diff --git a/source/Lib/CommonLib/InterPrediction.cpp b/source/Lib/CommonLib/InterPrediction.cpp
index 115a5dcb6044553dfc85c36bf24c81889b080d3b..5b73a47645f03f637f56b6429f249eeb2ef6dc4b 100644
--- a/source/Lib/CommonLib/InterPrediction.cpp
+++ b/source/Lib/CommonLib/InterPrediction.cpp
@@ -11437,13 +11437,21 @@ void InterPrediction::sortIbcAdaptiveMergeMbvdCandidates(PredictionUnit &pu, Me
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (uiCost < m_mbvdCandCostList[endEncIdx - 1] && m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -11591,14 +11599,22 @@ void InterPrediction::sortIbcMergeMbvdCandidates(PredictionUnit &pu, MergeCtx&
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -12519,14 +12535,22 @@ void InterPrediction::sortInterMergeMMVDCandidates(PredictionUnit &pu, MergeCtx
}
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -17191,14 +17215,22 @@ void InterPrediction::adjustAffineMergeCandidates(PredictionUnit &pu, AffineMer
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -17673,13 +17705,21 @@ void InterPrediction::adjustAffineMergeCandidates(PredictionUnit &pu, AffineMerg
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -17780,13 +17820,21 @@ void InterPrediction::adjustAffineMergeCandidates(PredictionUnit &pu, AffineMerg
);
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -17885,13 +17933,21 @@ void InterPrediction::adjustAffineMergeCandidates(PredictionUnit &pu, AffineMerg
);
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -19641,14 +19697,22 @@ void InterPrediction::adjustIBCMergeCandidates(PredictionUnit &pu, MergeCtx& mr
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop, pcBufPredRefTop, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop, pcBufPredRefTop, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft, pcBufPredRefLeft, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft, pcBufPredRefLeft, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -19825,14 +19889,22 @@ void InterPrediction::adjustIBCMergeCandidates(PredictionUnit &pu, MergeCtx& mr
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop, pcBufPredRefTop, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop, pcBufPredRefTop, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft, pcBufPredRefLeft, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft, pcBufPredRefLeft, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -19950,14 +20022,22 @@ void InterPrediction::adjustAffineMergeCandidatesOneGroup(PredictionUnit &pu, A
);
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -20140,7 +20220,11 @@ Distortion InterPrediction::getTempCost(const PredictionUnit &pu, const PelBuf &
Distortion uiCost;
DistParam cDistParam;
cDistParam.applyWeight = false;
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_pcRdCost->setDistParam(cDistParam, cur, org, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, cur, org, pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost = cDistParam.distFunc(cDistParam);
return uiCost;
}
@@ -31955,7 +32039,11 @@ void InterPrediction::deriveAffineMVDCandVecFromMotionInforPred(const Prediction
const bool res = getAffAMLRefTemplateMvdPredUni<1>(tmpPU, pcBufPredRefTop, pcBufPredRefLeft, pu.cs->sps->getUseFastSubTmvp(), tmp);
if (res)
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
else
@@ -31994,7 +32082,11 @@ void InterPrediction::deriveAffineMVDCandVecFromMotionInforPred(const Prediction
const bool res = getAffAMLRefTemplateMvdPredUni<2>(tmpPU, pcBufPredRefTop, pcBufPredRefLeft, pu.cs->sps->getUseFastSubTmvp(), tmp);
if (res)
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
else
@@ -32270,7 +32362,11 @@ void InterPrediction::reorderRefCombList(PredictionUnit &pu, std::vector<RefList
}
if (res)
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
else
@@ -32289,7 +32385,11 @@ void InterPrediction::reorderRefCombList(PredictionUnit &pu, std::vector<RefList
}
if (res)
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
else
@@ -32426,14 +32526,22 @@ void InterPrediction::reorderRefCombList(PredictionUnit &pu, std::vector<RefList
{
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -33351,9 +33459,17 @@ void InterPrediction::reorderRefPairList(PredictionUnit &pu, std::vector<RefPicP
if (m_bAMLTemplateAvailabe[0])
{
#if JVET_AD0140_MVD_PREDICTION
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), ((identicalMotion && !tmpPU.cu->licFlag)? pcBufPredRefTopIdMotion:pcBufPredRefTop).Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), ((identicalMotion && !tmpPU.cu->licFlag)? pcBufPredRefTopIdMotion:pcBufPredRefTop).Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
+#else
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
#endif
uiCost += cDistParam.distFunc(cDistParam);
@@ -33362,9 +33478,17 @@ void InterPrediction::reorderRefPairList(PredictionUnit &pu, std::vector<RefPicP
if (m_bAMLTemplateAvailabe[1])
{
#if JVET_AD0140_MVD_PREDICTION
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), ((identicalMotion && !tmpPU.cu->licFlag) ? pcBufPredRefLeftIdMotion : pcBufPredRefLeft).Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), ((identicalMotion && !tmpPU.cu->licFlag) ? pcBufPredRefLeftIdMotion : pcBufPredRefLeft).Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
+#else
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
#endif
uiCost += cDistParam.distFunc(cDistParam);
@@ -33721,7 +33845,11 @@ void InterPrediction::reorderRefPairList(PredictionUnit &pu, std::vector<RefPicP
}
#endif
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -33774,7 +33902,11 @@ void InterPrediction::reorderRefPairList(PredictionUnit &pu, std::vector<RefPicP
}
#endif
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
#if JVET_AD0140_MVD_PREDICTION
@@ -35666,14 +35798,22 @@ void InterPrediction::defineSignHypMatchAffine(PredictionUnit& pu, const RefPicL
uiCost = 0;
if (m_bAMLTemplateAvailabe[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (m_bAMLTemplateAvailabe[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
@@ -36463,12 +36603,20 @@ void InterPrediction::defineSignHypMatchAffine(PredictionUnit& pu, const RefPicL
if (numTemplate[0])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurTop.Y(), pcBufPredRefTop.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (numTemplate[1])
{
+#if JVET_AJ0096_SATD_REORDER_INTER
+ m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, true);
+#else
m_pcRdCost->setDistParam(cDistParam, pcBufPredCurLeft.Y(), pcBufPredRefLeft.Y(), pu.cs->sps->getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
#endif
diff --git a/source/Lib/CommonLib/IntraPrediction.cpp b/source/Lib/CommonLib/IntraPrediction.cpp
index f800a2da54e3c2fed3f3427f0e4fdfa19a5f0b62..e899340f1f517f4ec0eb17863dae9c613be2fbc6 100644
--- a/source/Lib/CommonLib/IntraPrediction.cpp
+++ b/source/Lib/CommonLib/IntraPrediction.cpp
@@ -3025,13 +3025,21 @@ Mv IntraPrediction::refineChromaBv(const ComponentID compId, const PredictionUni
if (topCanUse)
{
PelBuf tempRef = PelBuf(refPix + 1, uiWidth, Size(uiWidth, DBV_TEMPLATE_SIZE));
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, tempCurTop, tempRef, pu.cs->sps->getBitDepth(CHANNEL_TYPE_CHROMA), compId, uiWidth >= 4 && uiHeight >= 4 ? true : false);
+#else
m_dbvSadCost->setDistParam(cDistParam, tempCurTop, tempRef, pu.cs->sps->getBitDepth(CHANNEL_TYPE_CHROMA), compId, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
if (leftCanUse)
{
PelBuf tempRef = PelBuf(refPix + 1 + stride, uiHeight, Size(uiHeight, DBV_TEMPLATE_SIZE));
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, tempCurLeft, tempRef, pu.cs->sps->getBitDepth(CHANNEL_TYPE_CHROMA), compId, uiWidth >= 4 && uiHeight >= 4 ? true : false);
+#else
m_dbvSadCost->setDistParam(cDistParam, tempCurLeft, tempRef, pu.cs->sps->getBitDepth(CHANNEL_TYPE_CHROMA), compId, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
}
aBvCostVec.push_back(std::pair<Mv, Distortion>(*it, uiCost));
@@ -4793,7 +4801,11 @@ void IntraPrediction::geneChromaFusionPred(const ComponentID compId, PelBuf &piP
if (numSample > 0)
{
int bestSAD = cccmSAD < cclmSAD ? cccmSAD : cclmSAD;
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ if (bestSAD > 144 * numSample)
+#else
if (bestSAD > 64 * numSample)
+#endif
{
w0 = 3;
w1 = 1;
@@ -7906,21 +7918,39 @@ void IntraPrediction::deriveMPMSorted(const PredictionUnit& pu, uint8_t* mpm, in
distParamSad[1].useMR = false;
if (eTempType == LEFT_ABOVE_NEIGHBOR)
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_timdSatdCost->setTimdDistParam(distParamSad[0], piOrg + iTempWidth, piPred + iTempWidth, iOrgStride,
+ uiPredStride, channelBitDepth, COMPONENT_Y, uiWidth, iTempHeight, 0, 1, true);
+ m_timdSatdCost->setTimdDistParam(distParamSad[1], piOrg + iTempHeight * iOrgStride,
+ piPred + iTempHeight * uiPredStride, iOrgStride, uiPredStride, channelBitDepth,
+ COMPONENT_Y, iTempWidth, uiHeight, 0, 1, true);
+#else
m_timdSatdCost->setTimdDistParam(distParamSad[0], piOrg + iTempWidth, piPred + iTempWidth, iOrgStride,
uiPredStride, channelBitDepth, COMPONENT_Y, uiWidth, iTempHeight, 0, 1, false); // Use HAD (SATD) cost
m_timdSatdCost->setTimdDistParam(distParamSad[1], piOrg + iTempHeight * iOrgStride,
piPred + iTempHeight * uiPredStride, iOrgStride, uiPredStride, channelBitDepth,
COMPONENT_Y, iTempWidth, uiHeight, 0, 1, false);
+#endif
}
else if (eTempType == LEFT_NEIGHBOR)
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_timdSatdCost->setTimdDistParam(distParamSad[1], piOrg, piPred, iOrgStride, uiPredStride, channelBitDepth,
+ COMPONENT_Y, iTempWidth, uiHeight, 0, 1, true);
+#else
m_timdSatdCost->setTimdDistParam(distParamSad[1], piOrg, piPred, iOrgStride, uiPredStride, channelBitDepth,
COMPONENT_Y, iTempWidth, uiHeight, 0, 1, false);
+#endif
}
else if (eTempType == ABOVE_NEIGHBOR)
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_timdSatdCost->setTimdDistParam(distParamSad[0], piOrg, piPred, iOrgStride, uiPredStride, channelBitDepth,
+ COMPONENT_Y, uiWidth, iTempHeight, 0, 1, true);
+#else
m_timdSatdCost->setTimdDistParam(distParamSad[0], piOrg, piPred, iOrgStride, uiPredStride, channelBitDepth,
COMPONENT_Y, uiWidth, iTempHeight, 0, 1, false);
+#endif
}
initTimdIntraPatternLuma(*pu.cu, area, eTempType != ABOVE_NEIGHBOR ? iTempWidth : 0,
eTempType != LEFT_NEIGHBOR ? iTempHeight : 0, uiRefWidth, uiRefHeight);
@@ -21746,6 +21776,19 @@ uint32_t IntraPrediction::xCalculateCCLMcost(const PredictionUnit &pu, const Com
const SizeType cWidth = chromaArea.width;
const SizeType cHeight = chromaArea.height;
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ const ChannelType chType = toChannelType(compID);
+ DistParam cDistParam;
+ cDistParam.applyWeight = false;
+ static Pel predChromaA[MAX_CU_SIZE];
+ static Pel predChromaL[MAX_CU_SIZE];
+ PelBuf predTop(predChromaA, cWidth, 1);
+ PelBuf predLeft(predChromaL, 1, cHeight);
+ static Pel reconChromaA[MAX_CU_SIZE];
+ static Pel reconChromaL[MAX_CU_SIZE];
+ PelBuf reconTop(reconChromaA, cWidth, 1);
+ PelBuf reconLeft(reconChromaL, 1, cHeight);
+#endif
CodingStructure &cs = *(pu.cs);
const CodingUnit &cu = *(pu.cu);
@@ -21780,8 +21823,17 @@ uint32_t IntraPrediction::xCalculateCCLMcost(const PredictionUnit &pu, const Com
{
predChroma = ClipPel(rightShift(cclmModel.a2 * src[pos], cclmModel.shift2) + cclmModel.b2, pu.cs->slice->clpRng(compID));
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ predTop.at(pos, 0) = predChroma;
+ reconTop.at(pos, 0) = curChroma0[pos];
+#else
totalSAD += abs(predChroma - curChroma0[pos]);
+#endif
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, predTop, reconTop, pu.cs->sps->getBitDepth(chType), compID, cWidth >= 4 && cHeight >= 4 ? true : false);
+ totalSAD += (int)cDistParam.distFunc(cDistParam);
+#endif
}
else
#endif
@@ -21789,8 +21841,17 @@ uint32_t IntraPrediction::xCalculateCCLMcost(const PredictionUnit &pu, const Com
for (int pos = 0; pos < cWidth; pos++)
{
Pel predChroma = ClipPel(rightShift(cclmModel.a * src[pos], cclmModel.shift) + cclmModel.b, pu.cs->slice->clpRng(compID));
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ predTop.at(pos, 0) = predChroma;
+ reconTop.at(pos, 0) = curChroma0[pos];
+#else
totalSAD += abs(predChroma - curChroma0[pos]);
+#endif
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, predTop, reconTop, pu.cs->sps->getBitDepth(chType), compID, cWidth >= 4 && cHeight >= 4 ? true : false);
+ totalSAD += (int)cDistParam.distFunc(cDistParam);
+#endif
}
}
@@ -21814,8 +21875,17 @@ uint32_t IntraPrediction::xCalculateCCLMcost(const PredictionUnit &pu, const Com
{
predChroma = ClipPel(rightShift(cclmModel.a2 * src[pos * srcStride], cclmModel.shift2) + cclmModel.b2, pu.cs->slice->clpRng(compID));
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ predLeft.at(0, pos) = predChroma;
+ reconLeft.at(0, pos) = curChroma0[pos * curStride];
+#else
totalSAD += abs(predChroma - curChroma0[pos * curStride]);
+#endif
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, predLeft, reconLeft, pu.cs->sps->getBitDepth(chType), compID, cWidth >= 4 && cHeight >= 4 ? true : false);
+ totalSAD += (int)cDistParam.distFunc(cDistParam);
+#endif
}
else
#endif
@@ -21823,8 +21893,17 @@ uint32_t IntraPrediction::xCalculateCCLMcost(const PredictionUnit &pu, const Com
for (int pos = 0; pos < cHeight; pos++)
{
Pel predChroma = ClipPel(rightShift(cclmModel.a * src[pos * srcStride], cclmModel.shift) + cclmModel.b, pu.cs->slice->clpRng(compID));
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ predLeft.at(0, pos) = predChroma;
+ reconLeft.at(0, pos) = curChroma0[pos * curStride];
+#else
totalSAD += abs(predChroma - curChroma0[pos * curStride]);
+#endif
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, predLeft, reconLeft, pu.cs->sps->getBitDepth(chType), compID, cWidth >= 4 && cHeight >= 4 ? true : false);
+ totalSAD += (int)cDistParam.distFunc(cDistParam);
+#endif
}
}
@@ -21902,6 +21981,19 @@ uint32_t IntraPrediction::xCalculateCCCMcost(const PredictionUnit &pu, const Com
const SizeType cWidth = chromaArea.width;
const SizeType cHeight = chromaArea.height;
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ const ChannelType chType = toChannelType(compID);
+ DistParam cDistParam;
+ cDistParam.applyWeight = false;
+ static Pel predChromaA[MAX_CU_SIZE];
+ static Pel predChromaL[MAX_CU_SIZE];
+ PelBuf predTop(predChromaA, cWidth, 1);
+ PelBuf predLeft(predChromaL, 1, cHeight);
+ static Pel reconChromaA[MAX_CU_SIZE];
+ static Pel reconChromaL[MAX_CU_SIZE];
+ PelBuf reconTop(reconChromaA, cWidth, 1);
+ PelBuf reconLeft(reconChromaL, 1, cHeight);
+#endif
CodingStructure &cs = *(pu.cs);
const CodingUnit &cu = *(pu.cu);
@@ -21953,8 +22045,17 @@ uint32_t IntraPrediction::xCalculateCCCMcost(const PredictionUnit &pu, const Com
#endif
predChroma = ClipPel<Pel>(cccmModel[0].convolve(samples), clpRng);
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ predTop.at(pos, 0) = predChroma;
+ reconTop.at(pos, 0) = curChroma0[pos];
+#else
totalSAD += abs(predChroma - curChroma0[pos]);
+#endif
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, predTop, reconTop, pu.cs->sps->getBitDepth(chType), compID, cWidth >= 4 && cHeight >= 4 ? true : false);
+ totalSAD += (int)cDistParam.distFunc(cDistParam);
+#endif
}
if (checkLeft)
@@ -21989,8 +22090,17 @@ uint32_t IntraPrediction::xCalculateCCCMcost(const PredictionUnit &pu, const Com
else
#endif
predChroma = ClipPel<Pel>(cccmModel[0].convolve(samples), clpRng);
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ predLeft.at(0, pos) = predChroma;
+ reconLeft.at(0, pos) = curChroma0[pos * curStride];
+#else
totalSAD += abs(predChroma - curChroma0[pos * curStride]);
+#endif
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, predLeft, reconLeft, pu.cs->sps->getBitDepth(chType), compID, cWidth >= 4 && cHeight >= 4 ? true : false);
+ totalSAD += (int)cDistParam.distFunc(cDistParam);
+#endif
}
return totalSAD;
@@ -27274,12 +27384,22 @@ void IntraPrediction::getTmrlList(CodingUnit& cu)
distParamSad[1].applyWeight = false;
distParamSad[1].useMR = false;
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_timdSatdCost->setTimdDistParam(distParamSad[0], piOrg + tmrlInfo.uiTemplateLeft, piPred + tmrlInfo.uiTemplateLeft, iOrgStride, uiPredStride,
+ channelBitDepth, COMPONENT_Y, uiWidth, tmrlInfo.uiTemplateAbove, 0, 1, true);
+#else
m_timdSatdCost->setTimdDistParam(distParamSad[0], piOrg + tmrlInfo.uiTemplateLeft, piPred + tmrlInfo.uiTemplateLeft, iOrgStride, uiPredStride,
channelBitDepth, COMPONENT_Y, uiWidth, tmrlInfo.uiTemplateAbove, 0, 1, false);
+#endif
if (cu.lx())
{
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_timdSatdCost->setTimdDistParam(distParamSad[1], piOrg + tmrlInfo.uiTemplateAbove * iOrgStride, piPred + tmrlInfo.uiTemplateAbove * uiPredStride,
+ iOrgStride, uiPredStride, channelBitDepth, COMPONENT_Y, tmrlInfo.uiTemplateLeft, uiHeight, 0, 1, true);
+#else
m_timdSatdCost->setTimdDistParam(distParamSad[1], piOrg + tmrlInfo.uiTemplateAbove * iOrgStride, piPred + tmrlInfo.uiTemplateAbove * uiPredStride,
iOrgStride, uiPredStride, channelBitDepth, COMPONENT_Y, tmrlInfo.uiTemplateLeft, uiHeight, 0, 1, false);
+#endif
}
// step-2. define search range.
@@ -28253,7 +28373,11 @@ void IntraPrediction::reorderEipCands(const PredictionUnit& pu, static_vector<Ei
predTop.at(w, h) = ClipPel(cand.convolve(inputs), clipRng);
}
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, predTop, recoTop, pu.cs->sps->getBitDepth(chType), compId, true);
+#else
m_dbvSadCost->setDistParam(cDistParam, predTop, recoTop, pu.cs->sps->getBitDepth(chType), compId, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
for (int h = 0; h < blockHeight; h++)
@@ -28264,7 +28388,11 @@ void IntraPrediction::reorderEipCands(const PredictionUnit& pu, static_vector<Ei
predLeft.at(w, h) = ClipPel(cand.convolve(inputs), clipRng);
}
}
+#if JVET_AJ0096_SATD_REORDER_INTRA
+ m_dbvSadCost->setDistParam(cDistParam, predLeft, recoLeft, pu.cs->sps->getBitDepth(chType), compId, true);
+#else
m_dbvSadCost->setDistParam(cDistParam, predLeft, recoLeft, pu.cs->sps->getBitDepth(chType), compId, false);
+#endif
uiCost += cDistParam.distFunc(cDistParam);
updateCandList(model, uiCost, tmpCandList, candCostList, NUM_EIP_MERGE_SIGNAL);
diff --git a/source/Lib/CommonLib/RdCost.cpp b/source/Lib/CommonLib/RdCost.cpp
index 608a946e414b0145e39ffa646fef1c5a23cffe6a..da0d801cd2ab7c16a936ad2e34b4385ffb9fceb4 100644
--- a/source/Lib/CommonLib/RdCost.cpp
+++ b/source/Lib/CommonLib/RdCost.cpp
@@ -2346,6 +2346,196 @@ Distortion RdCost::xCalcHADs1xN(const Pel* piOrg, const Pel* piCur, int iStrideO
return satd;
}
#endif
+
+#if JVET_AJ0096_SATD_REORDER_INTRA || JVET_AJ0096_SATD_REORDER_INTER
+Distortion RdCost::xCalcHADs1x16( const Pel *piOrg, const Pel *piCur, int iStrideOrg, int iStrideCur, int iRows, int iCols)
+{
+ int j, sad = 0;
+ int diff[16], m1[16], m2[16];
+ if (iRows == 1)
+ {
+ diff[0] = piOrg[0] - piCur[0];
+ diff[1] = piOrg[1] - piCur[1];
+ diff[2] = piOrg[2] - piCur[2];
+ diff[3] = piOrg[3] - piCur[3];
+ diff[4] = piOrg[4] - piCur[4];
+ diff[5] = piOrg[5] - piCur[5];
+ diff[6] = piOrg[6] - piCur[6];
+ diff[7] = piOrg[7] - piCur[7];
+
+ diff[8] = piOrg[8] - piCur[8];
+ diff[9] = piOrg[9] - piCur[9];
+ diff[10] = piOrg[10] - piCur[10];
+ diff[11] = piOrg[11] - piCur[11];
+ diff[12] = piOrg[12] - piCur[12];
+ diff[13] = piOrg[13] - piCur[13];
+ diff[14] = piOrg[14] - piCur[14];
+ diff[15] = piOrg[15] - piCur[15];
+ }
+ else if (iCols == 1)
+ {
+ diff[0] = piOrg[0] - piCur[0];
+ diff[1] = piOrg[1 * iStrideOrg] - piCur[1 * iStrideCur];
+ diff[2] = piOrg[2 * iStrideOrg] - piCur[2 * iStrideCur];
+ diff[3] = piOrg[3 * iStrideOrg] - piCur[3 * iStrideCur];
+ diff[4] = piOrg[4 * iStrideOrg] - piCur[4 * iStrideCur];
+ diff[5] = piOrg[5 * iStrideOrg] - piCur[5 * iStrideCur];
+ diff[6] = piOrg[6 * iStrideOrg] - piCur[6 * iStrideCur];
+ diff[7] = piOrg[7 * iStrideOrg] - piCur[7 * iStrideCur];
+
+ diff[8] = piOrg[8 * iStrideOrg] - piCur[8 * iStrideCur];
+ diff[9] = piOrg[9 * iStrideOrg] - piCur[9 * iStrideCur];
+ diff[10] = piOrg[10 * iStrideOrg] - piCur[10 * iStrideCur];
+ diff[11] = piOrg[11 * iStrideOrg] - piCur[11 * iStrideCur];
+ diff[12] = piOrg[12 * iStrideOrg] - piCur[12 * iStrideCur];
+ diff[13] = piOrg[13 * iStrideOrg] - piCur[13 * iStrideCur];
+ diff[14] = piOrg[14 * iStrideOrg] - piCur[14 * iStrideCur];
+ diff[15] = piOrg[15 * iStrideOrg] - piCur[15 * iStrideCur];
+ }
+ else
+ {
+ CHECK(1, "shall not be here");
+ }
+
+ m2[0] = diff[0] + diff[8];
+ m2[1] = diff[1] + diff[9];
+ m2[2] = diff[2] + diff[10];
+ m2[3] = diff[3] + diff[11];
+ m2[4] = diff[4] + diff[12];
+ m2[5] = diff[5] + diff[13];
+ m2[6] = diff[6] + diff[14];
+ m2[7] = diff[7] + diff[15];
+ m2[8] = diff[0] - diff[8];
+ m2[9] = diff[1] - diff[9];
+ m2[10] = diff[2] - diff[10];
+ m2[11] = diff[3] - diff[11];
+ m2[12] = diff[4] - diff[12];
+ m2[13] = diff[5] - diff[13];
+ m2[14] = diff[6] - diff[14];
+ m2[15] = diff[7] - diff[15];
+
+ m1[0] = m2[0] + m2[4];
+ m1[1] = m2[1] + m2[5];
+ m1[2] = m2[2] + m2[6];
+ m1[3] = m2[3] + m2[7];
+ m1[4] = m2[0] - m2[4];
+ m1[5] = m2[1] - m2[5];
+ m1[6] = m2[2] - m2[6];
+ m1[7] = m2[3] - m2[7];
+ m1[8] = m2[8] + m2[12];
+ m1[9] = m2[9] + m2[13];
+ m1[10] = m2[10] + m2[14];
+ m1[11] = m2[11] + m2[15];
+ m1[12] = m2[8] - m2[12];
+ m1[13] = m2[9] - m2[13];
+ m1[14] = m2[10] - m2[14];
+ m1[15] = m2[11] - m2[15];
+
+ m2[0] = m1[0] + m1[2];
+ m2[1] = m1[1] + m1[3];
+ m2[2] = m1[0] - m1[2];
+ m2[3] = m1[1] - m1[3];
+ m2[4] = m1[4] + m1[6];
+ m2[5] = m1[5] + m1[7];
+ m2[6] = m1[4] - m1[6];
+ m2[7] = m1[5] - m1[7];
+ m2[8] = m1[8] + m1[10];
+ m2[9] = m1[9] + m1[11];
+ m2[10] = m1[8] - m1[10];
+ m2[11] = m1[9] - m1[11];
+ m2[12] = m1[12] + m1[14];
+ m2[13] = m1[13] + m1[15];
+ m2[14] = m1[12] - m1[14];
+ m2[15] = m1[13] - m1[15];
+
+ for( j = 0; j < 16; j++ )
+ {
+ sad += abs( m2[j] );
+ }
+
+#if JVET_R0164_MEAN_SCALED_SATD
+ sad -= abs(m2[0]);
+ sad += abs(m2[0]) >> 2;
+#endif
+ sad = ( int ) ( sad / sqrt( 16.0 * 1 ) * 2 );
+
+ return sad;
+}
+
+Distortion RdCost::xCalcHADs1x8( const Pel *piOrg, const Pel *piCur, int iStrideOrg, int iStrideCur, int iRows, int iCols)
+{
+ int j;
+ Distortion sad = 0;
+ int diff[8], m1[8], m2[8];
+ if (iRows == 1)
+ {
+ diff[0] = piOrg[0] - piCur[0];
+ diff[1] = piOrg[1] - piCur[1];
+ diff[2] = piOrg[2] - piCur[2];
+ diff[3] = piOrg[3] - piCur[3];
+ diff[4] = piOrg[4] - piCur[4];
+ diff[5] = piOrg[5] - piCur[5];
+ diff[6] = piOrg[6] - piCur[6];
+ diff[7] = piOrg[7] - piCur[7];
+ }
+ else if (iCols == 1)
+ {
+ diff[0] = piOrg[0] - piCur[0];
+ diff[1] = piOrg[1 * iStrideOrg] - piCur[1 * iStrideCur];
+ diff[2] = piOrg[2 * iStrideOrg] - piCur[2 * iStrideCur];
+ diff[3] = piOrg[3 * iStrideOrg] - piCur[3 * iStrideCur];
+ diff[4] = piOrg[4 * iStrideOrg] - piCur[4 * iStrideCur];
+ diff[5] = piOrg[5 * iStrideOrg] - piCur[5 * iStrideCur];
+ diff[6] = piOrg[6 * iStrideOrg] - piCur[6 * iStrideCur];
+ diff[7] = piOrg[7 * iStrideOrg] - piCur[7 * iStrideCur];
+ }
+ else
+ {
+ CHECK(1, "shall not be here");
+ }
+
+ m2[0] = diff[0] + diff[4];
+ m2[1] = diff[1] + diff[5];
+ m2[2] = diff[2] + diff[6];
+ m2[3] = diff[3] + diff[7];
+ m2[4] = diff[0] - diff[4];
+ m2[5] = diff[1] - diff[5];
+ m2[6] = diff[2] - diff[6];
+ m2[7] = diff[3] - diff[7];
+
+ m1[0] = m2[0] + m2[2];
+ m1[1] = m2[1] + m2[3];
+ m1[2] = m2[0] - m2[2];
+ m1[3] = m2[1] - m2[3];
+ m1[4] = m2[4] + m2[6];
+ m1[5] = m2[5] + m2[7];
+ m1[6] = m2[4] - m2[6];
+ m1[7] = m2[5] - m2[7];
+
+ m2[0] = m1[0] + m1[1];
+ m2[1] = m1[0] - m1[1];
+ m2[2] = m1[2] + m1[3];
+ m2[3] = m1[2] - m1[3];
+ m2[4] = m1[4] + m1[5];
+ m2[5] = m1[4] - m1[5];
+ m2[6] = m1[6] + m1[7];
+ m2[7] = m1[6] - m1[7];
+
+ for (j = 0; j < 8; j++)
+ {
+ sad += abs(m2[j]);
+ }
+
+#if JVET_R0164_MEAN_SCALED_SATD
+ sad -= abs(m2[0]);
+ sad += abs(m2[0]) >> 2;
+#endif
+ sad = ( int )( sad / sqrt( 8.0 * 1 ) * 2 );
+
+ return sad;
+}
+#endif
+
Distortion RdCost::xCalcHADs2x2( const Pel *piOrg, const Pel *piCur, int iStrideOrg, int iStrideCur, int iStep )
{
Distortion satd = 0;
@@ -3134,6 +3324,40 @@ Distortion RdCost::xGetHADs( const DistParam &rcDtParam )
piCur += iOffsetCur;
}
}
+#if JVET_AJ0096_SATD_REORDER_INTRA || JVET_AJ0096_SATD_REORDER_INTER
+ else if (iRows == 1 && iCols % 16 == 0)
+ {
+ for( x = 0; x < iCols; x += 16 )
+ {
+ uiSum += xCalcHADs1x16(&piOrg[x], &piCur[x], iStrideOrg, iStrideCur, iRows, 16);
+ }
+ }
+ else if (iCols == 1 && iRows % 16 == 0)
+ {
+ for( y = 0; y < iRows; y += 16 )
+ {
+ uiSum += xCalcHADs1x16( &piOrg[0], &piCur[0], iStrideOrg, iStrideCur, 16, iCols );
+ piOrg += (iStrideOrg << 4);
+ piCur += (iStrideCur << 4);
+ }
+ }
+ else if (iRows == 1 && iCols % 8 == 0)
+ {
+ for( x = 0; x < iCols; x += 8 )
+ {
+ uiSum += xCalcHADs1x8(&piOrg[x], &piCur[x], iStrideOrg, iStrideCur, iRows, 8);
+ }
+ }
+ else if (iCols == 1 && iRows % 8 == 0)
+ {
+ for( y = 0; y < iRows; y += 8 )
+ {
+ uiSum += xCalcHADs1x8(&piOrg[0], &piCur[0], iStrideOrg, iStrideCur, 8, iCols);
+ piOrg += (iStrideOrg << 3);
+ piCur += (iStrideCur << 3);
+ }
+ }
+#endif
#if JVET_AI0185_ADAPTIVE_COST_IN_MERGE_MODE
else if (iRows == 1 || iCols == 1)
{
diff --git a/source/Lib/CommonLib/RdCost.h b/source/Lib/CommonLib/RdCost.h
index 57dce9fb2a65c31e779ce35f9b03005c6c5e6df7..b767de0b2e8bed135ddba8100eed678d26274151 100644
--- a/source/Lib/CommonLib/RdCost.h
+++ b/source/Lib/CommonLib/RdCost.h
@@ -1097,6 +1097,10 @@ private:
static Distortion xGetHADs ( const DistParam& pcDtParam );
#if JVET_AI0185_ADAPTIVE_COST_IN_MERGE_MODE
static Distortion xCalcHADs1xN ( const Pel* piOrg, const Pel* piCurr, int iStrideOrg, int iStrideCur, int iRows, int iCols);
+#endif
+#if JVET_AJ0096_SATD_REORDER_INTRA || JVET_AJ0096_SATD_REORDER_INTER
+ static Distortion xCalcHADs1x16 ( const Pel *piOrg, const Pel *piCur, int iStrideOrg, int iStrideCur, int iRows, int iCols);
+ static Distortion xCalcHADs1x8 ( const Pel *piOrg, const Pel *piCur, int iStrideOrg, int iStrideCur, int iRows, int iCols);
#endif
static Distortion xCalcHADs2x2 ( const Pel *piOrg, const Pel *piCurr, int iStrideOrg, int iStrideCur, int iStep );
static Distortion xCalcHADs4x4 ( const Pel *piOrg, const Pel *piCurr, int iStrideOrg, int iStrideCur, int iStep );
diff --git a/source/Lib/CommonLib/TypeDef.h b/source/Lib/CommonLib/TypeDef.h
index 8d1d8515f3e526193f87fb87f5295c1cfa3bf573..58801c1d722b0b5c3cbf1d812f221799831d32b6 100644
--- a/source/Lib/CommonLib/TypeDef.h
+++ b/source/Lib/CommonLib/TypeDef.h
@@ -51,8 +51,6 @@
#include <cassert>
#include <cstdint>
-
-
#define BASE_ENCODER 1
#define BASE_NORMATIVE 1
#define TOOLS 1
@@ -258,6 +256,7 @@
#define JVET_AE0094_IBC_NONADJACENT_SPATIAL_CANDIDATES 1 // JVET-AE0094: IBC with non-adjacent spatial candidates
#define JVET_AG0091_ARBVP 1 // JVET-AG0091: Auto-relocated block vector prediction
#define JVET_AI0082_TEMPORAL_BV 1 // JVET-AI0081: Temporal BV for IBC merge list construction
+#define JVET_AJ0096_SATD_REORDER_INTRA 1 // JVET-AJ0096: SATD-based reordering for intra coding
#if JVET_AC0071_DBV && JVET_V0130_INTRA_TMP
#define JVET_AF0066_ENABLE_DBV_4_SINGLE_TREE 1 // JVET-AF0066: Enable DBV mode in single tree configuration
@@ -363,6 +362,7 @@
#define JVET_AI0185_ADAPTIVE_COST_IN_MERGE_MODE 1 // JVET-AI0185 adaptive cost function selection in merge mode
#define JVET_AI0183_MVP_EXTENSION 1 // JVET-AI0183 MVP extension
// Inter template matching tools
+#define JVET_AJ0096_SATD_REORDER_INTER 1 // JVET-AJ0096: SATD-based reordering for inter coding
#define ENABLE_INTER_TEMPLATE_MATCHING 1 // It controls whether template matching is enabled for inter prediction
#if ENABLE_INTER_TEMPLATE_MATCHING
#define TM_AMVP 1 // Add template matching to non-subblock inter to refine regular AMVP candidates
diff --git a/source/Lib/CommonLib/x86/RdCostX86.h b/source/Lib/CommonLib/x86/RdCostX86.h
index 8cac6fc5ca594bb3612633e03ad7f705135a879f..5a877a07ed810536e120974461d856b25f70f7a3 100644
--- a/source/Lib/CommonLib/x86/RdCostX86.h
+++ b/source/Lib/CommonLib/x86/RdCostX86.h
@@ -2102,6 +2102,146 @@ static uint32_t xCalcHAD16x8_AVX2( const Torg *piOrg, const Tcur *piCur, const i
return (sad);
}
+#if JVET_AJ0096_SATD_REORDER_INTRA || JVET_AJ0096_SATD_REORDER_INTER
+static uint32_t xCalcHADs1x16_SSE(const Torg* piOrg, const Tcur* piCur, int iStrideOrg, int iStrideCur, int iRows, int iCols)
+{
+ __m128i diff[4], m1[4], m2[4];
+
+ if (iRows == 1)
+ {
+ __m128i org0 = _mm_loadu_si128((__m128i*)&piOrg[0]);
+ __m128i cur0 = _mm_loadu_si128((__m128i*)&piCur[0]);
+ diff[0] = _mm_sub_epi16(org0, cur0);
+ diff[1] = _mm_cvtepi16_epi32( _mm_srli_si128( diff[0], 8 ) );
+ diff[0] = _mm_cvtepi16_epi32( diff[0] );
+
+ org0 = _mm_loadu_si128((__m128i*)&piOrg[8]);
+ cur0 = _mm_loadu_si128((__m128i*)&piCur[8]);
+ diff[2] = _mm_sub_epi16(org0, cur0);
+ diff[3] = _mm_cvtepi16_epi32( _mm_srli_si128( diff[2], 8 ) );
+ diff[2] = _mm_cvtepi16_epi32( diff[2] );
+ }
+ else if (iCols == 1)
+ {
+ Pel diffI[16];
+ for (int i = 0; i < 16; i++)
+ {
+ diffI[i] = piOrg[0] - piCur[0];
+ piOrg += iStrideOrg;
+ piCur += iStrideCur;
+ }
+ diff[0] = _mm_loadu_si128((__m128i*)&diffI[0]);
+ diff[1] = _mm_cvtepi16_epi32( _mm_srli_si128( diff[0], 8 ) );
+ diff[0] = _mm_cvtepi16_epi32( diff[0] );
+ diff[2] = _mm_loadu_si128((__m128i*)&diffI[8]);
+ diff[3] = _mm_cvtepi16_epi32( _mm_srli_si128( diff[2], 8 ) );
+ diff[2] = _mm_cvtepi16_epi32( diff[2] );
+ }
+ else
+ {
+ std::cerr << "shall not be here" << std::endl;
+ return -1;
+ }
+
+ m2[0] = _mm_add_epi32(diff[0], diff[2]);
+ m2[1] = _mm_add_epi32(diff[1], diff[3]);
+ m2[2] = _mm_sub_epi32(diff[0], diff[2]);
+ m2[3] = _mm_sub_epi32(diff[1], diff[3]);
+
+ m1[0] = _mm_add_epi32(m2[0], m2[1]);
+ m1[1] = _mm_sub_epi32(m2[0], m2[1]);
+ m1[2] = _mm_add_epi32(m2[2], m2[3]);
+ m1[3] = _mm_sub_epi32(m2[2], m2[3]);
+
+ m2[0] = _mm_unpacklo_epi32(m1[0], m1[1]);
+ m2[1] = _mm_unpackhi_epi32(m1[0], m1[1]);
+ m2[2] = _mm_abs_epi32(_mm_add_epi32(m2[0], m2[1]));
+#if JVET_R0164_MEAN_SCALED_SATD
+ uint32_t absDc = _mm_cvtsi128_si32(m2[2]);
+#endif
+ m2[3] = _mm_abs_epi32(_mm_sub_epi32(m2[0], m2[1]));
+ __m128i iSum = _mm_add_epi32(m2[2], m2[3]);
+ m2[0] = _mm_unpacklo_epi32(m1[2], m1[3]);
+ m2[1] = _mm_unpackhi_epi32(m1[2], m1[3]);
+ m2[2] = _mm_abs_epi32(_mm_add_epi32(m2[0], m2[1]));
+ m2[3] = _mm_abs_epi32(_mm_sub_epi32(m2[0], m2[1]));
+ iSum = _mm_add_epi32(iSum, m2[2]);
+ iSum = _mm_add_epi32(iSum, m2[3]);
+ iSum = _mm_add_epi32(iSum, _mm_shuffle_epi32(iSum, 0x4e)); // 01001110
+ iSum = _mm_add_epi32(iSum, _mm_shuffle_epi32(iSum, 0xb1)); // 10110001
+ uint32_t sad = _mm_cvtsi128_si32( iSum );
+
+#if JVET_R0164_MEAN_SCALED_SATD
+ sad -= absDc;
+ sad += absDc >> 2;
+#endif
+ sad = sad >> 1;
+
+ return sad;
+}
+
+static uint32_t xCalcHADs1x8_SSE(const Torg* piOrg, const Tcur* piCur, int iStrideOrg, int iStrideCur, int iRows, int iCols)
+{
+ __m128i diff[2], m1[2], m2[2];
+
+ if (iRows == 1)
+ {
+ __m128i org0 = _mm_loadu_si128((__m128i*)&piOrg[0]);
+ __m128i cur0 = _mm_loadu_si128((__m128i*)&piCur[0]);
+ diff[0] = _mm_sub_epi16(org0, cur0);
+ diff[1] = _mm_cvtepi16_epi32( _mm_srli_si128( diff[0], 8 ) );
+ diff[0] = _mm_cvtepi16_epi32( diff[0] );
+ }
+ else if (iCols == 1)
+ {
+ Pel diffI[8];
+ for (int i = 0; i < 8; i++)
+ {
+ diffI[i] = piOrg[0] - piCur[0];
+ piOrg += iStrideOrg;
+ piCur += iStrideCur;
+ }
+ diff[0] = _mm_loadu_si128((__m128i*)&diffI[0]);
+ diff[1] = _mm_cvtepi16_epi32( _mm_srli_si128( diff[0], 8 ) );
+ diff[0] = _mm_cvtepi16_epi32( diff[0] );
+ }
+ else
+ {
+ std::cerr << "shall not be here" << std::endl;
+ return -1;
+ }
+
+ m2[0] = _mm_add_epi32(diff[0], diff[1]);
+ m2[1] = _mm_sub_epi32(diff[0], diff[1]);
+
+ m1[0] = _mm_unpacklo_epi32(m2[0], m2[1]);
+ m1[1] = _mm_unpackhi_epi32(m2[0], m2[1]);
+ m2[0] = _mm_add_epi32(m1[0], m1[1]);
+ m2[1] = _mm_sub_epi32(m1[0], m1[1]);
+
+ m1[0] = _mm_unpacklo_epi32(m2[0], m2[1]);
+ m1[1] = _mm_unpackhi_epi32(m2[0], m2[1]);
+ m2[0] = _mm_abs_epi32(_mm_add_epi32(m1[0], m1[1]));
+#if JVET_R0164_MEAN_SCALED_SATD
+ uint32_t absDc = _mm_cvtsi128_si32(m2[0]);
+#endif
+ m2[1] = _mm_abs_epi32(_mm_sub_epi32(m1[0], m1[1]));
+
+ __m128i iSum = _mm_add_epi32(m2[0], m2[1]);
+ iSum = _mm_add_epi32(iSum, _mm_shuffle_epi32(iSum, 0x4e)); // 01001110
+ iSum = _mm_add_epi32(iSum, _mm_shuffle_epi32(iSum, 0xb1)); // 10110001
+ uint32_t sad = _mm_cvtsi128_si32( iSum );
+
+#if JVET_R0164_MEAN_SCALED_SATD
+ sad -= absDc;
+ sad += absDc >> 2;
+#endif
+ sad = ( int ) ( sad / sqrt( 8.0 * 1 ) * 2 );
+
+ return sad;
+}
+#endif
+
static uint32_t xCalcHAD8x16_AVX2( const Pel* piOrg, const Pel* piCur, const int iStrideOrg, const int iStrideCur, const int iBitDepth )
{
@@ -2570,6 +2710,40 @@ Distortion RdCost::xGetHADs_SIMD( const DistParam &rcDtParam )
piCur += iOffsetCur;
}
}
+#if JVET_AJ0096_SATD_REORDER_INTRA || JVET_AJ0096_SATD_REORDER_INTER
+ else if (iRows == 1 && iCols % 16 == 0)
+ {
+ for( x = 0; x < iCols; x += 16 )
+ {
+ uiSum += xCalcHADs1x16_SSE(&piOrg[x], &piCur[x], iStrideOrg, iStrideCur, iRows, 16);
+ }
+ }
+ else if (iCols == 1 && iRows % 16 == 0)
+ {
+ for( y = 0; y < iRows; y += 16 )
+ {
+ uiSum += xCalcHADs1x16_SSE( &piOrg[0], &piCur[0], iStrideOrg, iStrideCur, 16, iCols );
+ piOrg += (iStrideOrg << 4);
+ piCur += (iStrideCur << 4);
+ }
+ }
+ else if (iRows == 1 && iCols % 8 == 0)
+ {
+ for( x = 0; x < iCols; x += 8 )
+ {
+ uiSum += xCalcHADs1x8_SSE(&piOrg[x], &piCur[x], iStrideOrg, iStrideCur, iRows, 8);
+ }
+ }
+ else if (iCols == 1 && iRows % 8 == 0)
+ {
+ for( y = 0; y < iRows; y += 8 )
+ {
+ uiSum += xCalcHADs1x8_SSE(&piOrg[0], &piCur[0], iStrideOrg, iStrideCur, 8, iCols);
+ piOrg += (iStrideOrg << 3);
+ piCur += (iStrideCur << 3);
+ }
+ }
+#endif
#if JVET_AI0185_ADAPTIVE_COST_IN_MERGE_MODE
else if (iRows == 1 || iCols == 1)
{