT[19] * a[2][0] - t[1] + add) >> shift;
+ dst[18 * line] = ( iT[0] * a[0][5] + iT[11] * a[1][5] - iT[13] * a[3][5] - iT[24] * a[2][5] - iT[1] * a[1][0] - iT[10] * a[0][0] + iT[14] * a[2][0] + iT[23] * a[3][0] - iT[2] * a[5][1] + iT[9] * a[0][4] + iT[15] * a[4][1] + iT[22] * a[6][1] - iT[3] * a[8][1] - iT[8] * a[1][1] + iT[16] * a[4][4] + iT[21] * a[7][1] - iT[4] * a[9][2] - iT[7] * a[5][2] + iT[17] * a[2][3] - iT[20] * a[7][3] - iT[5] * a[9][3] - iT[6] * a[8][2] - iT[18] * a[3][2] - iT[19] * a[6][3] + t[1] + add) >> shift;
+ dst[20 * line] = ( - iT[0] * a[4][0] - iT[11] * a[6][0] - iT[13] * a[0][5] + iT[24] * a[5][0] + iT[1] * a[6][5] + iT[10] * a[4][5] - iT[14] * a[5][5] + iT[23] * a[0][0] - iT[2] * a[6][1] - iT[9] * a[3][4] - iT[15] * a[9][1] - iT[22] * a[8][4] + iT[3] * a[4][4] + iT[8] * a[7][1] - iT[16] * a[1][1] - iT[21] * a[8][1] - iT[4] * a[3][3] - iT[7] * a[2][3] + iT[17] * a[1][3] + iT[20] * a[0][3] + iT[5] * a[7][2] - iT[6] * a[2][2] + iT[18] * a[9][3] + iT[19] * a[5][3] + t[0] + add) >> shift;
+ dst[21 * line] = ( iT[0] * a[1][2] + iT[11] * a[8][2] - iT[13] * a[7][2] - iT[24] * a[4][3] + iT[1] * a[1][5] + iT[10] * a[8][5] - iT[14] * a[7][5] - iT[23] * a[4][0] + iT[2] * a[5][2] + iT[9] * a[9][2] + iT[15] * a[7][3] - iT[22] * a[2][3] + iT[3] * a[5][5] + iT[8] * a[9][5] + iT[16] * a[7][0] - iT[21] * a[2][0] + iT[4] * a[8][1] + iT[7] * a[9][4] + iT[17] * a[6][4] + iT[20] * a[3][1] + iT[5] * a[8][4] + iT[6] * a[9][1] + iT[18] * a[6][1] + iT[19] * a[3][4] + t[1] + add) >> shift;
+ dst[23 * line] = ( iT[0] * a[8][4] + iT[11] * a[9][1] + iT[13] * a[6][1] + iT[24] * a[3][4] - iT[1] * a[8][2] - iT[10] * a[1][2] + iT[14] * a[4][3] + iT[23] * a[7][2] - iT[2] * a[0][1] - iT[9] * a[1][1] + iT[15] * a[3][1] + iT[22] * a[2][1] + iT[3] * a[5][0] + iT[8] * a[9][0] + iT[16] * a[7][5] - iT[21] * a[2][5] - iT[4] * a[9][5] - iT[7] * a[8][0] - iT[17] * a[3][0] - iT[20] * a[6][5] + iT[5] * a[5][2] - iT[6] * a[0][3] - iT[18] * a[4][2] - iT[19] * a[6][2] - t[1] + add) >> shift;
+ dst[24 * line] = ( - iT[0] * a[2][3] + iT[11] * a[7][3] + iT[13] * a[5][2] + iT[24] * a[9][2] + iT[1] * a[4][1] + iT[10] * a[7][4] - iT[14] * a[1][4] - iT[23] * a[8][4] - iT[2] * a[4][5] - iT[9] * a[7][0] + iT[15] * a[1][0] + iT[22] * a[8][0] + iT[3] * a[4][3] + iT[8] * a[6][3] + iT[16] * a[0][2] - iT[21] * a[5][3] - iT[4] * a[2][5] - iT[7] * a[3][5] + iT[17] * a[0][5] + iT[20] * a[1][5] + iT[5] * a[2][1] + iT[6] * a[3][1] - iT[18] * a[0][1] - iT[19] * a[1][1] - t[0] + add) >> shift;
+ dst[25 * line] = ( - iT[0] * a[4][5] - iT[11] * a[6][5] - iT[13] * a[0][0] + iT[24] * a[5][5] - iT[1] * a[3][1] - iT[10] * a[2][1] + iT[14] * a[1][1] + iT[23] * a[0][1] + iT[2] * a[7][2] + iT[9] * a[4][3] - iT[15] * a[8][2] - iT[22] * a[1][2] + iT[3] * a[6][2] + iT[8] * a[3][3] + iT[16] * a[9][2] + iT[21] * a[8][3] + iT[4] * a[2][4] - iT[7] * a[7][4] - iT[17] * a[5][1] - iT[20] * a[9][1] - iT[5] * a[4][0] - iT[6] * a[6][0] - iT[18] * a[0][5] + iT[19] * a[5][0] - t[0] + add) >> shift;
+ dst[26 * line] = ( iT[0] * a[8][0] + iT[11] * a[1][0] - iT[13] * a[4][5] - iT[24] * a[7][0] + iT[1] * a[5][4] + iT[10] * a[9][4] + iT[14] * a[7][1] - iT[23] * a[2][1] - iT[2] * a[1][2] - iT[9] * a[0][2] + iT[15] * a[2][2] + iT[22] * a[3][2] - iT[3] * a[9][2] - iT[8] * a[8][3] - iT[16] * a[3][3] - iT[21] * a[6][2] + iT[4] * a[0][4] - iT[7] * a[5][1] + iT[17] * a[6][1] + iT[20] * a[4][1] + iT[5] * a[8][5] + iT[6] * a[1][5] - iT[18] * a[4][0] - iT[19] * a[7][5] - t[1] + add) >> shift;
+ dst[28 * line] = ( - iT[0] * a[5][1] - iT[11] * a[9][1] - iT[13] * a[7][4] + iT[24] * a[2][4] + iT[1] * a[8][2] + iT[10] * a[9][3] + iT[14] * a[6][3] + iT[23] * a[3][2] - iT[2] * a[9][4] - iT[9] * a[8][1] - iT[15] * a[3][1] - iT[22] * a[6][4] + iT[3] * a[9][0] + iT[8] * a[5][0] - iT[16] * a[2][5] + iT[21] * a[7][5] - iT[4] * a[5][5] + iT[7] * a[0][0] + iT[17] * a[4][5] + iT[20] * a[6][5] + iT[5] * a[1][3] + iT[6] * a[0][3] - iT[18] * a[2][3] - iT[19] * a[3][3] + t[1] + add) >> shift;
+ dst[29 * line] = ( iT[0] * a[6][4] + iT[11] * a[3][1] + iT[13] * a[9][4] + iT[24] * a[8][1] - iT[1] * a[7][3] - iT[10] * a[4][2] + iT[14] * a[8][3] + iT[23] * a[1][3] - iT[2] * a[3][5] - iT[9] * a[2][5] + iT[15] * a[1][5] + iT[22] * a[0][5] + iT[3] * a[2][4] + iT[8] * a[3][4] - iT[16] * a[0][4] - iT[21] * a[1][4] + iT[4] * a[4][3] + iT[7] * a[7][2] - iT[17] * a[1][2] - iT[20] * a[8][2] - iT[5] * a[3][0] - iT[6] * a[6][5] - iT[18] * a[8][0] - iT[19] * a[9][5] + t[0] + add) >> shift;
+ dst[30 * line] = ( - iT[0] * a[7][2] + iT[11] * a[2][2] - iT[13] * a[9][3] - iT[24] * a[5][3] - iT[1] * a[6][0] - iT[10] * a[4][0] + iT[14] * a[5][0] - iT[23] * a[0][5] - iT[2] * a[4][2] - iT[9] * a[6][2] - iT[15] * a[0][3] + iT[22] * a[5][2] + iT[3] * a[2][0] - iT[8] * a[7][0] - iT[16] * a[5][5] - iT[21] * a[9][5] + iT[4] * a[7][1] - iT[7] * a[2][1] + iT[17] * a[9][4] + iT[20] * a[5][4] + iT[5] * a[6][1] + iT[6] * a[4][1] - iT[18] * a[5][1] + iT[19] * a[0][4] + t[0] + add) >> shift;
+ dst[31 * line] = ( iT[0] * a[8][5] + iT[11] * a[1][5] - iT[13] * a[4][0] - iT[24] * a[7][5] - iT[1] * a[1][0] - iT[10] * a[8][0] + iT[14] * a[7][0] + iT[23] * a[4][5] - iT[2] * a[8][4] - iT[9] * a[1][4] + iT[15] * a[4][1] + iT[22] * a[7][4] + iT[3] * a[1][1] + iT[8] * a[8][1] - iT[16] * a[7][1] - iT[21] * a[4][4] + iT[4] * a[8][3] + iT[7] * a[1][3] - iT[17] * a[4][2] - iT[20] * a[7][3] - iT[5] * a[1][2] - iT[6] * a[8][2] + iT[18] * a[7][2] + iT[19] * a[4][3] + t[1] + add) >> shift;
+
+ dst[ 2 * line] = ( iT[ 4] * b[0] + iT[ 9] * b[1] + iT[14] * b[2] + iT[19] * b[3] + iT[24] * b[4] + iT[29] * b[5] + add) >> shift;
+ dst[ 7 * line] = ( - iT[14] * b[0] - iT[29] * b[1] - iT[19] * b[2] - iT[ 4] * b[3] + iT[ 9] * b[4] + iT[24] * b[5] + add) >> shift;
+ dst[12 * line] = ( iT[24] * b[0] + iT[14] * b[1] - iT[ 9] * b[2] - iT[29] * b[3] - iT[ 4] * b[4] + iT[19] * b[5] + add) >> shift;
+ dst[17 * line] = ( - iT[29] * b[0] + iT[ 4] * b[1] + iT[24] * b[2] - iT[ 9] * b[3] - iT[19] * b[4] + iT[14] * b[5] + add) >> shift;
+ dst[22 * line] = ( iT[19] * b[0] - iT[24] * b[1] + iT[ 4] * b[2] + iT[14] * b[3] - iT[29] * b[4] + iT[ 9] * b[5] + add) >> shift;
+ dst[27 * line] = ( - iT[ 9] * b[0] + iT[19] * b[1] - iT[29] * b[2] + iT[24] * b[3] - iT[14] * b[4] + iT[ 4] * b[5] + add) >> shift;
+
+ dst[ 6 * line] = ( iT[12] * c[0] + iT[25] * c[1] + add) >> shift;
+ dst[19 * line] = ( - iT[25] * c[0] + iT[12] * c[1] + add) >> shift;
+
+ src += 32;
+ dst++;
+ }
+
+ if (iSkipLine)
+ {
+ dst = pCoef + reducedLine;
+ for (j = 0; j < cutoff; j++)
+ {
+ memset(dst, 0, sizeof(TCoeff)*iSkipLine);
+ dst += line;
+ }
+ }
+
+ if (iSkipLine2)
+ {
+ dst = pCoef + line * cutoff;
+ memset(dst, 0, sizeof(TCoeff) * line * iSkipLine2);
+ }
+#else
_fastForwardMM< 32 >( src, dst, shift, line, iSkipLine, iSkipLine2, g_trCoreDCT8P32[TRANSFORM_FORWARD][0] );
+#endif
}
void fastInverseDCT8_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
{
+#if JVET_M0497_FAST_DST7 && !JVET_M0497_MATRIX_MULT
+ int j, k;
+ TCoeff a[10][6];
+ TCoeff t[2];
+ TCoeff b[6];
+ TCoeff c[2];
+ TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
+
+ const TMatrixCoeff *iT = g_trCoreDST7P32[TRANSFORM_INVERSE][0];
+
+ const int reducedLine = line - iSkipLine;
+
+ for (j = 0; j < reducedLine; j++)
+ {
+ for (k = 0; k < 6; k++)
+ {
+ a[0][k] = src[(31 - k)*line] - src[(20 + k)*line];
+ a[1][k] = src[(31 - k)*line] + src[(18 - k)*line];
+ a[2][k] = src[(31 - k)*line] + src[( 7 + k)*line];
+ a[3][k] = src[(31 - k)*line] - src[( 5 - k)*line];
+ a[4][k] = src[(25 - k)*line] + src[(13 + k)*line];
+ a[5][k] = src[(25 - k)*line] + src[(12 - k)*line];
+ a[6][k] = src[(25 - k)*line] - src[ k *line];
+ a[7][k] = src[(18 - k)*line] - src[( 7 + k)*line];
+ a[8][k] = src[(18 - k)*line] + src[( 5 - k)*line];
+ a[9][k] = src[(12 - k)*line] + src[ k *line];
+
+ b[k] = src[(31 - k)*line] + src[(20 + k)*line] - src[(18 - k)*line] - src[(7 + k)*line] + src[(5 - k)*line];
+ }
+
+ for (k = 0; k < 2; k++)
+ {
+ c[k] = src[(31 - k)*line] + src[(28 + k)*line] - src[(26 - k)*line] - src[(23 + k)*line] + src[(21 - k)*line] + src[(18 + k)*line] - src[(16 - k)*line] - src[(13 + k)*line] + src[(11 - k)*line] + src[(8 + k)*line] - src[(6 - k)*line] - src[(3 + k)*line] + src[(1 - k)*line];
+ }
+
+ t[0] = iT[12] * src[19 * line] + iT[25] * src[ 6 * line];
+ t[1] = iT[12] * src[ 6 * line] - iT[25] * src[19 * line];
+
+ dst[ 0] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[3][0] + iT[11] * a[6][5] + iT[13] * a[8][0] + iT[24] * a[9][5] + iT[1] * a[3][1] + iT[10] * a[6][4] + iT[14] * a[8][1] + iT[23] * a[9][4] + iT[2] * a[3][2] + iT[9] * a[6][3] + iT[15] * a[8][2] + iT[22] * a[9][3] + iT[3] * a[3][3] + iT[8] * a[6][2] + iT[16] * a[8][3] + iT[21] * a[9][2] + iT[4] * a[3][4] + iT[7] * a[6][1] + iT[17] * a[8][4] + iT[20] * a[9][1] + iT[5] * a[3][5] + iT[6] * a[6][0] + iT[18] * a[8][5] + iT[19] * a[9][0] + t[0] + add) >> shift);
+ dst[ 1] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[5][2] - iT[11] * a[0][3] - iT[13] * a[4][2] - iT[24] * a[6][2] - iT[1] * a[9][1] - iT[10] * a[8][4] - iT[14] * a[3][4] - iT[23] * a[6][1] - iT[2] * a[0][0] + iT[9] * a[5][5] - iT[15] * a[6][5] - iT[22] * a[4][5] + iT[3] * a[5][3] - iT[8] * a[0][2] - iT[16] * a[4][3] - iT[21] * a[6][3] - iT[4] * a[9][0] - iT[7] * a[8][5] - iT[17] * a[3][5] - iT[20] * a[6][0] - iT[5] * a[0][1] + iT[6] * a[5][4] - iT[18] * a[6][4] - iT[19] * a[4][4] + t[1] + add) >> shift);
+ dst[ 3] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[9][4] + iT[11] * a[5][4] - iT[13] * a[2][1] + iT[24] * a[7][1] + iT[1] * a[0][3] + iT[10] * a[1][3] - iT[14] * a[3][3] - iT[23] * a[2][3] - iT[2] * a[8][5] - iT[9] * a[9][0] - iT[15] * a[6][0] - iT[22] * a[3][5] + iT[3] * a[1][4] + iT[8] * a[0][4] - iT[16] * a[2][4] - iT[21] * a[3][4] + iT[4] * a[5][3] + iT[7] * a[9][3] + iT[17] * a[7][2] - iT[20] * a[2][2] - iT[5] * a[8][0] - iT[6] * a[1][0] + iT[18] * a[4][5] + iT[19] * a[7][0] - t[1] + add) >> shift);
+ dst[ 4] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[3][2] - iT[11] * a[2][2] + iT[13] * a[1][2] + iT[24] * a[0][2] + iT[1] * a[6][0] + iT[10] * a[3][5] + iT[14] * a[9][0] + iT[23] * a[8][5] - iT[2] * a[2][3] - iT[9] * a[3][3] + iT[15] * a[0][3] + iT[22] * a[1][3] - iT[3] * a[7][0] + iT[8] * a[2][0] - iT[16] * a[9][5] - iT[21] * a[5][5] + iT[4] * a[4][4] + iT[7] * a[6][4] + iT[17] * a[0][1] - iT[20] * a[5][4] - iT[5] * a[7][4] - iT[6] * a[4][1] + iT[18] * a[8][4] + iT[19] * a[1][4] - t[0] + add) >> shift);
+ dst[ 5] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[3][5] + iT[11] * a[6][0] + iT[13] * a[8][5] + iT[24] * a[9][0] - iT[1] * a[6][5] - iT[10] * a[3][0] - iT[14] * a[9][5] - iT[23] * a[8][0] + iT[2] * a[7][4] - iT[9] * a[2][4] + iT[15] * a[9][1] + iT[22] * a[5][1] + iT[3] * a[7][1] + iT[8] * a[4][4] - iT[16] * a[8][1] - iT[21] * a[1][1] - iT[4] * a[6][2] - iT[7] * a[4][2] + iT[17] * a[5][2] - iT[20] * a[0][3] + iT[5] * a[3][2] + iT[6] * a[2][2] - iT[18] * a[1][2] - iT[19] * a[0][2] - t[0] + add) >> shift);
+ dst[ 8] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[9][3] + iT[11] * a[8][2] + iT[13] * a[3][2] + iT[24] * a[6][3] + iT[1] * a[1][5] + iT[10] * a[0][5] - iT[14] * a[2][5] - iT[23] * a[3][5] - iT[2] * a[1][3] - iT[9] * a[8][3] + iT[15] * a[7][3] + iT[22] * a[4][2] - iT[3] * a[9][5] - iT[8] * a[5][5] + iT[16] * a[2][0] - iT[21] * a[7][0] - iT[4] * a[1][1] - iT[7] * a[0][1] + iT[17] * a[2][1] + iT[20] * a[3][1] + iT[5] * a[5][1] + iT[6] * a[9][1] + iT[18] * a[7][4] - iT[19] * a[2][4] + t[1] + add) >> shift);
+ dst[ 9] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[2][1] + iT[11] * a[3][1] - iT[13] * a[0][1] - iT[24] * a[1][1] - iT[1] * a[7][3] + iT[10] * a[2][3] - iT[14] * a[9][2] - iT[23] * a[5][2] - iT[2] * a[4][0] - iT[9] * a[7][5] + iT[15] * a[1][5] + iT[22] * a[8][5] - iT[3] * a[3][4] - iT[8] * a[2][4] + iT[16] * a[1][4] + iT[21] * a[0][4] - iT[4] * a[6][3] - iT[7] * a[3][2] - iT[17] * a[9][3] - iT[20] * a[8][2] - iT[5] * a[4][5] - iT[6] * a[6][5] - iT[18] * a[0][0] + iT[19] * a[5][5] + t[0] + add) >> shift);
+ dst[10] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[6][1] - iT[11] * a[4][1] + iT[13] * a[5][1] - iT[24] * a[0][4] + iT[1] * a[2][2] - iT[10] * a[7][2] - iT[14] * a[5][3] - iT[23] * a[9][3] + iT[2] * a[6][4] + iT[9] * a[4][4] - iT[15] * a[5][4] + iT[22] * a[0][1] - iT[3] * a[2][5] + iT[8] * a[7][5] + iT[16] * a[5][0] + iT[21] * a[9][0] - iT[4] * a[7][0] - iT[7] * a[4][5] + iT[17] * a[8][0] + iT[20] * a[1][0] + iT[5] * a[4][2] + iT[6] * a[7][3] - iT[18] * a[1][3] - iT[19] * a[8][3] + t[0] + add) >> shift);
+ dst[11] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[1][3] - iT[11] * a[0][3] + iT[13] * a[2][3] + iT[24] * a[3][3] - iT[1] * a[9][1] - iT[10] * a[5][1] + iT[14] * a[2][4] - iT[23] * a[7][4] - iT[2] * a[8][0] - iT[9] * a[9][5] - iT[15] * a[6][5] - iT[22] * a[3][0] + iT[3] * a[0][2] - iT[8] * a[5][3] + iT[16] * a[6][3] + iT[21] * a[4][3] + iT[4] * a[5][0] - iT[7] * a[0][5] - iT[17] * a[4][0] - iT[20] * a[6][0] + iT[5] * a[9][4] + iT[6] * a[5][4] - iT[18] * a[2][1] + iT[19] * a[7][1] + t[1] + add) >> shift);
+ dst[13] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[0][0] + iT[11] * a[1][0] - iT[13] * a[3][0] - iT[24] * a[2][0] + iT[1] * a[5][4] - iT[10] * a[0][1] - iT[14] * a[4][4] - iT[23] * a[6][4] - iT[2] * a[9][3] - iT[9] * a[5][3] + iT[15] * a[2][2] - iT[22] * a[7][2] + iT[3] * a[8][3] + iT[8] * a[9][2] + iT[16] * a[6][2] + iT[21] * a[3][3] - iT[4] * a[1][4] - iT[7] * a[8][4] + iT[17] * a[7][4] + iT[20] * a[4][1] + iT[5] * a[0][5] + iT[6] * a[1][5] - iT[18] * a[3][5] - iT[19] * a[2][5] - t[1] + add) >> shift);
+ dst[14] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[4][2] + iT[11] * a[7][3] - iT[13] * a[1][3] - iT[24] * a[8][3] + iT[1] * a[4][1] + iT[10] * a[6][1] + iT[14] * a[0][4] - iT[23] * a[5][1] - iT[2] * a[3][0] - iT[9] * a[2][0] + iT[15] * a[1][0] + iT[22] * a[0][0] - iT[3] * a[6][3] - iT[8] * a[4][3] + iT[16] * a[5][3] - iT[21] * a[0][2] - iT[4] * a[7][5] - iT[7] * a[4][0] + iT[17] * a[8][5] + iT[20] * a[1][5] + iT[5] * a[6][4] + iT[6] * a[3][1] + iT[18] * a[9][4] + iT[19] * a[8][1] - t[0] + add) >> shift);
+ dst[15] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[7][4] + iT[11] * a[4][1] - iT[13] * a[8][4] - iT[24] * a[1][4] - iT[1] * a[2][2] - iT[10] * a[3][2] + iT[14] * a[0][2] + iT[23] * a[1][2] - iT[2] * a[2][1] + iT[9] * a[7][1] + iT[15] * a[5][4] + iT[22] * a[9][4] + iT[3] * a[7][5] - iT[8] * a[2][5] + iT[16] * a[9][0] + iT[21] * a[5][0] + iT[4] * a[2][0] + iT[7] * a[3][0] - iT[17] * a[0][0] - iT[20] * a[1][0] + iT[5] * a[2][3] - iT[6] * a[7][3] - iT[18] * a[5][2] - iT[19] * a[9][2] - t[0] + add) >> shift);
+ dst[16] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[0][1] + iT[11] * a[5][4] - iT[13] * a[6][4] - iT[24] * a[4][4] + iT[1] * a[0][3] - iT[10] * a[5][2] + iT[14] * a[6][2] + iT[23] * a[4][2] - iT[2] * a[0][5] + iT[9] * a[5][0] - iT[15] * a[6][0] - iT[22] * a[4][0] - iT[3] * a[0][4] - iT[8] * a[1][4] + iT[16] * a[3][4] + iT[21] * a[2][4] + iT[4] * a[0][2] + iT[7] * a[1][2] - iT[17] * a[3][2] - iT[20] * a[2][2] - iT[5] * a[0][0] - iT[6] * a[1][0] + iT[18] * a[3][0] + iT[19] * a[2][0] - t[1] + add) >> shift);
+ dst[18] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[0][5] + iT[11] * a[1][5] - iT[13] * a[3][5] - iT[24] * a[2][5] - iT[1] * a[1][0] - iT[10] * a[0][0] + iT[14] * a[2][0] + iT[23] * a[3][0] - iT[2] * a[5][1] + iT[9] * a[0][4] + iT[15] * a[4][1] + iT[22] * a[6][1] - iT[3] * a[8][1] - iT[8] * a[1][1] + iT[16] * a[4][4] + iT[21] * a[7][1] - iT[4] * a[9][2] - iT[7] * a[5][2] + iT[17] * a[2][3] - iT[20] * a[7][3] - iT[5] * a[9][3] - iT[6] * a[8][2] - iT[18] * a[3][2] - iT[19] * a[6][3] + t[1] + add) >> shift);
+ dst[20] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[4][0] - iT[11] * a[6][0] - iT[13] * a[0][5] + iT[24] * a[5][0] + iT[1] * a[6][5] + iT[10] * a[4][5] - iT[14] * a[5][5] + iT[23] * a[0][0] - iT[2] * a[6][1] - iT[9] * a[3][4] - iT[15] * a[9][1] - iT[22] * a[8][4] + iT[3] * a[4][4] + iT[8] * a[7][1] - iT[16] * a[1][1] - iT[21] * a[8][1] - iT[4] * a[3][3] - iT[7] * a[2][3] + iT[17] * a[1][3] + iT[20] * a[0][3] + iT[5] * a[7][2] - iT[6] * a[2][2] + iT[18] * a[9][3] + iT[19] * a[5][3] + t[0] + add) >> shift);
+ dst[21] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[1][2] + iT[11] * a[8][2] - iT[13] * a[7][2] - iT[24] * a[4][3] + iT[1] * a[1][5] + iT[10] * a[8][5] - iT[14] * a[7][5] - iT[23] * a[4][0] + iT[2] * a[5][2] + iT[9] * a[9][2] + iT[15] * a[7][3] - iT[22] * a[2][3] + iT[3] * a[5][5] + iT[8] * a[9][5] + iT[16] * a[7][0] - iT[21] * a[2][0] + iT[4] * a[8][1] + iT[7] * a[9][4] + iT[17] * a[6][4] + iT[20] * a[3][1] + iT[5] * a[8][4] + iT[6] * a[9][1] + iT[18] * a[6][1] + iT[19] * a[3][4] + t[1] + add) >> shift);
+ dst[23] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[8][4] + iT[11] * a[9][1] + iT[13] * a[6][1] + iT[24] * a[3][4] - iT[1] * a[8][2] - iT[10] * a[1][2] + iT[14] * a[4][3] + iT[23] * a[7][2] - iT[2] * a[0][1] - iT[9] * a[1][1] + iT[15] * a[3][1] + iT[22] * a[2][1] + iT[3] * a[5][0] + iT[8] * a[9][0] + iT[16] * a[7][5] - iT[21] * a[2][5] - iT[4] * a[9][5] - iT[7] * a[8][0] - iT[17] * a[3][0] - iT[20] * a[6][5] + iT[5] * a[5][2] - iT[6] * a[0][3] - iT[18] * a[4][2] - iT[19] * a[6][2] - t[1] + add) >> shift);
+ dst[24] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[2][3] + iT[11] * a[7][3] + iT[13] * a[5][2] + iT[24] * a[9][2] + iT[1] * a[4][1] + iT[10] * a[7][4] - iT[14] * a[1][4] - iT[23] * a[8][4] - iT[2] * a[4][5] - iT[9] * a[7][0] + iT[15] * a[1][0] + iT[22] * a[8][0] + iT[3] * a[4][3] + iT[8] * a[6][3] + iT[16] * a[0][2] - iT[21] * a[5][3] - iT[4] * a[2][5] - iT[7] * a[3][5] + iT[17] * a[0][5] + iT[20] * a[1][5] + iT[5] * a[2][1] + iT[6] * a[3][1] - iT[18] * a[0][1] - iT[19] * a[1][1] - t[0] + add) >> shift);
+ dst[25] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[4][5] - iT[11] * a[6][5] - iT[13] * a[0][0] + iT[24] * a[5][5] - iT[1] * a[3][1] - iT[10] * a[2][1] + iT[14] * a[1][1] + iT[23] * a[0][1] + iT[2] * a[7][2] + iT[9] * a[4][3] - iT[15] * a[8][2] - iT[22] * a[1][2] + iT[3] * a[6][2] + iT[8] * a[3][3] + iT[16] * a[9][2] + iT[21] * a[8][3] + iT[4] * a[2][4] - iT[7] * a[7][4] - iT[17] * a[5][1] - iT[20] * a[9][1] - iT[5] * a[4][0] - iT[6] * a[6][0] - iT[18] * a[0][5] + iT[19] * a[5][0] - t[0] + add) >> shift);
+ dst[26] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[8][0] + iT[11] * a[1][0] - iT[13] * a[4][5] - iT[24] * a[7][0] + iT[1] * a[5][4] + iT[10] * a[9][4] + iT[14] * a[7][1] - iT[23] * a[2][1] - iT[2] * a[1][2] - iT[9] * a[0][2] + iT[15] * a[2][2] + iT[22] * a[3][2] - iT[3] * a[9][2] - iT[8] * a[8][3] - iT[16] * a[3][3] - iT[21] * a[6][2] + iT[4] * a[0][4] - iT[7] * a[5][1] + iT[17] * a[6][1] + iT[20] * a[4][1] + iT[5] * a[8][5] + iT[6] * a[1][5] - iT[18] * a[4][0] - iT[19] * a[7][5] - t[1] + add) >> shift);
+ dst[28] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[5][1] - iT[11] * a[9][1] - iT[13] * a[7][4] + iT[24] * a[2][4] + iT[1] * a[8][2] + iT[10] * a[9][3] + iT[14] * a[6][3] + iT[23] * a[3][2] - iT[2] * a[9][4] - iT[9] * a[8][1] - iT[15] * a[3][1] - iT[22] * a[6][4] + iT[3] * a[9][0] + iT[8] * a[5][0] - iT[16] * a[2][5] + iT[21] * a[7][5] - iT[4] * a[5][5] + iT[7] * a[0][0] + iT[17] * a[4][5] + iT[20] * a[6][5] + iT[5] * a[1][3] + iT[6] * a[0][3] - iT[18] * a[2][3] - iT[19] * a[3][3] + t[1] + add) >> shift);
+ dst[29] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[6][4] + iT[11] * a[3][1] + iT[13] * a[9][4] + iT[24] * a[8][1] - iT[1] * a[7][3] - iT[10] * a[4][2] + iT[14] * a[8][3] + iT[23] * a[1][3] - iT[2] * a[3][5] - iT[9] * a[2][5] + iT[15] * a[1][5] + iT[22] * a[0][5] + iT[3] * a[2][4] + iT[8] * a[3][4] - iT[16] * a[0][4] - iT[21] * a[1][4] + iT[4] * a[4][3] + iT[7] * a[7][2] - iT[17] * a[1][2] - iT[20] * a[8][2] - iT[5] * a[3][0] - iT[6] * a[6][5] - iT[18] * a[8][0] - iT[19] * a[9][5] + t[0] + add) >> shift);
+ dst[30] = Clip3(outputMinimum, outputMaximum, (int)( - iT[0] * a[7][2] + iT[11] * a[2][2] - iT[13] * a[9][3] - iT[24] * a[5][3] - iT[1] * a[6][0] - iT[10] * a[4][0] + iT[14] * a[5][0] - iT[23] * a[0][5] - iT[2] * a[4][2] - iT[9] * a[6][2] - iT[15] * a[0][3] + iT[22] * a[5][2] + iT[3] * a[2][0] - iT[8] * a[7][0] - iT[16] * a[5][5] - iT[21] * a[9][5] + iT[4] * a[7][1] - iT[7] * a[2][1] + iT[17] * a[9][4] + iT[20] * a[5][4] + iT[5] * a[6][1] + iT[6] * a[4][1] - iT[18] * a[5][1] + iT[19] * a[0][4] + t[0] + add) >> shift);
+ dst[31] = Clip3(outputMinimum, outputMaximum, (int)( iT[0] * a[8][5] + iT[11] * a[1][5] - iT[13] * a[4][0] - iT[24] * a[7][5] - iT[1] * a[1][0] - iT[10] * a[8][0] + iT[14] * a[7][0] + iT[23] * a[4][5] - iT[2] * a[8][4] - iT[9] * a[1][4] + iT[15] * a[4][1] + iT[22] * a[7][4] + iT[3] * a[1][1] + iT[8] * a[8][1] - iT[16] * a[7][1] - iT[21] * a[4][4] + iT[4] * a[8][3] + iT[7] * a[1][3] - iT[17] * a[4][2] - iT[20] * a[7][3] - iT[5] * a[1][2] - iT[6] * a[8][2] + iT[18] * a[7][2] + iT[19] * a[4][3] + t[1] + add) >> shift);
+
+ dst[ 2] = Clip3(outputMinimum, outputMaximum, (int)( iT[ 4] * b[0] + iT[ 9] * b[1] + iT[14] * b[2] + iT[19] * b[3] + iT[24] * b[4] + iT[29] * b[5] + add) >> shift);
+ dst[ 7] = Clip3(outputMinimum, outputMaximum, (int)( - iT[14] * b[0] - iT[29] * b[1] - iT[19] * b[2] - iT[ 4] * b[3] + iT[ 9] * b[4] + iT[24] * b[5] + add) >> shift);
+ dst[12] = Clip3(outputMinimum, outputMaximum, (int)( iT[24] * b[0] + iT[14] * b[1] - iT[ 9] * b[2] - iT[29] * b[3] - iT[ 4] * b[4] + iT[19] * b[5] + add) >> shift);
+ dst[17] = Clip3(outputMinimum, outputMaximum, (int)( - iT[29] * b[0] + iT[ 4] * b[1] + iT[24] * b[2] - iT[ 9] * b[3] - iT[19] * b[4] + iT[14] * b[5] + add) >> shift);
+ dst[22] = Clip3(outputMinimum, outputMaximum, (int)( iT[19] * b[0] - iT[24] * b[1] + iT[ 4] * b[2] + iT[14] * b[3] - iT[29] * b[4] + iT[ 9] * b[5] + add) >> shift);
+ dst[27] = Clip3(outputMinimum, outputMaximum, (int)( - iT[ 9] * b[0] + iT[19] * b[1] - iT[29] * b[2] + iT[24] * b[3] - iT[14] * b[4] + iT[ 4] * b[5] + add) >> shift);
+
+ dst[ 6] = Clip3(outputMinimum, outputMaximum, (int)( iT[12] * c[0] + iT[25] * c[1] + add) >> shift);
+ dst[19] = Clip3(outputMinimum, outputMaximum, (int)( - iT[25] * c[0] + iT[12] * c[1] + add) >> shift);
+
+ src++;
+ dst += 32;
+ }
+
+ if (iSkipLine)
+ {
+ memset(dst, 0, (iSkipLine * 32) * sizeof(TCoeff));
+ }
+#else
_fastInverseMM< 32 >( src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_trCoreDCT8P32[TRANSFORM_INVERSE][0] );
+#endif
}
diff --git a/source/Lib/CommonLib/TypeDef.h b/source/Lib/CommonLib/TypeDef.h
index 1d0452b39117b53d0fda1c2196cd0bb4e042e83a..7e15e14992f67adeb9ea2d1d6e8c0f46e1cda2e2 100644
--- a/source/Lib/CommonLib/TypeDef.h
+++ b/source/Lib/CommonLib/TypeDef.h
@@ -50,9 +50,53 @@
#include <assert.h>
#include <cassert>
+#define JVET_M0118_M0185_TRIANGLE_FLAG_FIX 1 // Avoid signaling triangle flag if a CU uses MMVD or CIIP
+
+#define JVET_M0487_INT_EXTEND 1 // CE9.1.1 b: integer reference samples as 1 extended samples
+
+#define JVET_M0444_SMVD 1 // SMVD mode
+
+#define JVET_M0170_MRG_SHARELIST 1
+#if JVET_M0170_MRG_SHARELIST
+#define MRG_SHARELIST_SHARSIZE 32
+#endif
+
+#define JVET_M0064_CCLM_SIMPLIFICATION 1
+
+#define JVET_M0142_CCLM_COLLOCATED_CHROMA 1 // Adding support for chroma sample location type 2 in CCLM
+
+
+#define JVET_M0479_18BITS_MV_CLIP 1
+
+#define JVET_M0497_FAST_DST7 1
+#if JVET_M0497_FAST_DST7
+#define JVET_M0497_MATRIX_MULT 0 // 0: Fast method; 1: Matrix multiplication
+#endif
+#define JVET_M0502_PRED_MODE_CTX 1
+
+#define JVET_M0407_IBC_RANGE 1 // extend IBC search range to some part of left CTU
+
+#define JVET_M0464_UNI_MTS 1
+#define JVET_M0068_M0171_MMVD_CLEANUP 1 // MMVD cleanup with 1) flip removal, 2) L1 zero vector fix, 3) bi-pred restriction after merge/MMVD
+
+#if JVET_M0464_UNI_MTS
+typedef std::pair<int, bool> TrMode;
+typedef std::pair<int, int> TrCost;
+#endif
+
+#define JVET_M0421_SPLIT_SIG 1
+
+#define JVET_M0173_MOVE_GT2_TO_FIRST_PASS 1 // Moving the gtr2 flag to the first coding pass
+
+#define REMOVE_BIN_DECISION_TREE 1
+
+#define JVET_M0446_M0888_M0905_VPDU_AT_PIC_BOUNDARY 1
+
// clang-format off
#define JVET_M0453_CABAC_ENGINE 1
+#define JVET_M0409_ATMVP_FIX 1
+
#define JVET_L0090_PAIR_AVG 1 // Add pairwise average candidates, replace HEVC combined candidates
#define REUSE_CU_RESULTS 1
// clang-format on
@@ -839,7 +883,7 @@ enum MergeType
{
MRG_TYPE_DEFAULT_N = 0, // 0
MRG_TYPE_SUBPU_ATMVP,
- MRG_TYPE_CPR,
+ MRG_TYPE_IBC,
NUM_MRG_TYPE // 5
};
@@ -850,6 +894,14 @@ enum TriangleSplit
TRIANGLE_DIR_NUM
};
+#if JVET_M0170_MRG_SHARELIST
+enum SharedMrgState
+{
+ NO_SHARE = 0,
+ GEN_ON_SHARED_BOUND = 1,
+ SHARING = 2
+};
+#endif
//////////////////////////////////////////////////////////////////////////
// Encoder modes to try out
//////////////////////////////////////////////////////////////////////////
diff --git a/source/Lib/CommonLib/Unit.cpp b/source/Lib/CommonLib/Unit.cpp
index 30286edd9bf4c88dd17cf26614ed33172eb188f3..ecaf89ca16990d33b31fcee0e1c6c86d1e0c0d15 100644
--- a/source/Lib/CommonLib/Unit.cpp
+++ b/source/Lib/CommonLib/Unit.cpp
@@ -261,7 +261,9 @@ CodingUnit& CodingUnit::operator=( const CodingUnit& other )
qp = other.qp;
chromaQpAdj = other.chromaQpAdj;
rootCbf = other.rootCbf;
+#if !JVET_M0464_UNI_MTS
emtFlag = other.emtFlag;
+#endif
#if HEVC_TILES_WPP
tileIdx = other.tileIdx;
#endif
@@ -270,7 +272,16 @@ CodingUnit& CodingUnit::operator=( const CodingUnit& other )
GBiIdx = other.GBiIdx;
for (int i = 0; i<2; i++)
refIdxBi[i] = other.refIdxBi[i];
- cpr = other.cpr;
+
+#if JVET_M0170_MRG_SHARELIST
+ shareParentPos = other.shareParentPos;
+ shareParentSize = other.shareParentSize;
+#endif
+ ibc = other.ibc;
+#if JVET_M0444_SMVD
+ smvdMode = other.smvdMode;
+#endif
+
return *this;
}
@@ -292,7 +303,9 @@ void CodingUnit::initData()
qp = 0;
chromaQpAdj = 0;
rootCbf = true;
+#if !JVET_M0464_UNI_MTS
emtFlag = 0;
+#endif
#if HEVC_TILES_WPP
tileIdx = 0;
#endif
@@ -301,7 +314,15 @@ void CodingUnit::initData()
GBiIdx = GBI_DEFAULT;
for (int i = 0; i < 2; i++)
refIdxBi[i] = -1;
- cpr = false;
+#if JVET_M0170_MRG_SHARELIST
+ shareParentPos = Position(-1, -1);
+ shareParentSize.width = -1;
+ shareParentSize.height = -1;
+#endif
+ ibc = false;
+#if JVET_M0444_SMVD
+ smvdMode = 0;
+#endif
}
@@ -345,6 +366,11 @@ void PredictionUnit::initData()
}
}
mhIntraFlag = false;
+#if JVET_M0170_MRG_SHARELIST
+ shareParentPos = Position(-1, -1);
+ shareParentSize.width = -1;
+ shareParentSize.height = -1;
+#endif
}
PredictionUnit& PredictionUnit::operator=(const IntraPredictionData& predData)
@@ -385,7 +411,10 @@ PredictionUnit& PredictionUnit::operator=(const InterPredictionData& predData)
}
}
mhIntraFlag = predData.mhIntraFlag;
-
+#if JVET_M0170_MRG_SHARELIST
+ shareParentPos = predData.shareParentPos;
+ shareParentSize = predData.shareParentSize;
+#endif
return *this;
}
@@ -422,7 +451,10 @@ PredictionUnit& PredictionUnit::operator=( const PredictionUnit& other )
}
}
mhIntraFlag = other.mhIntraFlag;
-
+#if JVET_M0170_MRG_SHARELIST
+ shareParentPos = other.shareParentPos;
+ shareParentSize = other.shareParentSize;
+#endif
return *this;
}
@@ -493,12 +525,17 @@ void TransformUnit::initData()
{
cbf[i] = 0;
rdpcm[i] = NUMBER_OF_RDPCM_MODES;
+#if !JVET_M0464_UNI_MTS
transformSkip[i] = false;
+#endif
compAlpha[i] = 0;
}
depth = 0;
+#if JVET_M0464_UNI_MTS
+ mtsIdx = 0;
+#else
emtIdx = 0;
-
+#endif
}
void TransformUnit::init(TCoeff **coeffs, Pel **pcmbuf)
@@ -528,11 +565,17 @@ TransformUnit& TransformUnit::operator=(const TransformUnit& other)
cbf[i] = other.cbf[i];
rdpcm[i] = other.rdpcm[i];
+#if !JVET_M0464_UNI_MTS
transformSkip[i] = other.transformSkip[i];
+#endif
compAlpha[i] = other.compAlpha[i];
}
depth = other.depth;
+#if JVET_M0464_UNI_MTS
+ mtsIdx = other.mtsIdx;
+#else
emtIdx = other.emtIdx;
+#endif
return *this;
}
@@ -549,15 +592,20 @@ void TransformUnit::copyComponentFrom(const TransformUnit& other, const Componen
cbf[i] = other.cbf[i];
rdpcm[i] = other.rdpcm[i];
+#if !JVET_M0464_UNI_MTS
transformSkip[i] = other.transformSkip[i];
+#endif
compAlpha[i] = other.compAlpha[i];
depth = other.depth;
-
+#if JVET_M0464_UNI_MTS
+ mtsIdx = isLuma( i ) ? other.mtsIdx : mtsIdx;
+#else
if( isLuma( i ) )
{
emtIdx = other.emtIdx;
}
+#endif
}
CoeffBuf TransformUnit::getCoeffs(const ComponentID id) { return CoeffBuf(m_coeffs[id], blocks[id]); }
diff --git a/source/Lib/CommonLib/Unit.h b/source/Lib/CommonLib/Unit.h
index 761da4cfe37a7b803d5d0a10aab70013849da36f..4eb057dde4a9cfe99bc1f4970d4f1e29e9f9613d 100644
--- a/source/Lib/CommonLib/Unit.h
+++ b/source/Lib/CommonLib/Unit.h
@@ -305,12 +305,21 @@ struct CodingUnit : public UnitArea
#if HEVC_TILES_WPP
uint32_t tileIdx;
#endif
+#if !JVET_M0464_UNI_MTS
uint8_t emtFlag;
+#endif
uint8_t GBiIdx;
int refIdxBi[2];
// needed for fast imv mode decisions
int8_t imvNumCand;
- bool cpr;
+#if JVET_M0170_MRG_SHARELIST
+ Position shareParentPos;
+ Size shareParentSize;
+#endif
+ bool ibc;
+#if JVET_M0444_SMVD
+ uint8_t smvdMode;
+#endif
CodingUnit() : chType( CH_L ) { }
CodingUnit(const UnitArea &unit);
@@ -361,8 +370,13 @@ struct InterPredictionData
Mv mvdAffi [NUM_REF_PIC_LIST_01][3];
Mv mvAffi[NUM_REF_PIC_LIST_01][3];
bool mhIntraFlag;
- Mv bv; // block vector for CPR
- Mv bvd; // block vector difference for CPR
+
+#if JVET_M0170_MRG_SHARELIST
+ Position shareParentPos;
+ Size shareParentSize;
+#endif
+ Mv bv; // block vector for IBC
+ Mv bvd; // block vector difference for IBC
};
struct PredictionUnit : public UnitArea, public IntraPredictionData, public InterPredictionData
@@ -384,6 +398,11 @@ struct PredictionUnit : public UnitArea, public IntraPredictionData, public Inte
PredictionUnit& operator=(const MotionInfo& mi);
unsigned idx;
+#if JVET_M0170_MRG_SHARELIST
+ Position shareParentPos;
+ Size shareParentSize;
+#endif
+
PredictionUnit *next;
// for accessing motion information, which can have higher resolution than PUs (should always be used, when accessing neighboring motion information)
@@ -410,11 +429,17 @@ struct TransformUnit : public UnitArea
ChannelType chType;
uint8_t depth;
+#if JVET_M0464_UNI_MTS
+ uint8_t mtsIdx;
+#else
uint8_t emtIdx;
- uint8_t cbf [ MAX_NUM_TBLOCKS ];
+#endif
+ uint8_t cbf [ MAX_NUM_TBLOCKS ];
RDPCMMode rdpcm [ MAX_NUM_TBLOCKS ];
+#if !JVET_M0464_UNI_MTS
bool transformSkip[ MAX_NUM_TBLOCKS ];
- int8_t compAlpha [ MAX_NUM_TBLOCKS ];
+#endif
+ int8_t compAlpha [ MAX_NUM_TBLOCKS ];
TransformUnit() : chType( CH_L ) { }
TransformUnit(const UnitArea& unit);
diff --git a/source/Lib/CommonLib/UnitPartitioner.cpp b/source/Lib/CommonLib/UnitPartitioner.cpp
index be921219b6e50cc53904cdfe371262b894b580cb..dae0769736f028afa368c3d303d82047aeb09b0b 100644
--- a/source/Lib/CommonLib/UnitPartitioner.cpp
+++ b/source/Lib/CommonLib/UnitPartitioner.cpp
@@ -297,8 +297,97 @@ void QTBTPartitioner::splitCurrArea( const PartSplit split, const CodingStructur
}
}
+#if JVET_M0421_SPLIT_SIG
+void QTBTPartitioner::canSplit( const CodingStructure &cs, bool& canNo, bool& canQt, bool& canBh, bool& canBv, bool& canTh, bool& canTv )
+{
+ const PartSplit implicitSplit = m_partStack.back().checkdIfImplicit ? m_partStack.back().implicitSplit : getImplicitSplit( cs );
+
+ const unsigned maxBTD = cs.pcv->getMaxBtDepth( *cs.slice, chType ) + currImplicitBtDepth;
+ const unsigned maxBtSize = cs.pcv->getMaxBtSize ( *cs.slice, chType );
+ const unsigned minBtSize = cs.pcv->getMinBtSize ( *cs.slice, chType );
+ const unsigned maxTtSize = cs.pcv->getMaxTtSize ( *cs.slice, chType );
+ const unsigned minTtSize = cs.pcv->getMinTtSize ( *cs.slice, chType );
+ const unsigned minQtSize = cs.pcv->getMinQtSize ( *cs.slice, chType );
+
+ canNo = canQt = canBh = canTh = canBv = canTv = true;
+ bool canBtt = currMtDepth < maxBTD;
+
+ // the minimal and maximal sizes are given in luma samples
+ const CompArea& area = currArea().Y();
+ PartLevel& level = m_partStack.back();
+
+ const PartSplit lastSplit = level.split;
+ const PartSplit parlSplit = lastSplit == CU_TRIH_SPLIT ? CU_HORZ_SPLIT : CU_VERT_SPLIT;
+
+ // don't allow QT-splitting below a BT split
+ if( lastSplit != CTU_LEVEL && lastSplit != CU_QUAD_SPLIT ) canQt = false;
+ if( area.width <= minQtSize ) canQt = false;
+
+ if( implicitSplit != CU_DONT_SPLIT )
+ {
+ canNo = canTh = canTv = false;
+
+ canBh = implicitSplit == CU_HORZ_SPLIT;
+ canBv = implicitSplit == CU_VERT_SPLIT;
+
+ return;
+ }
+
+ if( ( lastSplit == CU_TRIH_SPLIT || lastSplit == CU_TRIV_SPLIT ) && currPartIdx() == 1 )
+ {
+ canBh = parlSplit != CU_HORZ_SPLIT;
+ canBv = parlSplit != CU_VERT_SPLIT;
+ }
+
+ if( canBtt && ( area.width <= minBtSize && area.height <= minBtSize )
+ && ( ( area.width <= minTtSize && area.height <= minTtSize )
+ || cs.sps->getSpsNext().getMTTMode() == 0 ) )
+ {
+ canBtt = false;
+ }
+ if( canBtt && ( area.width > maxBtSize || area.height > maxBtSize )
+ && ( ( area.width > maxTtSize || area.height > maxTtSize )
+ || cs.sps->getSpsNext().getMTTMode() == 0 ) )
+ {
+ canBtt = false;
+ }
+
+ if( !canBtt )
+ {
+ canBh = canTh = canBv = canTv = false;
+
+ return;
+ }
+
+ // specific check for BT splits
+ if( area.height <= minBtSize || area.height > maxBtSize ) canBh = false;
+ if( area.width > MAX_TU_SIZE_FOR_PROFILE && area.height <= MAX_TU_SIZE_FOR_PROFILE ) canBh = false;
+
+ if( area.width <= minBtSize || area.width > maxBtSize ) canBv = false;
+ if( area.width <= MAX_TU_SIZE_FOR_PROFILE && area.height > MAX_TU_SIZE_FOR_PROFILE ) canBv = false;
+
+ if( ( cs.sps->getSpsNext().getMTTMode() & 1 ) != 1 ) canTh = false;
+ if( area.height <= 2 * minTtSize || area.height > maxTtSize || area.width > maxTtSize )
+ canTh = false;
+ if( area.width > MAX_TU_SIZE_FOR_PROFILE || area.height > MAX_TU_SIZE_FOR_PROFILE ) canTh = false;
+
+ if( ( cs.sps->getSpsNext().getMTTMode() & 1 ) != 1 ) canTv = false;
+ if( area.width <= 2 * minTtSize || area.width > maxTtSize || area.height > maxTtSize )
+ canTv = false;
+ if( area.width > MAX_TU_SIZE_FOR_PROFILE || area.height > MAX_TU_SIZE_FOR_PROFILE ) canTv = false;
+}
+
+#endif
bool QTBTPartitioner::canSplit( const PartSplit split, const CodingStructure &cs )
{
+#if JVET_M0421_SPLIT_SIG
+ const CompArea area = currArea().Y();
+ const unsigned maxTrSize = cs.sps->getMaxTrSize();
+
+ bool canNo, canQt, canBh, canTh, canBv, canTv;
+
+ canSplit( cs, canNo, canQt, canBh, canBv, canTh, canTv );
+#else
const PartSplit implicitSplit = getImplicitSplit( cs );
// the minimal and maximal sizes are given in luma samples
@@ -319,6 +408,7 @@ bool QTBTPartitioner::canSplit( const PartSplit split, const CodingStructure &cs
return false;
}
+#endif
switch( split )
{
case CTU_LEVEL:
@@ -328,6 +418,20 @@ bool QTBTPartitioner::canSplit( const PartSplit split, const CodingStructure &cs
case TU_MAX_TR_SPLIT:
return area.width > maxTrSize || area.height > maxTrSize;
break;
+#if JVET_M0421_SPLIT_SIG
+ case CU_QUAD_SPLIT:
+ return canQt;
+ case CU_DONT_SPLIT:
+ return canNo;
+ case CU_HORZ_SPLIT:
+ return canBh;
+ case CU_VERT_SPLIT:
+ return canBv;
+ case CU_TRIH_SPLIT:
+ return canTh;
+ case CU_TRIV_SPLIT:
+ return canTv;
+#else
case CU_QUAD_SPLIT:
{
// don't allow QT-splitting below a BT split
@@ -369,8 +473,15 @@ bool QTBTPartitioner::canSplit( const PartSplit split, const CodingStructure &cs
}
if( implicitSplit == split ) return true;
if( implicitSplit != CU_DONT_SPLIT && implicitSplit != split ) return false;
+#endif
case CU_MT_SPLIT:
+#if JVET_M0421_SPLIT_SIG
+ return ( canBh || canTh || canBv || canTv );
+#endif
case CU_BT_SPLIT:
+#if JVET_M0421_SPLIT_SIG
+ return ( canBh || canBv );
+#else
{
if( currMtDepth >= maxBTD ) return false;
if( ( area.width <= minBtSize && area.height <= minBtSize )
@@ -382,13 +493,14 @@ bool QTBTPartitioner::canSplit( const PartSplit split, const CodingStructure &cs
return false;
}
}
+#endif
break;
default:
THROW( "Unknown split mode" );
return false;
break;
}
-
+#if !JVET_M0421_SPLIT_SIG
// specific check for BT splits
switch( split )
{
@@ -413,6 +525,7 @@ bool QTBTPartitioner::canSplit( const PartSplit split, const CodingStructure &cs
default:
break;
}
+#endif
return true;
}
@@ -462,6 +575,12 @@ PartSplit QTBTPartitioner::getImplicitSplit( const CodingStructure &cs )
{
split = CU_QUAD_SPLIT;
}
+#if JVET_M0446_M0888_M0905_VPDU_AT_PIC_BOUNDARY
+ if ((!isBlInPic || !isTrInPic) && (currArea().Y().width > MAX_TU_SIZE_FOR_PROFILE || currArea().Y().height > MAX_TU_SIZE_FOR_PROFILE))
+ {
+ split = CU_QUAD_SPLIT;
+ }
+#endif
}
m_partStack.back().checkdIfImplicit = true;
diff --git a/source/Lib/CommonLib/UnitPartitioner.h b/source/Lib/CommonLib/UnitPartitioner.h
index 6b8e354f96a57456a2b3380f65d32fcef3d8a033..222fe99edeb0a83f5ef5ab5462a25f63d1bb1eb5 100644
--- a/source/Lib/CommonLib/UnitPartitioner.h
+++ b/source/Lib/CommonLib/UnitPartitioner.h
@@ -128,6 +128,9 @@ public:
virtual void copyState ( const Partitioner& other );
public:
+#if JVET_M0421_SPLIT_SIG
+ virtual void canSplit ( const CodingStructure &cs, bool& canNo, bool& canQt, bool& canBh, bool& canBv, bool& canTh, bool& canTv ) = 0;
+#endif
virtual bool canSplit ( const PartSplit split, const CodingStructure &cs ) = 0;
virtual bool isSplitImplicit ( const PartSplit split, const CodingStructure &cs ) = 0;
virtual PartSplit getImplicitSplit ( const CodingStructure &cs ) = 0;
@@ -147,7 +150,10 @@ public:
void exitCurrSplit ();
bool nextPart ( const CodingStructure &cs, bool autoPop = false );
bool hasNextPart ();
-
+
+#if JVET_M0421_SPLIT_SIG
+ void canSplit ( const CodingStructure &cs, bool& canNo, bool& canQt, bool& canBh, bool& canBv, bool& canTh, bool& canTv );
+#endif
bool canSplit ( const PartSplit split, const CodingStructure &cs );
bool isSplitImplicit ( const PartSplit split, const CodingStructure &cs );
PartSplit getImplicitSplit ( const CodingStructure &cs );
diff --git a/source/Lib/CommonLib/UnitTools.cpp b/source/Lib/CommonLib/UnitTools.cpp
index ee952d0c99bd5a5fabb49abf9d8c7a445f5d5e86..5bb6c03629e76d5b62c5bab6d180f5f04acbd58e 100644
--- a/source/Lib/CommonLib/UnitTools.cpp
+++ b/source/Lib/CommonLib/UnitTools.cpp
@@ -236,7 +236,11 @@ bool CU::hasNonTsCodedBlock( const CodingUnit& cu )
{
for( uint32_t i = 0; i < ::getNumberValidTBlocks( *cu.cs->pcv ); i++ )
{
+#if JVET_M0464_UNI_MTS
+ hasAnyNonTSCoded |= ( currTU.blocks[i].valid() && ( isLuma(ComponentID(i)) ? currTU.mtsIdx != 1 : true ) && TU::getCbf( currTU, ComponentID( i ) ) );
+#else
hasAnyNonTSCoded |= ( currTU.blocks[i].valid() && !currTU.transformSkip[i] && TU::getCbf( currTU, ComponentID( i ) ) );
+#endif
}
}
@@ -708,6 +712,9 @@ bool PU::xCheckSimilarMotion(const int mergeCandIndex, const int prevCnt, const
bool PU::addMergeHMVPCand(const Slice &slice, MergeCtx& mrgCtx, bool canFastExit, const int& mrgCandIdx, const uint32_t maxNumMergeCandMin1, int &cnt, const int prevCnt, bool isAvailableSubPu, unsigned subPuMvpPos
, int mmvdList
+#if JVET_M0170_MRG_SHARELIST
+ , bool isShared
+#endif
)
#else
@@ -716,7 +723,7 @@ bool PU::addMergeHMVPCand(const Slice &slice, MergeCtx& mrgCtx, bool isCandInter
)
#endif
{
- int mrgCandIdxCPR = mrgCandIdx;
+ int mrgCandIdxIBC = mrgCandIdx;
MotionInfo miNeighbor;
bool hasPruned[MRG_MAX_NUM_CANDS];
memset(hasPruned, 0, MRG_MAX_NUM_CANDS * sizeof(bool));
@@ -724,10 +731,19 @@ bool PU::addMergeHMVPCand(const Slice &slice, MergeCtx& mrgCtx, bool isCandInter
{
hasPruned[subPuMvpPos] = true;
}
+#if JVET_M0170_MRG_SHARELIST
+ int num_avai_candInLUT = (isShared ? slice.getAvailableLUTBkupMrgNum() : slice.getAvailableLUTMrgNum());
+#else
int num_avai_candInLUT = slice.getAvailableLUTMrgNum();
+#endif
+
for (int mrgIdx = 1; mrgIdx <= num_avai_candInLUT; mrgIdx++)
{
+#if JVET_M0170_MRG_SHARELIST
+ miNeighbor = isShared ? slice.getMotionInfoFromLUTBkup(num_avai_candInLUT - mrgIdx) : slice.getMotionInfoFromLUTs(num_avai_candInLUT - mrgIdx);
+#else
miNeighbor = slice.getMotionInfoFromLUTs(num_avai_candInLUT - mrgIdx);
+#endif
mrgCtx.interDirNeighbours[cnt] = miNeighbor.interDir;
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField(miNeighbor.mv[0], miNeighbor.refIdx[0]);
if (slice.isInterB())
@@ -741,11 +757,11 @@ bool PU::addMergeHMVPCand(const Slice &slice, MergeCtx& mrgCtx, bool isCandInter
#endif
if (miNeighbor.interDir == 1 && slice.getRefPic(REF_PIC_LIST_0, miNeighbor.refIdx[0])->getPOC() == slice.getPOC())
{
- mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_CPR;
+ mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_IBC;
if(mmvdList != 0 && mrgCandIdx != -1)
- mrgCandIdxCPR++;
+ mrgCandIdxIBC++;
}
- if (mrgCandIdxCPR == cnt && canFastExit)
+ if (mrgCandIdxIBC == cnt && canFastExit)
{
return true;
}
@@ -791,12 +807,17 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
int cnt = 0;
- int mrgCandIdxCPR = mrgCandIdx;
+ int mrgCandIdxIBC = mrgCandIdx;
+#if JVET_M0170_MRG_SHARELIST
+ const Position posLT = pu.shareParentPos;
+ const Position posRT = pu.shareParentPos.offset(pu.shareParentSize.width - 1, 0);
+ const Position posLB = pu.shareParentPos.offset(0, pu.shareParentSize.height - 1);
+#else
const Position posLT = pu.Y().topLeft();
const Position posRT = pu.Y().topRight();
const Position posLB = pu.Y().bottomLeft();
-
+#endif
MotionInfo miAbove, miLeft, miAboveLeft, miAboveRight, miBelowLeft;
//left
@@ -816,11 +837,11 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
mrgCtx.interDirNeighbours[cnt] = miLeft.interDir;
mrgCtx.GBiIdx[cnt] = (mrgCtx.interDirNeighbours[cnt] == 3) ? puLeft->cu->GBiIdx : GBI_DEFAULT;
// get Mv from Left
- if (puLeft->cu->cpr)
+ if (puLeft->cu->ibc)
{
- mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_CPR;
+ mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_IBC;
if (mmvdList != 0 && mrgCandIdx != -1)
- mrgCandIdxCPR++;
+ mrgCandIdxIBC++;
}
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField(miLeft.mv[0], miLeft.refIdx[0]);
@@ -828,7 +849,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
{
mrgCtx.mvFieldNeighbours[(cnt << 1) + 1].setMvField(miLeft.mv[1], miLeft.refIdx[1]);
}
- if (mrgCandIdxCPR == cnt && canFastExit)
+ if (mrgCandIdxIBC == cnt && canFastExit)
{
return;
}
@@ -862,11 +883,11 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
mrgCtx.interDirNeighbours[cnt] = miAbove.interDir;
// get Mv from Above
mrgCtx.GBiIdx[cnt] = (mrgCtx.interDirNeighbours[cnt] == 3) ? puAbove->cu->GBiIdx : GBI_DEFAULT;
- if (puAbove->cu->cpr)
+ if (puAbove->cu->ibc)
{
- mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_CPR;
+ mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_IBC;
if (mmvdList != 0 && mrgCandIdx != -1)
- mrgCandIdxCPR++;
+ mrgCandIdxIBC++;
}
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField( miAbove.mv[0], miAbove.refIdx[0] );
@@ -874,7 +895,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
{
mrgCtx.mvFieldNeighbours[( cnt << 1 ) + 1].setMvField( miAbove.mv[1], miAbove.refIdx[1] );
}
- if (mrgCandIdxCPR == cnt && canFastExit)
+ if (mrgCandIdxIBC == cnt && canFastExit)
{
return;
}
@@ -912,11 +933,11 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
mrgCtx.interDirNeighbours[cnt] = miAboveRight.interDir;
// get Mv from Above-right
mrgCtx.GBiIdx[cnt] = (mrgCtx.interDirNeighbours[cnt] == 3) ? puAboveRight->cu->GBiIdx : GBI_DEFAULT;
- if (puAboveRight->cu->cpr)
+ if (puAboveRight->cu->ibc)
{
- mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_CPR;
+ mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_IBC;
if (mmvdList != 0 && mrgCandIdx != -1)
- mrgCandIdxCPR++;
+ mrgCandIdxIBC++;
}
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField( miAboveRight.mv[0], miAboveRight.refIdx[0] );
@@ -925,7 +946,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
mrgCtx.mvFieldNeighbours[( cnt << 1 ) + 1].setMvField( miAboveRight.mv[1], miAboveRight.refIdx[1] );
}
- if (mrgCandIdxCPR == cnt && canFastExit)
+ if (mrgCandIdxIBC == cnt && canFastExit)
{
return;
}
@@ -962,11 +983,11 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
mrgCtx.interDirNeighbours[cnt] = miBelowLeft.interDir;
mrgCtx.GBiIdx[cnt] = (mrgCtx.interDirNeighbours[cnt] == 3) ? puLeftBottom->cu->GBiIdx : GBI_DEFAULT;
// get Mv from Bottom-Left
- if (puLeftBottom->cu->cpr)
+ if (puLeftBottom->cu->ibc)
{
- mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_CPR;
+ mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_IBC;
if (mmvdList != 0 && mrgCandIdx != -1)
- mrgCandIdxCPR++;
+ mrgCandIdxIBC++;
}
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField( miBelowLeft.mv[0], miBelowLeft.refIdx[0] );
@@ -975,7 +996,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
mrgCtx.mvFieldNeighbours[( cnt << 1 ) + 1].setMvField( miBelowLeft.mv[1], miBelowLeft.refIdx[1] );
}
- if (mrgCandIdxCPR == cnt && canFastExit)
+ if (mrgCandIdxIBC == cnt && canFastExit)
{
return;
}
@@ -1015,11 +1036,11 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
mrgCtx.interDirNeighbours[cnt] = miAboveLeft.interDir;
mrgCtx.GBiIdx[cnt] = (mrgCtx.interDirNeighbours[cnt] == 3) ? puAboveLeft->cu->GBiIdx : GBI_DEFAULT;
// get Mv from Above-Left
- if (puAboveLeft->cu->cpr)
+ if (puAboveLeft->cu->ibc)
{
- mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_CPR;
+ mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_IBC;
if (mmvdList != 0 && mrgCandIdx != -1)
- mrgCandIdxCPR++;
+ mrgCandIdxIBC++;
}
mrgCtx.mvFieldNeighbours[cnt << 1].setMvField( miAboveLeft.mv[0], miAboveLeft.refIdx[0] );
@@ -1028,7 +1049,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
mrgCtx.mvFieldNeighbours[( cnt << 1 ) + 1].setMvField( miAboveLeft.mv[1], miAboveLeft.refIdx[1] );
}
- if (mrgCandIdxCPR == cnt && canFastExit)
+ if (mrgCandIdxIBC == cnt && canFastExit)
{
return;
}
@@ -1047,13 +1068,24 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
{
//>> MTK colocated-RightBottom
// offset the pos to be sure to "point" to the same position the uiAbsPartIdx would've pointed to
+#if JVET_M0170_MRG_SHARELIST
+ Position posRB = pu.shareParentPos.offset(pu.shareParentSize.width-3, pu.shareParentSize.height - 3);
+#else
Position posRB = pu.Y().bottomRight().offset(-3, -3);
-
+#endif
const PreCalcValues& pcv = *cs.pcv;
Position posC0;
+#if JVET_M0170_MRG_SHARELIST
+ Position posC1 = pu.shareParentPos.offset((pu.shareParentSize.width/2), (pu.shareParentSize.height/2));
+#else
Position posC1 = pu.Y().center();
+#endif
+
bool C0Avail = false;
+#if JVET_M0170_MRG_SHARELIST
+ bool C1Avail = (posC1.x < pcv.lumaWidth) && (posC1.y < pcv.lumaHeight);
+#endif
if (((posRB.x + pcv.minCUWidth) < pcv.lumaWidth) && ((posRB.y + pcv.minCUHeight) < pcv.lumaHeight))
{
@@ -1089,7 +1121,11 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
int dir = 0;
unsigned uiArrayAddr = cnt;
bool bExistMV = ( C0Avail && getColocatedMVP(pu, REF_PIC_LIST_0, posC0, cColMv, iRefIdx ) )
+#if JVET_M0170_MRG_SHARELIST
+ || ( C1Avail && getColocatedMVP(pu, REF_PIC_LIST_0, posC1, cColMv, iRefIdx ));
+#else
|| getColocatedMVP(pu, REF_PIC_LIST_0, posC1, cColMv, iRefIdx );
+#endif
if (bExistMV)
{
@@ -1100,7 +1136,11 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
if (slice.isInterB())
{
bExistMV = ( C0Avail && getColocatedMVP(pu, REF_PIC_LIST_1, posC0, cColMv, iRefIdx ) )
+#if JVET_M0170_MRG_SHARELIST
+ || (C1Avail && getColocatedMVP(pu, REF_PIC_LIST_1, posC1, cColMv, iRefIdx ) );
+#else
|| getColocatedMVP(pu, REF_PIC_LIST_1, posC1, cColMv, iRefIdx );
+#endif
if (bExistMV)
{
dir |= 2;
@@ -1130,7 +1170,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
isCandInter [uiArrayAddr] = true;
#endif
mrgCtx.GBiIdx[uiArrayAddr] = GBI_DEFAULT;
- if (mrgCandIdxCPR == cnt && canFastExit)
+ if (mrgCandIdxIBC == cnt && canFastExit)
{
return;
}
@@ -1152,14 +1192,20 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
bool isAvailableSubPu = false;
unsigned subPuMvpPos = 0;
#if JVET_L0090_PAIR_AVG
+#if JVET_M0170_MRG_SHARELIST
+ bool isShared = ((pu.Y().lumaSize().width != pu.shareParentSize.width) || (pu.Y().lumaSize().height != pu.shareParentSize.height));
+#endif
bool bFound = addMergeHMVPCand(slice, mrgCtx, canFastExit
- , (mmvdList != 0 && mrgCandIdx != -1) ? (const int) mrgCandIdxCPR : mrgCandIdx
+ , (mmvdList != 0 && mrgCandIdx != -1) ? (const int) mrgCandIdxIBC : mrgCandIdx
, maxNumMergeCandMin1, cnt, cnt, isAvailableSubPu, subPuMvpPos
, mmvdList
+#if JVET_M0170_MRG_SHARELIST
+ , isShared
+#endif
);
#else
bool bFound = addMergeHMVPCand(slice, mrgCtx, isCandInter, canFastExit
- , (mmvdList != 0 && mrgCandIdx != -1) ? (const int)mrgCandIdxCPR : mrgCandIdx
+ , (mmvdList != 0 && mrgCandIdx != -1) ? (const int)mrgCandIdxIBC : mrgCandIdx
, maxNumMergeCandMin1, cnt, cnt, isAvailableSubPu, subPuMvpPos
, mmvdList
);
@@ -1190,7 +1236,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
// skip when one is BV and one is MV
- if (mrgCtx.mrgTypeNeighbours[i] != mrgCtx.mrgTypeNeighbours[j] && pu.cs->sps->getSpsNext().getCPRMode())
+ if (mrgCtx.mrgTypeNeighbours[i] != mrgCtx.mrgTypeNeighbours[j] && pu.cs->sps->getSpsNext().getIBCMode())
{
continue;
}
@@ -1220,9 +1266,9 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
- if (mrgCtx.mrgTypeNeighbours[i] == MRG_TYPE_CPR && mrgCtx.mrgTypeNeighbours[j] == MRG_TYPE_CPR && pu.cs->sps->getSpsNext().getCPRMode())
+ if (mrgCtx.mrgTypeNeighbours[i] == MRG_TYPE_IBC && mrgCtx.mrgTypeNeighbours[j] == MRG_TYPE_IBC && pu.cs->sps->getSpsNext().getIBCMode())
{
- mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_CPR;
+ mrgCtx.mrgTypeNeighbours[cnt] = MRG_TYPE_IBC;
avgMv.setHor((avgMv.getHor() / 16) << 4);
avgMv.setVer((avgMv.getVer() / 16) << 4);
}
@@ -1324,7 +1370,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
if ( mrgCtx.interDirNeighbours[uiArrayAddr] == 1 && pu.cs->slice->getRefPic(REF_PIC_LIST_0, mrgCtx.mvFieldNeighbours[uiArrayAddr << 1].refIdx)->getPOC() == pu.cs->slice->getPOC())
{
- mrgCtx.mrgTypeNeighbours[uiArrayAddr] = MRG_TYPE_CPR;
+ mrgCtx.mrgTypeNeighbours[uiArrayAddr] = MRG_TYPE_IBC;
}
uiArrayAddr++;
@@ -1341,7 +1387,7 @@ void PU::getInterMergeCandidates( const PredictionUnit &pu, MergeCtx& mrgCtx,
}
mrgCtx.numValidMergeCand = uiArrayAddr;
}
-// for cpr pu validation
+// for ibc pu validation
bool PU::isBlockVectorValid(PredictionUnit& pu, int xPos, int yPos, int width, int height, int picWidth, int picHeight, int xStartInCU, int yStartInCU, int xBv, int yBv, int ctuSize)
{
const int ctuSizeLog2 = g_aucLog2[ctuSize];
@@ -1385,9 +1431,28 @@ bool PU::isBlockVectorValid(PredictionUnit& pu, int xPos, int yPos, int width, i
}
// in the same CTU line
-
+#if JVET_M0407_IBC_RANGE
+ if ((refRightX >> ctuSizeLog2 <= xPos >> ctuSizeLog2) && (refLeftX >> ctuSizeLog2 >= (xPos >> ctuSizeLog2) - 1))
+#else
if ((refRightX >> ctuSizeLog2 <= xPos >> ctuSizeLog2) && (refLeftX >> ctuSizeLog2 >= (xPos >> ctuSizeLog2)))
+#endif
{
+
+#if JVET_M0407_IBC_RANGE
+ // in the same CTU, or left CTU
+ // if part of ref block is in the left CTU, some area can be referred from the not-yet updated local CTU buffer
+ if ((refLeftX >> ctuSizeLog2) == ((xPos >> ctuSizeLog2) - 1))
+ {
+ // ref block's collocated block in current CTU
+ const Position refPosCol = pu.Y().topLeft().offset(xBv + ctuSize, yBv);
+ int offset64x = (refPosCol.x >> (ctuSizeLog2 - 1)) << (ctuSizeLog2 - 1);
+ int offset64y = (refPosCol.y >> (ctuSizeLog2 - 1)) << (ctuSizeLog2 - 1);
+ const Position refPosCol64x64 = {offset64x, offset64y};
+ if (pu.cs->isDecomp(refPosCol64x64, toChannelType(COMPONENT_Y)))
+ return false;
+ }
+#endif
+#if !JVET_M0407_IBC_RANGE
// in the same CTU, check if the reference block is already coded
const Position refPosLT = pu.Y().topLeft().offset(xBv, yBv);
const Position refPosBR = pu.Y().bottomRight().offset(xBv, yBv);
@@ -1399,11 +1464,24 @@ bool PU::isBlockVectorValid(PredictionUnit& pu, int xPos, int yPos, int width, i
return false;
}
return true;
+#endif
}
else
return false;
-}// for cpr pu validation
+#if JVET_M0407_IBC_RANGE
+ // in the same CTU, or valid area from left CTU. Check if the reference block is already coded
+ const Position refPosLT = pu.Y().topLeft().offset(xBv, yBv);
+ const Position refPosBR = pu.Y().bottomRight().offset(xBv, yBv);
+ const ChannelType chType = toChannelType(COMPONENT_Y);
+ if (!pu.cs->isDecomp(refPosBR, chType))
+ return false;
+ if (!pu.cs->isDecomp(refPosLT, chType))
+ return false;
+ return true;
+#endif
+
+}// for ibc pu validation
static int xGetDistScaleFactor(const int &iCurrPOC, const int &iCurrRefPOC, const int &iColPOC, const int &iColRefPOC)
{
@@ -1469,8 +1547,14 @@ void PU::getInterMMVDMergeCandidates(const PredictionUnit &pu, MergeCtx& mrgCtx,
{
for (k = currBaseNum; k < MMVD_BASE_MV_NUM; k++)
{
+#if JVET_M0068_M0171_MMVD_CLEANUP
+ mrgCtx.mmvdBaseMv[k][0] = MvField(Mv(0, 0), 0);
+ const Slice &slice = *pu.cs->slice;
+ mrgCtx.mmvdBaseMv[k][1] = MvField(Mv(0, 0), (slice.isInterB() ? 0 : -1));
+#else
mrgCtx.mmvdBaseMv[k][0] = MvField(Mv(0, 0), 0);
mrgCtx.mmvdBaseMv[k][0] = MvField(Mv(0, 0), 0);
+#endif
}
}
}
@@ -1590,7 +1674,7 @@ bool PU::isDiffMER(const PredictionUnit &pu1, const PredictionUnit &pu2)
return false;
}
-void PU::getCprMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred)
+void PU::getIbcMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred)
{
//-- Get Spatial MV
@@ -1603,7 +1687,7 @@ void PU::getCprMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred)
//left
const PredictionUnit *neibLeftPU = NULL;
neibLeftPU = pu.cs->getPURestricted(posLB.offset(-1, 0), pu, pu.cs->chType);
- left = (neibLeftPU) ? neibLeftPU->cu->cpr : 0;
+ left = (neibLeftPU) ? neibLeftPU->cu->ibc : 0;
if (left)
{
@@ -1615,7 +1699,7 @@ void PU::getCprMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred)
//above
const PredictionUnit *neibAbovePU = NULL;
neibAbovePU = pu.cs->getPURestricted(posRT.offset(0, -1), pu, pu.cs->chType);
- above = (neibAbovePU) ? neibAbovePU->cu->cpr : 0;
+ above = (neibAbovePU) ? neibAbovePU->cu->ibc : 0;
if (above)
{
@@ -1627,7 +1711,7 @@ void PU::getCprMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred)
// Below Left predictor search
const PredictionUnit *neibBelowLeftPU = NULL;
neibBelowLeftPU = pu.cs->getPURestricted(posLB.offset(-1, 1), pu, pu.cs->chType);
- unsigned int belowLeft = (neibBelowLeftPU) ? neibBelowLeftPU->cu->cpr : 0;
+ unsigned int belowLeft = (neibBelowLeftPU) ? neibBelowLeftPU->cu->ibc : 0;
if (belowLeft)
{
@@ -1640,7 +1724,7 @@ void PU::getCprMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred)
// Above Right predictor search
const PredictionUnit *neibAboveRightPU = NULL;
neibAboveRightPU = pu.cs->getPURestricted(posRT.offset(1, -1), pu, pu.cs->chType);
- unsigned int aboveRight = (neibAboveRightPU) ? neibAboveRightPU->cu->cpr : 0;
+ unsigned int aboveRight = (neibAboveRightPU) ? neibAboveRightPU->cu->ibc : 0;
if (aboveRight)
{
@@ -1653,7 +1737,7 @@ void PU::getCprMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred)
// Above Left predictor search
const PredictionUnit *neibAboveLeftPU = NULL;
neibAboveLeftPU = pu.cs->getPURestricted(posLT.offset(-1, -1), pu, pu.cs->chType);
- unsigned int aboveLeft = (neibAboveLeftPU) ? neibAboveLeftPU->cu->cpr : 0;
+ unsigned int aboveLeft = (neibAboveLeftPU) ? neibAboveLeftPU->cu->ibc : 0;
if (aboveLeft)
{
@@ -1681,13 +1765,13 @@ bool PU::getDerivedBV(PredictionUnit &pu, const Mv& currentMv, Mv& derivedMv)
const PredictionUnit *neibRefPU = NULL;
neibRefPU = pu.cs->getPURestricted(pu.lumaPos().offset(offsetX, offsetY), pu, pu.cs->chType);
- bool isCPR = (neibRefPU) ? neibRefPU->cu->cpr : 0;
- if (isCPR)
+ bool isIBC = (neibRefPU) ? neibRefPU->cu->ibc : 0;
+ if (isIBC)
{
derivedMv = neibRefPU->bv;
derivedMv += currentMv;
}
- return isCPR;
+ return isIBC;
}
/** Constructs a list of candidates for AMVP (See specification, section "Derivation process for motion vector predictor candidates")
* \param uiPartIdx
@@ -1806,6 +1890,9 @@ void PU::fillMvpCand(PredictionUnit &pu, const RefPicList &eRefPicList, const in
Position posC0;
bool C0Avail = false;
Position posC1 = pu.Y().center();
+#if JVET_M0170_MRG_SHARELIST
+ bool C1Avail = ( posC1.x < pcv.lumaWidth ) && ( posC1.y < pcv.lumaHeight ) ;
+#endif
Mv cColMv;
@@ -1836,7 +1923,11 @@ void PU::fillMvpCand(PredictionUnit &pu, const RefPicList &eRefPicList, const in
}
}
+#if JVET_M0170_MRG_SHARELIST
+ if ((C0Avail && getColocatedMVP(pu, eRefPicList, posC0, cColMv, refIdx_Col)) || (C1Avail && getColocatedMVP(pu, eRefPicList, posC1, cColMv, refIdx_Col)))
+#else
if ((C0Avail && getColocatedMVP(pu, eRefPicList, posC0, cColMv, refIdx_Col)) || getColocatedMVP(pu, eRefPicList, posC1, cColMv, refIdx_Col))
+#endif
{
if (pu.cu->imv != 0)
{
@@ -2153,6 +2244,9 @@ void PU::fillAffineMvpCand(PredictionUnit &pu, const RefPicList &eRefPicList, co
Position posC0;
bool C0Avail = false;
Position posC1 = pu.Y().center();
+#if JVET_M0170_MRG_SHARELIST
+ bool C1Avail = ( posC1.x < pcv.lumaWidth ) && ( posC1.y < pcv.lumaHeight ) ;
+#endif
Mv cColMv;
if ( ((posRB.x + pcv.minCUWidth) < pcv.lumaWidth) && ((posRB.y + pcv.minCUHeight) < pcv.lumaHeight) )
@@ -2182,7 +2276,11 @@ void PU::fillAffineMvpCand(PredictionUnit &pu, const RefPicList &eRefPicList, co
}
}
+#if JVET_M0170_MRG_SHARELIST
+ if ( (C0Avail && getColocatedMVP( pu, eRefPicList, posC0, cColMv, refIdxCol )) || (C1Avail && getColocatedMVP( pu, eRefPicList, posC1, cColMv, refIdxCol ) ) )
+#else
if ( (C0Avail && getColocatedMVP( pu, eRefPicList, posC0, cColMv, refIdxCol )) || getColocatedMVP( pu, eRefPicList, posC1, cColMv, refIdxCol ) )
+#endif
{
cColMv.roundToPrecision(MV_PRECISION_INTERNAL, MV_PRECISION_QUARTER);
affiAMVPInfo.mvCandLT[affiAMVPInfo.numCand] = cColMv;
@@ -2628,11 +2726,13 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
affMrgCtx.numValidMergeCand = 0;
affMrgCtx.maxNumMergeCand = maxNumAffineMergeCand;
- bool enableSubPuMvp = slice.getSPS()->getSpsNext().getUseSubPuMvp() && !(slice.getPOC() == slice.getRefPic(REF_PIC_LIST_0, 0)->getPOC() && slice.isIRAP());
+ bool enableSubPuMvp = slice.getSPS()->getSBTMVPEnabledFlag() && !(slice.getPOC() == slice.getRefPic(REF_PIC_LIST_0, 0)->getPOC() && slice.isIRAP());
bool isAvailableSubPu = false;
if ( enableSubPuMvp && slice.getEnableTMVPFlag() )
{
- int cntCPR = 0;
+#if !JVET_M0409_ATMVP_FIX
+ int cntIBC = 0;
+#endif
MergeCtx mrgCtx = *affMrgCtx.mrgCtx;
bool tmpLICFlag = false;
@@ -2655,10 +2755,12 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
mrgCtx.interDirNeighbours[pos] = miLeft.interDir;
// get Mv from Left
- if (puLeft->cu->cpr)
+#if !JVET_M0409_ATMVP_FIX
+ if (puLeft->cu->ibc)
{
- cntCPR++;
+ cntIBC++;
}
+#endif
mrgCtx.mvFieldNeighbours[pos << 1].setMvField( miLeft.mv[0], miLeft.refIdx[0] );
if ( slice.isInterB() )
@@ -2680,10 +2782,12 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
// get Inter Dir
mrgCtx.interDirNeighbours[pos] = miAbove.interDir;
// get Mv from Left
- if (puAbove->cu->cpr)
+#if !JVET_M0409_ATMVP_FIX
+ if (puAbove->cu->ibc)
{
- cntCPR++;
+ cntIBC++;
}
+#endif
mrgCtx.mvFieldNeighbours[pos << 1].setMvField( miAbove.mv[0], miAbove.refIdx[0] );
if ( slice.isInterB() )
@@ -2711,10 +2815,12 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
// get Inter Dir
mrgCtx.interDirNeighbours[pos] = miAboveRight.interDir;
// get Mv from Left
- if (puAboveRight->cu->cpr)
+#if !JVET_M0409_ATMVP_FIX
+ if (puAboveRight->cu->ibc)
{
- cntCPR++;
+ cntIBC++;
}
+#endif
mrgCtx.mvFieldNeighbours[pos << 1].setMvField( miAboveRight.mv[0], miAboveRight.refIdx[0] );
if ( slice.isInterB() )
@@ -2742,10 +2848,12 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
// get Inter Dir
mrgCtx.interDirNeighbours[pos] = miBelowLeft.interDir;
// get Mv from Bottom-Left
- if (puLeftBottom->cu->cpr)
+#if !JVET_M0409_ATMVP_FIX
+ if (puLeftBottom->cu->ibc)
{
- cntCPR++;
+ cntIBC++;
}
+#endif
mrgCtx.mvFieldNeighbours[pos << 1].setMvField( miBelowLeft.mv[0], miBelowLeft.refIdx[0] );
if ( slice.isInterB() )
@@ -2759,7 +2867,9 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
isAvailableSubPu = getInterMergeSubPuMvpCand( pu, mrgCtx, tmpLICFlag, pos
, 0
- , cntCPR
+#if !JVET_M0409_ATMVP_FIX
+ , cntIBC
+#endif
);
if ( isAvailableSubPu )
{
@@ -2847,7 +2957,7 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
const PredictionUnit* puNeigh = cs.getPURestricted( pos, pu, pu.chType );
if ( puNeigh && CU::isInter( *puNeigh->cu )
- && !puNeigh->cu->cpr
+ && !puNeigh->cu->ibc
)
{
isAvailable[0] = true;
@@ -2865,7 +2975,7 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
if ( puNeigh && CU::isInter( *puNeigh->cu )
- && !puNeigh->cu->cpr
+ && !puNeigh->cu->ibc
)
{
isAvailable[1] = true;
@@ -2883,7 +2993,7 @@ void PU::getAffineMergeCand( const PredictionUnit &pu, AffineMergeCtx& affMrgCtx
if ( puNeigh && CU::isInter( *puNeigh->cu )
- && !puNeigh->cu->cpr
+ && !puNeigh->cu->ibc
)
{
isAvailable[2] = true;
@@ -3182,11 +3292,15 @@ void clipColPos(int& posX, int& posY, const PredictionUnit& pu)
bool PU::getInterMergeSubPuMvpCand(const PredictionUnit &pu, MergeCtx& mrgCtx, bool& LICFlag, const int count
, int mmvdList
- , const int countCPR
+#if !JVET_M0409_ATMVP_FIX
+ , const int countIBC
+#endif
)
{
- if (count == countCPR && pu.cs->slice->getSPS()->getSpsNext().getCPRMode())
+#if !JVET_M0409_ATMVP_FIX
+ if (count == countIBC && pu.cs->slice->getSPS()->getSpsNext().getIBCMode())
return false;
+#endif
const Slice &slice = *pu.cs->slice;
const unsigned scale = 4 * std::max<int>(1, 4 * AMVP_DECIMATION_FACTOR / 4);
const unsigned mask = ~(scale - 1);
@@ -3301,7 +3415,7 @@ bool PU::getInterMergeSubPuMvpCand(const PredictionUnit &pu, MergeCtx& mrgCtx, b
mi.isInter = true;
mi.sliceIdx = slice.getIndependentSliceIdx();
- if (colMi.isInter && !((colMi.interDir == 1 || colMi.interDir == 3) && (pColPic->cs->slice->getRefPOC(REF_PIC_LIST_0, colMi.refIdx[0]) == pColPic->cs->slice->getPOC()) && pu.cs->sps->getSpsNext().getCPRMode()))
+ if (colMi.isInter && !((colMi.interDir == 1 || colMi.interDir == 3) && (pColPic->cs->slice->getRefPOC(REF_PIC_LIST_0, colMi.refIdx[0]) == pColPic->cs->slice->getPOC()) && pu.cs->sps->getSpsNext().getIBCMode()))
{
for (unsigned currRefListId = 0; currRefListId < (bBSlice ? 2 : 1); currRefListId++)
{
@@ -3343,7 +3457,7 @@ void PU::spanMotionInfo( PredictionUnit &pu, const MergeCtx &mrgCtx )
MotionBuf mb = pu.getMotionBuf();
if( !pu.mergeFlag || pu.mergeType == MRG_TYPE_DEFAULT_N
- || pu.mergeType == MRG_TYPE_CPR
+ || pu.mergeType == MRG_TYPE_IBC
)
{
MotionInfo mi;
@@ -3435,7 +3549,7 @@ void PU::applyImv( PredictionUnit& pu, MergeCtx &mrgCtx, InterPrediction *interP
pu.mv[0].changePrecision(MV_PRECISION_QUARTER, MV_PRECISION_INTERNAL);
if (pu.interDir == 1 && pu.cs->slice->getRefPic(REF_PIC_LIST_0, pu.refIdx[REF_PIC_LIST_0])->getPOC() == pu.cs->slice->getPOC())
{
- pu.cu->cpr = true;
+ pu.cu->ibc = true;
}
}
@@ -3461,7 +3575,9 @@ void PU::applyImv( PredictionUnit& pu, MergeCtx &mrgCtx, InterPrediction *interP
PU::getInterMergeCandidates ( pu, mrgCtx
, 0
);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
PU::restrictBiPredMergeCands( pu, mrgCtx );
+#endif
mrgCtx.setMergeInfo( pu, pu.mergeIdx );
}
@@ -3501,6 +3617,22 @@ void PU::restrictBiPredMergeCands( const PredictionUnit &pu, MergeCtx& mergeCtx
}
}
+#if JVET_M0068_M0171_MMVD_CLEANUP
+void PU::restrictBiPredMergeCandsOne(PredictionUnit &pu)
+{
+ if (PU::isBipredRestriction(pu))
+ {
+ if (pu.interDir == 3)
+ {
+ pu.interDir = 1;
+ pu.refIdx[1] = -1;
+ pu.mv[1] = Mv(0, 0);
+ pu.cu->GBiIdx = GBI_DEFAULT;
+ }
+ }
+}
+#endif
+
void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangleMrgCtx )
{
const CodingStructure &cs = *pu.cs;
@@ -3530,7 +3662,7 @@ void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangl
//left
const PredictionUnit* puLeft = cs.getPURestricted( posLB.offset( -1, 0 ), pu, pu.chType );
const bool isAvailableA1 = puLeft && isDiffMER( pu, *puLeft ) && pu.cu != puLeft->cu && CU::isInter( *puLeft->cu )
- && !puLeft->cu->cpr
+ && !puLeft->cu->ibc
;
if( isAvailableA1 )
{
@@ -3547,7 +3679,7 @@ void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangl
// above
const PredictionUnit *puAbove = cs.getPURestricted( posRT.offset( 0, -1 ), pu, pu.chType );
bool isAvailableB1 = puAbove && isDiffMER( pu, *puAbove ) && pu.cu != puAbove->cu && CU::isInter( *puAbove->cu )
- && !puAbove->cu->cpr
+ && !puAbove->cu->ibc
;
if( isAvailableB1 )
{
@@ -3568,7 +3700,7 @@ void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangl
// above right
const PredictionUnit *puAboveRight = cs.getPURestricted( posRT.offset( 1, -1 ), pu, pu.chType );
bool isAvailableB0 = puAboveRight && isDiffMER( pu, *puAboveRight ) && CU::isInter( *puAboveRight->cu )
- && !puAboveRight->cu->cpr
+ && !puAboveRight->cu->ibc
;
if( isAvailableB0 )
@@ -3590,7 +3722,7 @@ void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangl
//left bottom
const PredictionUnit *puLeftBottom = cs.getPURestricted( posLB.offset( -1, 1 ), pu, pu.chType );
bool isAvailableA0 = puLeftBottom && isDiffMER( pu, *puLeftBottom ) && CU::isInter( *puLeftBottom->cu )
- && !puLeftBottom->cu->cpr
+ && !puLeftBottom->cu->ibc
;
if( isAvailableA0 )
{
@@ -3611,7 +3743,7 @@ void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangl
// above left
const PredictionUnit *puAboveLeft = cs.getPURestricted( posLT.offset( -1, -1 ), pu, pu.chType );
bool isAvailableB2 = puAboveLeft && isDiffMER( pu, *puAboveLeft ) && CU::isInter( *puAboveLeft->cu )
- && !puAboveLeft->cu->cpr
+ && !puAboveLeft->cu->ibc
;
if( isAvailableB2 )
@@ -3639,6 +3771,9 @@ void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangl
Position posC0;
Position posC1 = pu.Y().center();
bool isAvailableC0 = false;
+#if JVET_M0170_MRG_SHARELIST
+ bool isAvailableC1 = (posC1.x < pcv.lumaWidth) && (posC1.y < pcv.lumaHeight);
+#endif
if (((posRB.x + pcv.minCUWidth) < pcv.lumaWidth) && ((posRB.y + pcv.minCUHeight) < pcv.lumaHeight))
{
@@ -3701,7 +3836,11 @@ void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangl
// C1
temporalMv.interDir = 0;
+#if JVET_M0170_MRG_SHARELIST
+ existMV = isAvailableC1 && getColocatedMVP(pu, REF_PIC_LIST_0, posC1, cColMv, refIdx );
+#else
existMV = getColocatedMVP(pu, REF_PIC_LIST_0, posC1, cColMv, refIdx );
+#endif
if( existMV )
{
temporalMv.isInter = true;
@@ -3709,7 +3848,11 @@ void PU::getTriangleMergeCandidates( const PredictionUnit &pu, MergeCtx& triangl
temporalMv.mv[0] = cColMv;
temporalMv.refIdx[0] = refIdx;
}
+#if JVET_M0170_MRG_SHARELIST
+ existMV = isAvailableC1 && getColocatedMVP(pu, REF_PIC_LIST_1, posC1, cColMv, refIdx );
+#else
existMV = getColocatedMVP(pu, REF_PIC_LIST_1, posC1, cColMv, refIdx );
+#endif
if( existMV )
{
temporalMv.interDir |= 2;
@@ -4054,7 +4197,9 @@ void CU::resetMVDandMV2Int( CodingUnit& cu, InterPrediction *interPred )
PU::getInterMergeCandidates ( pu, mrgCtx
, 0
);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
PU::restrictBiPredMergeCands( pu, mrgCtx );
+#endif
mrgCtx.setMergeInfo( pu, pu.mergeIdx );
}
@@ -4248,6 +4393,31 @@ void TU::setCbfAtDepth(TransformUnit &tu, const ComponentID &compID, const unsig
tu.cbf[compID] |= ((cbf ? 1 : 0) << depth);
}
+#if JVET_M0464_UNI_MTS
+bool TU::isTSAllowed(const TransformUnit &tu, const ComponentID compID)
+{
+ bool tsAllowed = compID == COMPONENT_Y;
+ const int maxSize = tu.cs->pps->getPpsRangeExtension().getLog2MaxTransformSkipBlockSize();
+
+ tsAllowed &= tu.cs->pps->getUseTransformSkip();
+ tsAllowed &= !tu.cu->transQuantBypass;
+
+ SizeType transformSkipMaxSize = 1 << maxSize;
+ tsAllowed &= tu.lwidth() <= transformSkipMaxSize && tu.lheight() <= transformSkipMaxSize;
+
+ return tsAllowed;
+}
+
+bool TU::isMTSAllowed(const TransformUnit &tu, const ComponentID compID)
+{
+ bool mtsAllowed = compID == COMPONENT_Y;
+ const int maxSize = CU::isIntra( *tu.cu ) ? MTS_INTRA_MAX_CU_SIZE : MTS_INTER_MAX_CU_SIZE;
+
+ mtsAllowed &= CU::isIntra( *tu.cu ) ? tu.cs->sps->getSpsNext().getUseIntraMTS() : tu.cs->sps->getSpsNext().getUseInterMTS();
+ mtsAllowed &= ( tu.lwidth() <= maxSize && tu.lheight() <= maxSize );
+ return mtsAllowed;
+}
+#else
bool TU::hasTransformSkipFlag(const CodingStructure& cs, const CompArea& area)
{
uint32_t transformSkipLog2MaxSize = cs.pps->getPpsRangeExtension().getLog2MaxTransformSkipBlockSize();
@@ -4255,10 +4425,15 @@ bool TU::hasTransformSkipFlag(const CodingStructure& cs, const CompArea& area)
SizeType transformSkipMaxSize = 1 << transformSkipLog2MaxSize;
return area.width <= transformSkipMaxSize && area.height <= transformSkipMaxSize;
}
+#endif
uint32_t TU::getGolombRiceStatisticsIndex(const TransformUnit &tu, const ComponentID &compID)
{
+#if JVET_M0464_UNI_MTS
+ const bool transformSkip = tu.mtsIdx==1;
+#else
const bool transformSkip = tu.transformSkip[compID];
+#endif
const bool transquantBypass = tu.cu->transQuantBypass;
//--------
@@ -4340,7 +4515,11 @@ uint32_t TU::getNumNonZeroCoeffsNonTS( const TransformUnit& tu, const bool bLuma
uint32_t count = 0;
for( uint32_t i = 0; i < ::getNumberValidTBlocks( *tu.cs->pcv ); i++ )
{
+#if JVET_M0464_UNI_MTS
+ if( tu.blocks[i].valid() && ( isLuma(ComponentID(i)) ? tu.mtsIdx !=1 : true ) && TU::getCbf( tu, ComponentID( i ) ) )
+#else
if( tu.blocks[i].valid() && !tu.transformSkip[i] && TU::getCbf( tu, ComponentID( i ) ) )
+#endif
{
if( isLuma ( tu.blocks[i].compID ) && !bLuma ) continue;
if( isChroma( tu.blocks[i].compID ) && !bChroma ) continue;
diff --git a/source/Lib/CommonLib/UnitTools.h b/source/Lib/CommonLib/UnitTools.h
index b844652e5da8d9ebdce9fa509b2aff051053152b..0b7c97f00bbcb8f128e2befbc791a9d559ad5034 100644
--- a/source/Lib/CommonLib/UnitTools.h
+++ b/source/Lib/CommonLib/UnitTools.h
@@ -120,6 +120,9 @@ namespace PU
#if JVET_L0090_PAIR_AVG
bool addMergeHMVPCand(const Slice &slice, MergeCtx& mrgCtx, bool canFastExit, const int& mrgCandIdx, const uint32_t maxNumMergeCandMin1, int &cnt, const int prevCnt, bool isAvailableSubPu, unsigned subPuMvpPos
, int mmvdList
+#if JVET_M0170_MRG_SHARELIST
+ , bool isShared
+#endif
);
#else
bool addMergeHMVPCand(const Slice &slice, MergeCtx& mrgCtx, bool isCandInter[MRG_MAX_NUM_CANDS], bool canFastExit, const int& mrgCandIdx, const uint32_t maxNumMergeCandMin1, int &cnt, const int prevCnt, bool isAvailableSubPu, unsigned subPuMvpPos
@@ -139,11 +142,16 @@ namespace PU
);
bool getInterMergeSubPuMvpCand(const PredictionUnit &pu, MergeCtx &mrgCtx, bool& LICFlag, const int count
, int mmvdList
- , const int countCPR
+#if !JVET_M0409_ATMVP_FIX
+ , const int countIBC
+#endif
);
bool getInterMergeSubPuRecurCand(const PredictionUnit &pu, MergeCtx &mrgCtx, const int count);
bool isBiPredFromDifferentDir (const PredictionUnit &pu);
void restrictBiPredMergeCands (const PredictionUnit &pu, MergeCtx& mrgCtx);
+#if JVET_M0068_M0171_MMVD_CLEANUP
+ void restrictBiPredMergeCandsOne (PredictionUnit &pu);
+#endif
bool isLMCMode ( unsigned mode);
bool isLMCModeEnabled (const PredictionUnit &pu, unsigned mode);
@@ -155,7 +163,7 @@ namespace PU
bool getTriangleWeights (const PredictionUnit &pu, MergeCtx &triangleMrgCtx, const uint8_t candIdx0, const uint8_t candIdx1);
void spanTriangleMotionInfo ( PredictionUnit &pu, MergeCtx &triangleMrgCtx, const uint8_t mergeIdx, const bool splitDir, const uint8_t candIdx0, const uint8_t candIdx1);
int32_t mappingRefPic (const PredictionUnit &pu, int32_t refPicPoc, bool targetRefPicList);
- void getCprMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred);
+ void getIbcMVPsEncOnly(PredictionUnit &pu, Mv* MvPred, int& nbPred);
bool getDerivedBV(PredictionUnit &pu, const Mv& currentMv, Mv& derivedMv);
bool isBlockVectorValid(PredictionUnit& pu, int xPos, int yPos, int width, int height, int picWidth, int picHeight, int xStartInCU, int yStartInCU, int xBv, int yBv, int ctuSize);
}
@@ -171,7 +179,12 @@ namespace TU
bool getCbf (const TransformUnit &tu, const ComponentID &compID);
bool getCbfAtDepth (const TransformUnit &tu, const ComponentID &compID, const unsigned &depth);
void setCbfAtDepth ( TransformUnit &tu, const ComponentID &compID, const unsigned &depth, const bool &cbf);
+#if JVET_M0464_UNI_MTS
+ bool isTSAllowed (const TransformUnit &tu, const ComponentID compID);
+ bool isMTSAllowed (const TransformUnit &tu, const ComponentID compID);
+#else
bool hasTransformSkipFlag (const CodingStructure& cs, const CompArea& area);
+#endif
uint32_t getGolombRiceStatisticsIndex (const TransformUnit &tu, const ComponentID &compID);
#if HEVC_USE_MDCS
uint32_t getCoefScanIdx (const TransformUnit &tu, const ComponentID &compID);
diff --git a/source/Lib/CommonLib/dtrace_blockstatistics.cpp b/source/Lib/CommonLib/dtrace_blockstatistics.cpp
index 6e94dd90471545effdaddf4fe37b0ca9c3191bc3..5971ab4046563e91164918e0f8c3608e9921a4fa 100644
--- a/source/Lib/CommonLib/dtrace_blockstatistics.cpp
+++ b/source/Lib/CommonLib/dtrace_blockstatistics.cpp
@@ -668,7 +668,7 @@ void writeAllData(const CodingStructure& cs, const UnitArea& ctuArea)
DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::IMVMode), cu.imv);
DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::RootCbf), cu.rootCbf);
DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::GBIIndex), cu.GBiIdx);
- DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::CPRFlag), cu.cpr);
+ DTRACE_BLOCK_SCALAR(g_trace_ctx, D_BLOCK_STATISTICS_ALL, cu, GetBlockStatisticName(BlockStatistic::IBCFlag), cu.ibc);
}
break;
diff --git a/source/Lib/CommonLib/dtrace_blockstatistics.h b/source/Lib/CommonLib/dtrace_blockstatistics.h
index df69c1f2eba7b73428e739a2b6ec53a6e17ecc42..34d457098ab571aa683c1cd49efbba568eee2515 100644
--- a/source/Lib/CommonLib/dtrace_blockstatistics.h
+++ b/source/Lib/CommonLib/dtrace_blockstatistics.h
@@ -108,7 +108,7 @@ enum class BlockStatistic {
TriangleMVL0, //<< currently only uni-prediction enabled
TriangleMVL1, //<< currently only uni-prediction enabled
GBIIndex,
- CPRFlag,
+ IBCFlag,
// for dual tree
// general
Depth_Chroma,
@@ -192,7 +192,7 @@ static const std::map<BlockStatistic, std::tuple<std::string, BlockStatisticType
{ BlockStatistic::TriangleMVL0, std::tuple<std::string, BlockStatisticType, std::string>{"TriangleMVL0", BlockStatisticType::VectorPolygon, "Scale: 4"}},
{ BlockStatistic::TriangleMVL1, std::tuple<std::string, BlockStatisticType, std::string>{"TriangleMVL1", BlockStatisticType::VectorPolygon, "Scale: 4"}},
{ BlockStatistic::GBIIndex, std::tuple<std::string, BlockStatisticType, std::string>{"GBIIndex", BlockStatisticType::Integer, "[0, 4]"}},
- { BlockStatistic::CPRFlag, std::tuple<std::string, BlockStatisticType, std::string>{"CPRFlag", BlockStatisticType::Flag, ""}},
+ { BlockStatistic::IBCFlag, std::tuple<std::string, BlockStatisticType, std::string>{"IBCFlag", BlockStatisticType::Flag, ""}},
// for dual tree
{ 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?
diff --git a/source/Lib/CommonLib/x86/CprHashMapX86.h b/source/Lib/CommonLib/x86/IbcHashMapX86.h
similarity index 92%
rename from source/Lib/CommonLib/x86/CprHashMapX86.h
rename to source/Lib/CommonLib/x86/IbcHashMapX86.h
index 133fad437022b8728a04f820a02edc3ca87c31db..67e4c063d06741b1e4d7fe5dd7870326e6da83de 100644
--- a/source/Lib/CommonLib/x86/CprHashMapX86.h
+++ b/source/Lib/CommonLib/x86/IbcHashMapX86.h
@@ -33,11 +33,11 @@
/**
* \file
- * \brief Implementation of CprHashMap class
+ * \brief Implementation of IbcHashMap class
*/
#include "CommonDefX86.h"
-#include "../CprHashMap.h"
+#include "../IbcHashMap.h"
#ifdef TARGET_SIMD_X86
@@ -50,12 +50,12 @@ static uint32_t simdComputeCrc32c16bit(uint32_t crc, const Pel pel)
}
template <X86_VEXT vext>
-void CprHashMap::_initCprHashMapX86()
+void IbcHashMap::_initIbcHashMapX86()
{
m_computeCrc32c = simdComputeCrc32c16bit<vext>;
}
-template void CprHashMap::_initCprHashMapX86<SIMDX86>();
+template void IbcHashMap::_initIbcHashMapX86<SIMDX86>();
#endif //#ifdef TARGET_SIMD_X86
diff --git a/source/Lib/CommonLib/x86/InitX86.cpp b/source/Lib/CommonLib/x86/InitX86.cpp
index 14642e1611a5ad1b57d8d7e36dfe714c46626cf7..00f18ae12450d7db4c8c03c51cd6d3c31850b756 100644
--- a/source/Lib/CommonLib/x86/InitX86.cpp
+++ b/source/Lib/CommonLib/x86/InitX86.cpp
@@ -48,7 +48,7 @@
#include "CommonLib/AdaptiveLoopFilter.h"
-#include "CommonLib/CprHashMap.h"
+#include "CommonLib/IbcHashMap.h"
#ifdef TARGET_SIMD_X86
@@ -167,8 +167,8 @@ void AdaptiveLoopFilter::initAdaptiveLoopFilterX86()
}
#endif
-#if ENABLE_SIMD_OPT_CPR
-void CprHashMap::initCprHashMapX86()
+#if ENABLE_SIMD_OPT_IBC
+void IbcHashMap::initIbcHashMapX86()
{
auto vext = read_x86_extension_flags();
switch (vext)
@@ -177,7 +177,7 @@ void CprHashMap::initCprHashMapX86()
case AVX2:
case AVX:
case SSE42:
- _initCprHashMapX86<SSE42>();
+ _initIbcHashMapX86<SSE42>();
break;
case SSE41:
default:
diff --git a/source/Lib/CommonLib/x86/sse42/CprHashmap_sse42.cpp b/source/Lib/CommonLib/x86/sse42/CprHashmap_sse42.cpp
deleted file mode 100644
index 2a26c1c63909d81c48b0f4f7f07bf36e46b76f2e..0000000000000000000000000000000000000000
--- a/source/Lib/CommonLib/x86/sse42/CprHashmap_sse42.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "../CprHashMapX86.h"
diff --git a/source/Lib/CommonLib/x86/sse42/IbcHashmap_sse42.cpp b/source/Lib/CommonLib/x86/sse42/IbcHashmap_sse42.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..5aca413b98ebec0d6d55418864d6dce88427736e
--- /dev/null
+++ b/source/Lib/CommonLib/x86/sse42/IbcHashmap_sse42.cpp
@@ -0,0 +1 @@
+#include "../IbcHashMapX86.h"
diff --git a/source/Lib/DecoderLib/CABACReader.cpp b/source/Lib/DecoderLib/CABACReader.cpp
index 4105d63632773fd5d513262abe07b85c6110eb4a..6015b1921d223706e2740de44633a9e6e42c505d 100644
--- a/source/Lib/DecoderLib/CABACReader.cpp
+++ b/source/Lib/DecoderLib/CABACReader.cpp
@@ -149,7 +149,7 @@ bool CABACReader::coding_tree_unit( CodingStructure& cs, const UnitArea& area, i
AlfSliceParam& alfSliceParam = cs.slice->getAlfSliceParam();
- if( cs.sps->getUseALF() && ( alfSliceParam.enabledFlag[COMPONENT_Y] || alfSliceParam.enabledFlag[COMPONENT_Cb] || alfSliceParam.enabledFlag[COMPONENT_Cr] ) )
+ if( cs.sps->getALFEnabledFlag() && ( alfSliceParam.enabledFlag[COMPONENT_Y] || alfSliceParam.enabledFlag[COMPONENT_Cb] || alfSliceParam.enabledFlag[COMPONENT_Cr] ) )
{
const PreCalcValues& pcv = *cs.pcv;
@@ -231,7 +231,7 @@ void CABACReader::sao( CodingStructure& cs, unsigned ctuRsAddr )
{
const SPS& sps = *cs.sps;
- if( !sps.getUseSAO() )
+ if( !sps.getSAOEnabledFlag() )
{
return;
}
@@ -420,7 +420,18 @@ bool CABACReader::coding_tree( CodingStructure& cs, Partitioner& partitioner, CU
pCuCtxChroma->isChromaQpAdjCoded = false;
}
}
+#if JVET_M0170_MRG_SHARELIST
+ int startShareThisLevel = 0;
+#endif
+
+#if JVET_M0421_SPLIT_SIG
+ const PartSplit splitMode = split_cu_mode( cs, partitioner );
+
+ CHECK( !partitioner.canSplit( splitMode, cs ), "Got an invalid split!" );
+ if( splitMode != CU_DONT_SPLIT )
+ {
+#else
const PartSplit implicitSplit = partitioner.getImplicitSplit( cs );
// QT
@@ -440,6 +451,34 @@ bool CABACReader::coding_tree( CodingStructure& cs, Partitioner& partitioner, CU
// quad-tree split
if( qtSplit )
{
+#endif
+#if JVET_M0170_MRG_SHARELIST
+ const PartSplit split = splitMode;
+ int splitRatio = 1;
+ CHECK(!(split == CU_QUAD_SPLIT || split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT
+ || split == CU_TRIH_SPLIT || split == CU_TRIV_SPLIT), "invalid split type");
+ splitRatio = (split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT) ? 1 : 2;
+
+ bool isOneChildSmall = (((partitioner.currArea().lwidth())*(partitioner.currArea().lheight())) >> splitRatio) < MRG_SHARELIST_SHARSIZE;
+
+ if ((((partitioner.currArea().lwidth())*(partitioner.currArea().lheight())) > (MRG_SHARELIST_SHARSIZE * 1)))
+ {
+ shareStateDec = NO_SHARE;
+ }
+
+ if (shareStateDec == NO_SHARE)//init state
+ {
+ if (isOneChildSmall)
+ {
+ shareStateDec = SHARING;//share start state
+ startShareThisLevel = 1;
+
+ shareParentPos = partitioner.currArea().lumaPos();
+ shareParentSize.width = partitioner.currArea().lwidth();
+ shareParentSize.height = partitioner.currArea().lheight();
+ }
+ }
+#endif
if (CS::isDualITree(cs) && pPartitionerChroma != nullptr && (partitioner.currArea().lwidth() >= 64 || partitioner.currArea().lheight() >= 64))
{
partitioner.splitCurrArea(CU_QUAD_SPLIT, cs);
@@ -519,7 +558,11 @@ bool CABACReader::coding_tree( CodingStructure& cs, Partitioner& partitioner, CU
}
else
{
+#if JVET_M0421_SPLIT_SIG
+ partitioner.splitCurrArea( splitMode, cs );
+#else
partitioner.splitCurrArea( CU_QUAD_SPLIT, cs );
+#endif
do
{
if( !lastSegment && cs.area.blocks[partitioner.chType].contains( partitioner.currArea().blocks[partitioner.chType].pos() ) )
@@ -530,10 +573,17 @@ bool CABACReader::coding_tree( CodingStructure& cs, Partitioner& partitioner, CU
partitioner.exitCurrSplit();
}
+#if JVET_M0170_MRG_SHARELIST
+ if (startShareThisLevel == 1)
+ shareStateDec = NO_SHARE;
+#endif
return lastSegment;
+#if !JVET_M0421_SPLIT_SIG
}
+#endif
}
+#if !JVET_M0421_SPLIT_SIG
{
// MT
bool mtSplit = partitioner.canSplit( CU_MT_SPLIT, cs );
@@ -544,6 +594,33 @@ bool CABACReader::coding_tree( CodingStructure& cs, Partitioner& partitioner, CU
if( splitMode != CU_DONT_SPLIT )
{
+#if JVET_M0170_MRG_SHARELIST
+ const PartSplit split = splitMode;
+ int splitRatio = 1;
+ CHECK(!(split == CU_QUAD_SPLIT || split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT
+ || split == CU_TRIH_SPLIT || split == CU_TRIV_SPLIT), "invalid split type");
+ splitRatio = (split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT) ? 1 : 2;
+
+ bool isOneChildSmall = (((partitioner.currArea().lwidth())*(partitioner.currArea().lheight())) >> splitRatio) < MRG_SHARELIST_SHARSIZE;
+
+ if ((((partitioner.currArea().lwidth())*(partitioner.currArea().lheight())) > (MRG_SHARELIST_SHARSIZE * 1)))
+ {
+ shareStateDec = NO_SHARE;
+ }
+
+ if (shareStateDec == NO_SHARE)//init state
+ {
+ if (isOneChildSmall)
+ {
+ shareStateDec = SHARING;//share start state
+ startShareThisLevel = 1;
+
+ shareParentPos = partitioner.currArea().lumaPos();
+ shareParentSize.width = partitioner.currArea().lwidth();
+ shareParentSize.height = partitioner.currArea().lheight();
+ }
+ }
+#endif
partitioner.splitCurrArea( splitMode, cs );
do
@@ -555,12 +632,16 @@ bool CABACReader::coding_tree( CodingStructure& cs, Partitioner& partitioner, CU
} while( partitioner.nextPart( cs ) );
partitioner.exitCurrSplit();
+#if JVET_M0170_MRG_SHARELIST
+ if (startShareThisLevel == 1)
+ shareStateDec = NO_SHARE;
+#endif
return lastSegment;
}
}
}
-
+#endif
CodingUnit& cu = cs.addCU( CS::getArea( cs, currArea, partitioner.chType ), partitioner.chType );
partitioner.setCUData( cu );
@@ -588,13 +669,91 @@ bool CABACReader::coding_tree( CodingStructure& cs, Partitioner& partitioner, CU
cu.chromaQpAdj = cs.chromaQpAdj; //NOTE: CU chroma QP adjustment can be changed by adjustment signaling at TU level
// coding unit
+#if JVET_M0170_MRG_SHARELIST
+ cu.shareParentPos = (shareStateDec == SHARING) ? shareParentPos : partitioner.currArea().lumaPos();
+ cu.shareParentSize = (shareStateDec == SHARING) ? shareParentSize : partitioner.currArea().lumaSize();
+#endif
bool isLastCtu = coding_unit( cu, partitioner, cuCtx );
DTRACE( g_trace_ctx, D_QP, "x=%d, y=%d, w=%d, h=%d, qp=%d\n", cu.Y().x, cu.Y().y, cu.Y().width, cu.Y().height, cu.qp );
+#if JVET_M0170_MRG_SHARELIST
+ if (startShareThisLevel == 1)
+ shareStateDec = NO_SHARE;
+#endif
return isLastCtu;
}
+#if JVET_M0421_SPLIT_SIG
+PartSplit CABACReader::split_cu_mode( CodingStructure& cs, Partitioner &partitioner )
+{
+ RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET_SIZE2( STATS__CABAC_BITS__SPLIT_FLAG, partitioner.currArea().blocks[partitioner.chType].size(), partitioner.chType );
+
+ PartSplit mode = CU_DONT_SPLIT;
+
+ bool canNo, canQt, canBh, canBv, canTh, canTv;
+ partitioner.canSplit( cs, canNo, canQt, canBh, canBv, canTh, canTv );
+
+ bool canSpl[6] = { canNo, canQt, canBh, canBv, canTh, canTv };
+
+ unsigned ctxSplit = 0, ctxQtSplit = 0, ctxBttHV = 0, ctxBttH12 = 0, ctxBttV12;
+ DeriveCtx::CtxSplit( cs, partitioner, ctxSplit, ctxQtSplit, ctxBttHV, ctxBttH12, ctxBttV12, canSpl );
+
+ bool isSplit = canBh || canBv || canTh || canTv || canQt;
+
+ if( canNo && isSplit )
+ {
+ isSplit = m_BinDecoder.decodeBin( Ctx::SplitFlag( ctxSplit ) );
+ }
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode() ctx=%d split=%d\n", ctxSplit, isSplit );
+
+ if( !isSplit )
+ {
+ return CU_DONT_SPLIT;
+ }
+
+ const bool canBtt = canBh || canBv || canTh || canTv;
+ bool isQt = canQt;
+
+ if( isQt && canBtt )
+ {
+ isQt = m_BinDecoder.decodeBin( Ctx::SplitQtFlag( ctxQtSplit ) );
+ }
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode() ctx=%d qt=%d\n", ctxQtSplit, isQt );
+
+ if( isQt )
+ {
+ return CU_QUAD_SPLIT;
+ }
+
+ const bool canHor = canBh || canTh;
+ bool isVer = canBv || canTv;
+
+ if( isVer && canHor )
+ {
+ isVer = m_BinDecoder.decodeBin( Ctx::SplitHvFlag( ctxBttHV ) );
+ }
+
+ const bool can14 = isVer ? canTv : canTh;
+ bool is12 = isVer ? canBv : canBh;
+
+ if( is12 && can14 )
+ {
+ is12 = m_BinDecoder.decodeBin( Ctx::Split12Flag( isVer ? ctxBttV12 : ctxBttH12 ) );
+ }
+
+ if ( isVer && is12 ) mode = CU_VERT_SPLIT;
+ else if( isVer && !is12 ) mode = CU_TRIV_SPLIT;
+ else if( !isVer && is12 ) mode = CU_HORZ_SPLIT;
+ else mode = CU_TRIH_SPLIT;
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode() ctxHv=%d ctx12=%d mode=%d\n", ctxBttHV, isVer ? ctxBttV12 : ctxBttH12, mode );
+
+ return mode;
+}
+#else
PartSplit CABACReader::split_cu_mode_mt( CodingStructure& cs, Partitioner &partitioner )
{
RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET( STATS__CABAC_BITS__SPLIT_FLAG );
@@ -606,6 +765,50 @@ PartSplit CABACReader::split_cu_mode_mt( CodingStructure& cs, Partitioner &parti
unsigned width = partitioner.currArea().lumaSize().width;
unsigned height = partitioner.currArea().lumaSize().height;
+#if REMOVE_BIN_DECISION_TREE
+ unsigned btSCtxId = width == height ? 0 : ( width > height ? 1 : 2 );
+
+ const bool canNo = partitioner.canSplit( CU_DONT_SPLIT, cs );
+ const bool canBh = partitioner.canSplit( CU_HORZ_SPLIT, cs );
+ const bool canBv = partitioner.canSplit( CU_VERT_SPLIT, cs );
+ const bool canTh = partitioner.canSplit( CU_TRIH_SPLIT, cs );
+ const bool canTv = partitioner.canSplit( CU_TRIV_SPLIT, cs );
+
+ bool isSplit = canBh || canBv || canTh || canTv;
+
+ if( canNo && isSplit )
+ {
+ isSplit = m_BinDecoder.decodeBin( Ctx::BTSplitFlag( ctxIdBT ) );
+ }
+
+ if( !isSplit )
+ {
+ DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode_mt() ctx=%d split=%d\n", ctxIdBT, mode );
+
+ return mode;
+ }
+
+ const bool canHor = canBh || canTh;
+ bool isVer = canBv || canTv;
+
+ if( isVer && canHor )
+ {
+ isVer = m_BinDecoder.decodeBin( Ctx::BTSplitFlag( 12 + btSCtxId ) );
+ }
+
+ const bool can14 = isVer ? canTv : canTh;
+ bool is12 = isVer ? canBv : canBh;
+
+ if( is12 && can14 )
+ {
+ is12 = m_BinDecoder.decodeBin( Ctx::BTSplitFlag( 15 ) );
+ }
+
+ if ( isVer && is12 ) mode = CU_VERT_SPLIT;
+ else if( isVer && !is12 ) mode = CU_TRIV_SPLIT;
+ else if( !isVer && is12 ) mode = CU_HORZ_SPLIT;
+ else mode = CU_TRIH_SPLIT;
+#else
DecisionTree dt( g_mtSplitDTT );
dt.setAvail( DTT_SPLIT_BT_HORZ, partitioner.canSplit( CU_HORZ_SPLIT, cs ) );
@@ -626,6 +829,7 @@ PartSplit CABACReader::split_cu_mode_mt( CodingStructure& cs, Partitioner &parti
mode = id == DTT_SPLIT_NO_SPLIT ? CU_DONT_SPLIT : PartSplit( id );
+#endif
DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode_mt() ctx=%d split=%d\n", ctxIdBT, mode );
return mode;
@@ -649,6 +853,7 @@ bool CABACReader::split_cu_flag( CodingStructure& cs, Partitioner &partitioner )
return split;
}
+#endif
//================================================================================
// clause 7.3.8.5
@@ -690,6 +895,10 @@ bool CABACReader::coding_unit( CodingUnit &cu, Partitioner &partitioner, CUCtx&
{
cs.addTU ( cu, partitioner.chType );
PredictionUnit& pu = cs.addPU( cu, partitioner.chType );
+#if JVET_M0170_MRG_SHARELIST
+ pu.shareParentPos = cu.shareParentPos;
+ pu.shareParentSize = cu.shareParentSize;
+#endif
MergeCtx mrgCtx;
prediction_unit ( pu, mrgCtx );
return end_of_ctu( cu, cuCtx );
@@ -774,7 +983,7 @@ void CABACReader::imv_mode( CodingUnit& cu, MergeCtx& mrgCtx )
unsigned value = 0;
unsigned ctxId = DeriveCtx::CtxIMVFlag( cu );
- if (cu.firstPU->interDir == 1 && cu.cs->slice->getRefPic(REF_PIC_LIST_0, cu.firstPU->refIdx[REF_PIC_LIST_0])->getPOC() == cu.cs->slice->getPOC()) // the first bin of IMV flag does need to be signaled in CPR block
+ if (cu.firstPU->interDir == 1 && cu.cs->slice->getRefPic(REF_PIC_LIST_0, cu.firstPU->refIdx[REF_PIC_LIST_0])->getPOC() == cu.cs->slice->getPOC()) // the first bin of IMV flag does need to be signaled in IBC block
value = 1;
else
value = m_BinDecoder.decodeBin( Ctx::ImvFlag( ctxId ) );
@@ -795,7 +1004,11 @@ void CABACReader::pred_mode( CodingUnit& cu )
{
RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET( STATS__CABAC_BITS__PRED_MODE );
+#if JVET_M0502_PRED_MODE_CTX
+ if( cu.cs->slice->isIntra() || m_BinDecoder.decodeBin( Ctx::PredMode(DeriveCtx::CtxPredModeFlag(cu)) ) )
+#else
if( cu.cs->slice->isIntra() || m_BinDecoder.decodeBin( Ctx::PredMode() ) )
+#endif
{
cu.predMode = MODE_INTRA;
}
@@ -808,7 +1021,7 @@ void CABACReader::pred_mode( CodingUnit& cu )
void CABACReader::pcm_flag( CodingUnit& cu, Partitioner &partitioner )
{
const SPS& sps = *cu.cs->sps;
- if( !sps.getUsePCM() || partitioner.currArea().lwidth() > (1 << sps.getPCMLog2MaxSize()) || partitioner.currArea().lwidth() < (1 << sps.getPCMLog2MinSize())
+ if( !sps.getPCMEnabledFlag() || partitioner.currArea().lwidth() > (1 << sps.getPCMLog2MaxSize()) || partitioner.currArea().lwidth() < (1 << sps.getPCMLog2MinSize())
|| partitioner.currArea().lheight() > (1 << sps.getPCMLog2MaxSize()) || partitioner.currArea().lheight() < (1 << sps.getPCMLog2MinSize()) )
{
cu.ipcm = false;
@@ -829,13 +1042,17 @@ void CABACReader::cu_pred_data( CodingUnit &cu )
if (!cu.Y().valid()) // dual tree chroma CU
{
cu.predMode = MODE_INTER;
- cu.cpr = true;
+ cu.ibc = true;
return;
}
MergeCtx mrgCtx;
for( auto &pu : CU::traversePUs( cu ) )
{
+#if JVET_M0170_MRG_SHARELIST
+ pu.shareParentPos = cu.shareParentPos;
+ pu.shareParentSize = cu.shareParentSize;
+#endif
prediction_unit( pu, mrgCtx );
}
@@ -1206,6 +1423,9 @@ void CABACReader::prediction_unit( PredictionUnit& pu, MergeCtx& mrgCtx )
{
inter_pred_idc( pu );
affine_flag ( *pu.cu );
+#if JVET_M0444_SMVD
+ smvd_mode( pu );
+#endif
if( pu.interDir != 2 /* PRED_L1 */ )
{
@@ -1228,6 +1448,10 @@ void CABACReader::prediction_unit( PredictionUnit& pu, MergeCtx& mrgCtx )
if( pu.interDir != 1 /* PRED_L0 */ )
{
+#if JVET_M0444_SMVD
+ if ( pu.cu->smvdMode != 1 )
+ {
+#endif
ref_idx ( pu, REF_PIC_LIST_1 );
if( pu.cu->cs->slice->getMvdL1ZeroFlag() && pu.interDir == 3 /* PRED_BI */ )
{
@@ -1249,6 +1473,9 @@ void CABACReader::prediction_unit( PredictionUnit& pu, MergeCtx& mrgCtx )
{
mvd_coding( pu.mvd[REF_PIC_LIST_1] );
}
+#if JVET_M0444_SMVD
+ }
+#endif
mvp_flag ( pu, REF_PIC_LIST_1 );
}
}
@@ -1260,9 +1487,40 @@ void CABACReader::prediction_unit( PredictionUnit& pu, MergeCtx& mrgCtx )
pu.cu->GBiIdx = GBI_DEFAULT;
}
+#if JVET_M0444_SMVD
+ if ( pu.cu->smvdMode )
+ {
+ RefPicList eCurRefList = (RefPicList)(pu.cu->smvdMode - 1);
+ pu.mvd[1 - eCurRefList].set( -pu.mvd[eCurRefList].hor, -pu.mvd[eCurRefList].ver );
+ pu.refIdx[1 - eCurRefList] = pu.cs->slice->getSymRefIdx( 1 - eCurRefList );
+ }
+#endif
+
PU::spanMotionInfo( pu, mrgCtx );
}
+#if JVET_M0444_SMVD
+void CABACReader::smvd_mode( PredictionUnit& pu )
+{
+ pu.cu->smvdMode = 0;
+ if ( pu.interDir != 3 || pu.cu->affine )
+ {
+ return;
+ }
+
+ if ( pu.cs->slice->getBiDirPred() == false )
+ {
+ return;
+ }
+
+ RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET( STATS__CABAC_BITS__SYMMVD_FLAG );
+
+ pu.cu->smvdMode = m_BinDecoder.decodeBin( Ctx::SmvdFlag() ) ? 1 : 0;
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "symmvd_flag() symmvd=%d pos=(%d,%d) size=%dx%d\n", pu.cu->smvdMode ? 1 : 0, pu.lumaPos().x, pu.lumaPos().y, pu.lumaSize().width, pu.lumaSize().height );
+}
+#endif
+
void CABACReader::subblock_merge_flag( CodingUnit& cu )
{
if ( cu.firstPU->mergeFlag && (cu.firstPU->mmvdMergeFlag || cu.mmvdSkip) )
@@ -1270,7 +1528,7 @@ void CABACReader::subblock_merge_flag( CodingUnit& cu )
return;
}
- if ( !cu.cs->slice->isIntra() && (cu.cs->sps->getSpsNext().getUseAffine() || cu.cs->sps->getSpsNext().getUseSubPuMvp()) && cu.lumaSize().width >= 8 && cu.lumaSize().height >= 8 )
+ if ( !cu.cs->slice->isIntra() && (cu.cs->sps->getSpsNext().getUseAffine() || cu.cs->sps->getSBTMVPEnabledFlag()) && cu.lumaSize().width >= 8 && cu.lumaSize().height >= 8 )
{
RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET( STATS__CABAC_BITS__AFFINE_FLAG );
@@ -1342,7 +1600,7 @@ void CABACReader::merge_idx( PredictionUnit& pu )
{
if ( m_BinDecoder.decodeBin( Ctx::AffMergeIdx() ) )
{
- bool useExtCtx = pu.cs->sps->getSpsNext().getUseSubPuMvp();
+ bool useExtCtx = pu.cs->sps->getSBTMVPEnabledFlag();
pu.mergeIdx++;
for ( ; pu.mergeIdx < numCandminus1; pu.mergeIdx++ )
{
@@ -1508,6 +1766,14 @@ void CABACReader::ref_idx( PredictionUnit &pu, RefPicList eRefList )
{
RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET( STATS__CABAC_BITS__REF_FRM_IDX );
+#if JVET_M0444_SMVD
+ if ( pu.cu->smvdMode )
+ {
+ pu.refIdx[eRefList] = pu.cs->slice->getSymRefIdx( eRefList );
+ return;
+ }
+#endif
+
int numRef = pu.cs->slice->getNumRefIdx(eRefList);
if( numRef <= 1 || !m_BinDecoder.decodeBin( Ctx::RefPic() ) )
{
@@ -1685,6 +1951,18 @@ void CABACReader::triangle_mode( CodingUnit& cu )
return;
}
+#if JVET_M0118_M0185_TRIANGLE_FLAG_FIX
+ if ( cu.firstPU->mmvdMergeFlag || cu.mmvdSkip )
+ {
+ return;
+ }
+
+ if ( cu.firstPU->mhIntraFlag )
+ {
+ return;
+ }
+#endif
+
unsigned flag_idx = DeriveCtx::CtxTriangleFlag( cu );
cu.triangle = m_BinDecoder.decodeBin( Ctx::TriangleFlag(flag_idx) );
@@ -1773,7 +2051,9 @@ void CABACReader::transform_tree( CodingStructure &cs, Partitioner &partitioner,
if( split )
{
{
+#if !JVET_M0464_UNI_MTS
if( trDepth == 0 ) emt_cu_flag( cu );
+#endif
if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
{
@@ -1857,7 +2137,9 @@ void CABACReader::transform_tree( CodingStructure &cs, Partitioner &partitioner,
TU::setCbfAtDepth( tu, COMPONENT_Cr, trDepth, ( chromaCbfs.Cr ? 1 : 0 ) );
}
+#if !JVET_M0464_UNI_MTS
if( trDepth == 0 && TU::getCbfAtDepth( tu, COMPONENT_Y, 0 ) ) emt_cu_flag( cu );
+#endif
transform_unit( tu, cuCtx, chromaCbfs );
}
@@ -2054,14 +2336,22 @@ void CABACReader::residual_coding( TransformUnit& tu, ComponentID compID )
DTRACE( g_trace_ctx, D_SYNTAX, "residual_coding() etype=%d pos=(%d,%d) size=%dx%d predMode=%d\n", tu.blocks[compID].compID, tu.blocks[compID].x, tu.blocks[compID].y, tu.blocks[compID].width, tu.blocks[compID].height, cu.predMode );
// parse transform skip and explicit rdpcm mode
+#if JVET_M0464_UNI_MTS
+ mts_coding ( tu, compID );
+#else
transform_skip_flag( tu, compID );
+#endif
explicit_rdpcm_mode( tu, compID );
#if HEVC_USE_SIGN_HIDING
// determine sign hiding
bool signHiding = ( cu.cs->slice->getSignDataHidingEnabledFlag() && !cu.transQuantBypass && tu.rdpcm[compID] == RDPCM_OFF );
+#if JVET_M0464_UNI_MTS
+ if( signHiding && CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && tu.mtsIdx==1 )
+#else
if( signHiding && CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && tu.transformSkip[compID] )
+#endif
{
const ChannelType chType = toChannelType( compID );
const unsigned intraMode = PU::getFinalIntraMode( *cu.cs->getPU( tu.blocks[compID].pos(), chType ), chType );
@@ -2079,7 +2369,9 @@ void CABACReader::residual_coding( TransformUnit& tu, ComponentID compID )
CoeffCodingContext cctx ( tu, compID );
#endif
TCoeff* coeff = tu.getCoeffs( compID ).buf;
+#if !JVET_M0464_UNI_MTS
unsigned numSig = 0;
+#endif
// parse last coeff position
cctx.setScanPosLast( last_sig_coeff( cctx ) );
@@ -2088,21 +2380,25 @@ void CABACReader::residual_coding( TransformUnit& tu, ComponentID compID )
const int stateTransTab = ( tu.cs->slice->getDepQuantEnabledFlag() ? 32040 : 0 );
int state = 0;
+#if !JVET_M0464_UNI_MTS
bool useEmt = ( cu.cs->sps->getSpsNext().getUseIntraEMT() && cu.predMode == MODE_INTRA ) || ( cu.cs->sps->getSpsNext().getUseInterEMT() && cu.predMode != MODE_INTRA );
useEmt = useEmt && isLuma(compID);
+#endif
for( int subSetId = ( cctx.scanPosLast() >> cctx.log2CGSize() ); subSetId >= 0; subSetId--)
{
cctx.initSubblock ( subSetId );
residual_coding_subblock( cctx, coeff, stateTransTab, state );
+#if !JVET_M0464_UNI_MTS
if (useEmt)
{
numSig += cctx.emtNumSigCoeff();
cctx.setEmtNumSigCoeff( 0 );
}
+#endif
}
-
+#if !JVET_M0464_UNI_MTS
if( useEmt && !tu.transformSkip[compID] && compID == COMPONENT_Y && tu.cu->emtFlag )
{
if( CU::isIntra( *tu.cu ) )
@@ -2114,9 +2410,55 @@ void CABACReader::residual_coding( TransformUnit& tu, ComponentID compID )
emt_tu_index( tu );
}
}
+#endif
}
+#if JVET_M0464_UNI_MTS
+void CABACReader::mts_coding( TransformUnit& tu, ComponentID compID )
+{
+ const CodingUnit &cu = *tu.cu;
+ const bool tsAllowed = TU::isTSAllowed ( tu, compID );
+ const bool mtsAllowed = TU::isMTSAllowed( tu, compID );
+
+ if( !mtsAllowed && !tsAllowed ) return;
+
+ int symbol = 0;
+ int ctxIdx = 0;
+
+ if( tsAllowed )
+ {
+ ctxIdx = 6;
+ symbol = m_BinDecoder.decodeBin( Ctx::MTSIndex( ctxIdx ) );
+ tu.mtsIdx = 1-symbol; // 1 = TS
+ }
+
+ if( tu.mtsIdx != 1 )
+ {
+ if( mtsAllowed )
+ {
+ ctxIdx = std::min( (int)cu.qtDepth, 5 );
+ symbol = m_BinDecoder.decodeBin( Ctx::MTSIndex( ctxIdx ) );
+
+ if( symbol )
+ {
+ ctxIdx = 7;
+ tu.mtsIdx = 2; // mtsIdx = 2 -- 4
+ for( int i = 0; i < 3; i++, ctxIdx++ )
+ {
+ symbol = m_BinDecoder.decodeBin( Ctx::MTSIndex( ctxIdx ) );
+ tu.mtsIdx += symbol;
+ if( !symbol )
+ {
+ break;
+ }
+ }
+ }
+ }
+ }
+ DTRACE( g_trace_ctx, D_SYNTAX, "mts_coding() etype=%d pos=(%d,%d) mtsIdx=%d\n", COMPONENT_Y, cu.lx(), cu.ly(), tu.mtsIdx );
+}
+#else
void CABACReader::transform_skip_flag( TransformUnit& tu, ComponentID compID )
{
@@ -2195,7 +2537,7 @@ void CABACReader::emt_cu_flag( CodingUnit& cu )
DTRACE( g_trace_ctx, D_SYNTAX, "emt_cu_flag() etype=%d pos=(%d,%d) emtCuFlag=%d\n", COMPONENT_Y, cu.lx(), cu.ly(), ( int ) cu.emtFlag );
}
}
-
+#endif
void CABACReader::explicit_rdpcm_mode( TransformUnit& tu, ComponentID compID )
{
@@ -2203,7 +2545,11 @@ void CABACReader::explicit_rdpcm_mode( TransformUnit& tu, ComponentID compID )
tu.rdpcm[compID] = RDPCM_OFF;
+#if JVET_M0464_UNI_MTS
+ if( !CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && ( tu.mtsIdx==1 || cu.transQuantBypass ) )
+#else
if( !CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && ( tu.transformSkip[compID] || cu.transQuantBypass ) )
+#endif
{
RExt__DECODER_DEBUG_BIT_STATISTICS_CREATE_SET_SIZE( STATS__EXPLICIT_RDPCM_BITS, tu.blocks[tu.chType].lumaSize() );
@@ -2332,13 +2678,23 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
#endif
int numNonZero = 0;
bool is2x2subblock = ( cctx.log2CGSize() == 2 );
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ int remRegBins = ( is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK );
+#else
int remGt2Bins = ( is2x2subblock ? MAX_NUM_GT2_BINS_2x2SUBBLOCK : MAX_NUM_GT2_BINS_4x4SUBBLOCK );
int remRegBins = ( is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK ) - remGt2Bins;
+#endif
int firstPosMode2 = minSubPos - 1;
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
int firstPosMode1 = minSubPos - 1;
+#endif
int sigBlkPos[ 1 << MLS_CG_SIZE ];
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ for( ; nextSigPos >= minSubPos && remRegBins >= 4; nextSigPos-- )
+#else
for( ; nextSigPos >= minSubPos && remRegBins >= 3; nextSigPos-- )
+#endif
{
int blkPos = cctx.blockPos( nextSigPos );
unsigned sigFlag = ( !numNonZero && nextSigPos == inferSigPos );
@@ -2367,6 +2723,9 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
remRegBins--;
unsigned parFlag = 0;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ unsigned gt2Flag = 0;
+#endif
if( gt1Flag )
{
RExt__DECODER_DEBUG_BIT_STATISTICS_SET( ctype_par );
@@ -2374,19 +2733,33 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
DTRACE( g_trace_ctx, D_SYNTAX_RESI, "par_flag() bin=%d ctx=%d\n", parFlag, cctx.parityCtxIdAbs( ctxOff ) );
remRegBins--;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ RExt__DECODER_DEBUG_BIT_STATISTICS_SET(ctype_gt2);
+ gt2Flag = m_BinDecoder.decodeBin( cctx.greater2CtxIdAbs( ctxOff ) );
+ DTRACE( g_trace_ctx, D_SYNTAX_RESI, "gt2_flag() bin=%d ctx=%d\n", gt2Flag, cctx.greater2CtxIdAbs( ctxOff ) );
+ remRegBins--;
+#else
if( remGt2Bins && !--remGt2Bins )
{
firstPosMode1 = nextSigPos - 1;
}
+#endif
}
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ coeff[ blkPos ] += 1 + parFlag + gt1Flag + (gt2Flag << 1);
+#else
coeff[ blkPos ] += 1 + parFlag + gt1Flag;
+#endif
}
state = ( stateTransTable >> ((state<<2)+((coeff[blkPos]&1)<<1)) ) & 3;
}
firstPosMode2 = nextSigPos;
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
firstPosMode1 = ( firstPosMode1 > firstPosMode2 ? firstPosMode1 : firstPosMode2 );
+#endif
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
//===== 2nd PASS: gt2 =====
for( int scanPos = firstSigPos; scanPos > firstPosMode1; scanPos-- )
{
@@ -2400,10 +2773,17 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
tcoeff += (gt2Flag<<1);
}
}
+#endif
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ //===== 2nd PASS: Go-rice codes =====
+ unsigned ricePar = 0;
+ for( int scanPos = firstSigPos; scanPos > firstPosMode2; scanPos-- )
+#else
//===== 3rd PASS: Go-rice codes =====
unsigned ricePar = 0;
for( int scanPos = firstSigPos; scanPos > firstPosMode1; scanPos-- )
+#endif
{
TCoeff& tcoeff = coeff[ cctx.blockPos( scanPos ) ];
if( tcoeff >= 4 )
@@ -2418,6 +2798,7 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
}
}
}
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
for( int scanPos = firstPosMode1; scanPos > firstPosMode2; scanPos-- )
{
TCoeff& tcoeff = coeff[ cctx.blockPos( scanPos ) ];
@@ -2433,6 +2814,7 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
}
}
}
+#endif
//===== coeff bypass ====
for( int scanPos = firstPosMode2; scanPos >= minSubPos; scanPos-- )
@@ -2486,7 +2868,9 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
coeff[ sigBlkPos[k] ] = ( sumAbs & 1 ? -AbsCoeff : AbsCoeff );
}
#endif
+#if !JVET_M0464_UNI_MTS
cctx.setEmtNumSigCoeff( numNonZero );
+#endif
}
@@ -2577,7 +2961,7 @@ unsigned CABACReader::exp_golomb_eqprob( unsigned count )
return symbol;
}
-
+#if !REMOVE_BIN_DECISION_TREE
unsigned CABACReader::decode_sparse_dt( DecisionTree& dt )
{
dt.reduce();
@@ -2626,3 +3010,4 @@ unsigned CABACReader::decode_sparse_dt( DecisionTree& dt )
return dt.dtt.ids[offset];
}
+#endif
diff --git a/source/Lib/DecoderLib/CABACReader.h b/source/Lib/DecoderLib/CABACReader.h
index c1e2f3e75b73c95943a748ba3c4a7dee17bfab1e..8e5586e5007f95e5575812bfd5fbe899116e6aca 100644
--- a/source/Lib/DecoderLib/CABACReader.h
+++ b/source/Lib/DecoderLib/CABACReader.h
@@ -48,7 +48,11 @@
class CABACReader
{
public:
+#if JVET_M0170_MRG_SHARELIST
+ CABACReader(BinDecoderBase& binDecoder) : shareStateDec(0), m_BinDecoder(binDecoder), m_Bitstream(0) {}
+#else
CABACReader( BinDecoderBase& binDecoder ) : m_BinDecoder( binDecoder ), m_Bitstream( 0 ) {}
+#endif
virtual ~CABACReader() {}
public:
@@ -70,8 +74,12 @@ public:
// coding (quad)tree (clause 7.3.8.4)
bool coding_tree ( CodingStructure& cs, Partitioner& pm, CUCtx& cuCtx, Partitioner* pPartitionerChroma = nullptr, CUCtx* pCuCtxChroma = nullptr);
+#if JVET_M0421_SPLIT_SIG
+ PartSplit split_cu_mode ( CodingStructure& cs, Partitioner& pm );
+#else
bool split_cu_flag ( CodingStructure& cs, Partitioner& pm );
PartSplit split_cu_mode_mt ( CodingStructure& cs, Partitioner& pm );
+#endif
// coding unit (clause 7.3.8.5)
bool coding_unit ( CodingUnit& cu, Partitioner& pm, CUCtx& cuCtx );
@@ -105,6 +113,9 @@ public:
void MHIntra_flag ( PredictionUnit& pu );
void MHIntra_luma_pred_modes ( CodingUnit& cu );
void triangle_mode ( CodingUnit& cu );
+ #if JVET_M0444_SMVD
+ void smvd_mode ( PredictionUnit& pu );
+#endif
// pcm samples (clause 7.3.8.7)
void pcm_samples ( TransformUnit& tu );
@@ -123,9 +134,13 @@ public:
// residual coding (clause 7.3.8.11)
void residual_coding ( TransformUnit& tu, ComponentID compID );
+#if JVET_M0464_UNI_MTS
+ void mts_coding ( TransformUnit& tu, ComponentID compID );
+#else
void transform_skip_flag ( TransformUnit& tu, ComponentID compID );
void emt_tu_index ( TransformUnit& tu );
void emt_cu_flag ( CodingUnit& cu );
+#endif
void explicit_rdpcm_mode ( TransformUnit& tu, ComponentID compID );
int last_sig_coeff ( CoeffCodingContext& cctx );
void residual_coding_subblock ( CoeffCodingContext& cctx, TCoeff* coeff, const int stateTransTable, int& state );
@@ -137,11 +152,18 @@ private:
unsigned unary_max_symbol ( unsigned ctxId0, unsigned ctxIdN, unsigned maxSymbol );
unsigned unary_max_eqprob ( unsigned maxSymbol );
unsigned exp_golomb_eqprob ( unsigned count );
+#if !REMOVE_BIN_DECISION_TREE
unsigned decode_sparse_dt ( DecisionTree& dt );
+#endif
unsigned get_num_bits_read () { return m_BinDecoder.getNumBitsRead(); }
void xReadTruncBinCode(uint32_t& symbol, uint32_t maxSymbol);
-
+#if JVET_M0170_MRG_SHARELIST
+public:
+ int shareStateDec;
+ Position shareParentPos;
+ Size shareParentSize;
+#endif
private:
BinDecoderBase& m_BinDecoder;
InputBitstream* m_Bitstream;
diff --git a/source/Lib/DecoderLib/DecCu.cpp b/source/Lib/DecoderLib/DecCu.cpp
index c18a57b9b27fb195067add00b4e428163a1e2fa9..6913f67f690f184b7f47a0e3664118494d989372 100644
--- a/source/Lib/DecoderLib/DecCu.cpp
+++ b/source/Lib/DecoderLib/DecCu.cpp
@@ -81,14 +81,42 @@ void DecCu::init( TrQuant* pcTrQuant, IntraPrediction* pcIntra, InterPrediction*
void DecCu::decompressCtu( CodingStructure& cs, const UnitArea& ctuArea )
{
+
const int maxNumChannelType = cs.pcv->chrFormat != CHROMA_400 && CS::isDualITree( cs ) ? 2 : 1;
+#if JVET_M0170_MRG_SHARELIST
+ if (!cs.pcv->isEncoder)
+ {
+ m_shareStateDec = NO_SHARE;
+ }
+ bool sharePrepareCondition = ((!cs.pcv->isEncoder) && (!(cs.slice->isIntra())));
+#endif
for( int ch = 0; ch < maxNumChannelType; ch++ )
{
const ChannelType chType = ChannelType( ch );
+#if JVET_M0170_MRG_SHARELIST
+ Position prevTmpPos;
+ prevTmpPos.x = -1; prevTmpPos.y = -1;
+#endif
for( auto &currCU : cs.traverseCUs( CS::getArea( cs, ctuArea, chType ), chType ) )
{
+#if JVET_M0170_MRG_SHARELIST
+ if(sharePrepareCondition)
+ {
+ if ((currCU.shareParentPos.x >= 0) && (!(currCU.shareParentPos.x == prevTmpPos.x && currCU.shareParentPos.y == prevTmpPos.y)))
+ {
+ m_shareStateDec = GEN_ON_SHARED_BOUND;
+ cs.slice->copyMotionLUTs(cs.slice->getMotionLUTs(), cs.slice->m_MotionCandLuTsBkup);
+ }
+
+ if (currCU.shareParentPos.x < 0)
+ {
+ m_shareStateDec = 0;
+ }
+ prevTmpPos = currCU.shareParentPos;
+ }
+#endif
cs.chType = chType;
if (currCU.predMode != MODE_INTRA && currCU.Y().valid())
{
@@ -336,10 +364,10 @@ void DecCu::xReconInter(CodingUnit &cu)
m_pcIntraPred->geneIntrainterPred(cu);
// inter prediction
- CHECK(cu.cpr && cu.firstPU->mhIntraFlag, "CPR and MHIntra cannot be used together");
- CHECK(cu.cpr && cu.affine, "CPR and Affine cannot be used together");
- CHECK(cu.cpr && cu.triangle, "CPR and triangle cannot be used together");
- CHECK(cu.cpr && cu.firstPU->mmvdMergeFlag, "CPR and MMVD cannot be used together");
+ CHECK(cu.ibc && cu.firstPU->mhIntraFlag, "IBC and MHIntra cannot be used together");
+ CHECK(cu.ibc && cu.affine, "IBC and Affine cannot be used together");
+ CHECK(cu.ibc && cu.triangle, "IBC and triangle cannot be used together");
+ CHECK(cu.ibc && cu.firstPU->mmvdMergeFlag, "IBC and MMVD cannot be used together");
const bool luma = cu.Y().valid();
const bool chroma = cu.Cb().valid();
if (luma && chroma)
@@ -458,15 +486,21 @@ void DecCu::xDeriveCUMV( CodingUnit &cu )
if (pu.mmvdMergeFlag || pu.cu->mmvdSkip)
{
CHECK(pu.mhIntraFlag == true, "invalid MHIntra");
- if (pu.cs->sps->getSpsNext().getUseSubPuMvp())
+ if (pu.cs->sps->getSBTMVPEnabledFlag())
{
Size bufSize = g_miScaling.scale(pu.lumaSize());
mrgCtx.subPuMvpMiBuf = MotionBuf(m_SubPuMiBuf, bufSize);
}
int fPosBaseIdx = pu.mmvdMergeIdx / MMVD_MAX_REFINE_NUM;
+#if JVET_M0170_MRG_SHARELIST
+ pu.shareParentPos = cu.shareParentPos;
+ pu.shareParentSize = cu.shareParentSize;
+#endif
PU::getInterMergeCandidates(pu, mrgCtx, 1, fPosBaseIdx + 1);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
PU::restrictBiPredMergeCands(pu, mrgCtx);
+#endif
PU::getInterMMVDMergeCandidates(pu, mrgCtx,
pu.mmvdMergeIdx
);
@@ -486,7 +520,7 @@ void DecCu::xDeriveCUMV( CodingUnit &cu )
if( pu.cu->affine )
{
AffineMergeCtx affineMergeCtx;
- if ( pu.cs->sps->getSpsNext().getUseSubPuMvp() )
+ if ( pu.cs->sps->getSBTMVPEnabledFlag() )
{
Size bufSize = g_miScaling.scale( pu.lumaSize() );
mrgCtx.subPuMvpMiBuf = MotionBuf( m_SubPuMiBuf, bufSize );
@@ -520,10 +554,14 @@ void DecCu::xDeriveCUMV( CodingUnit &cu )
}
else
{
-
+#if JVET_M0170_MRG_SHARELIST
+ pu.shareParentPos = cu.shareParentPos;
+ pu.shareParentSize = cu.shareParentSize;
+#endif
PU::getInterMergeCandidates(pu, mrgCtx, 0, pu.mergeIdx);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
PU::restrictBiPredMergeCands(pu, mrgCtx);
-
+#endif
mrgCtx.setMergeInfo( pu, pu.mergeIdx );
PU::spanMotionInfo( pu, mrgCtx );
@@ -591,7 +629,7 @@ void DecCu::xDeriveCUMV( CodingUnit &cu )
Mv mvd = pu.mvd[eRefList];
if (eRefList == REF_PIC_LIST_0 && pu.cs->slice->getRefPic(eRefList, pu.refIdx[eRefList])->getPOC() == pu.cs->slice->getPOC())
{
- pu.cu->cpr = true;
+ pu.cu->ibc = true;
#if REUSE_CU_RESULTS
if (!cu.cs->pcv->isEncoder)
#endif
diff --git a/source/Lib/DecoderLib/DecCu.h b/source/Lib/DecoderLib/DecCu.h
index d5598b0dfa25662cf25b829d39a4645fb18527c1..67e5dc4fbc2a15faa777b7c1c34b35042e1d6fa0 100644
--- a/source/Lib/DecoderLib/DecCu.h
+++ b/source/Lib/DecoderLib/DecCu.h
@@ -69,6 +69,9 @@ public:
/// destroy internal buffers
void decompressCtu ( CodingStructure& cs, const UnitArea& ctuArea );
+#if JVET_M0170_MRG_SHARELIST
+ void setShareStateDec (int shareStateDecIn) { m_shareStateDec = shareStateDecIn; }
+#endif
/// reconstruct Ctu information
protected:
void xIntraRecQT ( CodingUnit& cu, const ChannelType chType );
@@ -90,6 +93,9 @@ private:
IntraPrediction* m_pcIntraPred;
InterPrediction* m_pcInterPred;
+#if JVET_M0170_MRG_SHARELIST
+ int m_shareStateDec;
+#endif
MotionInfo m_SubPuMiBuf[(MAX_CU_SIZE * MAX_CU_SIZE) >> (MIN_CU_LOG2 << 1)];
diff --git a/source/Lib/DecoderLib/DecLib.cpp b/source/Lib/DecoderLib/DecLib.cpp
index 0a1d88c69aa5b8584f0735d0204077a97892d3aa..09b493dd5c22f0168a000b5768cd947043a1d35d 100644
--- a/source/Lib/DecoderLib/DecLib.cpp
+++ b/source/Lib/DecoderLib/DecLib.cpp
@@ -168,12 +168,12 @@ bool tryDecodePicture( Picture* pcEncPic, const int expectedPoc, const std::stri
pcEncPic->cs->slice = pcEncPic->slices.back();
- if ( pic->cs->sps->getUseSAO() )
+ if ( pic->cs->sps->getSAOEnabledFlag() )
{
pcEncPic->copySAO( *pic, 0 );
}
- if( pic->cs->sps->getUseALF() )
+ if( pic->cs->sps->getALFEnabledFlag() )
{
for( int compIdx = 0; compIdx < MAX_NUM_COMPONENT; compIdx++ )
{
@@ -187,7 +187,7 @@ bool tryDecodePicture( Picture* pcEncPic, const int expectedPoc, const std::stri
}
pcDecLib->executeLoopFilters();
- if ( pic->cs->sps->getUseSAO() )
+ if ( pic->cs->sps->getSAOEnabledFlag() )
{
pcEncPic->copySAO( *pic, 1 );
}
@@ -511,12 +511,12 @@ void DecLib::executeLoopFilters()
// deblocking filter
m_cLoopFilter.loopFilterPic( cs );
- if( cs.sps->getUseSAO() )
+ if( cs.sps->getSAOEnabledFlag() )
{
m_cSAO.SAOProcess( cs, cs.picture->getSAO() );
}
- if( cs.sps->getUseALF() )
+ if( cs.sps->getALFEnabledFlag() )
{
m_cALF.ALFProcess( cs, cs.slice->getAlfSliceParam() );
}
@@ -772,7 +772,7 @@ void DecLib::xActivateParameterSets()
m_cSliceDecoder.create();
- if( sps->getUseALF() )
+ if( sps->getALFEnabledFlag() )
{
m_cALF.create( sps->getPicWidthInLumaSamples(), sps->getPicHeightInLumaSamples(), sps->getChromaFormatIdc(), sps->getMaxCUWidth(), sps->getMaxCUHeight(), sps->getMaxCodingDepth(), sps->getBitDepths().recon );
}
@@ -1116,6 +1116,84 @@ bool DecLib::xDecodeSlice(InputNALUnit &nalu, int &iSkipFrame, int iPOCLastDispl
pcSlice->setCheckLDC(bLowDelay);
}
+#if JVET_M0444_SMVD
+ if ( pcSlice->getCheckLDC() == false && pcSlice->getMvdL1ZeroFlag() == false )
+ {
+ int currPOC = pcSlice->getPOC();
+
+ int forwardPOC = currPOC;
+ int backwardPOC = currPOC;
+ int ref = 0;
+ int refIdx0 = -1;
+ int refIdx1 = -1;
+
+ // search nearest forward POC in List 0
+ for ( ref = 0; ref < pcSlice->getNumRefIdx( REF_PIC_LIST_0 ); ref++ )
+ {
+ int poc = pcSlice->getRefPic( REF_PIC_LIST_0, ref )->getPOC();
+ if ( poc < currPOC && (poc > forwardPOC || refIdx0 == -1) )
+ {
+ forwardPOC = poc;
+ refIdx0 = ref;
+ }
+ }
+
+ // search nearest backward POC in List 1
+ for ( ref = 0; ref < pcSlice->getNumRefIdx( REF_PIC_LIST_1 ); ref++ )
+ {
+ int poc = pcSlice->getRefPic( REF_PIC_LIST_1, ref )->getPOC();
+ if ( poc > currPOC && (poc < backwardPOC || refIdx1 == -1) )
+ {
+ backwardPOC = poc;
+ refIdx1 = ref;
+ }
+ }
+
+ if ( !(forwardPOC < currPOC && backwardPOC > currPOC) )
+ {
+ forwardPOC = currPOC;
+ backwardPOC = currPOC;
+ refIdx0 = -1;
+ refIdx1 = -1;
+
+ // search nearest backward POC in List 0
+ for ( ref = 0; ref < pcSlice->getNumRefIdx( REF_PIC_LIST_0 ); ref++ )
+ {
+ int poc = pcSlice->getRefPic( REF_PIC_LIST_0, ref )->getPOC();
+ if ( poc > currPOC && (poc < backwardPOC || refIdx0 == -1) )
+ {
+ backwardPOC = poc;
+ refIdx0 = ref;
+ }
+ }
+
+ // search nearest forward POC in List 1
+ for ( ref = 0; ref < pcSlice->getNumRefIdx( REF_PIC_LIST_1 ); ref++ )
+ {
+ int poc = pcSlice->getRefPic( REF_PIC_LIST_1, ref )->getPOC();
+ if ( poc < currPOC && (poc > forwardPOC || refIdx1 == -1) )
+ {
+ forwardPOC = poc;
+ refIdx1 = ref;
+ }
+ }
+ }
+
+ if ( forwardPOC < currPOC && backwardPOC > currPOC )
+ {
+ pcSlice->setBiDirPred( true, refIdx0, refIdx1 );
+ }
+ else
+ {
+ pcSlice->setBiDirPred( false, -1, -1 );
+ }
+ }
+ else
+ {
+ pcSlice->setBiDirPred( false, -1, -1 );
+ }
+#endif
+
//---------------
pcSlice->setRefPOCList();
@@ -1150,7 +1228,7 @@ bool DecLib::xDecodeSlice(InputNALUnit &nalu, int &iSkipFrame, int iPOCLastDispl
}
#endif
- if (pcSlice->getSPS()->getSpsNext().getCPRMode() && pcSlice->getEnableTMVPFlag())
+ if (pcSlice->getSPS()->getSpsNext().getIBCMode() && pcSlice->getEnableTMVPFlag())
{
CHECK(pcSlice->getRefPic(RefPicList(pcSlice->isInterB() ? 1 - pcSlice->getColFromL0Flag() : 0), pcSlice->getColRefIdx())->getPOC() == pcSlice->getPOC(), "curr ref picture cannot be collocated picture");
}
@@ -1160,7 +1238,7 @@ bool DecLib::xDecodeSlice(InputNALUnit &nalu, int &iSkipFrame, int iPOCLastDispl
m_cSliceDecoder.decompressSlice( pcSlice, &(nalu.getBitstream()) );
m_bFirstSliceInPicture = false;
- if (pcSlice->getSPS()->getSpsNext().getCPRMode())
+ if (pcSlice->getSPS()->getSpsNext().getIBCMode())
{
pcSlice->getPic()->longTerm = false;
}
diff --git a/source/Lib/DecoderLib/VLCReader.cpp b/source/Lib/DecoderLib/VLCReader.cpp
index e6813e513f4fe3a4eb1dea8065235b1051c7b69c..8e1725797c1a84d4eb0812d62a0b0356e20d60f7 100644
--- a/source/Lib/DecoderLib/VLCReader.cpp
+++ b/source/Lib/DecoderLib/VLCReader.cpp
@@ -789,20 +789,29 @@ void HLSyntaxReader::parseSPSNext( SPSNext& spsNext, const bool usePCM )
// tool enabling flags
READ_FLAG( symbol, "large_ctu_flag" ); spsNext.setUseLargeCTU ( symbol != 0 );
- READ_FLAG( symbol, "subpu_tmvp_flag" ); spsNext.setSubPuMvpMode (symbol);
READ_FLAG( symbol, "imv_enable_flag" ); spsNext.setUseIMV ( symbol != 0 );
- READ_FLAG( symbol, "bio_enable_flag" ); spsNext.setUseBIO ( symbol != 0 );
READ_FLAG( symbol, "disable_motion_compression_flag" ); spsNext.setDisableMotCompress ( symbol != 0 );
READ_FLAG( symbol, "lm_chroma_enabled_flag" ); spsNext.setUseLMChroma ( symbol != 0 );
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ if ( spsNext.getUseLMChroma() && spsNext.getSPS().getChromaFormatIdc() == CHROMA_420 )
+ {
+ READ_FLAG( symbol, "sps_cclm_collocated_chroma_flag" ); spsNext.setCclmCollocatedChromaFlag( symbol != 0 );
+ }
+#endif
+#if JVET_M0464_UNI_MTS
+ READ_FLAG( symbol, "mts_intra_enabled_flag" ); spsNext.setUseIntraMTS ( symbol != 0 );
+ READ_FLAG( symbol, "mts_inter_enabled_flag" ); spsNext.setUseInterMTS ( symbol != 0 );
+#else
READ_FLAG( symbol, "emt_intra_enabled_flag" ); spsNext.setUseIntraEMT ( symbol != 0 );
READ_FLAG( symbol, "emt_inter_enabled_flag" ); spsNext.setUseInterEMT ( symbol != 0 );
+#endif
READ_FLAG( symbol, "affine_flag" ); spsNext.setUseAffine ( symbol != 0 );
if ( spsNext.getUseAffine() )
{
READ_FLAG( symbol, "affine_type_flag" ); spsNext.setUseAffineType ( symbol != 0 );
}
READ_FLAG( symbol, "gbi_flag" ); spsNext.setUseGBi ( symbol != 0 );
- READ_FLAG( symbol, "cpr_flag"); spsNext.setCPRMode ( symbol != 0 );
+ READ_FLAG( symbol, "ibc_flag"); spsNext.setIBCMode ( symbol != 0 );
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)" );
@@ -816,14 +825,6 @@ void HLSyntaxReader::parseSPSNext( SPSNext& spsNext, const bool usePCM )
READ_FLAG( symbol, "reserved_flag" ); CHECK( symbol, "reserved flag not 0!" );
#endif
- // additional parameters
- if( spsNext.getUseSubPuMvp() )
- {
- int subPuMode = 1;
- spsNext.setSubPuMvpMode( subPuMode );
- }
-
-
if( spsNext.getUseIMV() )
{
READ_UVLC( symbol, "imv_mode_minus1" ); spsNext.setImvMode( ImvMode( symbol + 1 ) );
@@ -1014,8 +1015,39 @@ void HLSyntaxReader::parseSPS(SPS* pcSPS)
READ_UVLC( uiCode, "log2_diff_max_min_luma_transform_block_size" ); pcSPS->setQuadtreeTULog2MaxSize( uiCode + pcSPS->getQuadtreeTULog2MinSize() );
pcSPS->setMaxTrSize( 1<<(uiCode + pcSPS->getQuadtreeTULog2MinSize()) );
- READ_FLAG( uiCode, "sps_alf_enable_flag" ); pcSPS->setUseALF( uiCode );
+ READ_FLAG( uiCode, "sps_sao_enabled_flag" ); pcSPS->setSAOEnabledFlag ( uiCode ? true : false );
+ READ_FLAG( uiCode, "sps_alf_enabled_flag" ); pcSPS->setALFEnabledFlag ( uiCode ? true : false );
+
+ READ_FLAG( uiCode, "pcm_enabled_flag" ); pcSPS->setPCMEnabledFlag( uiCode ? true : false );
+ if( pcSPS->getPCMEnabledFlag() )
+ {
+ READ_CODE( 4, uiCode, "pcm_sample_bit_depth_luma_minus1" ); pcSPS->setPCMBitDepth ( CHANNEL_TYPE_LUMA, 1 + uiCode );
+ READ_CODE( 4, uiCode, "pcm_sample_bit_depth_chroma_minus1" ); pcSPS->setPCMBitDepth ( CHANNEL_TYPE_CHROMA, 1 + uiCode );
+ READ_UVLC( uiCode, "log2_min_pcm_luma_coding_block_size_minus3" ); pcSPS->setPCMLog2MinSize ( uiCode+3 );
+ READ_UVLC( uiCode, "log2_diff_max_min_pcm_luma_coding_block_size" ); pcSPS->setPCMLog2MaxSize ( uiCode+pcSPS->getPCMLog2MinSize() );
+ READ_FLAG( uiCode, "pcm_loop_filter_disable_flag" ); pcSPS->setPCMFilterDisableFlag ( uiCode ? true : false );
+ }
+
+ READ_FLAG(uiCode, "sps_ref_wraparound_enabled_flag"); pcSPS->setWrapAroundEnabledFlag( uiCode ? true : false );
+ if (pcSPS->getWrapAroundEnabledFlag())
+ {
+ READ_UVLC(uiCode, "sps_ref_wraparound_offset"); pcSPS->setWrapAroundOffset( uiCode );
+ }
+
+ READ_FLAG( uiCode, "sps_temporal_mvp_enabled_flag" ); pcSPS->setSPSTemporalMVPEnabledFlag(uiCode);
+
+ if ( pcSPS->getSPSTemporalMVPEnabledFlag() )
+ {
+ READ_FLAG( uiCode, "sps_sbtmvp_enabled_flag" ); pcSPS->setSBTMVPEnabledFlag ( uiCode != 0 );
+ }
+ else
+ {
+ pcSPS->setSBTMVPEnabledFlag(false);
+ }
+
+ READ_FLAG( uiCode, "sps_bdof_enable_flag" ); pcSPS->setBDOFEnabledFlag ( uiCode != 0 );
+
#if HEVC_USE_SCALING_LISTS
READ_FLAG( uiCode, "scaling_list_enabled_flag" ); pcSPS->setScalingListFlag ( uiCode );
if(pcSPS->getScalingListFlag())
@@ -1027,24 +1059,6 @@ void HLSyntaxReader::parseSPS(SPS* pcSPS)
}
}
#endif
- READ_FLAG( uiCode, "amp_enabled_flag" ); pcSPS->setUseAMP( uiCode );
- READ_FLAG( uiCode, "sample_adaptive_offset_enabled_flag" ); pcSPS->setUseSAO ( uiCode ? true : false );
-
- READ_FLAG( uiCode, "pcm_enabled_flag" ); pcSPS->setUsePCM( uiCode ? true : false );
- if( pcSPS->getUsePCM() )
- {
- READ_CODE( 4, uiCode, "pcm_sample_bit_depth_luma_minus1" ); pcSPS->setPCMBitDepth ( CHANNEL_TYPE_LUMA, 1 + uiCode );
- READ_CODE( 4, uiCode, "pcm_sample_bit_depth_chroma_minus1" ); pcSPS->setPCMBitDepth ( CHANNEL_TYPE_CHROMA, 1 + uiCode );
- READ_UVLC( uiCode, "log2_min_pcm_luma_coding_block_size_minus3" ); pcSPS->setPCMLog2MinSize (uiCode+3);
- READ_UVLC( uiCode, "log2_diff_max_min_pcm_luma_coding_block_size" ); pcSPS->setPCMLog2MaxSize ( uiCode+pcSPS->getPCMLog2MinSize() );
- READ_FLAG( uiCode, "pcm_loop_filter_disable_flag" ); pcSPS->setPCMFilterDisableFlag ( uiCode ? true : false );
- }
-
- READ_FLAG(uiCode, "ref_wraparound_enabled_flag"); pcSPS->setUseWrapAround( uiCode ? true : false );
- if (pcSPS->getUseWrapAround())
- {
- READ_UVLC(uiCode, "ref_wraparound_offset"); pcSPS->setWrapAroundOffset( uiCode );
- }
READ_UVLC( uiCode, "num_short_term_ref_pic_sets" );
CHECK(uiCode > 64, "Invalid code");
@@ -1071,8 +1085,6 @@ void HLSyntaxReader::parseSPS(SPS* pcSPS)
pcSPS->setUsedByCurrPicLtSPSFlag(k, uiCode?1:0);
}
}
- READ_FLAG( uiCode, "sps_temporal_mvp_enabled_flag" ); pcSPS->setSPSTemporalMVPEnabledFlag(uiCode);
-
#if HEVC_USE_INTRA_SMOOTHING_T32 || HEVC_USE_INTRA_SMOOTHING_T64
READ_FLAG( uiCode, "strong_intra_smoothing_enable_flag" ); pcSPS->setUseStrongIntraSmoothing(uiCode);
@@ -1137,7 +1149,7 @@ void HLSyntaxReader::parseSPS(SPS* pcSPS)
case SPS_EXT__NEXT:
{
CHECK( !pcSPS->getSpsNext().nextToolsEnabled(), "Got SPS Next extension in non NEXT profile" );
- parseSPSNext( pcSPS->getSpsNext(), pcSPS->getUsePCM() );
+ parseSPSNext( pcSPS->getSpsNext(), pcSPS->getPCMEnabledFlag() );
break;
}
default:
@@ -1512,7 +1524,7 @@ void HLSyntaxReader::parseSliceHeader (Slice* pcSlice, ParameterSetManager *para
pcSlice->setEnableTMVPFlag(false);
}
}
- if(sps->getUseSAO())
+ if(sps->getSAOEnabledFlag())
{
READ_FLAG(uiCode, "slice_sao_luma_flag"); pcSlice->setSaoEnabledFlag(CHANNEL_TYPE_LUMA, (bool)uiCode);
@@ -1522,7 +1534,7 @@ void HLSyntaxReader::parseSliceHeader (Slice* pcSlice, ParameterSetManager *para
}
}
- if( sps->getUseALF() )
+ if( sps->getALFEnabledFlag() )
{
alf( pcSlice->getAlfSliceParam() );
}
@@ -1741,11 +1753,11 @@ void HLSyntaxReader::parseSliceHeader (Slice* pcSlice, ParameterSetManager *para
READ_UVLC(uiCode, "six_minus_max_num_merge_cand");
pcSlice->setMaxNumMergeCand(MRG_MAX_NUM_CANDS - uiCode);
- if ( sps->getSpsNext().getUseSubPuMvp() && !sps->getSpsNext().getUseAffine() ) // ATMVP only
+ if ( sps->getSBTMVPEnabledFlag() && !sps->getSpsNext().getUseAffine() ) // ATMVP only
{
pcSlice->setMaxNumAffineMergeCand( 1 );
}
- else if ( !sps->getSpsNext().getUseSubPuMvp() && !sps->getSpsNext().getUseAffine() ) // both off
+ else if ( !sps->getSBTMVPEnabledFlag() && !sps->getSpsNext().getUseAffine() ) // both off
{
pcSlice->setMaxNumAffineMergeCand( 0 );
}
@@ -1833,7 +1845,7 @@ void HLSyntaxReader::parseSliceHeader (Slice* pcSlice, ParameterSetManager *para
pcSlice->setDeblockingFilterTcOffsetDiv2 ( 0 );
}
- bool isSAOEnabled = sps->getUseSAO() && (pcSlice->getSaoEnabledFlag(CHANNEL_TYPE_LUMA) || (bChroma && pcSlice->getSaoEnabledFlag(CHANNEL_TYPE_CHROMA)));
+ bool isSAOEnabled = sps->getSAOEnabledFlag() && (pcSlice->getSaoEnabledFlag(CHANNEL_TYPE_LUMA) || (bChroma && pcSlice->getSaoEnabledFlag(CHANNEL_TYPE_CHROMA)));
bool isDBFEnabled = (!pcSlice->getDeblockingFilterDisable());
if(pps->getLoopFilterAcrossSlicesEnabledFlag() && ( isSAOEnabled || isDBFEnabled ))
diff --git a/source/Lib/EncoderLib/Analyze.h b/source/Lib/EncoderLib/Analyze.h
index 69aaad7cef931082bd60885cb237805c93118d90..086c834bf6c9bcd3a4306d526c5e40c2a5c904fe 100644
--- a/source/Lib/EncoderLib/Analyze.h
+++ b/source/Lib/EncoderLib/Analyze.h
@@ -60,10 +60,6 @@
// Class definition
// ====================================================================================================================
-#if ENABLE_QPA
- #define FRAME_WEIGHTING 0 // WPSNR temporal weighting according to hierarchical coding structure; only for GOP size 16
-#endif
-
/// encoder analyzer class
class Analyze
{
@@ -73,10 +69,6 @@ private:
uint32_t m_uiNumPic;
double m_dFrmRate; //--CFG_KDY
double m_MSEyuvframe[MAX_NUM_COMPONENT]; // sum of MSEs
-#if ENABLE_QPA && FRAME_WEIGHTING
- double m_sumWSSD[MAX_NUM_COMPONENT]; // weighted SSDs
- double m_sumW;
-#endif
#if EXTENSION_360_VIDEO
TExt360EncAnalyze m_ext360;
#endif
@@ -101,13 +93,7 @@ public:
m_uiNumPic++;
}
#if ENABLE_QPA
- #if FRAME_WEIGHTING
- void addWeightedSSD(const double dWeightedSSD, const ComponentID compID) { m_sumWSSD[compID] += dWeightedSSD; }
- void addWeight (const double dWeight) { m_sumW += dWeight; }
- double getWPSNR (const ComponentID compID) const { return (m_sumWSSD[compID] > 0.0 ? 10.0 * log10(m_sumW / m_sumWSSD[compID]) : 999.99); }
- #else
double getWPSNR (const ComponentID compID) const { return m_dPSNRSum[compID] / (double)m_uiNumPic; }
- #endif
#endif
double getPsnr(ComponentID compID) const { return m_dPSNRSum[compID]; }
double getBits() const { return m_dAddBits; }
@@ -125,13 +111,7 @@ public:
{
m_dPSNRSum[i] = 0;
m_MSEyuvframe[i] = 0;
-#if ENABLE_QPA && FRAME_WEIGHTING
- m_sumWSSD[i] = 0;
-#endif
}
-#if ENABLE_QPA && FRAME_WEIGHTING
- m_sumW = 0;
-#endif
m_uiNumPic = 0;
#if EXTENSION_360_VIDEO
m_ext360.clear();
diff --git a/source/Lib/EncoderLib/CABACWriter.cpp b/source/Lib/EncoderLib/CABACWriter.cpp
index d0d937ad47e37c4e6611d65d094bfbc8605cf2e2..a9795b6dbf2e120a85cfced81944161f270c5930 100644
--- a/source/Lib/EncoderLib/CABACWriter.cpp
+++ b/source/Lib/EncoderLib/CABACWriter.cpp
@@ -215,7 +215,7 @@ void CABACWriter::coding_tree_unit( CodingStructure& cs, const UnitArea& area, i
void CABACWriter::sao( const Slice& slice, unsigned ctuRsAddr )
{
const SPS& sps = *slice.getSPS();
- if( !sps.getUseSAO() )
+ if( !sps.getSAOEnabledFlag() )
{
return;
}
@@ -396,6 +396,16 @@ void CABACWriter::coding_tree(const CodingStructure& cs, Partitioner& partitione
}
}
+#if JVET_M0421_SPLIT_SIG
+ const PartSplit splitMode = CU::getSplitAtDepth( cu, partitioner.currDepth );
+
+ split_cu_mode( splitMode, cs, partitioner );
+
+ CHECK( !partitioner.canSplit( splitMode, cs ), "The chosen split mode is invalid!" );
+
+ if( splitMode != CU_DONT_SPLIT )
+ {
+#else
const PartSplit implicitSplit = partitioner.getImplicitSplit( cs );
// QT
@@ -415,6 +425,7 @@ void CABACWriter::coding_tree(const CodingStructure& cs, Partitioner& partitione
// quad-tree split
if( qtSplit )
{
+#endif
if (CS::isDualITree(cs) && pPartitionerChroma != nullptr && (partitioner.currArea().lwidth() >= 64 || partitioner.currArea().lheight() >= 64))
{
partitioner.splitCurrArea(CU_QUAD_SPLIT, cs);
@@ -459,7 +470,11 @@ void CABACWriter::coding_tree(const CodingStructure& cs, Partitioner& partitione
}
else
{
+#if JVET_M0421_SPLIT_SIG
+ partitioner.splitCurrArea( splitMode, cs );
+#else
partitioner.splitCurrArea( CU_QUAD_SPLIT, cs );
+#endif
do
{
@@ -472,9 +487,12 @@ void CABACWriter::coding_tree(const CodingStructure& cs, Partitioner& partitione
partitioner.exitCurrSplit();
}
return;
+#if !JVET_M0421_SPLIT_SIG
}
+#endif
}
+#if !JVET_M0421_SPLIT_SIG
{
bool mtSplit = partitioner.canSplit( CU_MT_SPLIT, cs );
@@ -501,6 +519,7 @@ void CABACWriter::coding_tree(const CodingStructure& cs, Partitioner& partitione
}
}
+#endif
// Predict QP on start of quantization group
if( pps.getUseDQP() && !cuCtx.isDQPCoded && CU::isQGStart( cu, partitioner ) )
{
@@ -515,6 +534,68 @@ void CABACWriter::coding_tree(const CodingStructure& cs, Partitioner& partitione
DTRACE_BLOCK_REC_COND( ( !isEncoding() ), cs.picture->getRecoBuf( cu ), cu, cu.predMode );
}
+#if JVET_M0421_SPLIT_SIG
+void CABACWriter::split_cu_mode( const PartSplit split, const CodingStructure& cs, Partitioner& partitioner )
+{
+ bool canNo, canQt, canBh, canBv, canTh, canTv;
+ partitioner.canSplit( cs, canNo, canQt, canBh, canBv, canTh, canTv );
+
+ bool canSpl[6] = { canNo, canQt, canBh, canBv, canTh, canTv };
+
+ unsigned ctxSplit = 0, ctxQtSplit = 0, ctxBttHV = 0, ctxBttH12 = 0, ctxBttV12;
+ DeriveCtx::CtxSplit( cs, partitioner, ctxSplit, ctxQtSplit, ctxBttHV, ctxBttH12, ctxBttV12, canSpl );
+
+ const bool canSplit = canBh || canBv || canTh || canTv || canQt;
+ const bool isNo = split == CU_DONT_SPLIT;
+
+ if( canNo && canSplit )
+ {
+ m_BinEncoder.encodeBin( !isNo, Ctx::SplitFlag( ctxSplit ) );
+ }
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode() ctx=%d split=%d\n", ctxSplit, !isNo );
+
+ if( isNo )
+ {
+ return;
+ }
+
+ const bool canBtt = canBh || canBv || canTh || canTv;
+ const bool isQt = split == CU_QUAD_SPLIT;
+
+ if( canQt && canBtt )
+ {
+ m_BinEncoder.encodeBin( isQt, Ctx::SplitQtFlag( ctxQtSplit ) );
+ }
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode() ctx=%d qt=%d\n", ctxQtSplit, isQt );
+
+ if( isQt )
+ {
+ return;
+ }
+
+ const bool canHor = canBh || canTh;
+ const bool canVer = canBv || canTv;
+ const bool isVer = split == CU_VERT_SPLIT || split == CU_TRIV_SPLIT;
+
+ if( canVer && canHor )
+ {
+ m_BinEncoder.encodeBin( isVer, Ctx::SplitHvFlag( ctxBttHV ) );
+ }
+
+ const bool can14 = isVer ? canTv : canTh;
+ const bool can12 = isVer ? canBv : canBh;
+ const bool is12 = isVer ? ( split == CU_VERT_SPLIT ) : ( split == CU_HORZ_SPLIT );
+
+ if( can12 && can14 )
+ {
+ m_BinEncoder.encodeBin( is12, Ctx::Split12Flag( isVer ? ctxBttV12 : ctxBttH12 ) );
+ }
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode() ctxHv=%d ctx12=%d mode=%d\n", ctxBttHV, isVer ? ctxBttV12 : ctxBttH12, split );
+}
+#else
void CABACWriter::split_cu_flag( bool split, const CodingStructure& cs, Partitioner& partitioner )
{
unsigned maxQTDepth = ( g_aucLog2[cs.sps->getCTUSize()] - g_aucLog2[cs.sps->getMinQTSize(cs.slice->getSliceType(), partitioner.chType)] );
@@ -535,6 +616,49 @@ void CABACWriter::split_cu_mode_mt(const PartSplit split, const CodingStructure&
unsigned width = partitioner.currArea().lumaSize().width;
unsigned height = partitioner.currArea().lumaSize().height;
+#if REMOVE_BIN_DECISION_TREE
+ unsigned btSCtxId = width == height ? 0 : ( width > height ? 1 : 2 );
+
+ const bool canNo = partitioner.canSplit( CU_DONT_SPLIT, cs );
+ const bool canBh = partitioner.canSplit( CU_HORZ_SPLIT, cs );
+ const bool canBv = partitioner.canSplit( CU_VERT_SPLIT, cs );
+ const bool canTh = partitioner.canSplit( CU_TRIH_SPLIT, cs );
+ const bool canTv = partitioner.canSplit( CU_TRIV_SPLIT, cs );
+
+ const bool canSplit = canBh || canBv || canTh || canTv;
+ const bool isNo = split == CU_DONT_SPLIT;
+
+ if( canNo && canSplit )
+ {
+ m_BinEncoder.encodeBin( !isNo, Ctx::BTSplitFlag( ctxIdBT ) );
+ }
+
+ if( isNo )
+ {
+ DTRACE( g_trace_ctx, D_SYNTAX, "split_cu_mode_mt() ctx=%d split=%d\n", ctxIdBT, split );
+
+ return;
+ }
+
+ const bool canHor = canBh || canTh;
+ const bool canVer = canBv || canTv;
+ const bool isVer = split == CU_VERT_SPLIT || split == CU_TRIV_SPLIT;
+
+ if( canVer && canHor )
+ {
+ m_BinEncoder.encodeBin( isVer, Ctx::BTSplitFlag( 12 + btSCtxId ) );
+ }
+
+ const bool can14 = isVer ? canTv : canTh;
+ const bool can12 = isVer ? canBv : canBh;
+ const bool is12 = isVer ? ( split == CU_VERT_SPLIT ) : ( split == CU_HORZ_SPLIT );
+
+
+ if( can12 && can14 )
+ {
+ m_BinEncoder.encodeBin( is12, Ctx::BTSplitFlag( 15 ) );
+ }
+#else
DecisionTree dt( g_mtSplitDTT );
dt.setAvail( DTT_SPLIT_BT_HORZ, partitioner.canSplit( CU_HORZ_SPLIT, cs ) );
@@ -553,10 +677,11 @@ void CABACWriter::split_cu_mode_mt(const PartSplit split, const CodingStructure&
encode_sparse_dt( dt, split == CU_DONT_SPLIT ? ( unsigned ) DTT_SPLIT_NO_SPLIT : ( unsigned ) split );
+#endif
DTRACE(g_trace_ctx, D_SYNTAX, "split_cu_mode_mt() ctx=%d split=%d\n", ctxIdBT, split);
}
-
+#endif
//================================================================================
// clause 7.3.8.5
@@ -658,7 +783,11 @@ void CABACWriter::pred_mode( const CodingUnit& cu )
{
return;
}
+#if JVET_M0502_PRED_MODE_CTX
+ m_BinEncoder.encodeBin( ( CU::isIntra( cu ) ), Ctx::PredMode( DeriveCtx::CtxPredModeFlag( cu ) ) );
+#else
m_BinEncoder.encodeBin( ( CU::isIntra( cu ) ), Ctx::PredMode() );
+#endif
}
void CABACWriter::pcm_data( const CodingUnit& cu, Partitioner& partitioner )
@@ -674,7 +803,7 @@ void CABACWriter::pcm_data( const CodingUnit& cu, Partitioner& partitioner )
void CABACWriter::pcm_flag( const CodingUnit& cu, Partitioner& partitioner )
{
const SPS& sps = *cu.cs->sps;
- if( !sps.getUsePCM() || partitioner.currArea().lwidth() > (1 << sps.getPCMLog2MaxSize()) || partitioner.currArea().lwidth() < (1 << sps.getPCMLog2MinSize())
+ if( !sps.getPCMEnabledFlag() || partitioner.currArea().lwidth() > (1 << sps.getPCMLog2MaxSize()) || partitioner.currArea().lwidth() < (1 << sps.getPCMLog2MinSize())
|| partitioner.currArea().lheight() > (1 << sps.getPCMLog2MaxSize()) || partitioner.currArea().lheight() < (1 << sps.getPCMLog2MinSize()) )
{
return;
@@ -1199,6 +1328,9 @@ void CABACWriter::prediction_unit( const PredictionUnit& pu )
{
inter_pred_idc( pu );
affine_flag ( *pu.cu );
+#if JVET_M0444_SMVD
+ smvd_mode( pu );
+#endif
if( pu.interDir != 2 /* PRED_L1 */ )
{
ref_idx ( pu, REF_PIC_LIST_0 );
@@ -1219,6 +1351,10 @@ void CABACWriter::prediction_unit( const PredictionUnit& pu )
}
if( pu.interDir != 1 /* PRED_L0 */ )
{
+#if JVET_M0444_SMVD
+ if ( pu.cu->smvdMode != 1 )
+ {
+#endif
ref_idx ( pu, REF_PIC_LIST_1 );
if( !pu.cs->slice->getMvdL1ZeroFlag() || pu.interDir != 3 /* PRED_BI */ )
{
@@ -1236,11 +1372,33 @@ void CABACWriter::prediction_unit( const PredictionUnit& pu )
mvd_coding( pu.mvd[REF_PIC_LIST_1], pu.cu->imv );
}
}
+#if JVET_M0444_SMVD
+ }
+#endif
mvp_flag ( pu, REF_PIC_LIST_1 );
}
}
}
+#if JVET_M0444_SMVD
+void CABACWriter::smvd_mode( const PredictionUnit& pu )
+{
+ if ( pu.interDir != 3 || pu.cu->affine )
+ {
+ return;
+ }
+
+ if ( pu.cs->slice->getBiDirPred() == false )
+ {
+ return;
+ }
+
+ m_BinEncoder.encodeBin( pu.cu->smvdMode ? 1 : 0, Ctx::SmvdFlag() );
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "symmvd_flag() symmvd=%d pos=(%d,%d) size=%dx%d\n", pu.cu->smvdMode ? 1 : 0, pu.lumaPos().x, pu.lumaPos().y, pu.lumaSize().width, pu.lumaSize().height );
+}
+#endif
+
void CABACWriter::subblock_merge_flag( const CodingUnit& cu )
{
if ( cu.firstPU->mergeFlag && (cu.firstPU->mmvdMergeFlag || cu.mmvdSkip) )
@@ -1248,7 +1406,7 @@ void CABACWriter::subblock_merge_flag( const CodingUnit& cu )
return;
}
- if ( !cu.cs->slice->isIntra() && (cu.cs->sps->getSpsNext().getUseAffine() || cu.cs->sps->getSpsNext().getUseATMVP()) && cu.lumaSize().width >= 8 && cu.lumaSize().height >= 8 )
+ if ( !cu.cs->slice->isIntra() && (cu.cs->sps->getSpsNext().getUseAffine() || cu.cs->sps->getSBTMVPEnabledFlag()) && cu.lumaSize().width >= 8 && cu.lumaSize().height >= 8 )
{
unsigned ctxId = DeriveCtx::CtxAffineFlag( cu );
m_BinEncoder.encodeBin( cu.affine, Ctx::AffineFlag( ctxId ) );
@@ -1301,7 +1459,7 @@ void CABACWriter::imv_mode( const CodingUnit& cu )
}
unsigned ctxId = DeriveCtx::CtxIMVFlag( cu );
- if (!(cu.firstPU->interDir == 1 && cu.cs->slice->getRefPic(REF_PIC_LIST_0, cu.firstPU->refIdx[REF_PIC_LIST_0])->getPOC() == cu.cs->slice->getPOC())) // the first bin of IMV flag does need to be signaled in CPR block
+ if (!(cu.firstPU->interDir == 1 && cu.cs->slice->getRefPic(REF_PIC_LIST_0, cu.firstPU->refIdx[REF_PIC_LIST_0])->getPOC() == cu.cs->slice->getPOC())) // the first bin of IMV flag does need to be signaled in IBC block
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 );
@@ -1330,7 +1488,7 @@ void CABACWriter::merge_idx( const PredictionUnit& pu )
}
else
{
- bool useExtCtx = pu.cs->sps->getSpsNext().getUseSubPuMvp();
+ bool useExtCtx = pu.cs->sps->getSBTMVPEnabledFlag();
m_BinEncoder.encodeBin( 1, Ctx::AffMergeIdx() );
for ( unsigned idx = 1; idx < numCandminus1; idx++ )
{
@@ -1478,6 +1636,14 @@ void CABACWriter::inter_pred_idc( const PredictionUnit& pu )
void CABACWriter::ref_idx( const PredictionUnit& pu, RefPicList eRefList )
{
+#if JVET_M0444_SMVD
+ if ( pu.cu->smvdMode )
+ {
+ CHECK( pu.refIdx[eRefList] != pu.cs->slice->getSymRefIdx( eRefList ), "Invalid reference index!\n" );
+ return;
+ }
+#endif
+
int numRef = pu.cs->slice->getNumRefIdx(eRefList);
if( numRef <= 1 )
{
@@ -1616,6 +1782,18 @@ void CABACWriter::triangle_mode( const CodingUnit& cu )
return;
}
+#if JVET_M0118_M0185_TRIANGLE_FLAG_FIX
+ if ( cu.firstPU->mmvdMergeFlag || cu.mmvdSkip )
+ {
+ return;
+ }
+
+ if ( cu.firstPU->mhIntraFlag )
+ {
+ return;
+ }
+#endif
+
unsigned flag_idx = DeriveCtx::CtxTriangleFlag( cu );
m_BinEncoder.encodeBin( cu.triangle, Ctx::TriangleFlag(flag_idx) );
@@ -1718,7 +1896,9 @@ void CABACWriter::transform_tree( const CodingStructure& cs, Partitioner& partit
chromaCbfs.Cb = TU::getCbfAtDepth( tu, COMPONENT_Cb, trDepth );
chromaCbfs.Cr = TU::getCbfAtDepth( tu, COMPONENT_Cr, trDepth );
}
+#if !JVET_M0464_UNI_MTS
if( trDepth == 0 ) emt_cu_flag( cu );
+#endif
if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
{
@@ -1756,7 +1936,9 @@ void CABACWriter::transform_tree( const CodingStructure& cs, Partitioner& partit
}
}
+#if !JVET_M0464_UNI_MTS
if( trDepth == 0 && TU::getCbfAtDepth( tu, COMPONENT_Y, 0 ) ) emt_cu_flag( cu );
+#endif
transform_unit( tu, cuCtx, chromaCbfs );
}
@@ -1961,13 +2143,21 @@ void CABACWriter::residual_coding( const TransformUnit& tu, ComponentID compID )
DTRACE( g_trace_ctx, D_SYNTAX, "residual_coding() etype=%d pos=(%d,%d) size=%dx%d predMode=%d\n", tu.blocks[compID].compID, tu.blocks[compID].x, tu.blocks[compID].y, tu.blocks[compID].width, tu.blocks[compID].height, cu.predMode );
// code transform skip and explicit rdpcm mode
+#if JVET_M0464_UNI_MTS
+ mts_coding ( tu, compID );
+#else
transform_skip_flag( tu, compID );
+#endif
explicit_rdpcm_mode( tu, compID );
#if HEVC_USE_SIGN_HIDING
// determine sign hiding
bool signHiding = ( cu.cs->slice->getSignDataHidingEnabledFlag() && !cu.transQuantBypass && tu.rdpcm[compID] == RDPCM_OFF );
+#if JVET_M0464_UNI_MTS
+ if( signHiding && CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && tu.mtsIdx==1 )
+#else
if( signHiding && CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && tu.transformSkip[compID] )
+#endif
{
const ChannelType chType = toChannelType( compID );
const unsigned intraMode = PU::getFinalIntraMode( *cu.cs->getPU( tu.blocks[compID].pos(), chType ), chType );
@@ -1985,7 +2175,9 @@ void CABACWriter::residual_coding( const TransformUnit& tu, ComponentID compID )
CoeffCodingContext cctx ( tu, compID );
#endif
const TCoeff* coeff = tu.getCoeffs( compID ).buf;
+#if !JVET_M0464_UNI_MTS
unsigned numSig = 0;
+#endif
// determine and set last coeff position and sig group flags
int scanPosLast = -1;
@@ -2008,22 +2200,27 @@ void CABACWriter::residual_coding( const TransformUnit& tu, ComponentID compID )
// code subblocks
const int stateTab = ( tu.cs->slice->getDepQuantEnabledFlag() ? 32040 : 0 );
int state = 0;
+#if !JVET_M0464_UNI_MTS
bool useEmt = ( cu.cs->sps->getSpsNext().getUseIntraEMT() && cu.predMode == MODE_INTRA ) || ( cu.cs->sps->getSpsNext().getUseInterEMT() && cu.predMode != MODE_INTRA );
useEmt = useEmt && isLuma(compID);
+#endif
for( int subSetId = ( cctx.scanPosLast() >> cctx.log2CGSize() ); subSetId >= 0; subSetId--)
{
cctx.initSubblock ( subSetId, sigGroupFlags[subSetId] );
residual_coding_subblock( cctx, coeff, stateTab, state );
+#if !JVET_M0464_UNI_MTS
if (useEmt)
{
numSig += cctx.emtNumSigCoeff();
cctx.setEmtNumSigCoeff( 0 );
}
+#endif
}
+#if !JVET_M0464_UNI_MTS
if( useEmt && !tu.transformSkip[compID] && compID == COMPONENT_Y && tu.cu->emtFlag )
{
if( CU::isIntra( *tu.cu ) )
@@ -2035,9 +2232,56 @@ void CABACWriter::residual_coding( const TransformUnit& tu, ComponentID compID )
emt_tu_index( tu );
}
}
+#endif
}
+#if JVET_M0464_UNI_MTS
+void CABACWriter::mts_coding( const TransformUnit& tu, ComponentID compID )
+{
+ const CodingUnit &cu = *tu.cu;
+ const bool tsAllowed = TU::isTSAllowed ( tu, compID );
+ const bool mtsAllowed = TU::isMTSAllowed( tu, compID );
+
+ if( !mtsAllowed && !tsAllowed ) return;
+
+ int symbol = 0;
+ int ctxIdx = 0;
+
+ if( tsAllowed )
+ {
+ symbol = 1 - ( tu.mtsIdx == 1 ? 1 : 0 );
+ ctxIdx = 6;
+ m_BinEncoder.encodeBin( symbol, Ctx::MTSIndex( ctxIdx ) );
+ }
+
+ if( tu.mtsIdx != 1 )
+ {
+ if( mtsAllowed )
+ {
+ symbol = tu.mtsIdx != 0 ? 1 : 0;
+ ctxIdx = std::min( (int)cu.qtDepth, 5 );
+ m_BinEncoder.encodeBin( symbol, Ctx::MTSIndex( ctxIdx ) );
+ if( symbol )
+ {
+ ctxIdx = 7;
+ for( int i = 0; i < 3; i++, ctxIdx++ )
+ {
+ symbol = tu.mtsIdx > i + 2 ? 1 : 0;
+ m_BinEncoder.encodeBin( symbol, Ctx::MTSIndex( ctxIdx ) );
+
+ if( !symbol )
+ {
+ break;
+ }
+ }
+ }
+ }
+ }
+
+ DTRACE( g_trace_ctx, D_SYNTAX, "mts_coding() etype=%d pos=(%d,%d) mtsIdx=%d\n", COMPONENT_Y, cu.lx(), cu.ly(), tu.mtsIdx );
+}
+#else
void CABACWriter::transform_skip_flag( const TransformUnit& tu, ComponentID compID )
{
if( !tu.cu->cs->pps->getUseTransformSkip() || tu.cu->transQuantBypass || !TU::hasTransformSkipFlag( *tu.cs, tu.blocks[compID] ) || ( isLuma( compID ) && tu.cu->emtFlag ) )
@@ -2099,13 +2343,17 @@ void CABACWriter::emt_cu_flag( const CodingUnit& cu )
DTRACE( g_trace_ctx, D_SYNTAX, "emt_cu_flag() etype=%d pos=(%d,%d) emtCuFlag=%d\n", COMPONENT_Y, cu.lx(), cu.ly(), ( int ) cu.emtFlag );
}
}
-
+#endif
void CABACWriter::explicit_rdpcm_mode( const TransformUnit& tu, ComponentID compID )
{
const CodingUnit& cu = *tu.cu;
+#if JVET_M0464_UNI_MTS
+ if( !CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && ( tu.mtsIdx==1 || cu.transQuantBypass ) )
+#else
if( !CU::isIntra(cu) && CU::isRDPCMEnabled(cu) && ( tu.transformSkip[compID] || cu.transQuantBypass ) )
+#endif
{
ChannelType chType = toChannelType( compID );
switch( tu.rdpcm[compID] )
@@ -2216,12 +2464,22 @@ void CABACWriter::residual_coding_subblock( CoeffCodingContext& cctx, const TCoe
int numNonZero = 0;
unsigned signPattern = 0;
bool is2x2subblock = ( cctx.log2CGSize() == 2 );
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ int remRegBins = ( is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK );
+#else
int remGt2Bins = ( is2x2subblock ? MAX_NUM_GT2_BINS_2x2SUBBLOCK : MAX_NUM_GT2_BINS_4x4SUBBLOCK );
int remRegBins = ( is2x2subblock ? MAX_NUM_REG_BINS_2x2SUBBLOCK : MAX_NUM_REG_BINS_4x4SUBBLOCK ) - remGt2Bins;
+#endif
int firstPosMode2 = minSubPos - 1;
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
int firstPosMode1 = minSubPos - 1;
+#endif
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ for( ; nextSigPos >= minSubPos && remRegBins >= 4; nextSigPos-- )
+#else
for( ; nextSigPos >= minSubPos && remRegBins >= 3; nextSigPos-- )
+#endif
{
TCoeff Coeff = coeff[ cctx.blockPos( nextSigPos ) ];
unsigned sigFlag = ( Coeff != 0 );
@@ -2260,19 +2518,29 @@ void CABACWriter::residual_coding_subblock( CoeffCodingContext& cctx, const TCoe
remAbsLevel >>= 1;
remRegBins--;
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ unsigned gt2 = !!remAbsLevel;
+ m_BinEncoder.encodeBin(gt2, cctx.greater2CtxIdAbs(ctxOff));
+ DTRACE(g_trace_ctx, D_SYNTAX_RESI, "gt2_flag() bin=%d ctx=%d\n", gt2, cctx.greater2CtxIdAbs(ctxOff));
+ remRegBins--;
+#else
if( remGt2Bins && !--remGt2Bins )
{
firstPosMode1 = nextSigPos - 1;
}
+#endif
}
}
state = ( stateTransTable >> ((state<<2)+((Coeff&1)<<1)) ) & 3;
}
firstPosMode2 = nextSigPos;
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
firstPosMode1 = ( firstPosMode1 > firstPosMode2 ? firstPosMode1 : firstPosMode2 );
+#endif
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
//===== 2nd PASS: gt2 =====
for( int scanPos = firstSigPos; scanPos > firstPosMode1; scanPos-- )
{
@@ -2285,10 +2553,17 @@ void CABACWriter::residual_coding_subblock( CoeffCodingContext& cctx, const TCoe
DTRACE( g_trace_ctx, D_SYNTAX_RESI, "gt2_flag() bin=%d ctx=%d\n", gt2, cctx.greater2CtxIdAbs(ctxOff) );
}
}
+#endif
+#if JVET_M0173_MOVE_GT2_TO_FIRST_PASS
+ //===== 2nd PASS: Go-rice codes =====
+ unsigned ricePar = 0;
+ for( int scanPos = firstSigPos; scanPos > firstPosMode2; scanPos-- )
+#else
//===== 3rd PASS: Go-rice codes =====
unsigned ricePar = 0;
for( int scanPos = firstSigPos; scanPos > firstPosMode1; scanPos-- )
+#endif
{
unsigned absLevel = abs( coeff[ cctx.blockPos( scanPos ) ] );
if( absLevel >= 4 )
@@ -2302,6 +2577,7 @@ void CABACWriter::residual_coding_subblock( CoeffCodingContext& cctx, const TCoe
}
}
}
+#if !JVET_M0173_MOVE_GT2_TO_FIRST_PASS
for( int scanPos = firstPosMode1; scanPos > firstPosMode2; scanPos-- )
{
unsigned absLevel = abs( coeff[ cctx.blockPos( scanPos ) ] );
@@ -2316,6 +2592,7 @@ void CABACWriter::residual_coding_subblock( CoeffCodingContext& cctx, const TCoe
}
}
}
+#endif
//===== coeff bypass ====
for( int scanPos = firstPosMode2; scanPos >= minSubPos; scanPos-- )
@@ -2352,7 +2629,9 @@ void CABACWriter::residual_coding_subblock( CoeffCodingContext& cctx, const TCoe
#else
m_BinEncoder.encodeBinsEP( signPattern, numNonZero );
#endif
+#if !JVET_M0464_UNI_MTS
cctx.setEmtNumSigCoeff(numNonZero);
+#endif
}
@@ -2468,6 +2747,7 @@ void CABACWriter::exp_golomb_eqprob( unsigned symbol, unsigned count )
m_BinEncoder.encodeBinsEP( bins, numBins );
}
+#if !REMOVE_BIN_DECISION_TREE
void CABACWriter::encode_sparse_dt( DecisionTree& dt, unsigned toCodeId )
{
// propagate the sparsity information from end-nodes to intermediate nodes
@@ -2517,6 +2797,7 @@ void CABACWriter::encode_sparse_dt( DecisionTree& dt, unsigned toCodeId )
return;
}
+#endif
void CABACWriter::codeAlfCtuEnableFlags( CodingStructure& cs, ChannelType channel, AlfSliceParam* alfParam)
{
if( isLuma( channel ) )
@@ -2546,7 +2827,7 @@ void CABACWriter::codeAlfCtuEnableFlag( CodingStructure& cs, uint32_t ctuRsAddr,
{
AlfSliceParam& alfSliceParam = alfParam ? (*alfParam) : cs.slice->getAlfSliceParam();
- if( cs.sps->getUseALF() && alfSliceParam.enabledFlag[compIdx] )
+ if( cs.sps->getALFEnabledFlag() && alfSliceParam.enabledFlag[compIdx] )
{
const PreCalcValues& pcv = *cs.pcv;
int frame_width_in_ctus = pcv.widthInCtus;
diff --git a/source/Lib/EncoderLib/CABACWriter.h b/source/Lib/EncoderLib/CABACWriter.h
index 6cc2e708e3109415ffed9f7f55a30015cb3b8191..10c97e26f7796ea71d546d39521533b9f4fa2c19 100644
--- a/source/Lib/EncoderLib/CABACWriter.h
+++ b/source/Lib/EncoderLib/CABACWriter.h
@@ -82,8 +82,12 @@ public:
void sao_offset_pars ( const SAOOffset& ctbPars, ComponentID compID, bool sliceEnabled, int bitDepth );
// coding (quad)tree (clause 7.3.8.4)
void coding_tree ( const CodingStructure& cs, Partitioner& pm, CUCtx& cuCtx, Partitioner* pPartitionerChroma = nullptr, CUCtx* pCuCtxChroma = nullptr);
+#if JVET_M0421_SPLIT_SIG
+ void split_cu_mode ( const PartSplit split, const CodingStructure& cs, Partitioner& pm );
+#else
void split_cu_flag ( bool split, const CodingStructure& cs, Partitioner& pm );
void split_cu_mode_mt ( const PartSplit split, const CodingStructure& cs, Partitioner& pm );
+#endif
// coding unit (clause 7.3.8.5)
void coding_unit ( const CodingUnit& cu, Partitioner& pm, CUCtx& cuCtx );
@@ -120,6 +124,9 @@ public:
void MHIntra_flag ( const PredictionUnit& pu );
void MHIntra_luma_pred_modes ( const CodingUnit& cu );
void triangle_mode ( const CodingUnit& cu );
+#if JVET_M0444_SMVD
+ void smvd_mode ( const PredictionUnit& pu );
+#endif
// pcm samples (clause 7.3.8.7)
void pcm_samples ( const TransformUnit& tu );
@@ -138,9 +145,13 @@ public:
// residual coding (clause 7.3.8.11)
void residual_coding ( const TransformUnit& tu, ComponentID compID );
+#if JVET_M0464_UNI_MTS
+ void mts_coding ( const TransformUnit& tu, ComponentID compID );
+#else
void transform_skip_flag ( const TransformUnit& tu, ComponentID compID );
void emt_tu_index ( const TransformUnit& tu );
void emt_cu_flag ( const CodingUnit& cu );
+#endif
void explicit_rdpcm_mode ( const TransformUnit& tu, ComponentID compID );
void last_sig_coeff ( CoeffCodingContext& cctx );
void residual_coding_subblock ( CoeffCodingContext& cctx, const TCoeff* coeff, const int stateTransTable, int& state );
@@ -156,7 +167,9 @@ private:
void unary_max_symbol ( unsigned symbol, unsigned ctxId0, unsigned ctxIdN, unsigned maxSymbol );
void unary_max_eqprob ( unsigned symbol, unsigned maxSymbol );
void exp_golomb_eqprob ( unsigned symbol, unsigned count );
+#if !REMOVE_BIN_DECISION_TREE
void encode_sparse_dt ( DecisionTree& dt, unsigned toCodeId );
+#endif
// statistic
unsigned get_num_written_bits() { return m_BinEncoder.getNumWrittenBits(); }
diff --git a/source/Lib/EncoderLib/EncCfg.h b/source/Lib/EncoderLib/EncCfg.h
index c60a7ad50613f0b9ac6c1ce0e3701f528b26a25d..0fdf461193b496844a0114d1dc5da9f382ac02bf 100644
--- a/source/Lib/EncoderLib/EncCfg.h
+++ b/source/Lib/EncoderLib/EncCfg.h
@@ -188,7 +188,6 @@ protected:
// TODO: We need to have a common sliding mechanism used by both the encoder and decoder
int m_maxTempLayer; ///< Max temporal layer
- bool m_useAMP;
unsigned m_CTUSize;
bool m_useSplitConsOverride;
unsigned m_uiMinQT[3]; //0: I slice; 1: P/B slice, 2: I slice chroma
@@ -202,10 +201,20 @@ protected:
unsigned m_log2DiffMaxMinCodingBlockSize;
int m_LMChroma;
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ bool m_cclmCollocatedChromaFlag;
+#endif
+#if JVET_M0464_UNI_MTS
+ int m_IntraMTS;
+ int m_InterMTS;
+ int m_IntraMTSMaxCand;
+ int m_InterMTSMaxCand;
+#else
int m_IntraEMT;
int m_InterEMT;
int m_FastIntraEMT;
int m_FastInterEMT;
+#endif
bool m_LargeCTU;
int m_SubPuMvpMode;
bool m_Affine;
@@ -230,13 +239,13 @@ protected:
bool m_MHIntra;
bool m_Triangle;
- unsigned m_CPRMode;
- unsigned m_CPRLocalSearchRangeX;
- unsigned m_CPRLocalSearchRangeY;
- unsigned m_CPRHashSearch;
- unsigned m_CPRHashSearchMaxCand;
- unsigned m_CPRHashSearchRange4SmallBlk;
- unsigned m_CPRFastMethod;
+ unsigned m_IBCMode;
+ unsigned m_IBCLocalSearchRangeX;
+ unsigned m_IBCLocalSearchRangeY;
+ unsigned m_IBCHashSearch;
+ unsigned m_IBCHashSearchMaxCand;
+ unsigned m_IBCHashSearchRange4SmallBlk;
+ unsigned m_IBCFastMethod;
bool m_wrapAround;
unsigned m_wrapAroundOffset;
@@ -674,6 +683,10 @@ public:
void setUseLMChroma ( int n ) { m_LMChroma = n; }
int getUseLMChroma() const { return m_LMChroma; }
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ void setCclmCollocatedChromaFlag ( bool b ) { m_cclmCollocatedChromaFlag = b; }
+ bool getCclmCollocatedChromaFlag () const { return m_cclmCollocatedChromaFlag; }
+#endif
void setSubPuMvpMode(int n) { m_SubPuMvpMode = n; }
bool getSubPuMvpMode() const { return m_SubPuMvpMode; }
@@ -695,6 +708,16 @@ public:
bool getUseAltDQPCoding () const { return m_AltDQPCoding; }
#endif
+#if JVET_M0464_UNI_MTS
+ void setIntraMTSMaxCand ( unsigned u ) { m_IntraMTSMaxCand = u; }
+ unsigned getIntraMTSMaxCand () const { return m_IntraMTSMaxCand; }
+ void setInterMTSMaxCand ( unsigned u ) { m_InterMTSMaxCand = u; }
+ unsigned getInterMTSMaxCand () const { return m_InterMTSMaxCand; }
+ void setIntraMTS ( bool b ) { m_IntraMTS = b; }
+ bool getIntraMTS () const { return m_IntraMTS; }
+ void setInterMTS ( bool b ) { m_InterMTS = b; }
+ bool getInterMTS () const { return m_InterMTS; }
+#else
void setFastIntraEMT ( bool b ) { m_FastIntraEMT = b; }
bool getFastIntraEMT () const { return m_FastIntraEMT; }
void setFastInterEMT ( bool b ) { m_FastInterEMT = b; }
@@ -703,6 +726,7 @@ public:
bool getIntraEMT () const { return m_IntraEMT; }
void setInterEMT ( bool b ) { m_InterEMT = b; }
bool getInterEMT () const { return m_InterEMT; }
+#endif
@@ -732,20 +756,20 @@ public:
bool getUseTriangle () const { return m_Triangle; }
- void setCPRMode (unsigned n) { m_CPRMode = n; }
- unsigned getCPRMode () const { return m_CPRMode; }
- void setCPRLocalSearchRangeX (unsigned n) { m_CPRLocalSearchRangeX = n; }
- unsigned getCPRLocalSearchRangeX () const { return m_CPRLocalSearchRangeX; }
- void setCPRLocalSearchRangeY (unsigned n) { m_CPRLocalSearchRangeY = n; }
- unsigned getCPRLocalSearchRangeY () const { return m_CPRLocalSearchRangeY; }
- void setCPRHashSearch (unsigned n) { m_CPRHashSearch = n; }
- unsigned getCPRHashSearch () const { return m_CPRHashSearch; }
- void setCPRHashSearchMaxCand (unsigned n) { m_CPRHashSearchMaxCand = n; }
- unsigned getCPRHashSearchMaxCand () const { return m_CPRHashSearchMaxCand; }
- void setCPRHashSearchRange4SmallBlk (unsigned n) { m_CPRHashSearchRange4SmallBlk = n; }
- unsigned getCPRHashSearchRange4SmallBlk () const { return m_CPRHashSearchRange4SmallBlk; }
- void setCPRFastMethod (unsigned n) { m_CPRFastMethod = n; }
- unsigned getCPRFastMethod () const { return m_CPRFastMethod; }
+ void setIBCMode (unsigned n) { m_IBCMode = n; }
+ unsigned getIBCMode () const { return m_IBCMode; }
+ void setIBCLocalSearchRangeX (unsigned n) { m_IBCLocalSearchRangeX = n; }
+ unsigned getIBCLocalSearchRangeX () const { return m_IBCLocalSearchRangeX; }
+ void setIBCLocalSearchRangeY (unsigned n) { m_IBCLocalSearchRangeY = n; }
+ unsigned getIBCLocalSearchRangeY () const { return m_IBCLocalSearchRangeY; }
+ void setIBCHashSearch (unsigned n) { m_IBCHashSearch = n; }
+ unsigned getIBCHashSearch () const { return m_IBCHashSearch; }
+ void setIBCHashSearchMaxCand (unsigned n) { m_IBCHashSearchMaxCand = n; }
+ unsigned getIBCHashSearchMaxCand () const { return m_IBCHashSearchMaxCand; }
+ void setIBCHashSearchRange4SmallBlk (unsigned n) { m_IBCHashSearchRange4SmallBlk = n; }
+ unsigned getIBCHashSearchRange4SmallBlk () const { return m_IBCHashSearchRange4SmallBlk; }
+ void setIBCFastMethod (unsigned n) { m_IBCFastMethod = n; }
+ unsigned getIBCFastMethod () const { return m_IBCFastMethod; }
void setUseWrapAround ( bool b ) { m_wrapAround = b; }
bool getUseWrapAround () const { return m_wrapAround; }
@@ -783,8 +807,6 @@ public:
void setQuadtreeTUMaxDepthInter ( uint32_t u ) { m_uiQuadtreeTUMaxDepthInter = u; }
void setQuadtreeTUMaxDepthIntra ( uint32_t u ) { m_uiQuadtreeTUMaxDepthIntra = u; }
- void setUseAMP( bool b ) { m_useAMP = b; }
-
//====== Loop/Deblock Filter ========
void setLoopFilterDisable ( bool b ) { m_bLoopFilterDisable = b; }
void setLoopFilterOffsetInPPS ( bool b ) { m_loopFilterOffsetInPPS = b; }
@@ -878,13 +900,8 @@ public:
#if X0038_LAMBDA_FROM_QP_CAPABILITY
int getIntraQPOffset () const { return m_intraQPOffset; }
int getLambdaFromQPEnable () const { return m_lambdaFromQPEnable; }
-#if ENABLE_QPA | JVET_L0646_GBI
public:
-#else
-protected:
-#endif
int getBaseQP () const { return m_iQP; } // public should use getQPForPicture.
-public:
int getQPForPicture (const uint32_t gopIndex, const Slice *pSlice) const; // Function actually defined in EncLib.cpp
#else
int getBaseQP () { return m_iQP; }
diff --git a/source/Lib/EncoderLib/EncCu.cpp b/source/Lib/EncoderLib/EncCu.cpp
index f74add71740e260dd27ea94d3353de02ddd5e820..2b41fcb861b2322aa96914164e4b886e52868414 100644
--- a/source/Lib/EncoderLib/EncCu.cpp
+++ b/source/Lib/EncoderLib/EncCu.cpp
@@ -268,6 +268,18 @@ void EncCu::init( EncLib* pcEncLib, const SPS& sps PARL_PARAM( const int tId ) )
m_pcEncLib = pcEncLib;
m_dataId = tId;
#endif
+#if JVET_M0170_MRG_SHARELIST
+ m_shareState = NO_SHARE;
+ m_pcInterSearch->setShareState(0);
+ setShareStateDec(0);
+#endif
+
+#if JVET_M0170_MRG_SHARELIST
+ m_shareBndPosX = -1;
+ m_shareBndPosY = -1;
+ m_shareBndSizeW = 0;
+ m_shareBndSizeH = 0;
+#endif
#if REUSE_CU_RESULTS
DecCu::init( m_pcTrQuant, m_pcIntraSearch, m_pcInterSearch );
@@ -280,9 +292,9 @@ void EncCu::init( EncLib* pcEncLib, const SPS& sps PARL_PARAM( const int tId ) )
::memset(m_subMergeBlkNum, 0, sizeof(m_subMergeBlkNum));
m_prevPOC = MAX_UINT;
- if (m_pcEncCfg->getCPRHashSearch() && m_pcEncCfg->getCPRMode())
+ if (m_pcEncCfg->getIBCHashSearch() && m_pcEncCfg->getIBCMode())
{
- m_cprHashMap.init(m_pcEncCfg->getSourceWidth(), m_pcEncCfg->getSourceHeight());
+ m_ibcHashMap.init(m_pcEncCfg->getSourceWidth(), m_pcEncCfg->getSourceHeight());
}
}
@@ -292,9 +304,9 @@ void EncCu::init( EncLib* pcEncLib, const SPS& sps PARL_PARAM( const int tId ) )
void EncCu::compressCtu( CodingStructure& cs, const UnitArea& area, const unsigned ctuRsAddr, const int prevQP[], const int currQP[] )
{
- if (m_pcEncCfg->getCPRHashSearch() && ctuRsAddr == 0 && cs.slice->getSPS()->getSpsNext().getCPRMode())
+ if (m_pcEncCfg->getIBCHashSearch() && ctuRsAddr == 0 && cs.slice->getSPS()->getSpsNext().getIBCMode())
{
- m_cprHashMap.rebuildPicHashMap(cs.picture->getOrigBuf());
+ m_ibcHashMap.rebuildPicHashMap(cs.picture->getOrigBuf());
}
m_modeCtrl->initCTUEncoding( *cs.slice );
@@ -317,24 +329,24 @@ 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 );
- if (m_pcEncCfg->getCPRMode())
+ if (m_pcEncCfg->getIBCMode())
{
m_pcInterSearch->resetCtuRecord();
- m_ctuCprSearchRangeX = m_pcEncCfg->getCPRLocalSearchRangeX();
- m_ctuCprSearchRangeY = m_pcEncCfg->getCPRLocalSearchRangeY();
+ m_ctuIbcSearchRangeX = m_pcEncCfg->getIBCLocalSearchRangeX();
+ m_ctuIbcSearchRangeY = m_pcEncCfg->getIBCLocalSearchRangeY();
}
- if (m_pcEncCfg->getCPRMode() && m_pcEncCfg->getCPRHashSearch() && (m_pcEncCfg->getCPRFastMethod() & CPR_FAST_METHOD_ADAPTIVE_SEARCHRANGE))
+ if (m_pcEncCfg->getIBCMode() && m_pcEncCfg->getIBCHashSearch() && (m_pcEncCfg->getIBCFastMethod() & IBC_FAST_METHOD_ADAPTIVE_SEARCHRANGE))
{
- const int hashHitRatio = m_cprHashMap.getHashHitRatio(area.Y()); // in percent
+ const int hashHitRatio = m_ibcHashMap.getHashHitRatio(area.Y()); // in percent
if (hashHitRatio < 5) // 5%
{
- m_ctuCprSearchRangeX >>= 1;
- m_ctuCprSearchRangeY >>= 1;
+ m_ctuIbcSearchRangeX >>= 1;
+ m_ctuIbcSearchRangeY >>= 1;
}
if (cs.slice->getNumRefIdx(REF_PIC_LIST_0) > 1)
{
- m_ctuCprSearchRangeX >>= 1;
- m_ctuCprSearchRangeY >>= 1;
+ m_ctuIbcSearchRangeX >>= 1;
+ m_ctuIbcSearchRangeY >>= 1;
}
}
// init current context pointer
@@ -573,6 +585,17 @@ void EncCu::xCompressCU( CodingStructure *&tempCS, CodingStructure *&bestCS, Par
, LutMotionCand *&bestMotCandLUTs
)
{
+#if JVET_M0170_MRG_SHARELIST
+ if (m_shareState == NO_SHARE)
+ {
+ tempCS->sharedBndPos = tempCS->area.Y().lumaPos();
+ tempCS->sharedBndSize.width = tempCS->area.lwidth();
+ tempCS->sharedBndSize.height = tempCS->area.lheight();
+ bestCS->sharedBndPos = bestCS->area.Y().lumaPos();
+ bestCS->sharedBndSize.width = bestCS->area.lwidth();
+ bestCS->sharedBndSize.height = bestCS->area.lheight();
+ }
+#endif
#if ENABLE_SPLIT_PARALLELISM
CHECK( m_dataId != tempCS->picture->scheduler.getDataId(), "Working in the wrong dataId!" );
@@ -635,6 +658,11 @@ void EncCu::xCompressCU( CodingStructure *&tempCS, CodingStructure *&bestCS, Par
DTRACE_UPDATE( g_trace_ctx, std::make_pair( "cuh", tempCS->area.lheight() ) );
DTRACE( g_trace_ctx, D_COMMON, "@(%4d,%4d) [%2dx%2d]\n", tempCS->area.lx(), tempCS->area.ly(), tempCS->area.lwidth(), tempCS->area.lheight() );
+
+#if JVET_M0170_MRG_SHARELIST
+ int startShareThisLevel = 0;
+#endif
+
do
{
EncTestMode currTestMode = m_modeCtrl->currTestMode();
@@ -716,13 +744,13 @@ void EncCu::xCompressCU( CodingStructure *&tempCS, CodingStructure *&bestCS, Par
{
xCheckIntraPCM( tempCS, bestCS, partitioner, currTestMode );
}
- else if (currTestMode.type == ETM_CPR)
+ else if (currTestMode.type == ETM_IBC)
{
- xCheckRDCostCPRMode(tempCS, bestCS, partitioner, currTestMode);
+ xCheckRDCostIBCMode(tempCS, bestCS, partitioner, currTestMode);
}
- else if (currTestMode.type == ETM_CPR_MERGE)
+ else if (currTestMode.type == ETM_IBC_MERGE)
{
- xCheckRDCostCPRModeMerge2Nx2N(tempCS, bestCS, partitioner, currTestMode);
+ xCheckRDCostIBCModeMerge2Nx2N(tempCS, bestCS, partitioner, currTestMode);
}
else if( isModeSplit( currTestMode ) )
{
@@ -739,6 +767,15 @@ void EncCu::xCompressCU( CodingStructure *&tempCS, CodingStructure *&bestCS, Par
}
} while( m_modeCtrl->nextMode( *tempCS, partitioner ) );
+#if JVET_M0170_MRG_SHARELIST
+ if(startShareThisLevel == 1)
+ {
+ m_shareState = NO_SHARE;
+ m_pcInterSearch->setShareState(m_shareState);
+ setShareStateDec(m_shareState);
+ }
+#endif
+
//////////////////////////////////////////////////////////////////////////
// Finishing CU
#if ENABLE_SPLIT_PARALLELISM
@@ -1043,10 +1080,19 @@ void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS,
m_CABACEstimator->getCtx() = m_CurrCtx->start;
const TempCtx ctxStartSP( m_CtxCache, SubCtx( Ctx::SplitFlag, m_CABACEstimator->getCtx() ) );
+#if JVET_M0421_SPLIT_SIG
+ const TempCtx ctxStartQt( m_CtxCache, SubCtx( Ctx::SplitQtFlag, m_CABACEstimator->getCtx() ) );
+ const TempCtx ctxStartHv( m_CtxCache, SubCtx( Ctx::SplitHvFlag, m_CABACEstimator->getCtx() ) );
+ const TempCtx ctxStart12( m_CtxCache, SubCtx( Ctx::Split12Flag, m_CABACEstimator->getCtx() ) );
+#else
const TempCtx ctxStartBT( m_CtxCache, SubCtx( Ctx::BTSplitFlag, m_CABACEstimator->getCtx() ) );
+#endif
m_CABACEstimator->resetBits();
+#if JVET_M0421_SPLIT_SIG
+ m_CABACEstimator->split_cu_mode( split, *tempCS, partitioner );
+#else
if( partitioner.getImplicitSplit( *tempCS ) != CU_QUAD_SPLIT )
{
if( partitioner.canSplit( CU_QUAD_SPLIT, *tempCS ) )
@@ -1058,12 +1104,19 @@ void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS,
m_CABACEstimator->split_cu_mode_mt( split, *tempCS, partitioner );
}
}
+#endif
const double factor = ( tempCS->currQP[partitioner.chType] > 30 ? 1.1 : 1.075 );
const double cost = m_pcRdCost->calcRdCost( uint64_t( m_CABACEstimator->getEstFracBits() + ( ( bestCS->fracBits ) / factor ) ), Distortion( bestCS->dist / factor ) );
m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitFlag, ctxStartSP );
+#if JVET_M0421_SPLIT_SIG
+ m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitQtFlag, ctxStartQt );
+ m_CABACEstimator->getCtx() = SubCtx( Ctx::SplitHvFlag, ctxStartHv );
+ m_CABACEstimator->getCtx() = SubCtx( Ctx::Split12Flag, ctxStart12 );
+#else
m_CABACEstimator->getCtx() = SubCtx( Ctx::BTSplitFlag, ctxStartBT );
+#endif
if( cost > bestCS->cost
#if ENABLE_QPA_SUB_CTU
@@ -1076,6 +1129,55 @@ void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS,
return;
}
+#if JVET_M0170_MRG_SHARELIST
+ if (!slice.isIntra()
+ && tempCS->chType == CHANNEL_TYPE_LUMA
+ )
+ {
+ tempCS->slice->copyMotionLUTs(tempMotCandLUTs, tempCS->slice->getMotionLUTs());
+ }
+
+ int startShareThisLevel = 0;
+ const uint32_t uiLPelX = tempCS->area.Y().lumaPos().x;
+ const uint32_t uiTPelY = tempCS->area.Y().lumaPos().y;
+
+ int splitRatio = 1;
+ CHECK(!(split == CU_QUAD_SPLIT || split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT
+ || split == CU_TRIH_SPLIT || split == CU_TRIV_SPLIT), "invalid split type");
+ splitRatio = (split == CU_HORZ_SPLIT || split == CU_VERT_SPLIT) ? 1 : 2;
+
+ bool isOneChildSmall = ((tempCS->area.lwidth())*(tempCS->area.lheight()) >> splitRatio) < MRG_SHARELIST_SHARSIZE;
+
+ if ((((tempCS->area.lwidth())*(tempCS->area.lheight())) > (MRG_SHARELIST_SHARSIZE * 1)))
+ {
+ m_shareState = NO_SHARE;
+ }
+
+ if (m_shareState == NO_SHARE)//init state
+ {
+ if (isOneChildSmall)
+ {
+ m_shareState = GEN_ON_SHARED_BOUND;//share start state
+ startShareThisLevel = 1;
+ }
+ }
+ if ((m_shareState == GEN_ON_SHARED_BOUND) && (!slice.isIntra()))
+ {
+#if JVET_M0170_MRG_SHARELIST
+ tempCS->slice->copyMotionLUTs(tempCS->slice->getMotionLUTs(), tempCS->slice->m_MotionCandLuTsBkup);
+ m_shareBndPosX = uiLPelX;
+ m_shareBndPosY = uiTPelY;
+ m_shareBndSizeW = tempCS->area.lwidth();
+ m_shareBndSizeH = tempCS->area.lheight();
+ m_shareState = SHARING;
+#endif
+ }
+
+
+ m_pcInterSearch->setShareState(m_shareState);
+ setShareStateDec(m_shareState);
+#endif
+
partitioner.splitCurrArea( split, *tempCS );
m_CurrCtx++;
@@ -1106,7 +1208,16 @@ void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS,
tempCS->slice->copyMotionLUTs(tempMotCandLUTs, tempSubMotCandLUTs);
tempCS->slice->copyMotionLUTs(tempMotCandLUTs, bestSubMotCandLUTs);
}
-
+#if JVET_M0170_MRG_SHARELIST
+ tempSubCS->sharedBndPos.x = (m_shareState == SHARING) ? m_shareBndPosX : tempSubCS->area.Y().lumaPos().x;
+ tempSubCS->sharedBndPos.y = (m_shareState == SHARING) ? m_shareBndPosY : tempSubCS->area.Y().lumaPos().y;
+ tempSubCS->sharedBndSize.width = (m_shareState == SHARING) ? m_shareBndSizeW : tempSubCS->area.lwidth();
+ tempSubCS->sharedBndSize.height = (m_shareState == SHARING) ? m_shareBndSizeH : tempSubCS->area.lheight();
+ bestSubCS->sharedBndPos.x = (m_shareState == SHARING) ? m_shareBndPosX : tempSubCS->area.Y().lumaPos().x;
+ bestSubCS->sharedBndPos.y = (m_shareState == SHARING) ? m_shareBndPosY : tempSubCS->area.Y().lumaPos().y;
+ bestSubCS->sharedBndSize.width = (m_shareState == SHARING) ? m_shareBndSizeW : tempSubCS->area.lwidth();
+ bestSubCS->sharedBndSize.height = (m_shareState == SHARING) ? m_shareBndSizeH : tempSubCS->area.lheight();
+#endif
xCompressCU( tempSubCS, bestSubCS, partitioner
, tempSubMotCandLUTs
, bestSubMotCandLUTs
@@ -1146,6 +1257,15 @@ void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS,
partitioner.exitCurrSplit();
+#if JVET_M0170_MRG_SHARELIST
+ if (startShareThisLevel == 1)
+ {
+ m_shareState = NO_SHARE;
+ m_pcInterSearch->setShareState(m_shareState);
+ setShareStateDec(m_shareState);
+ }
+#endif
+
m_CurrCtx--;
// Finally, generate split-signaling bits for RD-cost check
@@ -1178,6 +1298,9 @@ void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS,
{
m_CABACEstimator->resetBits();
+#if JVET_M0421_SPLIT_SIG
+ m_CABACEstimator->split_cu_mode( split, *tempCS, partitioner );
+#else
if( partitioner.canSplit( CU_QUAD_SPLIT, *tempCS ) )
{
m_CABACEstimator->split_cu_flag( split == CU_QUAD_SPLIT, *tempCS, partitioner );
@@ -1186,6 +1309,7 @@ void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS,
{
m_CABACEstimator->split_cu_mode_mt( split, *tempCS, partitioner );
}
+#endif
tempCS->fracBits += m_CABACEstimator->getEstFracBits(); // split bits
}
@@ -1261,19 +1385,25 @@ void EncCu::xCheckModeSplit(CodingStructure *&tempCS, CodingStructure *&bestCS,
void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
{
+#if !JVET_M0464_UNI_MTS
double bestInterCost = m_modeCtrl->getBestInterCost();
double costSize2Nx2NemtFirstPass = m_modeCtrl->getEmtSize2Nx2NFirstPassCost();
bool skipSecondEmtPass = m_modeCtrl->getSkipSecondEMTPass();
const SPS &sps = *tempCS->sps;
- const PPS &pps = *tempCS->pps;
+#endif
+ const PPS &pps = *tempCS->pps;
+#if !JVET_M0464_UNI_MTS
const CodingUnit *bestCU = bestCS->getCU( partitioner.chType );
const int maxSizeEMT = EMT_INTRA_MAX_CU_WITH_QTBT;
uint8_t considerEmtSecondPass = ( sps.getSpsNext().getUseIntraEMT() && isLuma( partitioner.chType ) && partitioner.currArea().lwidth() <= maxSizeEMT && partitioner.currArea().lheight() <= maxSizeEMT ) ? 1 : 0;
+#endif
Distortion interHad = m_modeCtrl->getInterHad();
-
+#if JVET_M0464_UNI_MTS
+ {
+#else
for( uint8_t emtCuFlag = 0; emtCuFlag <= considerEmtSecondPass; emtCuFlag++ )
{
//Possible early EMT tests interruptions
@@ -1287,6 +1417,7 @@ void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestC
{
continue;
}
+#endif
tempCS->initStructData( encTestMode.qp, encTestMode.lossless );
@@ -1304,7 +1435,9 @@ void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestC
cu.chromaQpAdj = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
cu.qp = encTestMode.qp;
//cu.ipcm = false;
+#if !JVET_M0464_UNI_MTS
cu.emtFlag = emtCuFlag;
+#endif
CU::addPUs( cu );
@@ -1320,7 +1453,11 @@ void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestC
interHad = 0;
// JEM assumes only perfect reconstructions can from now on beat the inter mode
m_modeCtrl->enforceInterHad( 0 );
+#if JVET_M0464_UNI_MTS
+ return;
+#else
continue;
+#endif
}
if( !CS::isDualITree( *tempCS ) )
@@ -1373,9 +1510,10 @@ void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestC
xCheckDQP( *tempCS, partitioner );
-
+#if !JVET_M0464_UNI_MTS
// we save the cost of the modes for the first EMT pass
if( !emtCuFlag ) static_cast< double& >( costSize2Nx2NemtFirstPass ) = tempCS->cost;
+#endif
#if WCG_EXT
DTRACE_MODE_COST( *tempCS, m_pcRdCost->getLambda( true ) );
@@ -1384,7 +1522,7 @@ void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestC
#endif
xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
-
+#if !JVET_M0464_UNI_MTS
//now we check whether the second pass of SIZE_2Nx2N and the whole Intra SIZE_NxN should be skipped or not
if( !emtCuFlag && !tempCS->slice->isIntra() && bestCU && bestCU->predMode != MODE_INTRA && m_pcEncCfg->getFastInterEMT() )
{
@@ -1396,7 +1534,7 @@ void EncCu::xCheckRDCostIntra( CodingStructure *&tempCS, CodingStructure *&bestC
break;
}
}
-
+#endif
} //for emtCuFlag
}
@@ -1567,7 +1705,7 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
MergeCtx mergeCtx;
const SPS &sps = *tempCS->sps;
- if( sps.getSpsNext().getUseSubPuMvp() )
+ if( sps.getSBTMVPEnabledFlag() )
{
Size bufSize = g_miScaling.scale( tempCS->area.lumaSize() );
mergeCtx.subPuMvpMiBuf = MotionBuf( m_SubPuMiBuf, bufSize );
@@ -1588,10 +1726,16 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
PredictionUnit pu( tempCS->area );
pu.cu = &cu;
pu.cs = tempCS;
+#if JVET_M0170_MRG_SHARELIST
+ pu.shareParentPos = tempCS->sharedBndPos;
+ pu.shareParentSize = tempCS->sharedBndSize;
+#endif
PU::getInterMergeCandidates(pu, mergeCtx
, 0
);
+#if !JVET_M0068_M0171_MMVD_CLEANUP
PU::restrictBiPredMergeCands(pu, mergeCtx);
+#endif
PU::getInterMMVDMergeCandidates(pu, mergeCtx);
}
bool candHasNoResidual[MRG_MAX_NUM_CANDS + MMVD_ADD_NUM];
@@ -1654,7 +1798,7 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
if( auto blkCache = dynamic_cast< CacheBlkInfoCtrl* >( m_modeCtrl ) )
{
- if (slice.getSPS()->getSpsNext().getCPRMode())
+ if (slice.getSPS()->getSpsNext().getIBCMode())
{
ComprCUCtx cuECtx = m_modeCtrl->getComprCUCtx();
bestIsSkip = blkCache->isSkip(tempCS->area) && cuECtx.bestCU;
@@ -1704,13 +1848,13 @@ 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) );
- uint32_t cprCand = 0;
+ uint32_t ibcCand = 0;
uint32_t numValidMv = mergeCtx.numValidMergeCand;
for( uint32_t uiMergeCand = 0; uiMergeCand < mergeCtx.numValidMergeCand; uiMergeCand++ )
{
if ((mergeCtx.interDirNeighbours[uiMergeCand] == 1 || mergeCtx.interDirNeighbours[uiMergeCand] == 3) && tempCS->slice->getRefPic(REF_PIC_LIST_0, mergeCtx.mvFieldNeighbours[uiMergeCand << 1].refIdx)->getPOC() == tempCS->slice->getPOC())
{
- cprCand++;
+ ibcCand++;
numValidMv--;
continue;
}
@@ -1752,7 +1896,7 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
swap(singleMergeTempBuffer, acMergeTempBuffer[insertPos]);
}
}
- CHECK(std::min(uiMergeCand + 1 - cprCand, uiNumMrgSATDCand) != RdModeList.size(), "");
+ CHECK(std::min(uiMergeCand + 1 - ibcCand, uiNumMrgSATDCand) != RdModeList.size(), "");
}
if (numValidMv < uiNumMrgSATDCand)
uiNumMrgSATDCand = numValidMv;
@@ -2073,10 +2217,14 @@ void EncCu::xCheckRDCostMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&
isTestSkipMerge[uiMergeCand] = true;
}
+#if JVET_M0464_UNI_MTS
+ xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, uiNoResidualPass, NULL, uiNoResidualPass == 0 ? &candHasNoResidual[uiMrgHADIdx] : NULL );
+#else
xEncodeInterResidual(tempCS, bestCS, partitioner, encTestMode, uiNoResidualPass
, NULL
, 1
, uiNoResidualPass == 0 ? &candHasNoResidual[uiMrgHADIdx] : NULL);
+#endif
if( m_pcEncCfg->getUseFastDecisionForMerge() && !bestIsSkip && !pu.mhIntraFlag)
{
@@ -2333,7 +2481,11 @@ void EncCu::xCheckRDCostMergeTriangle2Nx2N( CodingStructure *&tempCS, CodingStru
m_pcInterSearch->weightedTriangleBlk( pu, PU::getTriangleWeights(pu, triangleMrgCtx, candIdx0, candIdx1), splitDir, MAX_NUM_CHANNEL_TYPE, predBuf, triangleBuffer[candIdx0], triangleBuffer[candIdx1] );
}
+#if JVET_M0464_UNI_MTS
+ xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, noResidualPass, NULL, ( noResidualPass == 0 ? &trianglecandHasNoResidual[mergeCand] : NULL ) );
+#else
xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, noResidualPass, NULL, true, ( (noResidualPass == 0 ) ? &trianglecandHasNoResidual[mergeCand] : NULL ) );
+#endif
if (m_pcEncCfg->getUseFastDecisionForMerge() && !bestIsSkip)
{
@@ -2367,7 +2519,7 @@ void EncCu::xCheckRDCostAffineMerge2Nx2N( CodingStructure *&tempCS, CodingStruct
const SPS &sps = *tempCS->sps;
MergeCtx mrgCtx;
- if ( sps.getSpsNext().getUseSubPuMvp() )
+ if ( sps.getSBTMVPEnabledFlag() )
{
Size bufSize = g_miScaling.scale( tempCS->area.lumaSize() );
mrgCtx.subPuMvpMiBuf = MotionBuf( m_SubPuMiBuf, bufSize );
@@ -2580,7 +2732,11 @@ void EncCu::xCheckRDCostAffineMerge2Nx2N( CodingStructure *&tempCS, CodingStruct
m_pcInterSearch->motionCompensation( pu );
}
+#if JVET_M0464_UNI_MTS
+ xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, uiNoResidualPass, NULL, ( uiNoResidualPass == 0 ? &candHasNoResidual[uiMergeCand] : NULL ) );
+#else
xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, uiNoResidualPass, NULL, true, ((uiNoResidualPass == 0) ? &candHasNoResidual[uiMergeCand] : NULL) );
+#endif
if ( m_pcEncCfg->getUseFastDecisionForMerge() && !bestIsSkip )
{
@@ -2622,12 +2778,12 @@ void EncCu::xCheckRDCostAffineMerge2Nx2N( CodingStructure *&tempCS, CodingStruct
}
}
//////////////////////////////////////////////////////////////////////////////////////////////
-// cpr merge/skip mode check
-void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode)
+// ibc merge/skip mode check
+void EncCu::xCheckRDCostIBCModeMerge2Nx2N(CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode)
{
- assert(tempCS->chType != CHANNEL_TYPE_CHROMA); // chroma CPR is derived
+ assert(tempCS->chType != CHANNEL_TYPE_CHROMA); // chroma IBC is derived
- if (tempCS->area.lwidth() > CPR_MAX_CAND_SIZE || tempCS->area.lheight() > CPR_MAX_CAND_SIZE) // currently only check 32x32 and below block for cpr merge/skip
+ if (tempCS->area.lwidth() > IBC_MAX_CAND_SIZE || tempCS->area.lheight() > IBC_MAX_CAND_SIZE) // currently only check 32x32 and below block for ibc merge/skip
{
return;
}
@@ -2637,7 +2793,7 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
MergeCtx mergeCtx;
- if (sps.getSpsNext().getUseSubPuMvp())
+ if (sps.getSBTMVPEnabledFlag())
{
Size bufSize = g_miScaling.scale(tempCS->area.lumaSize());
mergeCtx.subPuMvpMiBuf = MotionBuf(m_SubPuMiBuf, bufSize);
@@ -2648,7 +2804,7 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
CodingUnit cu(tempCS->area);
cu.cs = tempCS;
cu.predMode = MODE_INTER;
- cu.cpr = true;
+ cu.ibc = true;
cu.slice = tempCS->slice;
#if HEVC_TILES_WPP
cu.tileIdx = tempCS->picture->tileMap->getTileIdxMap(tempCS->area.lumaPos());
@@ -2659,6 +2815,10 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
cu.mmvdSkip = false;
pu.mmvdMergeFlag = false;
cu.triangle = false;
+#if JVET_M0170_MRG_SHARELIST
+ pu.shareParentPos = tempCS->sharedBndPos;
+ pu.shareParentSize = tempCS->sharedBndSize;
+#endif
PU::getInterMergeCandidates(pu, mergeCtx
, 0
);
@@ -2693,7 +2853,7 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
#endif
cu.skip = false;
cu.predMode = MODE_INTER;
- cu.cpr = true;
+ cu.ibc = true;
cu.transQuantBypass = encTestMode.lossless;
cu.chromaQpAdj = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
cu.qp = encTestMode.qp;
@@ -2803,6 +2963,9 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
{
if (!(bestIsSkip && (numResidualPass == 0)))
{
+#if JVET_M0464_UNI_MTS
+ {
+#else
unsigned char considerEmtSecondPass = 0;
bool skipSecondEmtPass = true;
bool hasResidual[2] = { false, false };
@@ -2815,6 +2978,7 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
{
continue;
}
+#endif
// first get merge candidates
CodingUnit &cu = tempCS->addCU(CS::getArea(*tempCS, tempCS->area, (const ChannelType)partitioner.chType), (const ChannelType)partitioner.chType);
@@ -2826,23 +2990,24 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
#endif
cu.skip = false;
cu.predMode = MODE_INTER;
- cu.cpr = true;
+ cu.ibc = true;
cu.transQuantBypass = encTestMode.lossless;
cu.chromaQpAdj = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
cu.qp = encTestMode.qp;
-
+#if !JVET_M0464_UNI_MTS
cu.emtFlag = false;
+#endif
PredictionUnit &pu = tempCS->addPU(cu, partitioner.chType);// tempCS->addPU(cu);
- pu.intraDir[0] = DC_IDX; // set intra pred for cpr block
- pu.intraDir[1] = PLANAR_IDX; // set intra pred for cpr block
+ pu.intraDir[0] = DC_IDX; // set intra pred for ibc block
+ pu.intraDir[1] = PLANAR_IDX; // set intra pred for ibc block
cu.mmvdSkip = false;
pu.mmvdMergeFlag = false;
cu.triangle = false;
mergeCtx.setMergeInfo(pu, mergeCand);
PU::spanMotionInfo(pu, mergeCtx);
- assert(mergeCtx.mrgTypeNeighbours[mergeCand] == MRG_TYPE_CPR); // should be CPR candidate at this round
+ assert(mergeCtx.mrgTypeNeighbours[mergeCand] == MRG_TYPE_IBC); // should be IBC candidate at this round
const bool chroma = !(CS::isDualITree(*tempCS));
// MC
@@ -2862,20 +3027,24 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
}
#endif
+#if !JVET_M0464_UNI_MTS
hasResidual[emtCuFlag] = cu.rootCbf;
emtCost[emtCuFlag] = tempCS->cost;
+#endif
DTRACE_MODE_COST(*tempCS, m_pcRdCost->getLambda());
xCheckBestMode(tempCS, bestCS, partitioner, encTestMode);
tempCS->initStructData(encTestMode.qp, encTestMode.lossless);
}
+#if !JVET_M0464_UNI_MTS
if (numResidualPass == 0 && (emtCost[0] <= emtCost[1] ? !hasResidual[0] : !hasResidual[1]))
{
// If no residual when allowing for one, then set mark to not try case where residual is forced to 0
candHasNoResidual[mergeCand] = 1;
}
+#endif
if (m_pcEncCfg->getUseFastDecisionForMerge() && !bestIsSkip)
{
@@ -2891,7 +3060,7 @@ void EncCu::xCheckRDCostCPRModeMerge2Nx2N(CodingStructure *&tempCS, CodingStruct
}
-void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode)
+void EncCu::xCheckRDCostIBCMode(CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode)
{
tempCS->initStructData(encTestMode.qp, encTestMode.lossless);
@@ -2907,7 +3076,7 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
cu.transQuantBypass = encTestMode.lossless;
cu.chromaQpAdj = cu.transQuantBypass ? 0 : m_cuChromaQpOffsetIdxPlus1;
cu.qp = encTestMode.qp;
- cu.cpr = true;
+ cu.ibc = true;
cu.imv = 0;
CU::addPUs(cu);
@@ -2916,16 +3085,16 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
cu.mmvdSkip = false;
pu.mmvdMergeFlag = false;
- pu.intraDir[0] = DC_IDX; // set intra pred for cpr block
- pu.intraDir[1] = PLANAR_IDX; // set intra pred for cpr block
+ pu.intraDir[0] = DC_IDX; // set intra pred for ibc block
+ pu.intraDir[1] = PLANAR_IDX; // set intra pred for ibc block
- pu.interDir = 1; // use list 0 for CPR mode
+ pu.interDir = 1; // use list 0 for IBC mode
pu.refIdx[REF_PIC_LIST_0] = pu.cs->slice->getNumRefIdx(REF_PIC_LIST_0) - 1; // last idx in the list
if (partitioner.chType == CHANNEL_TYPE_LUMA)
{
- bool bValid = m_pcInterSearch->predCPRSearch(cu, partitioner, m_ctuCprSearchRangeX, m_ctuCprSearchRangeY, m_cprHashMap);
+ bool bValid = m_pcInterSearch->predIBCSearch(cu, partitioner, m_ctuIbcSearchRangeX, m_ctuIbcSearchRangeY, m_ibcHashMap);
if (bValid)
{
@@ -2934,6 +3103,9 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
// MC
m_pcInterSearch->motionCompensation(pu, REF_PIC_LIST_0, true, chroma);
+#if JVET_M0464_UNI_MTS
+ {
+#else
double bestCost = bestCS->cost;
unsigned char considerEmtSecondPass = 0;
bool skipSecondEmtPass = true;
@@ -2949,13 +3121,16 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
}
tempCS->getCU(tempCS->chType)->emtFlag = emtCuFlag;
+#endif
m_pcInterSearch->encodeResAndCalcRdInterCU(*tempCS, partitioner, false, true, chroma);
+#if !JVET_M0464_UNI_MTS
if (m_pcEncCfg->getFastInterEMT())
{
emtFirstPassCost = (!emtCuFlag) ? tempCS->cost : emtFirstPassCost;
}
+#endif
xEncodeDontSplit(*tempCS, partitioner);
#if ENABLE_QPA_SUB_CTU
@@ -2971,6 +3146,7 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
DTRACE_MODE_COST(*tempCS, m_pcRdCost->getLambda());
xCheckBestMode(tempCS, bestCS, partitioner, encTestMode);
+#if !JVET_M0464_UNI_MTS
//now we check whether the second pass should be skipped or not
if (!emtCuFlag && considerEmtSecondPass)
{
@@ -2999,6 +3175,7 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
tempCS->cost = MAX_DOUBLE;
}
}
+#endif
}
} // bValid
@@ -3009,13 +3186,13 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
tempCS->cost = MAX_DOUBLE;
}
}
- // chroma CU cpr comp
+ // chroma CU ibc comp
else
{
bool success = true;
// chroma tree, reuse luma bv at minimal block level
// enabled search only when each chroma sub-block has a BV from its luma sub-block
- assert(tempCS->getCprLumaCoverage(pu.Cb()) == CPR_LUMA_COVERAGE_FULL);
+ assert(tempCS->getIbcLumaCoverage(pu.Cb()) == IBC_LUMA_COVERAGE_FULL);
// check if each BV for the chroma sub-block is valid
//static const UInt unitArea = MIN_PU_SIZE * MIN_PU_SIZE;
const CompArea lumaArea = CompArea(COMPONENT_Y, pu.chromaFormat, pu.Cb().lumaPos(), recalcSize(pu.chromaFormat, CHANNEL_TYPE_CHROMA, CHANNEL_TYPE_LUMA, pu.Cb().size()));
@@ -3050,7 +3227,7 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
if (success)
{
- //pu.mergeType = MRG_TYPE_CPR;
+ //pu.mergeType = MRG_TYPE_IBC;
m_pcInterSearch->motionCompensation(pu, REF_PIC_LIST_0, false, true); // luma=0, chroma=1
m_pcInterSearch->encodeResAndCalcRdInterCU(*tempCS, partitioner, false, false, true);
@@ -3070,7 +3247,7 @@ void EncCu::xCheckRDCostCPRMode(CodingStructure *&tempCS, CodingStructure *&best
}
}
}
- // check cpr mode in encoder RD
+ // check ibc mode in encoder RD
//////////////////////////////////////////////////////////////////////////////////////////////
void EncCu::xCheckRDCostInter( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode )
@@ -3164,11 +3341,18 @@ void EncCu::xCheckRDCostInter( CodingStructure *&tempCS, CodingStructure *&bestC
}
}
+#if JVET_M0464_UNI_MTS
+ xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, 0
+ , m_pImvTempCS ? m_pImvTempCS[wIdx] : NULL
+ , 0
+ , &equGBiCost
+#else
xEncodeInterResidual( tempCS, bestCS, partitioner, encTestMode, 0
, m_pImvTempCS ? m_pImvTempCS[wIdx] : NULL
, 1
, 0
, &equGBiCost
+#endif
);
if( g_GbiSearchOrder[gbiLoopIdx] == GBI_DEFAULT )
@@ -3291,7 +3475,9 @@ bool EncCu::xCheckRDCostInterIMV( CodingStructure *&tempCS, CodingStructure *&be
}
cu.imv = iIMV > 1 ? 2 : 1;
+#if !JVET_M0464_UNI_MTS
cu.emtFlag = false;
+#endif
bool testGbi;
uint8_t gbiIdx;
@@ -3358,11 +3544,18 @@ bool EncCu::xCheckRDCostInterIMV( CodingStructure *&tempCS, CodingStructure *&be
return false;
}
+#if JVET_M0464_UNI_MTS
+ xEncodeInterResidual( tempCS, bestCS, partitioner, encTestModeBase, 0
+ , NULL
+ , 0
+ , &equGBiCost
+#else
xEncodeInterResidual( tempCS, bestCS, partitioner, encTestModeBase, 0
, NULL
, true
, 0
, &equGBiCost
+#endif
);
tempCS->initStructData(encTestMode.qp, encTestMode.lossless);
@@ -3390,11 +3583,22 @@ bool EncCu::xCheckRDCostInterIMV( CodingStructure *&tempCS, CodingStructure *&be
return true;
}
+#if JVET_M0464_UNI_MTS
+void EncCu::xEncodeInterResidual( CodingStructure *&tempCS
+ , CodingStructure *&bestCS
+ , Partitioner &partitioner
+ , const EncTestMode& encTestMode
+ , int residualPass
+ , CodingStructure* imvCS
+ , bool* bestHasNonResi
+ , double* equGBiCost
+#else
void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode, int residualPass
, CodingStructure* imvCS
, int emtMode
, bool* bestHasNonResi
, double* equGBiCost
+#endif
)
{
if( residualPass == 1 && encTestMode.lossless )
@@ -3405,14 +3609,18 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
CodingUnit* cu = tempCS->getCU( partitioner.chType );
double bestCostInternal = MAX_DOUBLE;
double bestCost = bestCS->cost;
+#if !JVET_M0464_UNI_MTS
const SPS& sps = *tempCS->sps;
const int maxSizeEMT = EMT_INTER_MAX_CU_WITH_QTBT;
+#endif
bool swapped = false; // avoid unwanted data copy
bool reloadCU = false;
+#if !JVET_M0464_UNI_MTS
const bool considerEmtSecondPass = emtMode && sps.getSpsNext().getUseInterEMT() && partitioner.currArea().lwidth() <= maxSizeEMT && partitioner.currArea().lheight() <= maxSizeEMT;
int minEMTMode = 0;
int maxEMTMode = (considerEmtSecondPass?1:0);
+#endif
// Not allow very big |MVd| to avoid CABAC crash caused by too large MVd. Normally no impact on coding performance.
const int maxMvd = 1 << 15;
@@ -3442,12 +3650,14 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
}
}
+#if !JVET_M0464_UNI_MTS
if( emtMode == 2 )
{
minEMTMode = maxEMTMode = (cu->emtFlag?1:0);
}
for( int curEmtMode = minEMTMode; curEmtMode <= maxEMTMode; curEmtMode++ )
+#endif
{
if( reloadCU )
{
@@ -3479,7 +3689,9 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
reloadCU = true; // enable cu reloading
cu->skip = false;
+#if !JVET_M0464_UNI_MTS
cu->emtFlag = curEmtMode;
+#endif
const bool skipResidual = residualPass == 1;
m_pcInterSearch->encodeResAndCalcRdInterCU( *tempCS, partitioner, skipResidual );
@@ -3518,7 +3730,9 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
}
}
+#if !JVET_M0464_UNI_MTS
double emtFirstPassCost = tempCS->cost;
+#endif
if( imvCS && (tempCS->cost < imvCS->cost) )
{
if( imvCS->cost != MAX_DOUBLE )
@@ -3550,6 +3764,7 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
#endif
xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
+#if !JVET_M0464_UNI_MTS
//now we check whether the second pass should be skipped or not
if( !curEmtMode && maxEMTMode )
{
@@ -3563,6 +3778,7 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS, CodingStructure *&be
maxEMTMode = 0; // do not test EMT
}
}
+#endif
} //end emt loop
}
@@ -3571,6 +3787,9 @@ void EncCu::xEncodeDontSplit( CodingStructure &cs, Partitioner &partitioner )
{
m_CABACEstimator->resetBits();
+#if JVET_M0421_SPLIT_SIG
+ m_CABACEstimator->split_cu_mode( CU_DONT_SPLIT, cs, partitioner );
+#else
{
if( partitioner.canSplit( CU_QUAD_SPLIT, cs ) )
{
@@ -3581,6 +3800,7 @@ void EncCu::xEncodeDontSplit( CodingStructure &cs, Partitioner &partitioner )
m_CABACEstimator->split_cu_mode_mt( CU_DONT_SPLIT, cs, partitioner );
}
}
+#endif
cs.fracBits += m_CABACEstimator->getEstFracBits(); // split bits
cs.cost = m_pcRdCost->calcRdCost( cs.fracBits, cs.dist );
@@ -3599,6 +3819,10 @@ void EncCu::xReuseCachedResult( CodingStructure *&tempCS, CodingStructure *&best
if( bestEncCache->setCsFrom( *tempCS, cachedMode, partitioner ) )
{
CodingUnit& cu = *tempCS->cus.front();
+#if JVET_M0170_MRG_SHARELIST
+ cu.shareParentPos = tempCS->sharedBndPos;
+ cu.shareParentSize = tempCS->sharedBndSize;
+#endif
partitioner.setCUData( cu );
if( CU::isIntra( cu ) )
diff --git a/source/Lib/EncoderLib/EncCu.h b/source/Lib/EncoderLib/EncCu.h
index 22e50e7a8aaa90999ca4bfb34be4e85c34369aaa..b239ed8baf3e9351f980d0f2c13265898150f27e 100644
--- a/source/Lib/EncoderLib/EncCu.h
+++ b/source/Lib/EncoderLib/EncCu.h
@@ -45,7 +45,7 @@
#include "CommonLib/TrQuant.h"
#include "CommonLib/Unit.h"
#include "CommonLib/UnitPartitioner.h"
-#include "CommonLib/CprHashMap.h"
+#include "CommonLib/IbcHashMap.h"
#if REUSE_CU_RESULTS
#include "DecoderLib/DecCu.h"
@@ -109,9 +109,17 @@ private:
CABACWriter* m_CABACEstimator;
RateCtrl* m_pcRateCtrl;
- CprHashMap m_cprHashMap;
+ IbcHashMap m_ibcHashMap;
CodingStructure **m_pImvTempCS;
EncModeCtrl *m_modeCtrl;
+#if JVET_M0170_MRG_SHARELIST
+ int m_shareState;
+ uint32_t m_shareBndPosX;
+ uint32_t m_shareBndPosY;
+ SizeType m_shareBndSizeW;
+ SizeType m_shareBndSizeH;
+#endif
+
PelStorage m_acMergeBuffer[MMVD_MRG_MAX_RD_BUF_NUM];
PelStorage m_acRealMergeBuffer[MRG_MAX_NUM_CANDS];
PelStorage m_acTriangleWeightedBuffer[TRIANGLE_MAX_NUM_CANDS]; // to store weighted prediction pixles
@@ -120,8 +128,8 @@ private:
unsigned int m_subMergeBlkSize[10];
unsigned int m_subMergeBlkNum[10];
unsigned int m_prevPOC;
- int m_ctuCprSearchRangeX;
- int m_ctuCprSearchRangeY;
+ int m_ctuIbcSearchRangeX;
+ int m_ctuIbcSearchRangeY;
#if ENABLE_SPLIT_PARALLELISM || ENABLE_WPP_PARALLELISM
EncLib* m_pcEncLib;
#endif
@@ -152,6 +160,10 @@ public:
void setMergeBestSATDCost(double cost) { m_mergeBestSATDCost = cost; }
double getMergeBestSATDCost() { return m_mergeBestSATDCost; }
+#if JVET_M0170_MRG_SHARELIST
+ Position shareParentPos;
+ Size shareParentSize;
+#endif
~EncCu();
protected:
@@ -190,12 +202,24 @@ protected:
void xCheckRDCostMergeTriangle2Nx2N( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &pm, const EncTestMode& encTestMode );
+#if JVET_M0464_UNI_MTS
+ void xEncodeInterResidual( CodingStructure *&tempCS
+ , CodingStructure *&bestCS
+ , Partitioner &partitioner
+ , const EncTestMode& encTestMode
+ , int residualPass = 0
+ , CodingStructure* imvCS = NULL
+ , bool* bestHasNonResi = NULL
+ , double* equGBiCost = NULL
+ );
+#else
void xEncodeInterResidual ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode, int residualPass = 0
, CodingStructure* imvCS = NULL
, int emtMode = 1
, bool* bestHasNonResi = NULL
, double* equGBiCost = NULL
);
+#endif
#if REUSE_CU_RESULTS
void xReuseCachedResult ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &Partitioner );
#endif
@@ -210,8 +234,8 @@ protected:
&& (abs(cu.slice->getPOC() - cu.slice->getRefPOC(REF_PIC_LIST_0, cu.refIdxBi[0])) == 1
|| abs(cu.slice->getPOC() - cu.slice->getRefPOC(REF_PIC_LIST_1, cu.refIdxBi[1])) == 1))));
}
- void xCheckRDCostCPRMode ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &pm, const EncTestMode& encTestMode );
- void xCheckRDCostCPRModeMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode );
+ void xCheckRDCostIBCMode ( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &pm, const EncTestMode& encTestMode );
+ void xCheckRDCostIBCModeMerge2Nx2N( CodingStructure *&tempCS, CodingStructure *&bestCS, Partitioner &partitioner, const EncTestMode& encTestMode );
};
//! \}
diff --git a/source/Lib/EncoderLib/EncGOP.cpp b/source/Lib/EncoderLib/EncGOP.cpp
index 56da75e88d36164dc175e84f936577ccd3bdad2b..0872980a6a75910484f743ffbbd9bd40ef719c8f 100644
--- a/source/Lib/EncoderLib/EncGOP.cpp
+++ b/source/Lib/EncoderLib/EncGOP.cpp
@@ -1456,7 +1456,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
{
pcSlice->setSliceType(I_SLICE);
}
- if (pcSlice->getSliceType() == I_SLICE && pcSlice->getSPS()->getSpsNext().getCPRMode())
+ if (pcSlice->getSliceType() == I_SLICE && pcSlice->getSPS()->getSpsNext().getIBCMode())
{
pcSlice->setSliceType(P_SLICE);
}
@@ -1630,7 +1630,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
if (pcPic->cs->sps->getSpsNext().getUseCompositeRef() && getPrepareLTRef()) {
arrangeCompositeReference(pcSlice, rcListPic, pocCurr);
}
- if (pcSlice->getSPS()->getSpsNext().getCPRMode())
+ if (pcSlice->getSPS()->getSpsNext().getIBCMode())
{
if (m_pcCfg->getIntraPeriod() > 0 && pcSlice->getPOC() % m_pcCfg->getIntraPeriod() == 0)
{
@@ -1696,7 +1696,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
{
pcSlice->setSliceType ( P_SLICE );
}
- if (pcSlice->getSPS()->getSpsNext().getCPRMode() && pcSlice->getNumRefIdx(REF_PIC_LIST_0) == 1)
+ if (pcSlice->getSPS()->getSpsNext().getIBCMode() && pcSlice->getNumRefIdx(REF_PIC_LIST_0) == 1)
{
m_pcSliceEncoder->setEncCABACTableIdx(P_SLICE);
}
@@ -1786,7 +1786,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
// disable TMVP when current picture is the only ref picture
- if (pcSlice->isIRAP() && pcSlice->getSPS()->getSpsNext().getCPRMode())
+ if (pcSlice->isIRAP() && pcSlice->getSPS()->getSpsNext().getIBCMode())
{
pcSlice->setEnableTMVPFlag(0);
}
@@ -1800,7 +1800,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
bool bGPBcheck=false;
if ( pcSlice->getSliceType() == B_SLICE)
{
- if (pcSlice->getSPS()->getSpsNext().getCPRMode())
+ if (pcSlice->getSPS()->getSpsNext().getIBCMode())
{
if (pcSlice->getNumRefIdx(RefPicList(0)) - 1 == pcSlice->getNumRefIdx(RefPicList(1)))
{
@@ -1842,6 +1842,81 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
pcPic->slices[pcSlice->getSliceSegmentIdx()]->setMvdL1ZeroFlag(pcSlice->getMvdL1ZeroFlag());
#endif
+#if JVET_M0444_SMVD
+ if ( pcSlice->getCheckLDC() == false && pcSlice->getMvdL1ZeroFlag() == false )
+ {
+ int currPOC = pcSlice->getPOC();
+
+ int forwardPOC = currPOC;
+ int backwardPOC = currPOC;
+ int ref = 0, refIdx0 = -1, refIdx1 = -1;
+
+ // search nearest forward POC in List 0
+ for ( ref = 0; ref < pcSlice->getNumRefIdx( REF_PIC_LIST_0 ); ref++ )
+ {
+ int poc = pcSlice->getRefPic( REF_PIC_LIST_0, ref )->getPOC();
+ if ( poc < currPOC && (poc > forwardPOC || refIdx0 == -1) )
+ {
+ forwardPOC = poc;
+ refIdx0 = ref;
+ }
+ }
+
+ // search nearest backward POC in List 1
+ for ( ref = 0; ref < pcSlice->getNumRefIdx( REF_PIC_LIST_1 ); ref++ )
+ {
+ int poc = pcSlice->getRefPic( REF_PIC_LIST_1, ref )->getPOC();
+ if ( poc > currPOC && (poc < backwardPOC || refIdx1 == -1) )
+ {
+ backwardPOC = poc;
+ refIdx1 = ref;
+ }
+ }
+
+ if ( !(forwardPOC < currPOC && backwardPOC > currPOC) )
+ {
+ forwardPOC = currPOC;
+ backwardPOC = currPOC;
+ refIdx0 = -1;
+ refIdx1 = -1;
+
+ // search nearest backward POC in List 0
+ for ( ref = 0; ref < pcSlice->getNumRefIdx( REF_PIC_LIST_0 ); ref++ )
+ {
+ int poc = pcSlice->getRefPic( REF_PIC_LIST_0, ref )->getPOC();
+ if ( poc > currPOC && (poc < backwardPOC || refIdx0 == -1) )
+ {
+ backwardPOC = poc;
+ refIdx0 = ref;
+ }
+ }
+
+ // search nearest forward POC in List 1
+ for ( ref = 0; ref < pcSlice->getNumRefIdx( REF_PIC_LIST_1 ); ref++ )
+ {
+ int poc = pcSlice->getRefPic( REF_PIC_LIST_1, ref )->getPOC();
+ if ( poc < currPOC && (poc > forwardPOC || refIdx1 == -1) )
+ {
+ forwardPOC = poc;
+ refIdx1 = ref;
+ }
+ }
+ }
+
+ if ( forwardPOC < currPOC && backwardPOC > currPOC )
+ {
+ pcSlice->setBiDirPred( true, refIdx0, refIdx1 );
+ }
+ else
+ {
+ pcSlice->setBiDirPred( false, -1, -1 );
+ }
+ }
+ else
+ {
+ pcSlice->setBiDirPred( false, -1, -1 );
+ }
+#endif
double lambda = 0.0;
int actualHeadBits = 0;
@@ -1991,14 +2066,14 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
#endif
#endif
- if (pcSlice->getSPS()->getUseSAO())
+ if (pcSlice->getSPS()->getSAOEnabledFlag())
{
pcPic->resizeSAO( numberOfCtusInFrame, 0 );
pcPic->resizeSAO( numberOfCtusInFrame, 1 );
}
// it is used for signalling during CTU mode decision, i.e. before ALF processing
- if( pcSlice->getSPS()->getUseALF() )
+ if( pcSlice->getSPS()->getALFEnabledFlag() )
{
pcPic->resizeAlfCtuEnableFlag( numberOfCtusInFrame );
std::memset( pcSlice->getAlfSliceParam().enabledFlag, false, sizeof( pcSlice->getAlfSliceParam().enabledFlag ) );
@@ -2095,7 +2170,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
pcSlice = pcPic->slices[0];
// SAO parameter estimation using non-deblocked pixels for CTU bottom and right boundary areas
- if( pcSlice->getSPS()->getUseSAO() && m_pcCfg->getSaoCtuBoundary() )
+ if( pcSlice->getSPS()->getSAOEnabledFlag() && m_pcCfg->getSaoCtuBoundary() )
{
m_pcSAO->getPreDBFStatistics( cs );
}
@@ -2121,7 +2196,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
DTRACE_UPDATE( g_trace_ctx, ( std::make_pair( "final", 1 ) ) );
- if( pcSlice->getSPS()->getUseSAO() )
+ if( pcSlice->getSPS()->getSAOEnabledFlag() )
{
bool sliceEnabled[MAX_NUM_COMPONENT];
m_pcSAO->initCABACEstimator( m_pcEncLib->getCABACEncoder(), m_pcEncLib->getCtxCache(), pcSlice );
@@ -2144,7 +2219,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
}
- if( pcSlice->getSPS()->getUseALF() )
+ if( pcSlice->getSPS()->getALFEnabledFlag() )
{
AlfSliceParam alfSliceParam;
m_pcALF->initCABACEstimator( m_pcEncLib->getCABACEncoder(), m_pcEncLib->getCtxCache(), pcSlice );
@@ -2180,7 +2255,7 @@ void EncGOP::compressGOP( int iPOCLast, int iNumPicRcvd, PicList& rcListPic,
}
}
#endif
- if( pcSlice->getSPS()->getUseSAO() )
+ if( pcSlice->getSPS()->getSAOEnabledFlag() )
{
m_pcSAO->disabledRate( *pcPic->cs, pcPic->getSAO(1), m_pcCfg->getSaoEncodingRate(), m_pcCfg->getSaoEncodingRateChroma());
}
@@ -2646,10 +2721,6 @@ void EncGOP::xGetBuffer( PicList& rcListPic,
#ifndef BETA
#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 bitDepth,
@@ -2753,9 +2824,7 @@ uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1,
}
double wmse = 0.0, sumAct = 0.0; // compute activity normalized SNR value
-#if !GLOBAL_AVERAGING
- double numAct = 0.0;
-#endif
+
for (y = 0; y < H; y += B)
{
for (x = 0; x < W; x += B)
@@ -2765,29 +2834,13 @@ uint64_t EncGOP::xFindDistortionPlane(const CPelBuf& pic0, const CPelBuf& pic1,
W, H,
x, y,
B, B);
-#if !GLOBAL_AVERAGING
- numAct += 1.0;
-#endif
}
}
// integer weighted distortion
-#if GLOBAL_AVERAGING
- 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 *= 2.0;
- }
+ sumAct = 16.0 * sqrt ((3840.0 * 2160.0) / double((W << chromaShift) * (H << chromaShift))) * double(1 << BD);
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);
-#endif
}
#endif // ENABLE_QPA
uiTotalDiff = 0;
@@ -2992,12 +3045,7 @@ void EncGOP::xCalculateAddPSNR(Picture* pcPic, PelUnitBuf cPicD, const AccessUni
const bool bPicIsField = pcPic->fieldPic;
const Slice* pcSlice = pcPic->slices[0];
-#if ENABLE_QPA && FRAME_WEIGHTING
- const uint32_t currDQP = (pcSlice->getPOC() % m_pcEncLib->getIntraPeriod()) == 0 ? 0 : DQP[pcSlice->getPOC() % m_pcEncLib->getGOPSize()];
- const double frameWeight = pow(2.0, (double)currDQP / -3.0);
- if (useWPSNR) m_gcAnalyzeAll.addWeight(frameWeight);
-#endif
for (int comp = 0; comp < ::getNumberValidComponents(formatD); comp++)
{
const ComponentID compID = ComponentID(comp);
@@ -3016,11 +3064,10 @@ void EncGOP::xCalculateAddPSNR(Picture* pcPic, PelUnitBuf cPicD, const AccessUni
const uint32_t bitDepth = sps.getBitDepth(toChannelType(compID));
#if ENABLE_QPA
const uint64_t uiSSDtemp = xFindDistortionPlane(recPB, orgPB, useWPSNR ? bitDepth : 0, ::getComponentScaleX(compID, format));
- const uint32_t maxval = /*useWPSNR ? (1 << bitDepth) - 1 :*/ 255 << (bitDepth - 8); // fix with WPSNR: 1023 (4095) instead of 1020 (4080) for bit-depth 10 (12)
#else
const uint64_t uiSSDtemp = xFindDistortionPlane(recPB, orgPB, 0);
- const uint32_t maxval = 255 << (bitDepth - 8);
#endif
+ const uint32_t maxval = 255 << (bitDepth - 8);
const uint32_t size = width * height;
const double fRefValue = (double)maxval * maxval * size;
dPSNR[comp] = uiSSDtemp ? 10.0 * log10(fRefValue / (double)uiSSDtemp) : 999.99;
@@ -3033,10 +3080,6 @@ void EncGOP::xCalculateAddPSNR(Picture* pcPic, PelUnitBuf cPicD, const AccessUni
MSEyuvframeWeighted[comp] = (double)uiSSDtempWeighted / size;
}
#endif
-
-#if ENABLE_QPA && FRAME_WEIGHTING
- if (useWPSNR) m_gcAnalyzeAll.addWeightedSSD(frameWeight * (double)uiSSDtemp / fRefValue, compID);
-#endif
}
#if EXTENSION_360_VIDEO
@@ -3224,13 +3267,7 @@ void EncGOP::xCalculateInterlacedAddPSNR( Picture* pcPicOrgFirstField, Picture*
CHECK(!(acPicRecFields[0].chromaFormat==acPicRecFields[1].chromaFormat), "Unspecified error");
const uint32_t numValidComponents = ::getNumberValidComponents( acPicRecFields[0].chromaFormat );
-#if ENABLE_QPA && FRAME_WEIGHTING
- const Slice* pcSlice = pcPicOrgFirstField->slices[0];
- const uint32_t currDQP = (pcSlice->getPOC() % m_pcEncLib->getIntraPeriod()) == 0 ? 0 : DQP[pcSlice->getPOC() % m_pcEncLib->getGOPSize()];
- const double frameWeight = pow(2.0, (double)currDQP / -3.0);
- if (useWPSNR) m_gcAnalyzeAll_in.addWeight(frameWeight);
-#endif
for (int chan = 0; chan < numValidComponents; chan++)
{
const ComponentID ch=ComponentID(chan);
@@ -3251,18 +3288,11 @@ void EncGOP::xCalculateInterlacedAddPSNR( Picture* pcPicOrgFirstField, Picture*
uiSSDtemp += xFindDistortionPlane( acPicRecFields[fieldNum].get(ch), apcPicOrgFields[fieldNum]->getOrigBuf().get(ch), 0 );
#endif
}
-#if ENABLE_QPA
- const uint32_t maxval = /*useWPSNR ? (1 << bitDepth) - 1 :*/ 255 << (bitDepth - 8); // fix with WPSNR: 1023 (4095) instead of 1020 (4080) for bit-depth 10 (12)
-#else
const uint32_t maxval = 255 << (bitDepth - 8);
-#endif
const uint32_t size = width * height * 2;
const double fRefValue = (double)maxval * maxval * size;
dPSNR[ch] = uiSSDtemp ? 10.0 * log10(fRefValue / (double)uiSSDtemp) : 999.99;
MSEyuvframe[ch] = (double)uiSSDtemp / size;
-#if ENABLE_QPA && FRAME_WEIGHTING
- if (useWPSNR) m_gcAnalyzeAll_in.addWeightedSSD(frameWeight * (double)uiSSDtemp / fRefValue, ch);
-#endif
}
uint32_t uibits = 0; // the number of bits for the pair is not calculated here - instead the overall total is used elsewhere.
diff --git a/source/Lib/EncoderLib/EncLib.cpp b/source/Lib/EncoderLib/EncLib.cpp
index 68fdd2caf1ed82d37dcca64fd22a19d19fb1b3a1..56f49fc8620590b9f275d78e319747df63f7f826 100644
--- a/source/Lib/EncoderLib/EncLib.cpp
+++ b/source/Lib/EncoderLib/EncLib.cpp
@@ -797,7 +797,11 @@ void EncLib::xInitSPS(SPS &sps)
sps.setNoSbtmvpConstraintFlag(m_SubPuMvpMode ? false : true);
sps.setNoAmvrConstraintFlag(!m_bNoAmvrConstraintFlag);
sps.setNoAffineMotionConstraintFlag(!m_Affine);
+#if JVET_M0464_UNI_MTS
+ sps.setNoMtsConstraintFlag((m_IntraMTS || m_InterMTS) ? false : true);
+#else
sps.setNoMtsConstraintFlag((m_IntraEMT || m_InterEMT) ? false : true);
+#endif
sps.setNoLadfConstraintFlag(!m_LadfEnabled);
sps.setNoDepQuantConstraintFlag(!m_DepQuantEnabledFlag);
sps.setNoSignDataHidingConstraintFlag(!m_SignDataHidingEnabledFlag);
@@ -847,20 +851,28 @@ void EncLib::xInitSPS(SPS &sps)
sps.getSpsNext().setUseLargeCTU ( m_LargeCTU );
sps.setMaxBTDepth ( m_uiMaxBTDepth, m_uiMaxBTDepthI, m_uiMaxBTDepthIChroma );
sps.setUseDualITree ( m_dualITree );
- sps.getSpsNext().setSubPuMvpMode ( m_SubPuMvpMode );
+ sps.setSBTMVPEnabledFlag ( m_SubPuMvpMode );
sps.getSpsNext().setImvMode ( ImvMode(m_ImvMode) );
sps.getSpsNext().setUseIMV ( m_ImvMode != IMV_OFF );
- sps.getSpsNext().setUseBIO ( m_BIO );
+ sps.setBDOFEnabledFlag ( m_BIO );
sps.getSpsNext().setUseAffine ( m_Affine );
sps.getSpsNext().setUseAffineType ( m_AffineType );
sps.getSpsNext().setDisableMotCompress ( m_DisableMotionCompression );
sps.getSpsNext().setMTTMode ( m_MTTMode );
sps.getSpsNext().setUseLMChroma ( m_LMChroma ? true : false );
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ sps.getSpsNext().setCclmCollocatedChromaFlag( m_cclmCollocatedChromaFlag );
+#endif
#if ENABLE_WPP_PARALLELISM
sps.getSpsNext().setUseNextDQP ( m_AltDQPCoding );
#endif
+#if JVET_M0464_UNI_MTS
+ sps.getSpsNext().setUseIntraMTS ( m_IntraMTS );
+ sps.getSpsNext().setUseInterMTS ( m_InterMTS );
+#else
sps.getSpsNext().setUseIntraEMT ( m_IntraEMT );
sps.getSpsNext().setUseInterEMT ( m_InterEMT );
+#endif
sps.getSpsNext().setUseCompositeRef ( m_compositeRefEnabled );
sps.getSpsNext().setUseGBi ( m_GBi );
#if LUMA_ADAPTIVE_DEBLOCKING_FILTER_QP_OFFSET
@@ -880,9 +892,9 @@ void EncLib::xInitSPS(SPS &sps)
sps.getSpsNext().setUseMHIntra ( m_MHIntra );
sps.getSpsNext().setUseTriangle ( m_Triangle );
- sps.getSpsNext().setCPRMode ( m_CPRMode );
+ sps.getSpsNext().setIBCMode ( m_IBCMode );
- sps.setUseWrapAround ( m_wrapAround );
+ sps.setWrapAroundEnabledFlag ( m_wrapAround );
sps.setWrapAroundOffset ( m_wrapAroundOffset );
// ADD_NEW_TOOL : (encoder lib) set tool enabling flags and associated parameters here
@@ -897,7 +909,7 @@ void EncLib::xInitSPS(SPS &sps)
sps.setLog2MinCodingBlockSize(log2MinCUSize);
sps.setPCMLog2MinSize (m_uiPCMLog2MinSize);
- sps.setUsePCM ( m_usePCM );
+ sps.setPCMEnabledFlag ( m_usePCM );
sps.setPCMLog2MaxSize( m_pcmLog2MaxSize );
sps.setQuadtreeTULog2MaxSize( m_uiQuadtreeTULog2MaxSize );
@@ -909,8 +921,6 @@ void EncLib::xInitSPS(SPS &sps)
sps.setMaxTrSize ( 1 << m_uiQuadtreeTULog2MaxSize );
- sps.setUseAMP ( m_useAMP );
-
for (uint32_t channelType = 0; channelType < MAX_NUM_CHANNEL_TYPE; channelType++)
{
sps.setBitDepth (ChannelType(channelType), m_bitDepth[channelType] );
@@ -918,7 +928,7 @@ void EncLib::xInitSPS(SPS &sps)
sps.setPCMBitDepth (ChannelType(channelType), m_PCMBitDepth[channelType] );
}
- sps.setUseSAO( m_bUseSAO );
+ sps.setSAOEnabledFlag( m_bUseSAO );
sps.setMaxTLayers( m_maxTempLayer );
sps.setTemporalIdNestingFlag( ( m_maxTempLayer == 1 ) ? true : false );
@@ -936,7 +946,7 @@ void EncLib::xInitSPS(SPS &sps)
#if HEVC_USE_INTRA_SMOOTHING_T32 || HEVC_USE_INTRA_SMOOTHING_T64
sps.setUseStrongIntraSmoothing( m_useStrongIntraSmoothing );
#endif
- sps.setUseALF( m_alf );
+ sps.setALFEnabledFlag( m_alf );
sps.setVuiParametersPresentFlag(getVuiParametersPresentFlag());
if (sps.getVuiParametersPresentFlag())
@@ -1385,7 +1395,7 @@ void EncLib::xInitPPS(PPS &pps, const SPS &sps)
}
}
CHECK(!(bestPos <= 15), "Unspecified error");
- if (sps.getSpsNext().getCPRMode())
+ if (sps.getSpsNext().getIBCMode())
{
pps.setNumRefIdxL0DefaultActive(bestPos + 1);
}
diff --git a/source/Lib/EncoderLib/EncModeCtrl.cpp b/source/Lib/EncoderLib/EncModeCtrl.cpp
index f5032831e5bdf4c74c8e47dabfe943fe56c5e7e4..c9a6863301a9e2c439e7fafdb209a4f9ce99ee2d 100644
--- a/source/Lib/EncoderLib/EncModeCtrl.cpp
+++ b/source/Lib/EncoderLib/EncModeCtrl.cpp
@@ -528,7 +528,11 @@ uint8_t CacheBlkInfoCtrl::getGbiIdx(const UnitArea& area)
}
#if REUSE_CU_RESULTS
-static bool isTheSameNbHood( const CodingUnit &cu, const CodingStructure& cs, const Partitioner &partitioner )
+static bool isTheSameNbHood( const CodingUnit &cu, const CodingStructure& cs, const Partitioner &partitioner
+#if JVET_M0170_MRG_SHARELIST
+ , const PredictionUnit &pu, int picW, int picH
+#endif
+ )
{
if( cu.chType != partitioner.chType )
{
@@ -549,6 +553,32 @@ static bool isTheSameNbHood( const CodingUnit &cu, const CodingStructure& cs, co
const UnitArea &cmnAnc = ps[i - 1].parts[ps[i - 1].idx];
const UnitArea cuArea = CS::getArea( cs, cu, partitioner.chType );
+#if JVET_M0170_MRG_SHARELIST
+ bool sharedListReuseMode = true;
+ if(
+ pu.mergeFlag == true &&
+ cu.affine == false &&
+ cu.predMode == MODE_INTER
+ )
+ {
+ sharedListReuseMode = false;
+
+ if ((cu.lumaSize().width*cu.lumaSize().height) >= MRG_SHARELIST_SHARSIZE)
+ {
+ sharedListReuseMode = true;
+ }
+
+ if (((cmnAnc.lumaSize().width)*(cmnAnc.lumaSize().height) <= MRG_SHARELIST_SHARSIZE))
+ {
+ sharedListReuseMode = true;
+ }
+ }
+ else
+ {
+ sharedListReuseMode = true;
+ }
+//#endif
+#endif
for( int i = 0; i < cmnAnc.blocks.size(); i++ )
{
@@ -557,6 +587,13 @@ static bool isTheSameNbHood( const CodingUnit &cu, const CodingStructure& cs, co
return false;
}
}
+#if JVET_M0170_MRG_SHARELIST
+ if(!sharedListReuseMode)
+ {
+ return false;
+ }
+#endif
+
return true;
}
@@ -754,8 +791,12 @@ bool BestEncInfoCache::isValid( const CodingStructure& cs, const Partitioner& pa
if( encInfo.cu.qp != qp )
return false;
- if( cs.picture->poc != encInfo.poc || CS::getArea( cs, cs.area, partitioner.chType ) != CS::getArea( cs, encInfo.cu, partitioner.chType ) || !isTheSameNbHood( encInfo.cu, cs, partitioner )
- || encInfo.cu.cpr
+ if( cs.picture->poc != encInfo.poc || CS::getArea( cs, cs.area, partitioner.chType ) != CS::getArea( cs, encInfo.cu, partitioner.chType ) || !isTheSameNbHood( encInfo.cu, cs, partitioner
+#if JVET_M0170_MRG_SHARELIST
+ , encInfo.pu, (cs.picture->Y().width), (cs.picture->Y().height)
+#endif
+)
+ || encInfo.cu.ibc
)
{
return false;
@@ -773,7 +814,11 @@ bool BestEncInfoCache::setCsFrom( CodingStructure& cs, EncTestMode& testMode, co
BestEncodingInfo& encInfo = *m_bestEncInfo[idx1][idx2][idx3][idx4];
- if( cs.picture->poc != encInfo.poc || CS::getArea( cs, cs.area, partitioner.chType ) != CS::getArea( cs, encInfo.cu, partitioner.chType ) || !isTheSameNbHood( encInfo.cu, cs, partitioner ) )
+ if( cs.picture->poc != encInfo.poc || CS::getArea( cs, cs.area, partitioner.chType ) != CS::getArea( cs, encInfo.cu, partitioner.chType ) || !isTheSameNbHood( encInfo.cu, cs, partitioner
+#if JVET_M0170_MRG_SHARELIST
+ , encInfo.pu, (cs.picture->Y().width), (cs.picture->Y().height)
+#endif
+) )
{
return false;
}
@@ -938,7 +983,7 @@ void EncModeCtrlMTnoRQT::initCULevel( Partitioner &partitioner, const CodingStru
{
if (m_pcEncCfg->getUseAdaptiveQP())
{
- baseQP = Clip3 (-cs.sps->getQpBDOffset (CHANNEL_TYPE_LUMA), MAX_QP, baseQP + xComputeDQP (cs, partitioner));
+ baseQP = Clip3(-cs.sps->getQpBDOffset(CHANNEL_TYPE_LUMA), MAX_QP, baseQP + xComputeDQP (cs, partitioner));
}
#if ENABLE_QPA_SUB_CTU
else if (m_pcEncCfg->getUsePerceptQPA() && !m_pcEncCfg->getUseRateCtrl() && cs.pps->getUseDQP() && cs.pps->getMaxCuDQPDepth() > 0)
@@ -968,25 +1013,25 @@ void EncModeCtrlMTnoRQT::initCULevel( Partitioner &partitioner, const CodingStru
int maxQP = baseQP;
xGetMinMaxQP( minQP, maxQP, cs, partitioner, baseQP, *cs.sps, *cs.pps, true );
- bool checkCpr = true;
+ bool checkIbc = true;
if (cs.chType == CHANNEL_TYPE_CHROMA)
{
- CprLumaCoverage cprLumaCoverage = cs.getCprLumaCoverage(cs.area.Cb());
- switch (cprLumaCoverage)
+ IbcLumaCoverage ibcLumaCoverage = cs.getIbcLumaCoverage(cs.area.Cb());
+ switch (ibcLumaCoverage)
{
- case CPR_LUMA_COVERAGE_FULL:
- // check CPR
+ case IBC_LUMA_COVERAGE_FULL:
+ // check IBC
break;
- case CPR_LUMA_COVERAGE_PARTIAL:
- // do not check CPR
- checkCpr = false;
+ case IBC_LUMA_COVERAGE_PARTIAL:
+ // do not check IBC
+ checkIbc = false;
break;
- case CPR_LUMA_COVERAGE_NONE:
- // do not check CPR
- checkCpr = false;
+ case IBC_LUMA_COVERAGE_NONE:
+ // do not check IBC
+ checkIbc = false;
break;
default:
- THROW("Unknown CPR luma coverage type");
+ THROW("Unknown IBC luma coverage type");
}
}
// Add coding modes here
@@ -1093,13 +1138,13 @@ void EncModeCtrlMTnoRQT::initCULevel( Partitioner &partitioner, const CodingStru
// add intra modes
m_ComprCUCtxList.back().testModes.push_back( { ETM_IPCM, ETO_STANDARD, qp, lossless } );
m_ComprCUCtxList.back().testModes.push_back( { ETM_INTRA, ETO_STANDARD, qp, lossless } );
- // add cpr mode to intra path
- if (cs.sps->getSpsNext().getCPRMode() && checkCpr )
+ // add ibc mode to intra path
+ if (cs.sps->getSpsNext().getIBCMode() && checkIbc )
{
- m_ComprCUCtxList.back().testModes.push_back({ ETM_CPR, ETO_STANDARD, qp, lossless });
+ m_ComprCUCtxList.back().testModes.push_back({ ETM_IBC, ETO_STANDARD, qp, lossless });
if (cs.chType == CHANNEL_TYPE_LUMA)
{
- m_ComprCUCtxList.back().testModes.push_back({ ETM_CPR_MERGE, ETO_STANDARD, qp, lossless });
+ m_ComprCUCtxList.back().testModes.push_back({ ETM_IBC_MERGE, ETO_STANDARD, qp, lossless });
}
}
}
@@ -1129,7 +1174,7 @@ void EncModeCtrlMTnoRQT::initCULevel( Partitioner &partitioner, const CodingStru
m_ComprCUCtxList.back().testModes.push_back( { ETM_MERGE_TRIANGLE, ETO_STANDARD, qp, lossless } );
}
m_ComprCUCtxList.back().testModes.push_back( { ETM_MERGE_SKIP, ETO_STANDARD, qp, lossless } );
- if ( cs.sps->getSpsNext().getUseAffine() || cs.sps->getSpsNext().getUseSubPuMvp() )
+ if ( cs.sps->getSpsNext().getUseAffine() || cs.sps->getSBTMVPEnabledFlag() )
{
m_ComprCUCtxList.back().testModes.push_back( { ETM_AFFINE, ETO_STANDARD, qp, lossless } );
}
@@ -1143,7 +1188,7 @@ void EncModeCtrlMTnoRQT::initCULevel( Partitioner &partitioner, const CodingStru
m_ComprCUCtxList.back().testModes.push_back( { ETM_MERGE_TRIANGLE, ETO_STANDARD, qp, lossless } );
}
m_ComprCUCtxList.back().testModes.push_back( { ETM_MERGE_SKIP, ETO_STANDARD, qp, lossless } );
- if ( cs.sps->getSpsNext().getUseAffine() || cs.sps->getSpsNext().getUseSubPuMvp() )
+ if ( cs.sps->getSpsNext().getUseAffine() || cs.sps->getSBTMVPEnabledFlag() )
{
m_ComprCUCtxList.back().testModes.push_back( { ETM_AFFINE, ETO_STANDARD, qp, lossless } );
}
@@ -1258,7 +1303,7 @@ bool EncModeCtrlMTnoRQT::tryMode( const EncTestMode& encTestmode, const CodingSt
}
// INTRA MODES
- if (cs.sps->getSpsNext().getCPRMode() && !cuECtx.bestTU)
+ if (cs.sps->getSpsNext().getIBCMode() && !cuECtx.bestTU)
return true;
CHECK( !slice.isIntra() && !cuECtx.bestTU, "No possible non-intra encoding for a P- or B-slice found" );
@@ -1271,8 +1316,8 @@ bool EncModeCtrlMTnoRQT::tryMode( const EncTestMode& encTestmode, const CodingSt
{
return false;
}
- if ((m_pcEncCfg->getCPRFastMethod() & CPR_FAST_METHOD_NOINTRA_CPRCBF0)
- && (bestMode.type == ETM_CPR || bestMode.type == ETM_CPR_MERGE)
+ if ((m_pcEncCfg->getIBCFastMethod() & IBC_FAST_METHOD_NOINTRA_IBCCBF0)
+ && (bestMode.type == ETM_IBC || bestMode.type == ETM_IBC_MERGE)
&& (!cuECtx.bestCU->Y().valid() || cuECtx.bestTU->cbf[0] == 0)
&& (!cuECtx.bestCU->Cb().valid() || cuECtx.bestTU->cbf[1] == 0)
&& (!cuECtx.bestCU->Cr().valid() || cuECtx.bestTU->cbf[2] == 0))
@@ -1311,12 +1356,12 @@ bool EncModeCtrlMTnoRQT::tryMode( const EncTestMode& encTestmode, const CodingSt
}
// PCM MODES
- return sps.getUsePCM() && width <= ( 1 << sps.getPCMLog2MaxSize() ) && width >= ( 1 << sps.getPCMLog2MinSize() );
+ return sps.getPCMEnabledFlag() && width <= ( 1 << sps.getPCMLog2MaxSize() ) && width >= ( 1 << sps.getPCMLog2MinSize() );
}
- else if (encTestmode.type == ETM_CPR || encTestmode.type == ETM_CPR_MERGE)
+ else if (encTestmode.type == ETM_IBC || encTestmode.type == ETM_IBC_MERGE)
{
- // CPR MODES
- return sps.getSpsNext().getCPRMode() && width <= CPR_MAX_CAND_SIZE && partitioner.currArea().lumaSize().height <= CPR_MAX_CAND_SIZE;
+ // IBC MODES
+ return sps.getSpsNext().getIBCMode() && width <= IBC_MAX_CAND_SIZE && partitioner.currArea().lumaSize().height <= IBC_MAX_CAND_SIZE;
}
else if( isModeInter( encTestmode ) )
{
@@ -1623,6 +1668,7 @@ bool EncModeCtrlMTnoRQT::useModeResult( const EncTestMode& encTestmode, CodingSt
{
cuECtx.set( BEST_TRIV_SPLIT_COST, tempCS->cost );
}
+#if !JVET_M0464_UNI_MTS
else if( encTestmode.type == ETM_INTRA )
{
const CodingUnit cu = *tempCS->getCU( partitioner.chType );
@@ -1632,6 +1678,7 @@ bool EncModeCtrlMTnoRQT::useModeResult( const EncTestMode& encTestmode, CodingSt
cuECtx.bestEmtSize2Nx2N1stPass = tempCS->cost;
}
}
+#endif
if( m_pcEncCfg->getIMV4PelFast() && m_pcEncCfg->getIMV() && encTestmode.type == ETM_INTER_ME )
{
diff --git a/source/Lib/EncoderLib/EncModeCtrl.h b/source/Lib/EncoderLib/EncModeCtrl.h
index b320d6e036a179418c9edd12ca62ee28e30d78a7..422e4c80840c1bd6b3960c1b1abe75129e6c0249 100644
--- a/source/Lib/EncoderLib/EncModeCtrl.h
+++ b/source/Lib/EncoderLib/EncModeCtrl.h
@@ -70,8 +70,8 @@ enum EncTestModeType
ETM_RECO_CACHED,
#endif
ETM_TRIGGER_IMV_LIST,
- ETM_CPR, // cpr mode
- ETM_CPR_MERGE, // cpr merge mode
+ ETM_IBC, // ibc mode
+ ETM_IBC_MERGE, // ibc merge mode
ETM_INVALID
};
@@ -184,10 +184,12 @@ struct ComprCUCtx
, extraFeatures ( )
, extraFeaturesd( )
, bestInterCost ( MAX_DOUBLE )
+#if !JVET_M0464_UNI_MTS
, bestEmtSize2Nx2N1stPass
( MAX_DOUBLE )
, skipSecondEMTPass
( false )
+#endif
, interHad (std::numeric_limits<Distortion>::max())
#if ENABLE_SPLIT_PARALLELISM
, isLevelSplitParallel
@@ -216,8 +218,10 @@ struct ComprCUCtx
static_vector<int64_t, 30> extraFeatures;
static_vector<double, 30> extraFeaturesd;
double bestInterCost;
+#if !JVET_M0464_UNI_MTS
double bestEmtSize2Nx2N1stPass;
bool skipSecondEMTPass;
+#endif
Distortion interHad;
#if ENABLE_SPLIT_PARALLELISM
bool isLevelSplitParallel;
@@ -298,9 +302,11 @@ public:
double getBestInterCost () const { return m_ComprCUCtxList.back().bestInterCost; }
Distortion getInterHad () const { return m_ComprCUCtxList.back().interHad; }
void enforceInterHad ( Distortion had ) { m_ComprCUCtxList.back().interHad = had; }
+#if !JVET_M0464_UNI_MTS
double getEmtSize2Nx2NFirstPassCost () const { return m_ComprCUCtxList.back().bestEmtSize2Nx2N1stPass; }
bool getSkipSecondEMTPass () const { return m_ComprCUCtxList.back().skipSecondEMTPass; }
void setSkipSecondEMTPass ( bool b ) { m_ComprCUCtxList.back().skipSecondEMTPass = b; }
+#endif
protected:
void xExtractFeatures ( const EncTestMode encTestmode, CodingStructure& cs );
diff --git a/source/Lib/EncoderLib/EncSlice.cpp b/source/Lib/EncoderLib/EncSlice.cpp
index 1fa2f008aaaf0a19e12e575e3be2ea5909869a9e..2843c2696bea129ad063b6216fcce17a90eed295 100644
--- a/source/Lib/EncoderLib/EncSlice.cpp
+++ b/source/Lib/EncoderLib/EncSlice.cpp
@@ -193,18 +193,9 @@ static void filterAndCalculateAverageEnergies (const Pel* pSrc, const int iSrcS
#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
+ const double hpEnerPic = 16.0 * sqrt ((3840.0 * 2160.0) / double(picOrig.width * picOrig.height)) * double(1 << uBitDepth);
- 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;
+ return sqrt (hpEnerPic); // square-root of a_pic value
}
#endif
diff --git a/source/Lib/EncoderLib/InterSearch.cpp b/source/Lib/EncoderLib/InterSearch.cpp
index 1a30af140b175a9f84bc6b4f6db7eed7976b3e51..cad55ecf97b431ed446a4f85cea1a22aa389b9a3 100644
--- a/source/Lib/EncoderLib/InterSearch.cpp
+++ b/source/Lib/EncoderLib/InterSearch.cpp
@@ -695,9 +695,9 @@ Distortion InterSearch::xGetInterPredictionError( PredictionUnit& pu, PelUnitBuf
return (Distortion)cDistParam.distFunc( cDistParam );
}
-/// add cpr search functions here
+/// add ibc search functions here
-void InterSearch::xCPRSearchMVCandUpdate(Distortion sad, int x, int y, Distortion* sadBestCand, Mv* cMVCand)
+void InterSearch::xIBCSearchMVCandUpdate(Distortion sad, int x, int y, Distortion* sadBestCand, Mv* cMVCand)
{
int j = CHROMA_REFINEMENT_CANDIDATES - 1;
@@ -720,7 +720,7 @@ void InterSearch::xCPRSearchMVCandUpdate(Distortion sad, int x, int y, Distorti
}
}
-int InterSearch::xCPRSearchMVChromaRefine(PredictionUnit& pu,
+int InterSearch::xIBCSearchMVChromaRefine(PredictionUnit& pu,
int roiWidth,
int roiHeight,
int cuPelX,
@@ -814,7 +814,7 @@ int InterSearch::xCPRSearchMVChromaRefine(PredictionUnit& pu,
static unsigned int xMergeCandLists(Mv *dst, unsigned int dn, Mv *src, unsigned int sn)
{
- for (unsigned int cand = 0; cand < sn && dn<CPR_NUM_CANDIDATES; cand++)
+ for (unsigned int cand = 0; cand < sn && dn<IBC_NUM_CANDIDATES; cand++)
{
bool found = false;
for (int j = 0; j<dn; j++)
@@ -897,7 +897,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
Mv cMvPredEncOnly[16];
int nbPreds = 0;
- PU::getCprMVPsEncOnly(pu, cMvPredEncOnly, nbPreds);
+ PU::getIbcMVPsEncOnly(pu, cMvPredEncOnly, nbPreds);
m_numBVs = xMergeCandLists(m_acBVs, m_numBVs, cMvPredEncOnly, nbPreds);
for (unsigned int cand = 0; cand < m_numBVs; cand++)
@@ -917,7 +917,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
m_cDistParam.cur.buf = piRefSrch + cStruct.iRefStride * yPred + xPred;
sad += m_cDistParam.distFunc(m_cDistParam);
- xCPRSearchMVCandUpdate(sad, xPred, yPred, sadBestCand, cMVCand);
+ xIBCSearchMVCandUpdate(sad, xPred, yPred, sadBestCand, cMVCand);
}
}
}
@@ -939,7 +939,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
m_cDistParam.cur.buf = piRefSrch + cStruct.iRefStride * y;
sad += m_cDistParam.distFunc(m_cDistParam);
- xCPRSearchMVCandUpdate(sad, 0, y, sadBestCand, cMVCand);
+ xIBCSearchMVCandUpdate(sad, 0, y, sadBestCand, cMVCand);
tempSadBest = sadBestCand[0];
if (sadBestCand[0] <= 3)
{
@@ -965,7 +965,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
sad += m_cDistParam.distFunc(m_cDistParam);
- xCPRSearchMVCandUpdate(sad, x, 0, sadBestCand, cMVCand);
+ xIBCSearchMVCandUpdate(sad, x, 0, sadBestCand, cMVCand);
tempSadBest = sadBestCand[0];
if (sadBestCand[0] <= 3)
{
@@ -984,7 +984,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
if ((!bestX && !bestY) || (sadBest - m_pcRdCost->getBvCostMultiplePreds(bestX, bestY, pu.cs->sps->getSpsNext().getImvMode() == IMV_4PEL) <= 32))
{
//chroma refine
- bestCandIdx = xCPRSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
+ bestCandIdx = xIBCSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
bestX = cMVCand[bestCandIdx].getHor();
bestY = cMVCand[bestCandIdx].getVer();
sadBest = sadBestCand[bestCandIdx];
@@ -1015,7 +1015,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
m_cDistParam.cur.buf = piRefSrch + cStruct.iRefStride * y + x;
sad += m_cDistParam.distFunc(m_cDistParam);
- xCPRSearchMVCandUpdate(sad, x, y, sadBestCand, cMVCand);
+ xIBCSearchMVCandUpdate(sad, x, y, sadBestCand, cMVCand);
}
}
@@ -1025,7 +1025,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
if (sadBest - m_pcRdCost->getBvCostMultiplePreds(bestX, bestY, pu.cs->sps->getSpsNext().getImvMode() == IMV_4PEL) <= 16)
{
//chroma refine
- bestCandIdx = xCPRSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
+ bestCandIdx = xIBCSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
bestX = cMVCand[bestCandIdx].getHor();
bestY = cMVCand[bestCandIdx].getVer();
@@ -1056,11 +1056,11 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
sad += m_cDistParam.distFunc(m_cDistParam);
- xCPRSearchMVCandUpdate(sad, x, y, sadBestCand, cMVCand);
+ xIBCSearchMVCandUpdate(sad, x, y, sadBestCand, cMVCand);
if (sadBestCand[0] <= 5)
{
//chroma refine & return
- bestCandIdx = xCPRSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
+ bestCandIdx = xIBCSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
bestX = cMVCand[bestCandIdx].getHor();
bestY = cMVCand[bestCandIdx].getVer();
sadBest = sadBestCand[bestCandIdx];
@@ -1078,7 +1078,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
if ((sadBest >= tempSadBest) || ((sadBest - m_pcRdCost->getBvCostMultiplePreds(bestX, bestY, pu.cs->sps->getSpsNext().getImvMode() == IMV_4PEL)) <= 32))
{
//chroma refine
- bestCandIdx = xCPRSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
+ bestCandIdx = xIBCSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
bestX = cMVCand[bestCandIdx].getHor();
bestY = cMVCand[bestCandIdx].getVer();
sadBest = sadBestCand[bestCandIdx];
@@ -1113,11 +1113,11 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
sad += m_cDistParam.distFunc(m_cDistParam);
- xCPRSearchMVCandUpdate(sad, x, y, sadBestCand, cMVCand);
+ xIBCSearchMVCandUpdate(sad, x, y, sadBestCand, cMVCand);
if (sadBestCand[0] <= 5)
{
//chroma refine & return
- bestCandIdx = xCPRSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
+ bestCandIdx = xIBCSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
bestX = cMVCand[bestCandIdx].getHor();
bestY = cMVCand[bestCandIdx].getVer();
sadBest = sadBestCand[bestCandIdx];
@@ -1130,7 +1130,7 @@ void InterSearch::xIntraPatternSearch(PredictionUnit& pu, IntTZSearchStruct& cS
}
}
- bestCandIdx = xCPRSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
+ bestCandIdx = xIBCSearchMVChromaRefine(pu, roiWidth, roiHeight, cuPelX, cuPelY, sadBestCand, cMVCand);
bestX = cMVCand[bestCandIdx].getHor();
bestY = cMVCand[bestCandIdx].getVer();
@@ -1155,14 +1155,14 @@ end:
// based on xMotionEstimation
-void InterSearch::xCPREstimation(PredictionUnit& pu, PelUnitBuf& origBuf,
+void InterSearch::xIBCEstimation(PredictionUnit& pu, PelUnitBuf& origBuf,
Mv *pcMvPred,
Mv &rcMv,
Distortion &ruiCost, const int localSearchRangeX, const int localSearchRangeY
)
{
bool buffered = false;
- if (m_pcEncCfg->getCPRFastMethod() & CPR_FAST_METHOD_BUFFERBV)
+ if (m_pcEncCfg->getIBCFastMethod() & IBC_FAST_METHOD_BUFFERBV)
{
ruiCost = MAX_UINT;
const int iPicWidth = pu.cs->slice->getSPS()->getPicWidthInLumaSamples();
@@ -1280,10 +1280,10 @@ void InterSearch::xSetIntraSearchRange(PredictionUnit& pu, int iRoiWidth, int iR
rcMvSrchRngRB >>= 2;
}
-bool InterSearch::predCPRSearch(CodingUnit& cu, Partitioner& partitioner, const int localSearchRangeX, const int localSearchRangeY, CprHashMap& cprHashMap)
+bool InterSearch::predIBCSearch(CodingUnit& cu, Partitioner& partitioner, const int localSearchRangeX, const int localSearchRangeY, IbcHashMap& ibcHashMap)
{
- // check only no greater than CPR_MAX_CAND_SIZE
- if (cu.Y().width > CPR_MAX_CAND_SIZE || cu.Y().height > CPR_MAX_CAND_SIZE)
+ // check only no greater than IBC_MAX_CAND_SIZE
+ if (cu.Y().width > IBC_MAX_CAND_SIZE || cu.Y().height > IBC_MAX_CAND_SIZE)
return false;
Mv cMvSrchRngLT;
Mv cMvSrchRngRB;
@@ -1297,7 +1297,7 @@ bool InterSearch::predCPRSearch(CodingUnit& cu, Partitioner& partitioner, const
CHECK(pu.cu != &cu, "PU is contained in another CU");
//////////////////////////////////////////////////////////
- /// cpr search
+ /// ibc search
pu.cu->imv = 2;
AMVPInfo amvpInfo4Pel;
PU::fillMvpCand(pu, REF_PIC_LIST_0, pu.refIdx[REF_PIC_LIST_0], amvpInfo4Pel);
@@ -1315,23 +1315,23 @@ bool InterSearch::predCPRSearch(CodingUnit& cu, Partitioner& partitioner, const
cMv.setZero();
Distortion cost = 0;
- if (m_pcEncCfg->getCPRHashSearch())
+ if (m_pcEncCfg->getIBCHashSearch())
{
- xxCPRHashSearch(pu, cMvPred, iBvpNum, cMv, bvpIdxBest, cprHashMap);
+ xxIBCHashSearch(pu, cMvPred, iBvpNum, cMv, bvpIdxBest, ibcHashMap);
}
if (cMv.getHor() == 0 && cMv.getVer() == 0)
{
// if hash search does not work or is not enabled
PelUnitBuf origBuf = pu.cs->getOrgBuf(pu);
- xCPREstimation(pu, origBuf, cMvPred, cMv, cost, localSearchRangeX, localSearchRangeY);
+ xIBCEstimation(pu, origBuf, cMvPred, cMv, cost, localSearchRangeX, localSearchRangeY);
}
if (cMv.getHor() == 0 && cMv.getVer() == 0)
{
return false;
}
- /// cpr search
+ /// ibc search
/////////////////////////////////////////////////////////
unsigned int bitsBVPBest, bitsBVPTemp;
bitsBVPBest = MAX_INT;
@@ -1412,13 +1412,13 @@ bool InterSearch::predCPRSearch(CodingUnit& cu, Partitioner& partitioner, const
return true;
}
-void InterSearch::xxCPRHashSearch(PredictionUnit& pu, Mv* mvPred, int numMvPred, Mv &mv, int& idxMvPred, CprHashMap& cprHashMap)
+void InterSearch::xxIBCHashSearch(PredictionUnit& pu, Mv* mvPred, int numMvPred, Mv &mv, int& idxMvPred, IbcHashMap& ibcHashMap)
{
mv.setZero();
m_pcRdCost->setCostScale(0);
std::vector<Position> candPos;
- if (cprHashMap.cprHashMatch(pu.Y(), candPos, *pu.cs, m_pcEncCfg->getCPRHashSearchMaxCand(), m_pcEncCfg->getCPRHashSearchRange4SmallBlk()))
+ if (ibcHashMap.ibcHashMatch(pu.Y(), candPos, *pu.cs, m_pcEncCfg->getIBCHashSearchMaxCand(), m_pcEncCfg->getIBCHashSearchRange4SmallBlk()))
{
unsigned int minCost = MAX_UINT;
@@ -1523,6 +1523,10 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
uint32_t uiLastModeTemp = 0;
int iRefStart, iRefEnd;
+#if JVET_M0444_SMVD
+ int symMode = 0;
+#endif
+
int bestBiPRefIdxL1 = 0;
int bestBiPMvpL1 = 0;
Distortion biPDistTemp = std::numeric_limits<Distortion>::max();
@@ -1543,7 +1547,7 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
CHECK(pu.cu != &cu, "PU is contained in another CU");
- if (cu.cs->sps->getSpsNext().getUseSubPuMvp())
+ if (cu.cs->sps->getSBTMVPEnabledFlag())
{
Size bufSize = g_miScaling.scale(pu.lumaSize());
mergeCtx.subPuMvpMiBuf = MotionBuf(m_SubPuMiBuf, bufSize);
@@ -1584,7 +1588,7 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
{
RefPicList eRefPicList = ( iRefList ? REF_PIC_LIST_1 : REF_PIC_LIST_0 );
int refPicNumber = cs.slice->getNumRefIdx(eRefPicList);
- if (cs.slice->getSPS()->getSpsNext().getCPRMode() && eRefPicList == REF_PIC_LIST_0)
+ if (cs.slice->getSPS()->getSpsNext().getIBCMode() && eRefPicList == REF_PIC_LIST_0)
{
refPicNumber--;
}
@@ -1790,7 +1794,7 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
iRefStart = 0;
iRefEnd = cs.slice->getNumRefIdx(eRefPicList)-1;
- if (cs.slice->getSPS()->getSpsNext().getCPRMode() && eRefPicList == REF_PIC_LIST_0)
+ if (cs.slice->getSPS()->getSpsNext().getIBCMode() && eRefPicList == REF_PIC_LIST_0)
{
iRefEnd--;
}
@@ -1813,7 +1817,12 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
}
}
uiBitsTemp += m_auiMVPIdxCost[aaiMvpIdxBi[iRefList][iRefIdxTemp]][AMVP_MAX_NUM_CANDS];
-
+#if JVET_M0444_SMVD
+ if ( cs.slice->getBiDirPred() )
+ {
+ uiBitsTemp += 1; // add one bit for symmetrical MVD mode
+ }
+#endif
// call ME
xCopyAMVPInfo(&aacAMVPInfo[iRefList][iRefIdxTemp], &amvp[eRefPicList] );
xMotionEstimation ( pu, origBuf, eRefPicList, cMvPredBi[iRefList][iRefIdxTemp], iRefIdxTemp, cMvTemp[iRefList][iRefIdxTemp], aaiMvpIdxBi[iRefList][iRefIdxTemp], uiBitsTemp, uiCostTemp, amvp[eRefPicList], true );
@@ -1860,6 +1869,113 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
} // for loop-iter
cu.refIdxBi[0] = iRefIdxBi[0];
cu.refIdxBi[1] = iRefIdxBi[1];
+
+#if JVET_M0444_SMVD
+ if ( cs.slice->getBiDirPred() )
+ {
+ Distortion symCost;
+ Mv cMvPredSym[2];
+ int mvpIdxSym[2];
+
+ int curRefList = REF_PIC_LIST_0;
+ int tarRefList = 1 - curRefList;
+ RefPicList eCurRefList = (curRefList ? REF_PIC_LIST_1 : REF_PIC_LIST_0);
+ int refIdxCur = cs.slice->getSymRefIdx( curRefList );
+ int refIdxTar = cs.slice->getSymRefIdx( tarRefList );
+
+ MvField cCurMvField, cTarMvField;
+ Distortion costStart = std::numeric_limits<Distortion>::max();
+ for ( int i = 0; i < aacAMVPInfo[curRefList][refIdxCur].numCand; i++ )
+ {
+ for ( int j = 0; j < aacAMVPInfo[tarRefList][refIdxTar].numCand; j++ )
+ {
+ cCurMvField.setMvField( aacAMVPInfo[curRefList][refIdxCur].mvCand[i], refIdxCur );
+ cTarMvField.setMvField( aacAMVPInfo[tarRefList][refIdxTar].mvCand[j], refIdxTar );
+ Distortion cost = xGetSymmetricCost( pu, origBuf, eCurRefList, cCurMvField, cTarMvField, gbiIdx );
+ if ( cost < costStart )
+ {
+ costStart = cost;
+ cMvPredSym[curRefList] = aacAMVPInfo[curRefList][refIdxCur].mvCand[i];
+ cMvPredSym[tarRefList] = aacAMVPInfo[tarRefList][refIdxTar].mvCand[j];
+ mvpIdxSym[curRefList] = i;
+ mvpIdxSym[tarRefList] = j;
+ }
+ }
+ }
+ cCurMvField.mv = cMvPredSym[curRefList];
+ cTarMvField.mv = cMvPredSym[tarRefList];
+
+ m_pcRdCost->setCostScale(0);
+ m_pcRdCost->setPredictor(cMvPredSym[curRefList]);
+ uint32_t bits = m_pcRdCost->getBitsOfVectorWithPredictor(cCurMvField.mv.hor, cCurMvField.mv.ver, (pu.cu->imv << 1));
+ bits += m_auiMVPIdxCost[mvpIdxSym[curRefList]][AMVP_MAX_NUM_CANDS];
+ bits += m_auiMVPIdxCost[mvpIdxSym[tarRefList]][AMVP_MAX_NUM_CANDS];
+ costStart += m_pcRdCost->getCost(bits);
+
+ std::vector<Mv> symmvdCands;
+ symmvdCands.push_back(cMvTemp[curRefList][refIdxCur]);
+ if (iRefIdxBi[curRefList] == refIdxCur && cMvBi[curRefList] != cMvTemp[curRefList][refIdxCur])
+ {
+ symmvdCands.push_back(cMvBi[curRefList]);
+ }
+
+ for (auto mvStart : symmvdCands)
+ {
+ bool checked = false; //if it has been checkin in the mvPred.
+ for (int i = 0; i < aacAMVPInfo[curRefList][refIdxCur].numCand && !checked; i++)
+ {
+ checked |= (mvStart == aacAMVPInfo[curRefList][refIdxCur].mvCand[i]);
+ }
+ if (checked)
+ break;
+
+ Distortion bestCost = costStart;
+ symmvdCheckBestMvp(pu, origBuf, mvStart, (RefPicList)curRefList, aacAMVPInfo, gbiIdx, cMvPredSym, mvpIdxSym, costStart);
+ if (costStart < bestCost)
+ {
+ cCurMvField.setMvField(mvStart, refIdxCur);
+ cTarMvField.setMvField(mvStart.getSymmvdMv(cMvPredSym[curRefList], cMvPredSym[tarRefList]), refIdxTar);
+ }
+ }
+ Mv startPtMv = cCurMvField.mv;
+
+ Distortion mvpCost = m_pcRdCost->getCost(m_auiMVPIdxCost[mvpIdxSym[curRefList]][AMVP_MAX_NUM_CANDS] + m_auiMVPIdxCost[mvpIdxSym[tarRefList]][AMVP_MAX_NUM_CANDS]);
+ symCost = costStart - mvpCost;
+
+ // ME
+ xSymmetricMotionEstimation( pu, origBuf, cMvPredSym[curRefList], cMvPredSym[tarRefList], eCurRefList, cCurMvField, cTarMvField, symCost, gbiIdx );
+
+ symCost += mvpCost;
+
+ if (startPtMv != cCurMvField.mv)
+ { // if ME change MV, run a final check for best MVP.
+ symmvdCheckBestMvp(pu, origBuf, cCurMvField.mv, (RefPicList)curRefList, aacAMVPInfo, gbiIdx, cMvPredSym, mvpIdxSym, symCost, true);
+ }
+
+ bits = uiMbBits[2];
+ bits += 1; // add one bit for symmetrical MVD mode
+ bits += ((cs.slice->getSPS()->getSpsNext().getUseGBi() == true) ? getWeightIdxBits(gbiIdx) : 0);
+ symCost += m_pcRdCost->getCost(bits);
+ cTarMvField.setMvField(cCurMvField.mv.getSymmvdMv(cMvPredSym[curRefList], cMvPredSym[tarRefList]), refIdxTar);
+
+ // save results
+ if ( symCost < uiCostBi )
+ {
+ uiCostBi = symCost;
+ symMode = 1 + curRefList;
+
+ cMvBi[curRefList] = cCurMvField.mv;
+ iRefIdxBi[curRefList] = cCurMvField.refIdx;
+ aaiMvpIdxBi[curRefList][cCurMvField.refIdx] = mvpIdxSym[curRefList];
+ cMvPredBi[curRefList][iRefIdxBi[curRefList]] = cMvPredSym[curRefList];
+
+ cMvBi[tarRefList] = cTarMvField.mv;
+ iRefIdxBi[tarRefList] = cTarMvField.refIdx;
+ aaiMvpIdxBi[tarRefList][cTarMvField.refIdx] = mvpIdxSym[tarRefList];
+ cMvPredBi[tarRefList][iRefIdxBi[tarRefList]] = cMvPredSym[tarRefList];
+ }
+ }
+#endif
} // if (B_SLICE)
@@ -1906,6 +2022,10 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
pu.mvpNum[REF_PIC_LIST_0] = aaiMvpNum[0][iRefIdxBi[0]];
pu.mvpNum[REF_PIC_LIST_1] = aaiMvpNum[1][iRefIdxBi[1]];
pu.interDir = 3;
+
+#if JVET_M0444_SMVD
+ pu.cu->smvdMode = symMode;
+#endif
}
else if ( uiCost[0] <= uiCost[1] )
{
@@ -1945,6 +2065,9 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
uint32_t uiMRGIndex = pu.mergeIdx;
bool bMergeFlag = pu.mergeFlag;
uint32_t uiInterDir = pu.interDir;
+#if JVET_M0444_SMVD
+ int iSymMode = cu.smvdMode;
+#endif
Mv cMvd[2];
uint32_t uiMvpIdx[2], uiMvpNum[2];
@@ -2042,6 +2165,9 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
pu.mergeFlag = bMergeFlag;
pu.mergeIdx = uiMRGIndex;
pu.interDir = uiInterDir;
+#if JVET_M0444_SMVD
+ cu.smvdMode = iSymMode;
+#endif
pu.mv [REF_PIC_LIST_0] = cHevcMvField[0].mv;
pu.refIdx[REF_PIC_LIST_0] = cHevcMvField[0].refIdx;
pu.mv [REF_PIC_LIST_1] = cHevcMvField[1].mv;
@@ -2055,6 +2181,9 @@ void InterSearch::predInterSearch(CodingUnit& cu, Partitioner& partitioner)
}
else
{
+#if JVET_M0444_SMVD
+ cu.smvdMode = 0;
+#endif
CHECK( !cu.affine, "Wrong." );
uiLastMode = uiLastModeTemp;
}
@@ -3133,6 +3262,153 @@ void InterSearch::xPatternSearchFracDIF(
ruiCost = xPatternRefinement( cStruct.pcPatternKey, baseRefMv, 1, rcMvQter, !bIsLosslessCoded );
}
+#if JVET_M0444_SMVD
+Distortion InterSearch::xGetSymmetricCost( PredictionUnit& pu, PelUnitBuf& origBuf, RefPicList eCurRefPicList, const MvField& cCurMvField, MvField& cTarMvField, int gbiIdx )
+{
+ Distortion cost = std::numeric_limits<Distortion>::max();
+ RefPicList eTarRefPicList = (RefPicList)(1 - (int)eCurRefPicList);
+
+ // get prediction of eCurRefPicList
+ PelUnitBuf predBufA = m_tmpPredStorage[eCurRefPicList].getBuf( UnitAreaRelative( *pu.cu, pu ) );
+ const Picture* picRefA = pu.cu->slice->getRefPic( eCurRefPicList, cCurMvField.refIdx );
+ Mv mvA = cCurMvField.mv;
+ mvA.changePrecision(MV_PRECISION_QUARTER, MV_PRECISION_INTERNAL);
+ clipMv( mvA, pu.cu->lumaPos(), pu.cu->lumaSize(), *pu.cs->sps );
+ xPredInterBlk( COMPONENT_Y, pu, picRefA, mvA, predBufA, true, pu.cu->slice->clpRng( COMPONENT_Y ), false, false );
+
+ // get prediction of eTarRefPicList
+ PelUnitBuf predBufB = m_tmpPredStorage[eTarRefPicList].getBuf( UnitAreaRelative( *pu.cu, pu ) );
+ const Picture* picRefB = pu.cu->slice->getRefPic( eTarRefPicList, cTarMvField.refIdx );
+ Mv mvB = cTarMvField.mv;
+ mvB.changePrecision(MV_PRECISION_QUARTER, MV_PRECISION_INTERNAL);
+ clipMv( mvB, pu.cu->lumaPos(), pu.cu->lumaSize(), *pu.cs->sps );
+ xPredInterBlk( COMPONENT_Y, pu, picRefB, mvB, predBufB, true, pu.cu->slice->clpRng( COMPONENT_Y ), false, false );
+
+ PelUnitBuf bufTmp = m_tmpStorageLCU.getBuf( UnitAreaRelative( *pu.cu, pu ) );
+ if (gbiIdx != GBI_DEFAULT)
+ bufTmp.Y().addWeightedAvg(predBufA.Y(), predBufB.Y(), pu.cu->slice->clpRng(COMPONENT_Y), gbiIdx);
+ else
+ bufTmp.Y().addAvg( predBufA.Y(), predBufB.Y(), pu.cu->slice->clpRng( COMPONENT_Y ) );
+
+ // calc distortion
+ cost = m_pcRdCost->getDistPart(bufTmp.Y(), origBuf.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, DF_HAD);
+
+ return(cost);
+}
+
+Distortion InterSearch::xSymmeticRefineMvSearch( PredictionUnit &pu, PelUnitBuf& origBuf, Mv& rcMvCurPred, Mv& rcMvTarPred
+ , RefPicList eRefPicList, MvField& rCurMvField, MvField& rTarMvField, Distortion uiMinCost, int SearchPattern, int nSearchStepShift, uint32_t uiMaxSearchRounds, int gbiIdx )
+{
+ const Mv mvSearchOffsetCross[4] = { Mv( 0 , 1 ) , Mv( 1 , 0 ) , Mv( 0 , -1 ) , Mv( -1 , 0 ) };
+ const Mv mvSearchOffsetSquare[8] = { Mv( -1 , 1 ) , Mv( 0 , 1 ) , Mv( 1 , 1 ) , Mv( 1 , 0 ) , Mv( 1 , -1 ) , Mv( 0 , -1 ) , Mv( -1 , -1 ) , Mv( -1 , 0 ) };
+ const Mv mvSearchOffsetDiamond[8] = { Mv( 0 , 2 ) , Mv( 1 , 1 ) , Mv( 2 , 0 ) , Mv( 1 , -1 ) , Mv( 0 , -2 ) , Mv( -1 , -1 ) , Mv( -2 , 0 ) , Mv( -1 , 1 ) };
+ const Mv mvSearchOffsetHexagon[6] = { Mv( 2 , 0 ) , Mv( 1 , 2 ) , Mv( -1 , 2 ) , Mv( -2 , 0 ) , Mv( -1 , -2 ) , Mv( 1 , -2 ) };
+
+ int nDirectStart = 0, nDirectEnd = 0, nDirectRounding = 0, nDirectMask = 0;
+ const Mv * pSearchOffset;
+ if ( SearchPattern == 0 )
+ {
+ nDirectEnd = 3;
+ nDirectRounding = 4;
+ nDirectMask = 0x03;
+ pSearchOffset = mvSearchOffsetCross;
+ }
+ else if ( SearchPattern == 1 )
+ {
+ nDirectEnd = 7;
+ nDirectRounding = 8;
+ nDirectMask = 0x07;
+ pSearchOffset = mvSearchOffsetSquare;
+ }
+ else if ( SearchPattern == 2 )
+ {
+ nDirectEnd = 7;
+ nDirectRounding = 8;
+ nDirectMask = 0x07;
+ pSearchOffset = mvSearchOffsetDiamond;
+ }
+ else if ( SearchPattern == 3 )
+ {
+ nDirectEnd = 5;
+ pSearchOffset = mvSearchOffsetHexagon;
+ }
+ else
+ {
+ THROW( "Invalid search pattern" );
+ }
+
+ int nBestDirect;
+ for ( uint32_t uiRound = 0; uiRound < uiMaxSearchRounds; uiRound++ )
+ {
+ nBestDirect = -1;
+ MvField mvCurCenter = rCurMvField;
+ for ( int nIdx = nDirectStart; nIdx <= nDirectEnd; nIdx++ )
+ {
+ int nDirect;
+ if ( SearchPattern == 3 )
+ {
+ nDirect = nIdx < 0 ? nIdx + 6 : nIdx >= 6 ? nIdx - 6 : nIdx;
+ }
+ else
+ {
+ nDirect = (nIdx + nDirectRounding) & nDirectMask;
+ }
+
+ Mv mvOffset = pSearchOffset[nDirect];
+ mvOffset <<= nSearchStepShift;
+ MvField mvCand = mvCurCenter, mvPair;
+ mvCand.mv += mvOffset;
+
+ // get MVD cost
+ m_pcRdCost->setPredictor( rcMvCurPred );
+ m_pcRdCost->setCostScale( 0 );
+ uint32_t uiMvBits = m_pcRdCost->getBitsOfVectorWithPredictor( mvCand.mv.getHor(), mvCand.mv.getVer(), (pu.cu->imv << 1) );
+ Distortion uiCost = m_pcRdCost->getCost( uiMvBits );
+
+ // get MVD pair and set target MV
+ mvPair.refIdx = rTarMvField.refIdx;
+ mvPair.mv.set( rcMvTarPred.hor - (mvCand.mv.hor - rcMvCurPred.hor), rcMvTarPred.ver - (mvCand.mv.ver - rcMvCurPred.ver) );
+ uiCost += xGetSymmetricCost( pu, origBuf, eRefPicList, mvCand, mvPair, gbiIdx );
+ if ( uiCost < uiMinCost )
+ {
+ uiMinCost = uiCost;
+ rCurMvField = mvCand;
+ rTarMvField = mvPair;
+ nBestDirect = nDirect;
+ }
+ }
+
+ if ( nBestDirect == -1 )
+ {
+ break;
+ }
+ int nStep = 1;
+ if ( SearchPattern == 1 || SearchPattern == 2 )
+ {
+ nStep = 2 - (nBestDirect & 0x01);
+ }
+ nDirectStart = nBestDirect - nStep;
+ nDirectEnd = nBestDirect + nStep;
+ }
+
+ return(uiMinCost);
+}
+
+
+void InterSearch::xSymmetricMotionEstimation( PredictionUnit& pu, PelUnitBuf& origBuf, Mv& rcMvCurPred, Mv& rcMvTarPred, RefPicList eRefPicList, MvField& rCurMvField, MvField& rTarMvField, Distortion& ruiCost, int gbiIdx )
+{
+ // Refine Search
+ int nSearchStepShift = 0;
+ int nDiamondRound = 8;
+ int nCrossRound = 1;
+
+ nSearchStepShift += (pu.cu->imv << 1);
+ nDiamondRound >>= pu.cu->imv;
+
+ ruiCost = xSymmeticRefineMvSearch( pu, origBuf, rcMvCurPred, rcMvTarPred, eRefPicList, rCurMvField, rTarMvField, ruiCost, 2, nSearchStepShift, nDiamondRound, gbiIdx );
+ ruiCost = xSymmeticRefineMvSearch( pu, origBuf, rcMvCurPred, rcMvTarPred, eRefPicList, rCurMvField, rTarMvField, ruiCost, 0, nSearchStepShift, nCrossRound, gbiIdx );
+}
+#endif // JVET_M0444_SMVD
void InterSearch::xPredAffineInterSearch( PredictionUnit& pu,
PelUnitBuf& origBuf,
@@ -3218,7 +3494,7 @@ void InterSearch::xPredAffineInterSearch( PredictionUnit& pu,
{
RefPicList eRefPicList = ( iRefList ? REF_PIC_LIST_1 : REF_PIC_LIST_0 );
int refPicNumber = slice.getNumRefIdx(eRefPicList);
- if (slice.getSPS()->getSpsNext().getCPRMode() && eRefPicList == REF_PIC_LIST_0)
+ if (slice.getSPS()->getSpsNext().getIBCMode() && eRefPicList == REF_PIC_LIST_0)
{
refPicNumber--;
}
@@ -3591,7 +3867,7 @@ void InterSearch::xPredAffineInterSearch( PredictionUnit& pu,
iRefStart = 0;
iRefEnd = slice.getNumRefIdx(eRefPicList) - 1;
- if (slice.getSPS()->getSpsNext().getCPRMode() && eRefPicList == REF_PIC_LIST_0)
+ if (slice.getSPS()->getSpsNext().getIBCMode() && eRefPicList == REF_PIC_LIST_0)
{
iRefEnd--;
}
@@ -4160,8 +4436,13 @@ void InterSearch::xAffineMotionEstimation( PredictionUnit& pu,
for ( int i = 0; i < mvNum; i++ )
{
acMvTemp[i] += acDeltaMv[i];
+#if JVET_M0479_18BITS_MV_CLIP
+ acMvTemp[i].hor = Clip3( -131072, 131071, acMvTemp[i].hor );
+ acMvTemp[i].ver = Clip3( -131072, 131071, acMvTemp[i].ver );
+#else
acMvTemp[i].hor = Clip3( -32768, 32767, acMvTemp[i].hor );
acMvTemp[i].ver = Clip3( -32768, 32767, acMvTemp[i].ver );
+#endif
acMvTemp[i].roundToPrecision(MV_PRECISION_INTERNAL, MV_PRECISION_QUARTER);
clipMv(acMvTemp[i], pu.cu->lumaPos(),
pu.cu->lumaSize(),
@@ -4721,7 +5002,9 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
{
const UnitArea& currArea = partitioner.currArea();
const SPS &sps = *cs.sps;
+#if !JVET_M0464_UNI_MTS
const PPS &pps = *cs.pps;
+#endif
const uint32_t numValidComp = getNumberValidComponents( sps.getChromaFormatIdc() );
const uint32_t numTBlocks = getNumberValidTBlocks ( *cs.pcv );
const CodingUnit &cu = *cs.getCU(partitioner.chType);
@@ -4753,10 +5036,16 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
{
TransformUnit &tu = csFull->addTU(CS::isDualITree(cs) ? cu : currArea, partitioner.chType);
tu.depth = currDepth;
+#if JVET_M0464_UNI_MTS
+ tu.mtsIdx = 0;
+#else
tu.emtIdx = 0;
+#endif
double minCost [MAX_NUM_TBLOCKS];
+#if !JVET_M0464_UNI_MTS
bool checkTransformSkip [MAX_NUM_TBLOCKS];
+#endif
m_CABACEstimator->resetBits();
@@ -4783,19 +5072,22 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
continue;
const CompArea& compArea = tu.blocks[compID];
const int channelBitDepth = sps.getBitDepth(toChannelType(compID));
-
+#if !JVET_M0464_UNI_MTS
checkTransformSkip[compID] = false;
+#endif
if( !tu.blocks[compID].valid() )
{
continue;
}
+#if !JVET_M0464_UNI_MTS
checkTransformSkip[compID] = pps.getUseTransformSkip() && TU::hasTransformSkipFlag( *tu.cs, tu.blocks[compID] ) && !cs.isLossless;
if( isLuma(compID) )
{
checkTransformSkip[compID] &= !tu.cu->emtFlag;
}
+#endif
const bool isCrossCPredictionAvailable = TU::hasCrossCompPredInfo( tu, compID );
@@ -4808,10 +5100,32 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
preCalcAlpha = xCalcCrossComponentPredictionAlpha( tu, compID, m_pcEncCfg->getUseReconBasedCrossCPredictionEstimate() );
}
+#if JVET_M0464_UNI_MTS
+ const bool tsAllowed = TU::isTSAllowed ( tu, compID );
+ const bool mtsAllowed = TU::isMTSAllowed( tu, compID );
+ uint8_t nNumTransformCands = 1 + ( tsAllowed ? 1 : 0 ) + ( mtsAllowed ? 4 : 0 ); // DCT + TS + 4 MTS = 6 tests
+ std::vector<TrMode> trModes;
+ trModes.push_back( TrMode( 0, true ) ); //DCT2
+ if( tsAllowed )
+ {
+ trModes.push_back( TrMode( 1, true ) );
+ }
+ if( mtsAllowed )
+ {
+ for( int i = 2; i < 6; i++ )
+ {
+ trModes.push_back( TrMode( i, true ) );
+ }
+ }
+#endif
const int crossCPredictionModesToTest = preCalcAlpha != 0 ? 2 : 1;
+#if JVET_M0464_UNI_MTS
+ const int numTransformCandidates = nNumTransformCands;
+#else
const int numEmtTransformCandidates = isLuma(compID) && tu.cu->emtFlag && sps.getSpsNext().getUseInterEMT() ? 4 : 1;
const int numTransformCandidates = checkTransformSkip[compID] ? ( numEmtTransformCandidates + 1 ) : numEmtTransformCandidates;
int lastTransformModeIndex = numTransformCandidates - 1; //lastTransformModeIndex is the mode for transformSkip (if transformSkip is active)
+#endif
const bool isOneMode = crossCPredictionModesToTest == 1 && numTransformCandidates == 1;
bool isLastBest = isOneMode;
@@ -4829,8 +5143,23 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
m_CABACEstimator->getCtx() = ctxStart;
m_CABACEstimator->resetBits();
+#if JVET_M0464_UNI_MTS
+ if( isLuma( compID ) )
+ {
+ if( bestTU.mtsIdx == 1 && m_pcEncCfg->getUseTransformSkipFast() )
+ {
+ continue;
+ }
+ if( !trModes[transformMode].second )
+ {
+ continue;
+ }
+ tu.mtsIdx = trModes[transformMode].first;
+ }
+#else
if( isLuma( compID ) ) tu.emtIdx = transformMode;
tu.transformSkip[compID] = checkTransformSkip[compID] && transformMode == lastTransformModeIndex;
+#endif
tu.compAlpha[compID] = bUseCrossCPrediction ? preCalcAlpha : 0;
const QpParam cQP(tu, compID); // note: uses tu.transformSkip[compID]
@@ -4853,7 +5182,23 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
crossComponentPrediction( tu, compID, lumaResi, resiBuf, resiBuf, false );
}
+#if JVET_M0464_UNI_MTS
+ if( nNumTransformCands > 1 )
+ {
+ if( transformMode == 0 )
+ {
+ m_pcTrQuant->transformNxN( tu, compID, cQP, &trModes, CU::isIntra( *tu.cu ) ? m_pcEncCfg->getIntraMTSMaxCand() : m_pcEncCfg->getInterMTSMaxCand() );
+ tu.mtsIdx = trModes[0].first;
+ }
+ m_pcTrQuant->transformNxN( tu, compID, cQP, currAbsSum, m_CABACEstimator->getCtx(), true );
+ }
+ else
+ {
+ m_pcTrQuant->transformNxN( tu, compID, cQP, currAbsSum, m_CABACEstimator->getCtx() );
+ }
+#else
m_pcTrQuant->transformNxN(tu, compID, cQP, currAbsSum, m_CABACEstimator->getCtx());
+#endif
if (isFirstMode || (currAbsSum == 0))
{
@@ -4937,7 +5282,11 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
nonCoeffCost = MAX_DOUBLE;
}
}
+#if JVET_M0464_UNI_MTS
+ else if( transformMode > 0 && !bUseCrossCPrediction )
+#else
else if( ( transformMode == lastTransformModeIndex ) && checkTransformSkip[compID] && !bUseCrossCPrediction )
+#endif
{
currCompCost = MAX_DOUBLE;
}
@@ -4951,7 +5300,11 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
}
// evaluate
+#if JVET_M0464_UNI_MTS
+ if( ( currCompCost < minCost[compID] ) || ( transformMode == 1 && currCompCost == minCost[compID] ) )
+#else
if( ( currCompCost < minCost[compID] ) || ( transformMode == lastTransformModeIndex && checkTransformSkip[compID] && currCompCost == minCost[compID] ) )
+#endif
{
// copy component
if (isFirstMode && ((nonCoeffCost < currCompCost) || (currAbsSum == 0))) // check for forced null
@@ -5073,10 +5426,12 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
);
csSplit->cost = m_pcRdCost->calcRdCost( csSplit->fracBits, csSplit->dist );
+#if !JVET_M0464_UNI_MTS
if( csFull && csSplit->cost >= csFull->cost && m_pcEncCfg->getFastInterEMT() )
{
break;
}
+#endif
} while( partitioner.nextPart( *csSplit ) );
partitioner.exitCurrSplit();
@@ -5084,8 +5439,12 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
unsigned anyCbfSet = 0;
unsigned compCbf[3] = { 0, 0, 0 };
+#if JVET_M0464_UNI_MTS
+ if( !bCheckFull )
+#else
bool isSplit = bCheckFull ? false : true;
if( !bCheckFull || ( csSplit->cost < csFull->cost && m_pcEncCfg->getFastInterEMT() ) || !m_pcEncCfg->getFastInterEMT() )
+#endif
{
for( auto &currTU : csSplit->traverseTUs( currArea, partitioner.chType ) )
{
@@ -5137,10 +5496,13 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
{
cs.useSubStructure( *csSplit, partitioner.chType, currArea, false, false, false, true );
cs.cost = csSplit->cost;
+#if !JVET_M0464_UNI_MTS
isSplit = true;
+#endif
}
}
+#if !JVET_M0464_UNI_MTS
if( ( !isSplit && m_pcEncCfg->getFastInterEMT() ) || ( !m_pcEncCfg->getFastInterEMT() && !( !bCheckFull || ( anyCbfSet && csSplit->cost < csFull->cost ) ) ) )
{
CHECK( !bCheckFull, "Error!" );
@@ -5148,6 +5510,7 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
cs.cost = csFull->cost;
m_CABACEstimator->getCtx() = ctxBest;
}
+#endif
if( csSplit && csFull )
{
@@ -5475,3 +5838,83 @@ void InterSearch::initWeightIdxBits()
}
}
+#if JVET_M0444_SMVD
+void InterSearch::symmvdCheckBestMvp(
+ PredictionUnit& pu,
+ PelUnitBuf& origBuf,
+ Mv curMv,
+ RefPicList curRefList,
+ AMVPInfo amvpInfo[2][33],
+ int32_t gbiIdx,
+ Mv cMvPredSym[2],
+ int32_t mvpIdxSym[2],
+ Distortion& bestCost,
+ bool skip
+)
+{
+ RefPicList tarRefList = (RefPicList)(1 - curRefList);
+ int32_t refIdxCur = pu.cu->slice->getSymRefIdx(curRefList);
+ int32_t refIdxTar = pu.cu->slice->getSymRefIdx(tarRefList);
+
+ MvField cCurMvField, cTarMvField;
+ cCurMvField.setMvField(curMv, refIdxCur);
+ AMVPInfo& amvpCur = amvpInfo[curRefList][refIdxCur];
+ AMVPInfo& amvpTar = amvpInfo[tarRefList][refIdxTar];
+ m_pcRdCost->setCostScale(0);
+
+
+ // get prediction of eCurRefPicList
+ PelUnitBuf predBufA = m_tmpPredStorage[curRefList].getBuf(UnitAreaRelative(*pu.cu, pu));
+ const Picture* picRefA = pu.cu->slice->getRefPic(curRefList, cCurMvField.refIdx);
+ Mv mvA = cCurMvField.mv;
+ mvA.changePrecision(MV_PRECISION_QUARTER, MV_PRECISION_INTERNAL);
+ clipMv(mvA, pu.cu->lumaPos(), pu.cu->lumaSize(), *pu.cs->sps);
+ xPredInterBlk(COMPONENT_Y, pu, picRefA, mvA, predBufA, true, pu.cu->slice->clpRng(COMPONENT_Y), false, false);
+
+ int32_t skipMvpIdx[2];
+ skipMvpIdx[0] = skip ? mvpIdxSym[0] : -1;
+ skipMvpIdx[1] = skip ? mvpIdxSym[1] : -1;
+
+ for (int i = 0; i < amvpCur.numCand; i++)
+ {
+ for (int j = 0; j < amvpTar.numCand; j++)
+ {
+ if (skipMvpIdx[curRefList] == i && skipMvpIdx[tarRefList] == j)
+ continue;
+
+ cTarMvField.setMvField(curMv.getSymmvdMv(amvpCur.mvCand[i], amvpTar.mvCand[j]), refIdxTar);
+
+ // get prediction of eTarRefPicList
+ PelUnitBuf predBufB = m_tmpPredStorage[tarRefList].getBuf(UnitAreaRelative(*pu.cu, pu));
+ const Picture* picRefB = pu.cu->slice->getRefPic(tarRefList, cTarMvField.refIdx);
+ Mv mvB = cTarMvField.mv;
+ mvB.changePrecision(MV_PRECISION_QUARTER, MV_PRECISION_INTERNAL);
+ clipMv(mvB, pu.cu->lumaPos(), pu.cu->lumaSize(), *pu.cs->sps);
+ xPredInterBlk(COMPONENT_Y, pu, picRefB, mvB, predBufB, true, pu.cu->slice->clpRng(COMPONENT_Y), false, false);
+
+ PelUnitBuf bufTmp = m_tmpStorageLCU.getBuf(UnitAreaRelative(*pu.cu, pu));
+ if (gbiIdx != GBI_DEFAULT)
+ bufTmp.Y().addWeightedAvg(predBufA.Y(), predBufB.Y(), pu.cu->slice->clpRng(COMPONENT_Y), gbiIdx);
+ else
+ bufTmp.Y().addAvg(predBufA.Y(), predBufB.Y(), pu.cu->slice->clpRng(COMPONENT_Y));
+
+ // calc distortion
+ Distortion cost = m_pcRdCost->getDistPart(bufTmp.Y(), origBuf.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, DF_HAD);
+
+ m_pcRdCost->setPredictor(amvpCur.mvCand[i]);
+ uint32_t bits = m_pcRdCost->getBitsOfVectorWithPredictor(curMv.hor, curMv.ver, (pu.cu->imv << 1));
+ bits += m_auiMVPIdxCost[i][AMVP_MAX_NUM_CANDS];
+ bits += m_auiMVPIdxCost[j][AMVP_MAX_NUM_CANDS];
+ cost += m_pcRdCost->getCost(bits);
+ if (cost < bestCost)
+ {
+ bestCost = cost;
+ cMvPredSym[curRefList] = amvpCur.mvCand[i];
+ cMvPredSym[tarRefList] = amvpTar.mvCand[j];
+ mvpIdxSym[curRefList] = i;
+ mvpIdxSym[tarRefList] = j;
+ }
+ }
+ }
+}
+#endif
\ No newline at end of file
diff --git a/source/Lib/EncoderLib/InterSearch.h b/source/Lib/EncoderLib/InterSearch.h
index 05dcc7fa8aa389cf925e95727e6e4d0f7ba774a8..f9edc9438352622bd17b04cddf91a8c1a235074a 100644
--- a/source/Lib/EncoderLib/InterSearch.h
+++ b/source/Lib/EncoderLib/InterSearch.h
@@ -51,7 +51,7 @@
#include "CommonLib/RdCost.h"
#include "CommonLib/AffineGradientSearch.h"
-#include "CommonLib/CprHashMap.h"
+#include "CommonLib/IbcHashMap.h"
#include <unordered_map>
#include <vector>
//! \ingroup EncoderLib
@@ -132,7 +132,7 @@ protected:
bool m_isInitialized;
unsigned int m_numBVs, m_numBV16s;
- Mv m_acBVs[CPR_NUM_CANDIDATES];
+ Mv m_acBVs[IBC_NUM_CANDIDATES];
public:
InterSearch();
virtual ~InterSearch();
@@ -238,12 +238,12 @@ 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; }
- bool predCPRSearch ( CodingUnit& cu, Partitioner& partitioner, const int localSearchRangeX, const int localSearchRangeY, CprHashMap& cprHashMap);
+ bool predIBCSearch ( CodingUnit& cu, Partitioner& partitioner, const int localSearchRangeX, const int localSearchRangeY, IbcHashMap& ibcHashMap);
void xIntraPatternSearch ( PredictionUnit& pu, IntTZSearchStruct& cStruct, Mv& rcMv, Distortion& ruiCost, Mv* cMvSrchRngLT, Mv* cMvSrchRngRB, Mv* pcMvPred);
void xSetIntraSearchRange ( PredictionUnit& pu, int iRoiWidth, int iRoiHeight, const int localSearchRangeX, const int localSearchRangeY, Mv& rcMvSrchRngLT, Mv& rcMvSrchRngRB);
- void xCPREstimation ( PredictionUnit& pu, PelUnitBuf& origBuf, Mv *pcMvPred, Mv &rcMv, Distortion &ruiCost, const int localSearchRangeX, const int localSearchRangeY);
- void xCPRSearchMVCandUpdate ( Distortion uiSad, int x, int y, Distortion* uiSadBestCand, Mv* cMVCand);
- int xCPRSearchMVChromaRefine( PredictionUnit& pu, int iRoiWidth, int iRoiHeight, int cuPelX, int cuPelY, Distortion* uiSadBestCand, Mv* cMVCand);
+ void xIBCEstimation ( PredictionUnit& pu, PelUnitBuf& origBuf, Mv *pcMvPred, Mv &rcMv, Distortion &ruiCost, const int localSearchRangeX, const int localSearchRangeY);
+ void xIBCSearchMVCandUpdate ( Distortion uiSad, int x, int y, Distortion* uiSadBestCand, Mv* cMVCand);
+ int xIBCSearchMVChromaRefine( PredictionUnit& pu, int iRoiWidth, int iRoiHeight, int cuPelX, int cuPelY, Distortion* uiSadBestCand, Mv* cMVCand);
protected:
// -------------------------------------------------------------------------------------------------------------------
@@ -399,6 +399,15 @@ 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 );
+#if JVET_M0444_SMVD
+ Distortion xGetSymmetricCost( PredictionUnit& pu, PelUnitBuf& origBuf, RefPicList eCurRefPicList, const MvField& cCurMvField, MvField& cTarMvField , int gbiIdx );
+
+ Distortion xSymmeticRefineMvSearch( PredictionUnit& pu, PelUnitBuf& origBuf, Mv& rcMvCurPred, Mv& rcMvTarPred
+ , RefPicList eRefPicList, MvField& rCurMvField, MvField& rTarMvField, Distortion uiMinCost, int searchPattern, int nSearchStepShift, uint32_t uiMaxSearchRounds , int gbiIdx );
+
+ void xSymmetricMotionEstimation( PredictionUnit& pu, PelUnitBuf& origBuf, Mv& rcMvCurPred, Mv& rcMvTarPred, RefPicList eRefPicList, MvField& rCurMvField, MvField& rTarMvField, Distortion& ruiCost, int gbiIdx );
+#endif
+
bool xReadBufferedAffineUniMv ( PredictionUnit& pu, RefPicList eRefPicList, int32_t iRefIdx, Mv acMvPred[3], Mv acMv[3], uint32_t& ruiBits, Distortion& ruiCost);
double xGetMEDistortionWeight ( uint8_t gbiIdx, RefPicList eRefPicList);
bool xReadBufferedUniMv ( PredictionUnit& pu, RefPicList eRefPicList, int32_t iRefIdx, Mv& pcMvPred, Mv& rcMv, uint32_t& ruiBits, Distortion& ruiCost);
@@ -406,6 +415,20 @@ public:
void resetBufferedUniMotions () { m_uniMotions.reset(); }
uint32_t getWeightIdxBits ( uint8_t gbiIdx ) { return m_estWeightIdxBits[gbiIdx]; }
void initWeightIdxBits ();
+#if JVET_M0444_SMVD
+ void symmvdCheckBestMvp(
+ PredictionUnit& pu,
+ PelUnitBuf& origBuf,
+ Mv curMv,
+ RefPicList curRefList,
+ AMVPInfo amvpInfo[2][33],
+ int32_t gbiIdx,
+ Mv cMvPredSym[2],
+ int32_t mvpIdxSym[2],
+ Distortion& bestCost,
+ bool skip = false
+ );
+#endif
protected:
void xExtDIFUpSamplingH ( CPelBuf* pcPattern );
@@ -417,7 +440,7 @@ protected:
void setWpScalingDistParam ( int iRefIdx, RefPicList eRefPicListCur, Slice *slice );
private:
- void xxCPRHashSearch(PredictionUnit& pu, Mv* mvPred, int numMvPred, Mv &mv, int& idxMvPred, CprHashMap& cprHashMap);
+ void xxIBCHashSearch(PredictionUnit& pu, Mv* mvPred, int numMvPred, Mv &mv, int& idxMvPred, IbcHashMap& ibcHashMap);
public:
void encodeResAndCalcRdInterCU (CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual
diff --git a/source/Lib/EncoderLib/IntraSearch.cpp b/source/Lib/EncoderLib/IntraSearch.cpp
index d8ebcd96c05cd8dd54bd97ac443144e0f9d48752..ac23b0a313c1459a5601ecc27059bbc6c00ab1d0 100644
--- a/source/Lib/EncoderLib/IntraSearch.cpp
+++ b/source/Lib/EncoderLib/IntraSearch.cpp
@@ -285,7 +285,7 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
uint32_t extraModes = 0; // add two extra modes, which would be used after uiMode <= DC_IDX is removed for cu.nsstIdx == 3
-
+#if !JVET_M0464_UNI_MTS
const int width = partitioner.currArea().lwidth();
const int height = partitioner.currArea().lheight();
@@ -306,6 +306,7 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
{
emtUsageFlag = 0; //this forces the recalculation of the candidates list. Why is this necessary? (to be checked)
}
+#endif
static_vector<uint32_t, FAST_UDI_MAX_RDMODE_NUM> uiHadModeList;
static_vector<double, FAST_UDI_MAX_RDMODE_NUM> CandCostList;
@@ -315,7 +316,9 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
static_vector<int, FAST_UDI_MAX_RDMODE_NUM>* nullList = NULL;
auto &pu = *cu.firstPU;
+#if !JVET_M0464_UNI_MTS
int puIndex = 0;
+#endif
{
CandHadList.clear();
CandCostList.clear();
@@ -335,8 +338,9 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
numModesForFullRD = numModesAvailable;
#endif
-
+#if !JVET_M0464_UNI_MTS
if( emtUsageFlag != 2 )
+#endif
{
// this should always be true
CHECK( !pu.Y().valid(), "PU is not valid" );
@@ -547,6 +551,7 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
uiRdModeList.push_back( i );
}
}
+#if !JVET_M0464_UNI_MTS
if( emtUsageFlag == 1 )
{
// Store the modes to be checked with RD
@@ -554,7 +559,9 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
std::copy_n( uiRdModeList.begin(), numModesForFullRD, m_savedRdModeList[puIndex] );
std::copy_n(extendRefList.begin(), numModesForFullRD, m_savedExtendRefList[puIndex]);
}
+#endif
}
+#if !JVET_M0464_UNI_MTS
else //emtUsage = 2 (here we potentially reduce the number of modes that will be full-RD checked)
{
if( isAllIntra && m_pcEncCfg->getFastIntraEMT() )
@@ -585,6 +592,7 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
std::copy_n(m_savedExtendRefList[puIndex], m_savedNumRdModes[puIndex], extendRefList.begin());
}
}
+#endif
CHECK( numModesForFullRD != uiRdModeList.size(), "Inconsistent state!" );
@@ -592,7 +600,11 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
// after this point, don't use numModesForFullRD
// PBINTRA fast
+#if JVET_M0464_UNI_MTS
+ if( m_pcEncCfg->getUsePbIntraFast() && !cs.slice->isIntra() && uiRdModeList.size() < numModesAvailable )
+#else
if( m_pcEncCfg->getUsePbIntraFast() && !cs.slice->isIntra() && uiRdModeList.size() < numModesAvailable && emtUsageFlag != 2 )
+#endif
{
if( CandHadList.size() < 3 || CandHadList[2] > cs.interHad * PBINTRA_RATIO )
{
@@ -649,11 +661,12 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
xRecurIntraCodingLumaQT( *csTemp, partitioner );
+#if !JVET_M0464_UNI_MTS
if( emtUsageFlag == 1 && m_pcEncCfg->getFastIntraEMT() )
{
m_modeCostStore[puIndex][uiMode] = csTemp->cost; //cs.cost;
}
-
+#endif
DTRACE( g_trace_ctx, D_INTRA_COST, "IntraCost T %f (%d) \n", csTemp->cost, uiOrgMode );
@@ -664,11 +677,12 @@ void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
uiBestPUMode = uiOrgMode;
bestExtendRef = multiRefIdx;
-
+#if !JVET_M0464_UNI_MTS
if( ( emtUsageFlag == 1 ) && m_pcEncCfg->getFastIntraEMT() )
{
m_bestModeCostStore[puIndex] = csBest->cost; //cs.cost;
}
+#endif
}
csTemp->releaseIntermediateData();
@@ -1048,7 +1062,9 @@ void IntraSearch::xEncSubdivCbfQT(CodingStructure &cs, Partitioner &partitioner,
{
const UnitArea &currArea = partitioner.currArea();
TransformUnit &currTU = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType );
+#if !JVET_M0464_UNI_MTS
CodingUnit &currCU = *currTU.cu;
+#endif
uint32_t currDepth = partitioner.currTrDepth;
const bool subdiv = currTU.depth > currDepth;
@@ -1081,7 +1097,9 @@ void IntraSearch::xEncSubdivCbfQT(CodingStructure &cs, Partitioner &partitioner,
if (subdiv)
{
+#if !JVET_M0464_UNI_MTS
if( currDepth == 0 && bLuma ) m_CABACEstimator->emt_cu_flag( currCU );
+#endif
if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
{
@@ -1099,8 +1117,9 @@ void IntraSearch::xEncSubdivCbfQT(CodingStructure &cs, Partitioner &partitioner,
}
else
{
+#if !JVET_M0464_UNI_MTS
if( currDepth == 0 && bLuma && TU::getCbfAtDepth( currTU, COMPONENT_Y, 0 ) ) m_CABACEstimator->emt_cu_flag( currCU );
-
+#endif
//===== Cbfs =====
if (bLuma)
{
@@ -1185,7 +1204,11 @@ uint64_t IntraSearch::xGetIntraFracBitsQTChroma(TransformUnit& currTU, const Com
return fracBits;
}
+#if JVET_M0464_UNI_MTS
+void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &compID, const bool &checkCrossCPrediction, Distortion& ruiDist, const int &default0Save1Load2, uint32_t* numSig, std::vector<TrMode>* trModes, const bool loadTr)
+#else
void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &compID, const bool &checkCrossCPrediction, Distortion& ruiDist, const int &default0Save1Load2, uint32_t* numSig )
+#endif
{
if (!tu.blocks[compID].valid())
{
@@ -1280,8 +1303,16 @@ void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &comp
m_pcTrQuant->selectLambda(compID);
#endif
-
+#if JVET_M0464_UNI_MTS
+ if( trModes )
+ {
+ m_pcTrQuant->transformNxN( tu, compID, cQP, trModes, CU::isIntra( *tu.cu ) ? m_pcEncCfg->getIntraMTSMaxCand() : m_pcEncCfg->getInterMTSMaxCand() );
+ tu.mtsIdx = trModes->at(0).first;
+ }
+ m_pcTrQuant->transformNxN(tu, compID, cQP, uiAbsSum, m_CABACEstimator->getCtx(), loadTr);
+#else
m_pcTrQuant->transformNxN(tu, compID, cQP, uiAbsSum, m_CABACEstimator->getCtx());
+#endif
DTRACE( g_trace_ctx, D_TU_ABS_SUM, "%d: comp=%d, abssum=%d\n", DTRACE_GET_COUNTER( g_trace_ctx, D_TU_ABS_SUM ), compID, uiAbsSum );
@@ -1322,8 +1353,10 @@ void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &comp
void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &partitioner )
{
const UnitArea &currArea = partitioner.currArea();
+#if !JVET_M0464_UNI_MTS
const CodingUnit &cu = *cs.getCU(currArea.lumaPos(), partitioner.chType);
- uint32_t currDepth = partitioner.currTrDepth;
+#endif
+ uint32_t currDepth = partitioner.currTrDepth;
const PPS &pps = *cs.pps;
const bool keepResi = pps.getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() || KEEP_PRED_AND_RESI_SIGNALS;
bool bCheckFull = true;
@@ -1333,6 +1366,12 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
uint32_t numSig = 0;
+#if JVET_M0464_UNI_MTS
+ double dSingleCost = MAX_DOUBLE;
+ Distortion uiSingleDistLuma = 0;
+ uint64_t singleFracBits = 0;
+ int bestModeId[MAX_NUM_COMPONENT] = { 0, 0, 0 };
+#else
bool checkInitTrDepth = false, checkInitTrDepthTransformSkipWinner = false;
double dSingleCost = MAX_DOUBLE;
@@ -1344,6 +1383,7 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
bool isAllIntra = m_pcEncCfg->getIntraPeriod() == 1;
uint8_t numTransformIndexCands = nNumTransformCands;
+#endif
const TempCtx ctxStart ( m_CtxCache, m_CABACEstimator->getCtx() );
TempCtx ctxBest ( m_CtxCache );
@@ -1367,6 +1407,26 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
TransformUnit &tu = csFull->addTU( CS::getArea( *csFull, currArea, partitioner.chType ), partitioner.chType );
tu.depth = currDepth;
+#if JVET_M0464_UNI_MTS
+ const bool tsAllowed = TU::isTSAllowed ( tu, COMPONENT_Y );
+ const bool mtsAllowed = TU::isMTSAllowed( tu, COMPONENT_Y );
+ uint8_t nNumTransformCands = 1 + ( tsAllowed ? 1 : 0 ) + ( mtsAllowed ? 4 : 0 ); // DCT + TS + 4 MTS = 6 tests
+ std::vector<TrMode> trModes;
+ trModes.push_back( TrMode( 0, true ) ); //DCT2
+ if( tsAllowed )
+ {
+ trModes.push_back( TrMode( 1, true ) );
+ }
+ if( mtsAllowed )
+ {
+ for( int i = 2; i < 6; i++ )
+ {
+ trModes.push_back( TrMode( i, true) );
+ }
+ }
+
+ CHECK( !tu.Y().valid(), "Invalid TU" );
+#else
checkTransformSkip &= TU::hasTransformSkipFlag( *tu.cs, tu.Y() );
checkTransformSkip &= !cu.transQuantBypass;
checkTransformSkip &= !cu.emtFlag;
@@ -1378,6 +1438,7 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
CHECK( checkInitTrDepthTransformSkipWinner && !checkTransformSkip, "Transform Skip must be enabled if it was the winner in the previous call of xRecurIntraCodingLumaQT!" );
+#endif
CodingStructure &saveCS = *m_pSaveCS[0];
@@ -1388,9 +1449,14 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
double singleCostTmp = 0;
int firstCheckId = 0;
+#if JVET_M0464_UNI_MTS
+ int lastCheckId = trModes[nNumTransformCands-1].first;
+ bool isNotOnlyOneMode = nNumTransformCands != 1;
+#else
//we add the EMT candidates to the loop. TransformSkip will still be the last one to be checked (when modeId == lastCheckId) as long as checkTransformSkip is true
int lastCheckId = numTransformIndexCands - ( firstCheckId + 1 ) + ( int ) checkTransformSkip;
bool isNotOnlyOneMode = lastCheckId != firstCheckId && !checkInitTrDepthTransformSkipWinner;
+#endif
if( isNotOnlyOneMode )
{
@@ -1401,9 +1467,25 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
tmpTU = &saveCS.addTU(currArea, partitioner.chType);
}
+#if JVET_M0464_UNI_MTS
+ bool cbfDCT2 = true;
+#else
bool cbfBestMode = false;
+#endif
-
+#if JVET_M0464_UNI_MTS
+ for( int modeId = firstCheckId; modeId < nNumTransformCands; modeId++ )
+ {
+ if( !cbfDCT2 || ( m_pcEncCfg->getUseTransformSkipFast() && bestModeId[COMPONENT_Y] == 1 ) )
+ {
+ break;
+ }
+ if( !trModes[modeId].second )
+ {
+ continue;
+ }
+ tu.mtsIdx = trModes[modeId].first;
+#else
for( int modeId = firstCheckId; modeId <= lastCheckId; modeId++ )
{
if( checkInitTrDepthTransformSkipWinner )
@@ -1426,6 +1508,7 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
continue;
}
}
+#endif
if ((modeId != firstCheckId) && isNotOnlyOneMode)
{
@@ -1435,7 +1518,11 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
int default0Save1Load2 = 0;
singleDistTmpLuma = 0;
+#if JVET_M0464_UNI_MTS
+ if( modeId == firstCheckId && nNumTransformCands > 1 )
+#else
if (modeId == firstCheckId && modeId != lastCheckId && !checkInitTrDepthTransformSkipWinner )
+#endif
{
default0Save1Load2 = 1;
}
@@ -1443,7 +1530,26 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
{
default0Save1Load2 = 2;
}
-
+#if JVET_M0464_UNI_MTS
+ if( nNumTransformCands > 1 )
+ {
+ xIntraCodingTUBlock( tu, COMPONENT_Y, false, singleDistTmpLuma, default0Save1Load2, &numSig, modeId == 0 ? &trModes : nullptr, true );
+ if( modeId == 0 )
+ {
+ for( int i = 0; i < nNumTransformCands; i++ )
+ {
+ if( trModes[i].second )
+ {
+ lastCheckId = trModes[i].first;
+ }
+ }
+ }
+ }
+ else
+ {
+ xIntraCodingTUBlock( tu, COMPONENT_Y, false, singleDistTmpLuma, default0Save1Load2, &numSig );
+ }
+#else
if (cu.emtFlag)
{
tu.emtIdx = transformIndex;
@@ -1458,10 +1564,15 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
}
xIntraCodingTUBlock( tu, COMPONENT_Y, false, singleDistTmpLuma, default0Save1Load2, &numSig );
+#endif
//----- determine rate and r-d cost -----
+#if JVET_M0464_UNI_MTS
+ if( ( trModes[modeId].first != 0 && !TU::getCbfAtDepth( tu, COMPONENT_Y, currDepth ) ) )
+#else
//the condition (transformIndex != DCT2_EMT) seems to be irrelevant, since DCT2_EMT=7 and the highest value of transformIndex is 4
if( ( modeId == lastCheckId && checkTransformSkip && !TU::getCbfAtDepth( tu, COMPONENT_Y, currDepth ) ) )
+#endif
{
//In order not to code TS flag when cbf is zero, the case for TS with cbf being zero is forbidden.
singleCostTmp = MAX_DOUBLE;
@@ -1478,9 +1589,16 @@ void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &par
uiSingleDistLuma = singleDistTmpLuma;
singleFracBits = singleTmpFracBits;
+#if JVET_M0464_UNI_MTS
+ bestModeId[COMPONENT_Y] = trModes[modeId].first;
+ if( trModes[modeId].first == 0 )
+ {
+ cbfDCT2 = TU::getCbfAtDepth( tu, COMPONENT_Y, currDepth );
+ }
+#else
bestModeId[COMPONENT_Y] = modeId;
cbfBestMode = TU::getCbfAtDepth( tu, COMPONENT_Y, currDepth );
-
+#endif
if( bestModeId[COMPONENT_Y] != lastCheckId )
{
@@ -1604,7 +1722,9 @@ ChromaCbfs IntraSearch::xRecurIntraChromaCodingQT(CodingStructure &cs, Partition
TransformUnit &currTU = *cs.getTU( currArea.chromaPos(), CHANNEL_TYPE_CHROMA );
const PredictionUnit &pu = *cs.getPU( currArea.chromaPos(), CHANNEL_TYPE_CHROMA );
+#if !JVET_M0464_UNI_MTS
const TransformUnit &currTULuma = CS::isDualITree( cs ) ? *cs.picture->cs->getTU( currArea.lumaPos(), CHANNEL_TYPE_LUMA ) : currTU;
+#endif
uint32_t currDepth = partitioner.currTrDepth;
const PPS &pps = *cs.pps;
@@ -1617,6 +1737,7 @@ ChromaCbfs IntraSearch::xRecurIntraChromaCodingQT(CodingStructure &cs, Partition
return cbfs;
}
+#if !JVET_M0464_UNI_MTS
bool checkTransformSkip = pps.getUseTransformSkip();
checkTransformSkip &= TU::hasTransformSkipFlag( *currTU.cs, partitioner.currArea().Cb() );
@@ -1638,6 +1759,7 @@ ChromaCbfs IntraSearch::xRecurIntraChromaCodingQT(CodingStructure &cs, Partition
checkTransformSkip &= ( nbLumaSkip > 0 );
}
}
+#endif
CodingStructure &saveCS = *m_pSaveCS[1];
saveCS.pcv = cs.pcv;
@@ -1666,8 +1788,12 @@ ChromaCbfs IntraSearch::xRecurIntraChromaCodingQT(CodingStructure &cs, Partition
const bool checkCrossComponentPrediction = PU::isChromaIntraModeCrossCheckMode( pu ) && pps.getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() && TU::getCbf( currTU, COMPONENT_Y );
const int crossCPredictionModesToTest = checkCrossComponentPrediction ? 2 : 1;
+#if JVET_M0464_UNI_MTS
+ const int totalModesToTest = crossCPredictionModesToTest;
+#else
const int transformSkipModesToTest = checkTransformSkip ? 2 : 1;
const int totalModesToTest = crossCPredictionModesToTest * transformSkipModesToTest;
+#endif
const bool isOneMode = (totalModesToTest == 1);
int currModeId = 0;
@@ -1681,12 +1807,16 @@ ChromaCbfs IntraSearch::xRecurIntraChromaCodingQT(CodingStructure &cs, Partition
ctxStart = m_CABACEstimator->getCtx();
}
+#if !JVET_M0464_UNI_MTS
for (int transformSkipModeId = 0; transformSkipModeId < transformSkipModesToTest; transformSkipModeId++)
+#endif
{
for (int crossCPredictionModeId = 0; crossCPredictionModeId < crossCPredictionModesToTest; crossCPredictionModeId++)
{
currTU.compAlpha [compID] = 0;
+#if !JVET_M0464_UNI_MTS
currTU.transformSkip[compID] = transformSkipModeId;
+#endif
currModeId++;
@@ -1697,7 +1827,11 @@ ChromaCbfs IntraSearch::xRecurIntraChromaCodingQT(CodingStructure &cs, Partition
{
default0Save1Load2 = 0;
}
+#if JVET_M0464_UNI_MTS
+ else if (!isOneMode && (crossCPredictionModeId == 0))
+#else
else if (!isOneMode && (transformSkipModeId == 0) && (crossCPredictionModeId == 0))
+#endif
{
default0Save1Load2 = 1; //save prediction on first mode
}
@@ -1715,7 +1849,11 @@ ChromaCbfs IntraSearch::xRecurIntraChromaCodingQT(CodingStructure &cs, Partition
xIntraCodingTUBlock( currTU, compID, crossCPredictionModeId != 0, singleDistCTmp, default0Save1Load2 );
+#if JVET_M0464_UNI_MTS
+ if( ( ( crossCPredictionModeId == 1 ) && ( currTU.compAlpha[compID] == 0 ) ) ) //In order not to code TS flag when cbf is zero, the case for TS with cbf being zero is forbidden.
+#else
if( ( ( crossCPredictionModeId == 1 ) && ( currTU.compAlpha[compID] == 0 ) ) || ( ( transformSkipModeId == 1 ) && !TU::getCbf( currTU, compID ) ) ) //In order not to code TS flag when cbf is zero, the case for TS with cbf being zero is forbidden.
+#endif
{
singleCostTmp = MAX_DOUBLE;
}
diff --git a/source/Lib/EncoderLib/IntraSearch.h b/source/Lib/EncoderLib/IntraSearch.h
index 92448ab87457a0f0ef76183f100679f7adef3e60..3c1ce2b8137592b7a868bf67ed1d2ee70c9b9003 100644
--- a/source/Lib/EncoderLib/IntraSearch.h
+++ b/source/Lib/EncoderLib/IntraSearch.h
@@ -75,10 +75,12 @@ private:
CodingStructure **m_pSaveCS;
//cost variables for the EMT algorithm and new modes list
+#if !JVET_M0464_UNI_MTS
double m_bestModeCostStore[4]; // RD cost of the best mode for each PU using DCT2
double m_modeCostStore [4][NUM_LUMA_MODE]; // RD cost of each mode for each PU using DCT2
uint32_t m_savedRdModeList [4][NUM_LUMA_MODE], m_savedNumRdModes[4];
int m_savedExtendRefList[4][NUM_LUMA_MODE];
+#endif
protected:
// interface to option
@@ -142,7 +144,11 @@ protected:
void xEncCoeffQT (CodingStructure &cs, Partitioner& pm, const ComponentID &compID);
+#if JVET_M0464_UNI_MTS
+ void xIntraCodingTUBlock (TransformUnit &tu, const ComponentID &compID, const bool &checkCrossCPrediction, Distortion& ruiDist, const int &default0Save1Load2 = 0, uint32_t* numSig = nullptr, std::vector<TrMode>* trModes=nullptr, const bool loadTr=false );
+#else
void xIntraCodingTUBlock (TransformUnit &tu, const ComponentID &compID, const bool &checkCrossCPrediction, Distortion& ruiDist, const int &default0Save1Load2 = 0, uint32_t* numSig = nullptr );
+#endif
ChromaCbfs xRecurIntraChromaCodingQT (CodingStructure &cs, Partitioner& pm);
diff --git a/source/Lib/EncoderLib/VLCWriter.cpp b/source/Lib/EncoderLib/VLCWriter.cpp
index cbb7bbdc2b005623228053b7ad74defad7fa7256..34a8b308db02f6ede5d36002a8614f3214b78a6f 100644
--- a/source/Lib/EncoderLib/VLCWriter.cpp
+++ b/source/Lib/EncoderLib/VLCWriter.cpp
@@ -529,20 +529,29 @@ void HLSWriter::codeSPSNext( const SPSNext& spsNext, const bool usePCM )
{
// tool enabling flags
WRITE_FLAG( spsNext.getUseLargeCTU() ? 1 : 0, "large_ctu_flag" );
- WRITE_FLAG(spsNext.getUseSubPuMvp() ? 1 : 0, "subpu_tmvp_flag");
WRITE_FLAG( spsNext.getUseIMV() ? 1 : 0, "imv_enable_flag" );
- WRITE_FLAG( spsNext.getUseBIO() ? 1 : 0, "bio_enable_flag" );
WRITE_FLAG( spsNext.getDisableMotCompress() ? 1 : 0, "disable_motion_compression_flag" );
WRITE_FLAG( spsNext.getUseLMChroma() ? 1 : 0, "lm_chroma_enabled_flag" );
+#if JVET_M0142_CCLM_COLLOCATED_CHROMA
+ if ( spsNext.getUseLMChroma() && spsNext.getSPS().getChromaFormatIdc() == CHROMA_420 )
+ {
+ WRITE_FLAG( spsNext.getCclmCollocatedChromaFlag() ? 1 : 0, "sps_cclm_collocated_chroma_flag" );
+ }
+#endif
+#if JVET_M0464_UNI_MTS
+ WRITE_FLAG( spsNext.getUseIntraMTS() ? 1 : 0, "mts_intra_enabled_flag" );
+ WRITE_FLAG( spsNext.getUseInterMTS() ? 1 : 0, "mts_inter_enabled_flag" );
+#else
WRITE_FLAG( spsNext.getUseIntraEMT() ? 1 : 0, "emt_intra_enabled_flag" );
WRITE_FLAG( spsNext.getUseInterEMT() ? 1 : 0, "emt_inter_enabled_flag" );
+#endif
WRITE_FLAG( spsNext.getUseAffine() ? 1 : 0, "affine_flag" );
if ( spsNext.getUseAffine() )
{
WRITE_FLAG( spsNext.getUseAffineType() ? 1 : 0, "affine_type_flag" );
}
WRITE_FLAG( spsNext.getUseGBi() ? 1 : 0, "gbi_flag" );
- WRITE_FLAG(spsNext.getCPRMode() ? 1 : 0, "cpr_flag" );
+ WRITE_FLAG(spsNext.getIBCMode() ? 1 : 0, "ibc_flag" );
for( int k = 0; k < SPSNext::NumReservedFlags; k++ )
{
WRITE_FLAG( 0, "reserved_flag" );
@@ -684,23 +693,12 @@ void HLSWriter::codeSPS( const SPS* pcSPS )
}
WRITE_UVLC( pcSPS->getQuadtreeTULog2MinSize() - 2, "log2_min_luma_transform_block_size_minus2" );
WRITE_UVLC( pcSPS->getQuadtreeTULog2MaxSize() - pcSPS->getQuadtreeTULog2MinSize(), "log2_diff_max_min_luma_transform_block_size" );
- WRITE_FLAG( pcSPS->getUseALF(), "sps_alf_enable_flag" );
-#if HEVC_USE_SCALING_LISTS
- WRITE_FLAG( pcSPS->getScalingListFlag() ? 1 : 0, "scaling_list_enabled_flag" );
- if(pcSPS->getScalingListFlag())
- {
- WRITE_FLAG( pcSPS->getScalingListPresentFlag() ? 1 : 0, "sps_scaling_list_data_present_flag" );
- if(pcSPS->getScalingListPresentFlag())
- {
- codeScalingList( pcSPS->getScalingList() );
- }
- }
-#endif
- WRITE_FLAG( pcSPS->getUseAMP() ? 1 : 0, "amp_enabled_flag" );
- WRITE_FLAG( pcSPS->getUseSAO() ? 1 : 0, "sample_adaptive_offset_enabled_flag");
- WRITE_FLAG( pcSPS->getUsePCM() ? 1 : 0, "pcm_enabled_flag");
- if( pcSPS->getUsePCM() )
+ WRITE_FLAG( pcSPS->getSAOEnabledFlag(), "sps_sao_enabled_flag");
+ WRITE_FLAG( pcSPS->getALFEnabledFlag(), "sps_alf_enabled_flag" );
+
+ WRITE_FLAG( pcSPS->getPCMEnabledFlag() ? 1 : 0, "pcm_enabled_flag");
+ if( pcSPS->getPCMEnabledFlag() )
{
WRITE_CODE( pcSPS->getPCMBitDepth(CHANNEL_TYPE_LUMA) - 1, 4, "pcm_sample_bit_depth_luma_minus1" );
WRITE_CODE( chromaEnabled ? (pcSPS->getPCMBitDepth(CHANNEL_TYPE_CHROMA) - 1) : 0, 4, "pcm_sample_bit_depth_chroma_minus1" );
@@ -709,12 +707,33 @@ void HLSWriter::codeSPS( const SPS* pcSPS )
WRITE_FLAG( pcSPS->getPCMFilterDisableFlag()?1 : 0, "pcm_loop_filter_disable_flag");
}
- WRITE_FLAG( pcSPS->getUseWrapAround() ? 1 : 0, "ref_wraparound_enabled_flag" );
- if( pcSPS->getUseWrapAround() )
+ WRITE_FLAG( pcSPS->getWrapAroundEnabledFlag() ? 1 : 0, "sps_ref_wraparound_enabled_flag" );
+ if( pcSPS->getWrapAroundEnabledFlag() )
{
- WRITE_UVLC( pcSPS->getWrapAroundOffset(), "ref_wraparound_offset" );
+ WRITE_UVLC( pcSPS->getWrapAroundOffset(), "sps_ref_wraparound_offset" );
}
+ WRITE_FLAG( pcSPS->getSPSTemporalMVPEnabledFlag() ? 1 : 0, "sps_temporal_mvp_enabled_flag" );
+
+ if ( pcSPS->getSPSTemporalMVPEnabledFlag() )
+ {
+ WRITE_FLAG( pcSPS->getSBTMVPEnabledFlag() ? 1 : 0, "sps_sbtmvp_enabled_flag");
+ }
+
+ WRITE_FLAG( pcSPS->getBDOFEnabledFlag() ? 1 : 0, "sps_bdof_enabled_flag" );
+
+#if HEVC_USE_SCALING_LISTS
+ WRITE_FLAG( pcSPS->getScalingListFlag() ? 1 : 0, "scaling_list_enabled_flag" );
+ if(pcSPS->getScalingListFlag())
+ {
+ WRITE_FLAG( pcSPS->getScalingListPresentFlag() ? 1 : 0, "sps_scaling_list_data_present_flag" );
+ if(pcSPS->getScalingListPresentFlag())
+ {
+ codeScalingList( pcSPS->getScalingList() );
+ }
+ }
+#endif
+
CHECK( pcSPS->getMaxTLayers() == 0, "Maximum number of T-layers is '0'" );
const RPSList* rpsList = pcSPS->getRPSList();
@@ -735,7 +754,6 @@ void HLSWriter::codeSPS( const SPS* pcSPS )
WRITE_FLAG( pcSPS->getUsedByCurrPicLtSPSFlag(k), "used_by_curr_pic_lt_sps_flag[i]");
}
}
- WRITE_FLAG( pcSPS->getSPSTemporalMVPEnabledFlag() ? 1 : 0, "sps_temporal_mvp_enabled_flag" );
#if HEVC_USE_INTRA_SMOOTHING_T32 || HEVC_USE_INTRA_SMOOTHING_T64
WRITE_FLAG( pcSPS->getUseStrongIntraSmoothing(), "strong_intra_smoothing_enable_flag" );
@@ -805,7 +823,7 @@ void HLSWriter::codeSPS( const SPS* pcSPS )
}
case SPS_EXT__NEXT:
{
- codeSPSNext( pcSPS->getSpsNext(), pcSPS->getUsePCM() );
+ codeSPSNext( pcSPS->getSpsNext(), pcSPS->getPCMEnabledFlag() );
break;
}
default:
@@ -1083,7 +1101,7 @@ void HLSWriter::codeSliceHeader ( Slice* pcSlice )
WRITE_FLAG( pcSlice->getEnableTMVPFlag() ? 1 : 0, "slice_temporal_mvp_enabled_flag" );
}
}
- if( pcSlice->getSPS()->getUseSAO() )
+ if( pcSlice->getSPS()->getSAOEnabledFlag() )
{
WRITE_FLAG( pcSlice->getSaoEnabledFlag( CHANNEL_TYPE_LUMA ), "slice_sao_luma_flag" );
if( chromaEnabled )
@@ -1092,7 +1110,7 @@ void HLSWriter::codeSliceHeader ( Slice* pcSlice )
}
}
- if( pcSlice->getSPS()->getUseALF() )
+ if( pcSlice->getSPS()->getALFEnabledFlag() )
{
alf( pcSlice->getAlfSliceParam() );
}
@@ -1252,12 +1270,12 @@ void HLSWriter::codeSliceHeader ( Slice* pcSlice )
CHECK( pcSlice->getMaxNumMergeCand() > MRG_MAX_NUM_CANDS, "More merge candidates signalled than supported" );
WRITE_UVLC( MRG_MAX_NUM_CANDS - pcSlice->getMaxNumMergeCand(), "six_minus_max_num_merge_cand" );
- if ( pcSlice->getSPS()->getSpsNext().getUseSubPuMvp() && !pcSlice->getSPS()->getSpsNext().getUseAffine() ) // ATMVP only
+ if ( pcSlice->getSPS()->getSBTMVPEnabledFlag() && !pcSlice->getSPS()->getSpsNext().getUseAffine() ) // ATMVP only
{
CHECK( pcSlice->getMaxNumAffineMergeCand() != 1, "Sub-block merge can number should be 1" );
}
else
- if ( !pcSlice->getSPS()->getSpsNext().getUseSubPuMvp() && !pcSlice->getSPS()->getSpsNext().getUseAffine() ) // both off
+ if ( !pcSlice->getSPS()->getSBTMVPEnabledFlag() && !pcSlice->getSPS()->getSpsNext().getUseAffine() ) // both off
{
CHECK( pcSlice->getMaxNumAffineMergeCand() != 0, "Sub-block merge can number should be 0" );
}
@@ -1305,7 +1323,7 @@ void HLSWriter::codeSliceHeader ( Slice* pcSlice )
}
}
- bool isSAOEnabled = pcSlice->getSPS()->getUseSAO() && (pcSlice->getSaoEnabledFlag(CHANNEL_TYPE_LUMA) || (chromaEnabled && pcSlice->getSaoEnabledFlag(CHANNEL_TYPE_CHROMA)));
+ bool isSAOEnabled = pcSlice->getSPS()->getSAOEnabledFlag() && (pcSlice->getSaoEnabledFlag(CHANNEL_TYPE_LUMA) || (chromaEnabled && pcSlice->getSaoEnabledFlag(CHANNEL_TYPE_CHROMA)));
bool isDBFEnabled = (!pcSlice->getDeblockingFilterDisable());
if(pcSlice->getPPS()->getLoopFilterAcrossSlicesEnabledFlag() && ( isSAOEnabled || isDBFEnabled ))