diff --git a/cfg/encoder_lowdelay_P_ecm.cfg b/cfg/encoder_lowdelay_P_ecm.cfg
index 54500d5bd1f512bed1ecf1fbb86024f80add8a2e..575047c4edd26a14737f1d5c3b35b925e487be91 100644
--- a/cfg/encoder_lowdelay_P_ecm.cfg
+++ b/cfg/encoder_lowdelay_P_ecm.cfg
@@ -100,7 +100,7 @@ MaxMTTHierarchyDepth : 3
MaxMTTHierarchyDepthISliceL : 3
MaxMTTHierarchyDepthISliceC : 3
-MTS : 1
+MTS : 3
MTSIntraMaxCand : 3
MTSInterMaxCand : 4
SBT : 1
diff --git a/cfg/encoder_lowdelay_ecm.cfg b/cfg/encoder_lowdelay_ecm.cfg
index c82ae77afcfd935a3496762c90dd20e392a45ae0..f949e38353877347f3fd3779b113dc49e4ccd9b4 100644
--- a/cfg/encoder_lowdelay_ecm.cfg
+++ b/cfg/encoder_lowdelay_ecm.cfg
@@ -100,7 +100,7 @@ MaxMTTHierarchyDepth : 3
MaxMTTHierarchyDepthISliceL : 3
MaxMTTHierarchyDepthISliceC : 3
-MTS : 1
+MTS : 3
MTSIntraMaxCand : 3
MTSInterMaxCand : 4
SBT : 1
diff --git a/cfg/encoder_randomaccess_ecm.cfg b/cfg/encoder_randomaccess_ecm.cfg
index 58c6de3442c0a8e98f6673765e1decde798be227..e20ad326cded71c4868a671917bc5fc2fe788c3d 100644
--- a/cfg/encoder_randomaccess_ecm.cfg
+++ b/cfg/encoder_randomaccess_ecm.cfg
@@ -127,7 +127,7 @@ MaxMTTHierarchyDepth : 3
MaxMTTHierarchyDepthISliceL : 3
MaxMTTHierarchyDepthISliceC : 3
-MTS : 1
+MTS : 3
MTSIntraMaxCand : 4
MTSInterMaxCand : 4
SBT : 1
diff --git a/source/App/EncoderApp/EncApp.cpp b/source/App/EncoderApp/EncApp.cpp
index a815a0cc86168613c83a93758251f6540328104a..76fd1f61365192974d7c829703e6f738073e58d8 100644
--- a/source/App/EncoderApp/EncApp.cpp
+++ b/source/App/EncoderApp/EncApp.cpp
@@ -814,6 +814,9 @@ void EncApp::xInitLibCfg()
}
}
#endif
+#if JVET_AA0133_INTER_MTS_OPT
+ m_cEncLib.setInterMTSMaxSize(m_interMTSMaxSize);
+#endif
#if ENABLE_DIMD
m_cEncLib.setUseDimd ( m_dimd );
#endif
diff --git a/source/App/EncoderApp/EncAppCfg.cpp b/source/App/EncoderApp/EncAppCfg.cpp
index a78e951a6614e3dfede64c4cc91b913aae244df6..3bd83877dd86dda4d02e6ae9873355cc7a2288c4 100644
--- a/source/App/EncoderApp/EncAppCfg.cpp
+++ b/source/App/EncoderApp/EncAppCfg.cpp
@@ -1059,6 +1059,9 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
("LadfQpOffset", cfg_LadfQpOffset, cfg_LadfQpOffset, "LADF QP offset")
("LadfIntervalLowerBound", cfg_LadfIntervalLowerBound, cfg_LadfIntervalLowerBound, "LADF lower bound for 2nd lowest interval")
#endif
+#if JVET_AA0133_INTER_MTS_OPT
+ ("InterMTSMaxSize", m_interMTSMaxSize, 32, "InterMTSMaxSize")
+#endif
#if ENABLE_DIMD
( "DIMD", m_dimd, true, "Enable decoder side intra mode derivation\n" )
#endif
@@ -2800,7 +2803,13 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
}
}
#endif
-
+#if JVET_AA0133_INTER_MTS_OPT
+#if JVET_AA0146_WRAP_AROUND_FIX
+ m_interMTSMaxSize = (m_sourceHeight > 1080) ? 32 : 16;
+#else
+ m_interMTSMaxSize = (m_iSourceHeight > 1080)? 32 : 16;
+#endif
+#endif
if (m_chromaFormatIDC != CHROMA_420)
{
if (!m_horCollocatedChromaFlag)
@@ -4879,6 +4888,12 @@ void EncAppCfg::xPrintParameter()
msg( VERBOSE, "HorCollocatedChroma:%d ", m_horCollocatedChromaFlag );
msg( VERBOSE, "VerCollocatedChroma:%d ", m_verCollocatedChromaFlag );
msg( VERBOSE, "MTS: %1d(intra) %1d(inter) ", m_MTS & 1, ( m_MTS >> 1 ) & 1 );
+#if JVET_AA0133_INTER_MTS_OPT
+ if ((m_MTS >> 1) & 1)
+ {
+ msg(VERBOSE, "InterMTSMaxSize: %d ", m_interMTSMaxSize);
+ }
+#endif
msg( VERBOSE, "SBT:%d ", m_SBT );
msg( VERBOSE, "ISP:%d ", m_ISP );
msg( VERBOSE, "SMVD:%d ", m_SMVD );
diff --git a/source/App/EncoderApp/EncAppCfg.h b/source/App/EncoderApp/EncAppCfg.h
index 7b0d539e19974f16303466bb4a5a53f2dc52d600..4170416e34b9a67ab0e1cda53f45e7f7a351fe0a 100644
--- a/source/App/EncoderApp/EncAppCfg.h
+++ b/source/App/EncoderApp/EncAppCfg.h
@@ -424,6 +424,9 @@ protected:
std::vector<int> m_LadfQpOffset;
int m_LadfIntervalLowerBound[MAX_LADF_INTERVALS];
#endif
+#if JVET_AA0133_INTER_MTS_OPT
+ int m_interMTSMaxSize;
+#endif
#if ENABLE_DIMD
bool m_dimd;
#endif
diff --git a/source/Lib/CommonLib/Rom.cpp b/source/Lib/CommonLib/Rom.cpp
index 475a5bc33ee5764455010dd29ba56fd99bc55952..b69de50e4d8427cc1767dd8d28027949b04cbd86 100644
--- a/source/Lib/CommonLib/Rom.cpp
+++ b/source/Lib/CommonLib/Rom.cpp
@@ -4572,6 +4572,110 @@ void initROM()
}
c <<= 1;
}
+
+////////////////////////////////////////////////////////////////////////////////////////////////
+#if JVET_AA0133_INTER_MTS_OPT
+ TMatrixCoeff KLT4[2][4][4] = {
+ {
+ { -79, -88, -46, -17},
+ { 74, -10, -86, -58},
+ { -59, 69, 4, -91},
+ { -35, 62, -82, 67},
+ },
+ {
+ { 3, 19, 71, 105},
+ { 27, 90, 64, -60},
+ { -91, -45, 69, -36},
+ { 86, -77, 51, -23},
+ },
+ };
+ TMatrixCoeff KLT8[2][8][8] = {
+ {
+ { 55, 81, 92, 87, 68, 43, 25, 15},
+ { -84, -87, -28, 46, 82, 72, 49, 31},
+ { 75, 32, -63, -79, -2, 71, 84, 63},
+ { 80, -15, -88, 15, 88, 11, -68, -75},
+ { 65, -52, -37, 83, -13, -92, 20, 92},
+ { 63, -83, 33, 39, -82, 53, 53, -83},
+ { -45, 76, -75, 53, -12, -46, 96, -71},
+ { 22, -44, 61, -75, 83, -85, 75, -38},
+ },
+ {
+ { 8, 16, 31, 53, 74, 88, 93, 83},
+ { 23, 43, 70, 90, 71, 7, -64, -91},
+ { -53, -83, -83, -19, 67, 77, -6, -72},
+ { -73, -79, -4, 86, 31, -82, -36, 70},
+ { -93, -33, 89, 22, -83, 27, 62, -55},
+ { -88, 45, 58, -82, 35, 41, -86, 52},
+ { -77, 95, -43, -13, 54, -77, 75, -37},
+ { -42, 75, -83, 82, -75, 62, -43, 19},
+ },
+ };
+ TMatrixCoeff KLT16[2][16][16] = {
+ {
+ { 42, 57, 71, 82, 90, 94, 93, 88, 79, 64, 50, 38, 28, 20, 15, 11},
+ { -72, -91, -95, -83, -55, -16, 26, 61, 80, 83, 75, 63, 50, 39, 31, 23},
+ { 71, 78, 56, 12, -39, -77, -87, -65, -20, 31, 68, 84, 84, 75, 63, 49},
+ { 87, 77, 22, -47, -89, -76, -15, 57, 88, 67, 17, -31, -62, -75, -74, -61},
+ { 80, 51, -24, -78, -62, 12, 77, 73, 3, -73, -90, -49, 12, 62, 88, 84},
+ { 89, 29, -68, -85, -1, 84, 64, -37, -87, -31, 53, 84, 48, -16, -68, -82},
+ { 80, -1, -84, -36, 68, 62, -44, -82, 10, 90, 41, -58, -89, -35, 51, 93},
+ { 83, -31, -91, 24, 91, -21, -90, 20, 85, -15, -83, -15, 73, 67, -19, -82},
+ { 66, -50, -57, 69, 34, -82, -5, 88, -25, -87, 41, 88, -29, -96, -10, 87},
+ { 67, -72, -27, 91, -35, -67, 83, 7, -85, 52, 56, -78, -37, 82, 38, -74},
+ { 57, -80, 16, 62, -77, 17, 59, -83, 27, 63, -89, 5, 93, -54, -71, 76},
+ { 55, -92, 57, 17, -77, 82, -32, -38, 81, -68, 4, 72, -81, -3, 91, -66},
+ { 43, -82, 77, -37, -19, 68, -87, 73, -28, -32, 79, -81, 25, 58, -100, 58},
+ { -28, 60, -71, 64, -37, 0, 43, -79, 98, -96, 71, -21, -40, 84, -88, 43},
+ { 30, -67, 92, -106, 108, -101, 88, -71, 55, -40, 25, -11, 0, 7, -8, 4},
+ { 6, -13, 17, -19, 17, -10, 0, 16, -36, 63, -91, 115, -125, 115, -84, 36},
+ },
+ {
+ { 6, 10, 15, 22, 30, 39, 51, 63, 73, 81, 86, 90, 91, 89, 83, 74},
+ { -19, -28, -39, -50, -63, -75, -84, -87, -77, -53, -20, 18, 53, 80, 95, 93},
+ { 43, 60, 75, 85, 86, 73, 41, -5, -49, -80, -86, -64, -21, 28, 68, 81},
+ { 58, 75, 78, 66, 36, -12, -61, -88, -69, -9, 57, 92, 75, 15, -54, -89},
+ { 83, 92, 66, 14, -48, -89, -75, -7, 64, 83, 34, -42, -81, -50, 26, 78},
+ { 83, 72, 17, -49, -84, -52, 32, 88, 47, -51, -89, -25, 70, 85, -2, -83},
+ { 92, 54, -35, -90, -54, 45, 90, 14, -78, -54, 49, 80, -13, -88, -24, 73},
+ { 83, 23, -68, -74, 22, 87, 12, -87, -25, 87, 36, -83, -44, 80, 50, -71},
+ { 87, -10, -95, -22, 91, 34, -89, -23, 89, 1, -85, 21, 78, -43, -62, 58},
+ { 76, -37, -80, 42, 74, -61, -47, 86, -5, -86, 60, 48, -89, 10, 82, -58},
+ { 79, -71, -52, 94, -3, -86, 68, 24, -85, 57, 26, -81, 53, 29, -81, 48},
+ { 68, -90, 7, 78, -78, 5, 66, -84, 42, 34, -84, 71, -6, -64, 89, -45},
+ { 59, -98, 60, 19, -79, 84, -41, -21, 70, -88, 68, -15, -43, 80, -78, 36},
+ { 44, -87, 86, -47, -9, 61, -91, 97, -80, 43, 3, -44, 70, -75, 59, -25},
+ { 5, -8, 6, 2, -13, 28, -43, 59, -76, 92, -105, 110, -103, 86, -59, 23},
+ { 36, -81, 111, -122, 114, -94, 67, -41, 18, 0, -14, 22, -25, 22, -16, 7},
+ },
+ };
+
+ for (int i = 0; i < 4; i++)
+ {
+ for (int j = 0; j < 4; j++)
+ {
+ g_aiTr4[KLT0][i][j] = KLT4[0][i][j];
+ g_aiTr4[KLT1][i][j] = KLT4[1][i][j];
+ }
+ }
+ for (int i = 0; i < 8; i++)
+ {
+ for (int j = 0; j < 8; j++)
+ {
+ g_aiTr8[KLT0][i][j] = KLT8[0][i][j];
+ g_aiTr8[KLT1][i][j] = KLT8[1][i][j];
+ }
+ }
+ for (int i = 0; i < 16; i++)
+ {
+ for (int j = 0; j < 16; j++)
+ {
+ g_aiTr16[KLT0][i][j] = KLT16[0][i][j];
+ g_aiTr16[KLT1][i][j] = KLT16[1][i][j];
+ }
+ }
+#endif
+////////////////////////////////////////////////////////////////////////////////////////////////
#endif
#if JVET_AA0107_RMVF_AFFINE_MERGE_DERIVATION
g_rmvfMultApproxTbl[0] = 0;
diff --git a/source/Lib/CommonLib/Slice.cpp b/source/Lib/CommonLib/Slice.cpp
index ab0b8a20f9a8cff488b6c2d06cf70fec8a326ce9..08eea3167478e203e4dedb42516039880786033e 100644
--- a/source/Lib/CommonLib/Slice.cpp
+++ b/source/Lib/CommonLib/Slice.cpp
@@ -3604,7 +3604,9 @@ SPS::SPS()
, m_LadfQpOffset { 0 }
, m_LadfIntervalLowerBound { 0 }
#endif
-
+#if JVET_AA0133_INTER_MTS_OPT
+, m_interMTSMaxSize ( 32 )
+#endif
#if MULTI_HYP_PRED
, m_InterMultiHyp(false)
, m_maxNumAddHyps(0)
diff --git a/source/Lib/CommonLib/Slice.h b/source/Lib/CommonLib/Slice.h
index 477d3b69177c3c85c77e039009c9dc2cfd4bc31d..1ca1eeebbc16eba3af3dea0c4ecce3e5c2bd663a 100644
--- a/source/Lib/CommonLib/Slice.h
+++ b/source/Lib/CommonLib/Slice.h
@@ -1715,7 +1715,9 @@ private:
int m_LadfQpOffset[MAX_LADF_INTERVALS];
int m_LadfIntervalLowerBound[MAX_LADF_INTERVALS];
#endif
-
+#if JVET_AA0133_INTER_MTS_OPT
+ int m_interMTSMaxSize;
+#endif
#if MULTI_HYP_PRED
bool m_InterMultiHyp; // multi hypothesis inter prediction
int m_maxNumAddHyps;
@@ -2173,7 +2175,10 @@ void setCCALFEnabledFlag( bool b )
void setLadfIntervalLowerBound( int value, int idx ) { m_LadfIntervalLowerBound[ idx ] = value; }
int getLadfIntervalLowerBound( int idx ) const { return m_LadfIntervalLowerBound[ idx ]; }
#endif
-
+#if JVET_AA0133_INTER_MTS_OPT
+ void setInterMTSMaxSize(int size) { m_interMTSMaxSize = size; }
+ int getInterMTSMaxSize() const { return m_interMTSMaxSize; }
+#endif
#if MULTI_HYP_PRED
bool getUseInterMultiHyp() const { return m_InterMultiHyp; }
int getMaxNumAddHyps() const { return m_maxNumAddHyps; }
diff --git a/source/Lib/CommonLib/TrQuant.cpp b/source/Lib/CommonLib/TrQuant.cpp
index 05bb11fabb968624fa6073d7bd39ee3d49486e8b..f8b39827d48ff6fe169d73854f7c3c7af5261caa 100644
--- a/source/Lib/CommonLib/TrQuant.cpp
+++ b/source/Lib/CommonLib/TrQuant.cpp
@@ -245,6 +245,10 @@ void TrQuant::init( const Quant* otherQuant,
{ nullptr, fastForwardDST4_B4, fastForwardDST4_B8, fastForwardDST4_B16, fastForwardDST4_B32, fastForwardDST4_B64, fastForwardDST4_B128, fastForwardDST4_B256 },
{ nullptr, fastForwardDST1_B4, fastForwardDST1_B8, fastForwardDST1_B16, fastForwardDST1_B32, fastForwardDST1_B64, fastForwardDST1_B128, fastForwardDST1_B256 },
{ nullptr, fastForwardIDTR_B4, fastForwardIDTR_B8, fastForwardIDTR_B16, fastForwardIDTR_B32, fastForwardIDTR_B64, fastForwardIDTR_B128, fastForwardIDTR_B256 },
+#if JVET_AA0133_INTER_MTS_OPT
+ {nullptr, fastForwardKLT0_B4, fastForwardKLT0_B8, fastForwardKLT0_B16, nullptr, nullptr, nullptr, nullptr },
+ {nullptr, fastForwardKLT1_B4, fastForwardKLT1_B8, fastForwardKLT1_B16, nullptr, nullptr, nullptr, nullptr },
+#endif
#endif
} };
@@ -258,6 +262,10 @@ void TrQuant::init( const Quant* otherQuant,
{ nullptr, fastInverseDST4_B4, fastInverseDST4_B8, fastInverseDST4_B16, fastInverseDST4_B32, fastInverseDST4_B64, fastInverseDST4_B128, fastInverseDST4_B256 },
{ nullptr, fastInverseDST1_B4, fastInverseDST1_B8, fastInverseDST1_B16, fastInverseDST1_B32, fastInverseDST1_B64, fastInverseDST1_B128, fastInverseDST1_B256 },
{ nullptr, fastInverseIDTR_B4, fastInverseIDTR_B8, fastInverseIDTR_B16, fastInverseIDTR_B32, fastInverseIDTR_B64, fastInverseIDTR_B128, fastInverseIDTR_B256 },
+#if JVET_AA0133_INTER_MTS_OPT
+ {nullptr, fastInverseKLT0_B4, fastInverseKLT0_B8, fastInverseKLT0_B16, nullptr, nullptr, nullptr, nullptr },
+ {nullptr, fastInverseKLT1_B4, fastInverseKLT1_B8, fastInverseKLT1_B16, nullptr, nullptr, nullptr, nullptr },
+#endif
#endif
} };
#else
@@ -1167,6 +1175,16 @@ void TrQuant::getTrTypes(const TransformUnit tu, const ComponentID compID, int &
int indVer = (tu.mtsIdx[compID] - MTS_DST7_DST7) >> 1;
trTypeHor = indHor ? DCT8 : DST7;
trTypeVer = indVer ? DCT8 : DST7;
+#if JVET_AA0133_INTER_MTS_OPT
+ uint32_t width = tu.blocks[compID].width;
+ uint32_t height = tu.blocks[compID].height;
+ CHECK(width < 4 || height < 4, "width < 4 || height < 4 for KLT");
+ if (width <= 16 && height <= 16)
+ {
+ trTypeHor = indHor ? KLT1 : KLT0;
+ trTypeVer = indVer ? KLT1 : KLT0;
+ }
+#endif
}
}
}
@@ -1409,10 +1427,12 @@ void TrQuant::transformNxN( TransformUnit& tu, const ComponentID& compID, const
#else
const double facBB[] = { 1.2, 1.3, 1.3, 1.4, 1.5 };
#endif
+
while( it != trModes->end() )
{
tu.mtsIdx[compID] = it->first;
CoeffBuf tempCoeff( m_mtsCoeffs[tu.mtsIdx[compID]], rect);
+
if( tu.noResidual )
{
int sumAbs = 0;
@@ -1420,7 +1440,6 @@ void TrQuant::transformNxN( TransformUnit& tu, const ComponentID& compID, const
it++;
continue;
}
-
if ( tu.mtsIdx[compID] == MTS_SKIP )
{
xTransformSkip( tu, compID, resiBuf, tempCoeff.buf );
@@ -1451,7 +1470,6 @@ void TrQuant::transformNxN( TransformUnit& tu, const ComponentID& compID, const
tu.cu->slice->getSPS()->getMaxLog2TrDynamicRange(toChannelType(compID)));
scaleSAD *= pow(2, trShift);
}
-
#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT_VS
trCosts.push_back( TrCost( int(std::min<double>(sumAbs*scaleSAD, std::numeric_limits<int>::max())), pos++ ) );
#else
@@ -1459,7 +1477,25 @@ void TrQuant::transformNxN( TransformUnit& tu, const ComponentID& compID, const
#endif
it++;
}
-
+#if JVET_AA0133_INTER_MTS_OPT
+ if (CU::isInter(*tu.cu) && tu.cu->mtsFlag && compID == COMPONENT_Y)
+ {
+ std::stable_sort(trCosts.begin(), trCosts.end(), [](const TrCost l, const TrCost r) {return l.first < r.first; });
+ std::vector<TrMode> trModesTemp;
+ trModesTemp.resize(trModes->size());
+ for (int i = 0; i < trModes->size(); i++)
+ {
+ trModesTemp[i] = trModes->at(i);
+ }
+ for (int i = 0; i < trModes->size(); i++)
+ {
+ int index = trCosts[i].second;
+ trModes->at(i) = trModesTemp[index];
+ }
+ trModesTemp.resize(0);
+ return;
+ }
+#endif
int numTests = 0;
std::vector<TrCost>::iterator itC = trCosts.begin();
const double fac = facBB[std::max(0, floorLog2(std::max(width, height)) - 2)];
diff --git a/source/Lib/CommonLib/TrQuant_EMT.cpp b/source/Lib/CommonLib/TrQuant_EMT.cpp
index ef0050b970e5e459b0454840ed5c66968b6650ad..10af6a81c9a766a44186bb9cadbca34922dae917 100644
--- a/source/Lib/CommonLib/TrQuant_EMT.cpp
+++ b/source/Lib/CommonLib/TrQuant_EMT.cpp
@@ -2258,4 +2258,66 @@ void fastInverseIDTR_B256(const TCoeff *src, TCoeff *dst, int shift, int line, i
{
_fastInverseMM< 256 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr256[IDTR][0]);
}
+#if JVET_AA0133_INTER_MTS_OPT
+//KLT0
+void fastForwardKLT0_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr4[KLT0][0]);
+}
+
+void fastInverseKLT0_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr4[KLT0][0]);
+}
+
+void fastForwardKLT0_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr8[KLT0][0]);
+}
+
+void fastInverseKLT0_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr8[KLT0][0]);
+}
+
+void fastForwardKLT0_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr16[KLT0][0]);
+}
+
+void fastInverseKLT0_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr16[KLT0][0]);
+}
+//KLT1
+void fastForwardKLT1_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr4[KLT1][0]);
+}
+
+void fastInverseKLT1_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 4 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr4[KLT1][0]);
+}
+
+void fastForwardKLT1_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr8[KLT1][0]);
+}
+
+void fastInverseKLT1_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 8 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr8[KLT1][0]);
+}
+
+void fastForwardKLT1_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
+{
+ _fastForwardMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, g_aiTr16[KLT1][0]);
+}
+
+void fastInverseKLT1_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
+{
+ _fastInverseMM< 16 >(src, dst, shift, line, iSkipLine, iSkipLine2, outputMinimum, outputMaximum, g_aiTr16[KLT1][0]);
+}
+#endif
#endif
diff --git a/source/Lib/CommonLib/TrQuant_EMT.h b/source/Lib/CommonLib/TrQuant_EMT.h
index 5a97b41b6907b466f41e52b5d316b24832b2bd62..a2a71b8fea09d371b5b0734ad21e9b511326babd 100644
--- a/source/Lib/CommonLib/TrQuant_EMT.h
+++ b/source/Lib/CommonLib/TrQuant_EMT.h
@@ -169,6 +169,23 @@ void fastInverseIDTR_B128(const TCoeff *src, TCoeff *dst, int shift, int line, i
void fastForwardIDTR_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
void fastInverseIDTR_B256(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
#endif
+#if JVET_AA0133_INTER_MTS_OPT
+//KLT0 transforms
+void fastForwardKLT0_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseKLT0_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardKLT0_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseKLT0_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardKLT0_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseKLT0_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+
+//KLT1 transforms
+void fastForwardKLT1_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseKLT1_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardKLT1_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseKLT1_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+void fastForwardKLT1_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2);
+void fastInverseKLT1_B16(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum);
+#endif
#endif
#endif // __TRQUANT__
diff --git a/source/Lib/CommonLib/TypeDef.h b/source/Lib/CommonLib/TypeDef.h
index 4dff1f1e01fa7f52c2d7dbca3d74d69e427fea8e..486ccfba4aa6e4dd55bd3a06daf44cb769947051 100644
--- a/source/Lib/CommonLib/TypeDef.h
+++ b/source/Lib/CommonLib/TypeDef.h
@@ -236,6 +236,7 @@
#define JVET_Y0159_INTER_MTS 1 // JVET-Y0159: Inter MTS uses fixed 4 candidates
#endif
#endif
+#define JVET_AA0133_INTER_MTS_OPT 1 // JVET-AA0133: Inter MTS optimization
// Entropy Coding
#define EC_HIGH_PRECISION 1 // CABAC high precision
@@ -677,6 +678,19 @@ enum QuantFlags
enum TransType
{
#if JVET_W0103_INTRA_MTS
+#if JVET_AA0133_INTER_MTS_OPT
+ DCT2 = 0,
+ DCT8 = 1,
+ DST7 = 2,
+ DCT5 = 3,
+ DST4 = 4,
+ DST1 = 5,
+ IDTR = 6,
+ KLT0 = 7,
+ KLT1 = 8,
+ NUM_TRANS_TYPE = 9,
+ DCT2_EMT = 10
+#else
DCT2 = 0,
DCT8 = 1,
DST7 = 2,
@@ -686,6 +700,7 @@ enum TransType
IDTR = 6,
NUM_TRANS_TYPE = 7,
DCT2_EMT = 8
+#endif
#else
DCT2 = 0,
DCT8 = 1,
diff --git a/source/Lib/CommonLib/UnitTools.cpp b/source/Lib/CommonLib/UnitTools.cpp
index 510342dc78f48ee731935231f389bd4e347097e5..5a6a300a6a26b6e28d236ee1bc4d6abda961f980 100644
--- a/source/Lib/CommonLib/UnitTools.cpp
+++ b/source/Lib/CommonLib/UnitTools.cpp
@@ -13396,7 +13396,11 @@ bool CU::bdpcmAllowed( const CodingUnit& cu, const ComponentID compID )
bool CU::isMTSAllowed(const CodingUnit &cu, const ComponentID compID)
{
SizeType tsMaxSize = 1 << cu.cs->sps->getLog2MaxTransformSkipBlockSize();
+#if JVET_AA0133_INTER_MTS_OPT
+ const int maxSize = CU::isIntra(cu) ? MTS_INTRA_MAX_CU_SIZE : cu.cs->sps->getInterMTSMaxSize();
+#else
const int maxSize = CU::isIntra( cu ) ? MTS_INTRA_MAX_CU_SIZE : MTS_INTER_MAX_CU_SIZE;
+#endif
const int cuWidth = cu.blocks[0].lumaSize().width;
const int cuHeight = cu.blocks[0].lumaSize().height;
bool mtsAllowed = cu.chType == CHANNEL_TYPE_LUMA && compID == COMPONENT_Y;
diff --git a/source/Lib/CommonLib/x86/TrQuantX86.h b/source/Lib/CommonLib/x86/TrQuantX86.h
index 7dbf585b0f9d4fa7e7ee535573618c5035e3d678..0cc85f201e6b9db7853a016cdedecf1467e32ab5 100644
--- a/source/Lib/CommonLib/x86/TrQuantX86.h
+++ b/source/Lib/CommonLib/x86/TrQuantX86.h
@@ -640,6 +640,10 @@ void TrQuant::_initTrQuantX86()
{ nullptr, g_aiTr4[DST4][0], g_aiTr8[DST4][0], g_aiTr16[DST4][0], g_aiTr32[DST4][0], g_aiTr64[DST4][0], g_aiTr128[DST4][0], g_aiTr256[DST4][0] },
{ nullptr, g_aiTr4[DST1][0], g_aiTr8[DST1][0], g_aiTr16[DST1][0], g_aiTr32[DST1][0], g_aiTr64[DST1][0], g_aiTr128[DST1][0], g_aiTr256[DST1][0] },
{ nullptr, g_aiTr4[IDTR][0], g_aiTr8[IDTR][0], g_aiTr16[IDTR][0], g_aiTr32[IDTR][0], g_aiTr64[IDTR][0], g_aiTr128[IDTR][0], g_aiTr256[IDTR][0] },
+#if JVET_AA0133_INTER_MTS_OPT
+ { nullptr, g_aiTr4[KLT0][0], g_aiTr8[KLT0][0], g_aiTr16[KLT0][0], nullptr, nullptr, nullptr, nullptr },
+ { nullptr, g_aiTr4[KLT1][0], g_aiTr8[KLT1][0], g_aiTr16[KLT1][0], nullptr, nullptr, nullptr, nullptr },
+#endif
#endif
} };
@@ -653,6 +657,10 @@ void TrQuant::_initTrQuantX86()
{ nullptr, g_aiTr4[DST4][0], g_aiTr8[DST4][0], g_aiTr16[DST4][0], g_aiTr32[DST4][0], g_aiTr64[DST4][0], g_aiTr128[DST4][0], g_aiTr256[DST4][0] },
{ nullptr, g_aiTr4[DST1][0], g_aiTr8[DST1][0], g_aiTr16[DST1][0], g_aiTr32[DST1][0], g_aiTr64[DST1][0], g_aiTr128[DST1][0], g_aiTr256[DST1][0] },
{ nullptr, g_aiTr4[IDTR][0], g_aiTr8[IDTR][0], g_aiTr16[IDTR][0], g_aiTr32[IDTR][0], g_aiTr64[IDTR][0], g_aiTr128[IDTR][0], g_aiTr256[IDTR][0] },
+#if JVET_AA0133_INTER_MTS_OPT
+ { nullptr, g_aiTr4[KLT0][0], g_aiTr8[KLT0][0], g_aiTr16[KLT0][0], nullptr, nullptr, nullptr, nullptr },
+ { nullptr, g_aiTr4[KLT1][0], g_aiTr8[KLT1][0], g_aiTr16[KLT1][0], nullptr, nullptr, nullptr, nullptr },
+#endif
#endif
} };
@@ -719,6 +727,25 @@ void TrQuant::_initTrQuantX86()
fastFwdTrans[6][5] = fastForwardTransform_SIMD<IDTR, 64>;
fastFwdTrans[6][6] = fastForwardTransform_SIMD<IDTR, 128>;
fastFwdTrans[6][7] = fastForwardTransform_SIMD<IDTR, 256>;
+#if JVET_AA0133_INTER_MTS_OPT
+ fastFwdTrans[7][0] = nullptr;
+ fastFwdTrans[7][1] = fastForwardTransform_SIMD<KLT0, 4>;
+ fastFwdTrans[7][2] = fastForwardTransform_SIMD<KLT0, 8>;
+ fastFwdTrans[7][3] = fastForwardTransform_SIMD<KLT0, 16>;
+ fastFwdTrans[7][4] = nullptr;
+ fastFwdTrans[7][5] = nullptr;
+ fastFwdTrans[7][6] = nullptr;
+ fastFwdTrans[7][7] = nullptr;
+
+ fastFwdTrans[8][0] = nullptr;
+ fastFwdTrans[8][1] = fastForwardTransform_SIMD<KLT1, 4>;
+ fastFwdTrans[8][2] = fastForwardTransform_SIMD<KLT1, 8>;
+ fastFwdTrans[8][3] = fastForwardTransform_SIMD<KLT1, 16>;
+ fastFwdTrans[8][4] = nullptr;
+ fastFwdTrans[8][5] = nullptr;
+ fastFwdTrans[8][6] = nullptr;
+ fastFwdTrans[8][7] = nullptr;
+#endif
#endif
fastInvTrans[0][0] = fastInverseTransform_SIMD<DCT2, 2>;
@@ -784,6 +811,25 @@ void TrQuant::_initTrQuantX86()
fastInvTrans[6][5] = fastInverseTransform_SIMD<IDTR, 64>;
fastInvTrans[6][6] = fastInverseTransform_SIMD<IDTR, 128>;
fastInvTrans[6][7] = fastInverseTransform_SIMD<IDTR, 256>;
+#if JVET_AA0133_INTER_MTS_OPT
+ fastInvTrans[7][0] = nullptr;
+ fastInvTrans[7][1] = fastInverseTransform_SIMD<KLT0, 4>;
+ fastInvTrans[7][2] = fastInverseTransform_SIMD<KLT0, 8>;
+ fastInvTrans[7][3] = fastInverseTransform_SIMD<KLT0, 16>;
+ fastInvTrans[7][4] = nullptr;
+ fastInvTrans[7][5] = nullptr;
+ fastInvTrans[7][6] = nullptr;
+ fastInvTrans[7][7] = nullptr;
+
+ fastInvTrans[8][0] = nullptr;
+ fastInvTrans[8][1] = fastInverseTransform_SIMD<KLT1, 4>;
+ fastInvTrans[8][2] = fastInverseTransform_SIMD<KLT1, 8>;
+ fastInvTrans[8][3] = fastInverseTransform_SIMD<KLT1, 16>;
+ fastInvTrans[8][4] = nullptr;
+ fastInvTrans[8][5] = nullptr;
+ fastInvTrans[8][6] = nullptr;
+ fastInvTrans[8][7] = nullptr;
+#endif
#endif
#else
m_forwardTransformKernels =
diff --git a/source/Lib/DecoderLib/VLCReader.cpp b/source/Lib/DecoderLib/VLCReader.cpp
index a192a039b9abbc82dd7aaa7f264a6c5133f3bd88..d3539facebf902f6ad8106aabc73babf93c81611 100644
--- a/source/Lib/DecoderLib/VLCReader.cpp
+++ b/source/Lib/DecoderLib/VLCReader.cpp
@@ -2041,6 +2041,12 @@ void HLSyntaxReader::parseSPS(SPS* pcSPS)
{
READ_FLAG(uiCode, "sps_explicit_mts_intra_enabled_flag"); pcSPS->setUseIntraMTS(uiCode != 0);
READ_FLAG(uiCode, "sps_explicit_mts_inter_enabled_flag"); pcSPS->setUseInterMTS(uiCode != 0);
+#if JVET_AA0133_INTER_MTS_OPT
+ if (pcSPS->getUseInterMTS())
+ {
+ READ_FLAG(uiCode, "sps_inter_mts_max_size"); pcSPS->setInterMTSMaxSize((uiCode != 0) ? 16 : 32);
+ }
+#endif
}
READ_FLAG(uiCode, "sps_lfnst_enabled_flag"); pcSPS->setUseLFNST(uiCode != 0);
#endif
diff --git a/source/Lib/EncoderLib/EncCfg.h b/source/Lib/EncoderLib/EncCfg.h
index 85726c6e3c62609a12fd825e4dfc0ab22a6c514d..8a488b86cbd8aec3e0dc18db7a0c2c929535da0d 100644
--- a/source/Lib/EncoderLib/EncCfg.h
+++ b/source/Lib/EncoderLib/EncCfg.h
@@ -431,6 +431,9 @@ protected:
int m_LadfQpOffset[MAX_LADF_INTERVALS];
int m_LadfIntervalLowerBound[MAX_LADF_INTERVALS];
#endif
+#if JVET_AA0133_INTER_MTS_OPT
+ int m_interMTSMaxSize;
+#endif
#if ENABLE_DIMD
bool m_dimd;
#endif
@@ -1385,6 +1388,10 @@ public:
int getLadfIntervalLowerBound ( int idx ) const { return m_LadfIntervalLowerBound[ idx ]; }
#endif
+#if JVET_AA0133_INTER_MTS_OPT
+ void setInterMTSMaxSize(int size) { m_interMTSMaxSize = size; }
+ int getInterMTSMaxSize() const { return m_interMTSMaxSize; }
+#endif
#if ENABLE_DIMD
void setUseDimd ( bool b ) { m_dimd = b; }
diff --git a/source/Lib/EncoderLib/EncCu.cpp b/source/Lib/EncoderLib/EncCu.cpp
index 8c705f2c8dcaa00ae73807807100b60424029f4a..0c885ca42e873287a243f45e364f6dca33f90d26 100644
--- a/source/Lib/EncoderLib/EncCu.cpp
+++ b/source/Lib/EncoderLib/EncCu.cpp
@@ -741,6 +741,9 @@ bool EncCu::xCheckBestMode( CodingStructure *&tempCS, CodingStructure *&bestCS,
void EncCu::xCompressCU( CodingStructure*& tempCS, CodingStructure*& bestCS, Partitioner& partitioner, double maxCostAllowed )
{
CHECK(maxCostAllowed < 0, "Wrong value of maxCostAllowed!");
+#if JVET_AA0133_INTER_MTS_OPT
+ m_pcInterSearch->setBestCost(maxCostAllowed);
+#endif
#if ENABLE_SPLIT_PARALLELISM
CHECK( m_dataId != tempCS->picture->scheduler.getDataId(), "Working in the wrong dataId!" );
@@ -986,7 +989,9 @@ void EncCu::xCompressCU( CodingStructure*& tempCS, CodingStructure*& bestCS, Par
#endif
}
m_sbtCostSave[0] = m_sbtCostSave[1] = MAX_DOUBLE;
-
+#if JVET_AA0133_INTER_MTS_OPT
+ m_mtsCostSave = MAX_DOUBLE;
+#endif
m_CurrCtx->start = m_CABACEstimator->getCtx();
m_cuChromaQpOffsetIdxPlus1 = 0;
@@ -11625,6 +11630,9 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS
double bestCostBegin = bestCS->cost;
CodingUnit* prevBestCU = bestCS->getCU( partitioner.chType );
uint8_t prevBestSbt = ( prevBestCU == nullptr ) ? 0 : prevBestCU->sbtInfo;
+#if JVET_AA0133_INTER_MTS_OPT
+ bool prevBestMts = (prevBestCU == nullptr) ? 0 : (prevBestCU->firstTU->mtsIdx[COMPONENT_Y] > MTS_SKIP)? true : false ;
+#endif
bool swapped = false; // avoid unwanted data copy
bool reloadCU = false;
@@ -11698,7 +11706,13 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS
}
}
}
+#if JVET_AA0133_INTER_MTS_OPT
+ m_pcInterSearch->setBestCost(bestCS->cost);
+ cu->mtsFlag = false;
+ const bool mtsAllowed = tempCS->sps->getUseInterMTS() && CU::isInter(*cu) && partitioner.currArea().lwidth() <= tempCS->sps->getInterMTSMaxSize() && partitioner.currArea().lheight() <= tempCS->sps->getInterMTSMaxSize();
+#else
const bool mtsAllowed = tempCS->sps->getUseInterMTS() && CU::isInter( *cu ) && partitioner.currArea().lwidth() <= MTS_INTER_MAX_CU_SIZE && partitioner.currArea().lheight() <= MTS_INTER_MAX_CU_SIZE;
+#endif
uint8_t sbtAllowed = cu->checkAllowedSbt();
//SBT resolution-dependent fast algorithm: not try size-64 SBT in RDO for low-resolution sequences (now resolution below HD)
if( tempCS->pps->getPicWidthInLumaSamples() < (uint32_t)m_pcEncCfg->getSBTFast64WidthTh() )
@@ -11711,7 +11725,9 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS
double sbtOffCost = MAX_DOUBLE;
double currBestCost = MAX_DOUBLE;
bool doPreAnalyzeResi = ( sbtAllowed || mtsAllowed ) && residualPass == 0;
-
+#if JVET_AA0133_INTER_MTS_OPT
+ double mtsOffCost = MAX_DOUBLE;
+#endif
m_pcInterSearch->initTuAnalyzer();
if( doPreAnalyzeResi )
{
@@ -11790,7 +11806,11 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS
bestCS->tmpColorSpaceCost = tempCS->tmpColorSpaceCost;
bestCS->firstColorSpaceSelected = tempCS->firstColorSpaceSelected;
}
+#if JVET_AA0133_INTER_MTS_OPT
+ numRDOTried += 1;
+#else
numRDOTried += mtsAllowed ? 2 : 1;
+#endif
xEncodeDontSplit( *tempCS, partitioner );
xCheckDQP( *tempCS, partitioner );
@@ -11967,11 +11987,116 @@ void EncCu::xEncodeInterResidual( CodingStructure *&tempCS
#endif
xCheckBestMode( tempCS, bestCS, partitioner, encTestMode );
}
+#if JVET_AA0133_INTER_MTS_OPT
+ if (!skipResidual && mtsAllowed)
+ {
+ if (bestCost == bestCS->cost) //The first EMT pass didn't become the bestCS, so we clear the TUs generated
+ {
+ tempCS->clearTUs();
+ }
+ else if (false == swapped)
+ {
+ tempCS->initStructData(encTestMode.qp);
+ tempCS->copyStructure(*bestCS, partitioner.chType);
+ tempCS->getPredBuf().copyFrom(bestCS->getPredBuf());
+ bestCost = bestCS->cost;
+ cu = tempCS->getCU(partitioner.chType);
+ swapped = true;
+ }
+ else
+ {
+ tempCS->clearTUs();
+ bestCost = bestCS->cost;
+ cu = tempCS->getCU(partitioner.chType);
+ }
+
+ //we need to restart the distortion for the new tempCS, the bit count and the cost
+ tempCS->dist = 0;
+ tempCS->fracBits = 0;
+ tempCS->cost = MAX_DOUBLE;
+ tempCS->costDbOffset = 0;
+ cu->skip = false;
+ cu->sbtInfo = 0;
+ cu->mtsFlag = true;
+ m_pcInterSearch->setBestCost(bestCS->cost);
+ mtsOffCost = currBestCost;
+ bool testMts = true;
+ if (bestCost != MAX_DOUBLE && mtsOffCost != MAX_DOUBLE)
+ {
+ double th = 1.07;
+ if (!(prevBestMts == 0 || m_mtsCostSave == MAX_DOUBLE))
+ {
+ assert(m_sbtCostSave[1] <= m_mtsCostSave);
+ th *= (m_mtsCostSave / m_sbtCostSave[1]);
+ }
+ if (mtsOffCost > bestCost * th)
+ {
+ testMts = false;
+ }
+ }
+ if(testMts)
+ {
+ //try residual coding
+ bool isValid = m_pcInterSearch->encodeResAndCalcRdInterCU(*tempCS, partitioner, skipResidual);
+ if (isValid)
+ {
+ if (tempCS->slice->getSPS()->getUseColorTrans())
+ {
+ bestCS->tmpColorSpaceCost = tempCS->tmpColorSpaceCost;
+ bestCS->firstColorSpaceSelected = tempCS->firstColorSpaceSelected;
+ }
+ numRDOTried++;
+
+ xEncodeDontSplit(*tempCS, partitioner);
+
+ xCheckDQP(*tempCS, partitioner);
+ xCheckChromaQPOffset(*tempCS, partitioner);
+
+ if (NULL != bestHasNonResi && (bestCostInternal > tempCS->cost))
+ {
+ bestCostInternal = tempCS->cost;
+ if (!(tempCS->getPU(partitioner.chType)->ciipFlag))
+ {
+ *bestHasNonResi = !cu->rootCbf;
+ }
+ }
+ if (cu->rootCbf == false)
+ {
+ if (tempCS->getPU(partitioner.chType)->ciipFlag)
+ {
+ tempCS->cost = MAX_DOUBLE;
+ tempCS->costDbOffset = 0;
+ return;
+ }
+ }
+ if (tempCS->cost < currBestCost)
+ {
+ currBestCost = tempCS->cost;
+ sbtOffCost = tempCS->cost;
+ sbtOffDist = tempCS->dist;
+ sbtOffRootCbf = cu->rootCbf;
+ currBestSbt = CU::getSbtInfo(cu->firstTU->mtsIdx[COMPONENT_Y] > MTS_SKIP ? SBT_OFF_MTS : SBT_OFF_DCT, 0);
+ currBestTrs = cu->firstTU->mtsIdx[COMPONENT_Y];
+ }
+
+ #if WCG_EXT
+ DTRACE_MODE_COST(*tempCS, m_pcRdCost->getLambda(true));
+ #else
+ DTRACE_MODE_COST(*tempCS, m_pcRdCost->getLambda());
+ #endif
+ xCheckBestMode(tempCS, bestCS, partitioner, encTestMode);
+ }
+ }
+ }
+#endif
if( bestCostBegin != bestCS->cost )
{
m_sbtCostSave[0] = sbtOffCost;
m_sbtCostSave[1] = currBestCost;
+#if JVET_AA0133_INTER_MTS_OPT
+ m_mtsCostSave = mtsOffCost;
+#endif
}
} //end emt loop
diff --git a/source/Lib/EncoderLib/EncCu.h b/source/Lib/EncoderLib/EncCu.h
index 8274b2870df27fb0ca42b15c4e169045753e1bd7..22ba1bc679e91c7576696986a3f9a9e88e96daf7 100644
--- a/source/Lib/EncoderLib/EncCu.h
+++ b/source/Lib/EncoderLib/EncCu.h
@@ -370,6 +370,9 @@ private:
const bool updateRdCostLambda );
#endif
double m_sbtCostSave[2];
+#if JVET_AA0133_INTER_MTS_OPT
+ double m_mtsCostSave;
+#endif
#if JVET_W0097_GPM_MMVD_TM
MergeCtx m_mergeCand;
bool m_mergeCandAvail;
diff --git a/source/Lib/EncoderLib/EncLib.cpp b/source/Lib/EncoderLib/EncLib.cpp
index 35376a2d94409f445d531c01e082a5c50e543d87..1994d7e985a391b528bb3ab44c9c8d4f7eae83e3 100644
--- a/source/Lib/EncoderLib/EncLib.cpp
+++ b/source/Lib/EncoderLib/EncLib.cpp
@@ -1585,6 +1585,9 @@ void EncLib::xInitSPS( SPS& sps )
CHECK( m_LadfIntervalLowerBound[0] != 0, "abnormal value set to LadfIntervalLowerBound[0]" );
}
#endif
+#if JVET_AA0133_INTER_MTS_OPT
+ sps.setInterMTSMaxSize(m_interMTSMaxSize);
+#endif
#if ENABLE_DIMD
sps.setUseDimd ( m_dimd );
#endif
diff --git a/source/Lib/EncoderLib/InterSearch.cpp b/source/Lib/EncoderLib/InterSearch.cpp
index effcbe1f7f593d590a0c9ec1f9aee6265920a044..2ceaced3b9bf0c7af73e712699dbd9c94336838b 100644
--- a/source/Lib/EncoderLib/InterSearch.cpp
+++ b/source/Lib/EncoderLib/InterSearch.cpp
@@ -8684,11 +8684,17 @@ uint8_t InterSearch::skipSbtByRDCost( int width, int height, int mtDepth, uint8_
}
return MAX_UCHAR;
}
-
+#if JVET_AA0133_INTER_MTS_OPT
+bool InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &partitioner, Distortion *puiZeroDist /*= NULL*/
+ , const bool luma, const bool chroma
+ , PelUnitBuf* orgResi
+)
+#else
void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &partitioner, Distortion *puiZeroDist /*= NULL*/
, const bool luma, const bool chroma
, PelUnitBuf* orgResi
)
+#endif
{
const UnitArea& currArea = partitioner.currArea();
const SPS &sps = *cs.sps;
@@ -8798,10 +8804,21 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
}
- const bool tsAllowed = TU::isTSAllowed(tu, compID) && (isLuma(compID) || (isChroma(compID) && m_pcEncCfg->getUseChromaTS()));
+#if JVET_AA0133_INTER_MTS_OPT
+ const bool mtsAllowed = CU::isMTSAllowed(*tu.cu, compID) && cu.mtsFlag;
+ const bool tsAllowed = TU::isTSAllowed(tu, compID) && ((isLuma(compID) && !cu.mtsFlag) || (isChroma(compID) && m_pcEncCfg->getUseChromaTS()));
+#else
+ const bool tsAllowed = TU::isTSAllowed(tu, compID) && (isLuma(compID) || (isChroma(compID) && m_pcEncCfg->getUseChromaTS()));
const bool mtsAllowed = CU::isMTSAllowed( *tu.cu, compID );
+#endif
uint8_t nNumTransformCands = 1 + ( tsAllowed ? 1 : 0 ) + ( mtsAllowed ? 4 : 0 ); // DCT + TS + 4 MTS = 6 tests
+#if JVET_AA0133_INTER_MTS_OPT
+ if (cu.mtsFlag && compID == COMPONENT_Y)
+ {
+ nNumTransformCands = (mtsAllowed ? 4 : 0);
+ }
+#endif
std::vector<TrMode> trModes;
#if TU_256
if(tu.idx != cu.firstTU->idx)
@@ -8816,11 +8833,21 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
{
nNumTransformCands = 0;
}
+#if JVET_AA0133_INTER_MTS_OPT
+ else if(!(cu.mtsFlag && compID == COMPONENT_Y))
+#else
else
+#endif
{
trModes.push_back( TrMode( 0, true ) ); //DCT2
nNumTransformCands = 1;
}
+#if JVET_AA0133_INTER_MTS_OPT
+ else
+ {
+ nNumTransformCands = 0;
+ }
+#endif
//for a SBT-no-residual TU, the RDO process should be called once, in order to get the RD cost
if( tsAllowed && !tu.noResidual )
{
@@ -8866,11 +8893,16 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
#endif
m_pcRdCost->lambdaAdjustColorTrans(true, compID);
}
-
+#if JVET_AA0133_INTER_MTS_OPT
+ bool skipRemainingMTS = false;
+ bool skipMTSPass = false;
+ int countSkipMTSLoop = 0;
+#endif
const int numTransformCandidates = nNumTransformCands;
for( int transformMode = 0; transformMode < numTransformCandidates; transformMode++ )
{
const bool isFirstMode = transformMode == 0;
+
// copy the original residual into the residual buffer
#if JVET_S0234_ACT_CRS_FIX
if (colorTransFlag)
@@ -8898,6 +8930,12 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
}
tu.mtsIdx[compID] = trModes[transformMode].first;
}
+#if JVET_AA0133_INTER_MTS_OPT
+ if (compID == COMPONENT_Y && cu.mtsFlag && skipRemainingMTS)
+ {
+ break;
+ }
+#endif
QpParam cQP(tu, compID); // note: uses tu.transformSkip[compID]
#if RDOQ_CHROMA_LAMBDA
@@ -9185,7 +9223,11 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
#endif
}
}
+#if JVET_AA0133_INTER_MTS_OPT
+ else if ((cu.mtsFlag && compID == COMPONENT_Y) || (transformMode > 0))
+#else
else if( transformMode > 0 )
+#endif
{
currCompCost = MAX_DOUBLE;
}
@@ -9197,7 +9239,23 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
tu.cbf[compID] = 0;
}
-
+#if JVET_AA0133_INTER_MTS_OPT
+ if (compID == COMPONENT_Y && cu.mtsFlag && currCompCost < MAX_DOUBLE)
+ {
+ double globalMinCost = std::min(minCost[compID], m_bestDCT2PassLumaCost);
+ double fac = std::max((1.0 + 1.0 / sqrt(tu.lumaSize().width * tu.lumaSize().height)), 1.06);
+ if (currCompCost > fac*globalMinCost)
+ {
+ skipRemainingMTS = true;
+ if (countSkipMTSLoop == 0)
+ {
+ //skip MTS candidate condition fulfilled for first valid MTS candidate (count = 0), so skip chroma coding.
+ skipMTSPass = true;
+ }
+ }
+ countSkipMTSLoop++;
+ }
+ #endif
// evaluate
#if TU_256
if( isFirstMode || ( currCompCost < minCost[compID] ) || ( transformMode == 1 && currCompCost == minCost[compID] ) )
@@ -9230,7 +9288,29 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
CHECK( currCompFracBits > 0 || currAbsSum, "currCompFracBits > 0 when tu noResidual" );
}
}
-
+#if JVET_AA0133_INTER_MTS_OPT
+ if (compID == 0)
+ {
+ if (cu.mtsFlag)
+ {
+ if (minCost[compID] == MAX_DOUBLE || bestTU.cbf[0] == 0) //When checking only MTS cands, cbf can't be zero, or just only contain DC coefficients.
+ {
+ return false;
+ }
+ if (skipMTSPass) //Luma is not selecting any MTS (skipping), so skip chroma coding (Encoder speedup).
+ {
+ return false;
+ }
+ }
+ else
+ {
+ if (!cu.sbtInfo)
+ {
+ m_bestDCT2PassLumaCost = minCost[compID];
+ }
+ }
+ }
+#endif
// copy component
tu.copyComponentFrom( bestTU, compID );
csFull->getResiBuf( compArea ).copyFrom( saveCS.getResiBuf( compArea ) );
@@ -9756,11 +9836,19 @@ void InterSearch::xEstimateInterResidualQT(CodingStructure &cs, Partitioner &par
csFull ->releaseIntermediateData();
}
}
+#if JVET_AA0133_INTER_MTS_OPT
+ return true;
+#endif
}
-
+#if JVET_AA0133_INTER_MTS_OPT
+bool InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual
+ , const bool luma, const bool chroma
+)
+#else
void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual
, const bool luma, const bool chroma
)
+#endif
{
m_pcRdCost->setChromaFormat(cs.sps->getChromaFormatIdc());
@@ -9851,8 +9939,11 @@ void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &pa
cs.dist = distortion;
cs.fracBits = m_CABACEstimator->getEstFracBits();
cs.cost = m_pcRdCost->calcRdCost(cs.fracBits, cs.dist);
-
+#if JVET_AA0133_INTER_MTS_OPT
+ return true;
+#else
return;
+#endif
}
// Residual coding.
@@ -10027,7 +10118,15 @@ void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &pa
cs.getOrgResiBuf().bufs[1].copyFrom(orgResidual.bufs[1]);
cs.getOrgResiBuf().bufs[2].copyFrom(orgResidual.bufs[2]);
}
+#if JVET_AA0133_INTER_MTS_OPT
+ bool isValidReturn = xEstimateInterResidualQT(cs, partitioner, &zeroDistortion, luma, chroma);
+ if (cu.mtsFlag && !isValidReturn)
+ {
+ return false;
+ }
+#else
xEstimateInterResidualQT(cs, partitioner, &zeroDistortion, luma, chroma);
+#endif
}
TransformUnit &firstTU = *cs.getTU( partitioner.chType );
@@ -10387,6 +10486,9 @@ void InterSearch::encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &pa
}
CHECK(cs.tus.size() == 0, "No TUs present");
+#if JVET_AA0133_INTER_MTS_OPT
+ return true;
+#endif
}
uint64_t InterSearch::xGetSymbolFracBitsInter(CodingStructure &cs, Partitioner &partitioner)
diff --git a/source/Lib/EncoderLib/InterSearch.h b/source/Lib/EncoderLib/InterSearch.h
index 24b3db71aa15620a3de77282003f35ebca02d67e..7cd239f081b9905d7642d7973963ebfd8a15b18c 100644
--- a/source/Lib/EncoderLib/InterSearch.h
+++ b/source/Lib/EncoderLib/InterSearch.h
@@ -351,7 +351,10 @@ protected:
CABACWriter* m_CABACEstimator;
CtxCache* m_CtxCache;
DistParam m_cDistParam;
-
+#if JVET_AA0133_INTER_MTS_OPT
+ double m_globalBestLumaCost;
+ double m_bestDCT2PassLumaCost;
+#endif
RefPicList m_currRefPicList;
int m_currRefPicIndex;
bool m_skipFracME;
@@ -418,6 +421,7 @@ public:
m_geoMrgCtx = geoMrgCtx;
}
#endif
+
InterSearch();
virtual ~InterSearch();
@@ -458,6 +462,9 @@ public:
void resetCtuRecord () { m_ctuRecord.clear(); }
#if ENABLE_SPLIT_PARALLELISM
void copyState ( const InterSearch& other );
+#endif
+#if JVET_AA0133_INTER_MTS_OPT
+ void setBestCost(double cost) { m_globalBestLumaCost = cost; }
#endif
void setAffineModeSelected ( bool flag) { m_affineModeSelected = flag; }
void resetAffineMVList() { m_affMVListIdx = 0; m_affMVListSize = 0; }
@@ -984,15 +991,27 @@ protected:
private:
void xxIBCHashSearch(PredictionUnit& pu, Mv* mvPred, int numMvPred, Mv &mv, int& idxMvPred, IbcHashMap& ibcHashMap);
public:
-
+#if JVET_AA0133_INTER_MTS_OPT
+ bool encodeResAndCalcRdInterCU(CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual
+ , const bool luma = true, const bool chroma = true
+ );
+#else
void encodeResAndCalcRdInterCU (CodingStructure &cs, Partitioner &partitioner, const bool &skipResidual
, const bool luma = true, const bool chroma = true
);
+#endif
void xEncodeInterResidualQT (CodingStructure &cs, Partitioner &partitioner, const ComponentID &compID);
+#if JVET_AA0133_INTER_MTS_OPT
+ bool xEstimateInterResidualQT(CodingStructure &cs, Partitioner &partitioner, Distortion *puiZeroDist = NULL
+ , const bool luma = true, const bool chroma = true
+ , PelUnitBuf* orgResi = NULL
+ );
+#else
void xEstimateInterResidualQT (CodingStructure &cs, Partitioner &partitioner, Distortion *puiZeroDist = NULL
, const bool luma = true, const bool chroma = true
, PelUnitBuf* orgResi = NULL
);
+#endif
uint64_t xGetSymbolFracBitsInter (CodingStructure &cs, Partitioner &partitioner);
uint64_t xCalcPuMeBits (PredictionUnit& pu);
diff --git a/source/Lib/EncoderLib/VLCWriter.cpp b/source/Lib/EncoderLib/VLCWriter.cpp
index a7bfc77196d9970496be521bbd68f54aaaef0410..0611d87735f235c2e1c738f353f454da6c270974 100644
--- a/source/Lib/EncoderLib/VLCWriter.cpp
+++ b/source/Lib/EncoderLib/VLCWriter.cpp
@@ -1234,6 +1234,14 @@ void HLSWriter::codeSPS( const SPS* pcSPS )
{
WRITE_FLAG(pcSPS->getUseIntraMTS() ? 1 : 0, "sps_explicit_mts_intra_enabled_flag");
WRITE_FLAG(pcSPS->getUseInterMTS() ? 1 : 0, "sps_explicit_mts_inter_enabled_flag");
+#if JVET_AA0133_INTER_MTS_OPT
+ if (pcSPS->getUseInterMTS())
+ {
+ int interMTSMaxCU = pcSPS->getInterMTSMaxSize();
+ CHECK((interMTSMaxCU != 16 && interMTSMaxCU != 32), "interMTSMaxSize != 32 or 16");
+ WRITE_FLAG(interMTSMaxCU == 16? 1 : 0, "sps_inter_mts_max_size");
+ }
+#endif
}
WRITE_FLAG(pcSPS->getUseLFNST() ? 1 : 0, "sps_lfnst_enabled_flag");
#endif