From 8607019780a132f7e01d6c5c2c5a6b3748bdb01b Mon Sep 17 00:00:00 2001
From: Karl Sharman <karl.sharman@sony.com>
Date: Wed, 17 Jun 2020 01:10:08 +0200
Subject: [PATCH] JVET-R0351: High bit depth coding
---
source/Lib/CommonLib/AdaptiveLoopFilter.cpp | 25 +-
source/Lib/CommonLib/AdaptiveLoopFilter.h | 33 ++-
source/Lib/CommonLib/AlfParameters.h | 8 +
source/Lib/CommonLib/Buffer.cpp | 32 +++
source/Lib/CommonLib/Buffer.h | 5 +
source/Lib/CommonLib/CommonDef.h | 6 +
source/Lib/CommonLib/ContextModelling.cpp | 4 +
source/Lib/CommonLib/ContextModelling.h | 24 ++
source/Lib/CommonLib/DepQuant.cpp | 8 +
source/Lib/CommonLib/InterPrediction.cpp | 20 ++
source/Lib/CommonLib/InterpolationFilter.cpp | 16 ++
source/Lib/CommonLib/InterpolationFilter.h | 3 +
source/Lib/CommonLib/Rom.cpp | 8 +
source/Lib/CommonLib/Rom.h | 8 +
source/Lib/CommonLib/RomTr.cpp | 86 ++++++-
source/Lib/CommonLib/TrQuant.cpp | 41 ++-
source/Lib/CommonLib/TrQuant.h | 5 +
source/Lib/CommonLib/TrQuant_EMT.cpp | 237 +++++++++++++++++-
source/Lib/CommonLib/TypeDef.h | 7 +
source/Lib/CommonLib/WeightPrediction.cpp | 12 +
.../Lib/CommonLib/x86/AdaptiveLoopFilterX86.h | 8 +
.../CommonLib/x86/InterpolationFilterX86.h | 16 ++
source/Lib/DecoderLib/CABACReader.cpp | 25 ++
source/Lib/DecoderLib/VLCReader.cpp | 4 +
.../Lib/EncoderLib/EncAdaptiveLoopFilter.cpp | 92 +++++++
source/Lib/EncoderLib/EncAdaptiveLoopFilter.h | 8 +
source/Lib/EncoderLib/VLCWriter.cpp | 4 +
27 files changed, 733 insertions(+), 12 deletions(-)
diff --git a/source/Lib/CommonLib/AdaptiveLoopFilter.cpp b/source/Lib/CommonLib/AdaptiveLoopFilter.cpp
index c82551538..9e16dcac0 100644
--- a/source/Lib/CommonLib/AdaptiveLoopFilter.cpp
+++ b/source/Lib/CommonLib/AdaptiveLoopFilter.cpp
@@ -501,7 +501,11 @@ void AdaptiveLoopFilter::ALFProcess(CodingStructure& cs)
deriveClassification( m_classifier, buf.get(COMPONENT_Y), blkDst, blkSrc );
short filterSetIndex = alfCtuFilterIndex[ctuIdx];
short *coeff;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel *clip;
+#else
short *clip;
+#endif
if (filterSetIndex >= NUM_FIXED_FILTER_SETS)
{
coeff = m_coeffApsLuma[filterSetIndex - NUM_FIXED_FILTER_SETS];
@@ -565,7 +569,11 @@ void AdaptiveLoopFilter::ALFProcess(CodingStructure& cs)
deriveClassification( m_classifier, tmpYuv.get( COMPONENT_Y ), blk, blk );
short filterSetIndex = alfCtuFilterIndex[ctuIdx];
short *coeff;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel *clip;
+#else
short *clip;
+#endif
if (filterSetIndex >= NUM_FIXED_FILTER_SETS)
{
coeff = m_coeffApsLuma[filterSetIndex - NUM_FIXED_FILTER_SETS];
@@ -662,7 +670,11 @@ void AdaptiveLoopFilter::reconstructCoeff( AlfParam& alfParam, ChannelType chann
{
int numFilters = isLuma( channel ) ? alfParam.numLumaFilters : 1;
short* coeff = isLuma( channel ) ? alfParam.lumaCoeff : alfParam.chromaCoeff[altIdx];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel* clipp = isLuma( channel ) ? alfParam.lumaClipp : alfParam.chromaClipp[altIdx];
+#else
short* clipp = isLuma( channel ) ? alfParam.lumaClipp : alfParam.chromaClipp[altIdx];
+#endif
for( int filterIdx = 0; filterIdx < numFilters; filterIdx++ )
{
@@ -1058,7 +1070,11 @@ void AdaptiveLoopFilter::deriveClassificationBlk(AlfClassifier **classifier, int
template<AlfFilterType filtType>
void AdaptiveLoopFilter::filterBlk(AlfClassifier **classifier, const PelUnitBuf &recDst, const CPelUnitBuf &recSrc,
const Area &blkDst, const Area &blk, const ComponentID compId,
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const short *filterSet, const Pel *fClipSet, const ClpRng &clpRng,
+#else
const short *filterSet, const short *fClipSet, const ClpRng &clpRng,
+#endif
CodingStructure &cs, const int vbCTUHeight, int vbPos)
{
CHECK((vbCTUHeight & (vbCTUHeight - 1)) != 0, "vbCTUHeight must be a power of 2");
@@ -1087,8 +1103,11 @@ void AdaptiveLoopFilter::filterBlk(AlfClassifier **classifier, const PelUnitBuf
const Pel *pImg0, *pImg1, *pImg2, *pImg3, *pImg4, *pImg5, *pImg6;
const short *coef = filterSet;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel *clip = fClipSet;
+#else
const short *clip = fClipSet;
-
+#endif
const int shift = m_NUM_BITS - 1;
const int offset = 1 << ( shift - 1 );
@@ -1225,7 +1244,11 @@ void AdaptiveLoopFilter::filterBlk(AlfClassifier **classifier, const PelUnitBuf
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel sum = 0;
+#else
int sum = 0;
+#endif
const Pel curr = pImg0[+0];
if( filtType == ALF_FILTER_7 )
{
diff --git a/source/Lib/CommonLib/AdaptiveLoopFilter.h b/source/Lib/CommonLib/AdaptiveLoopFilter.h
index 74c9bdbd0..be9819ada 100644
--- a/source/Lib/CommonLib/AdaptiveLoopFilter.h
+++ b/source/Lib/CommonLib/AdaptiveLoopFilter.h
@@ -67,10 +67,17 @@ enum Direction
class AdaptiveLoopFilter
{
public:
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ static inline Pel clipALF(const Pel clip, const Pel ref, const Pel val0, const Pel val1)
+ {
+ return Clip3<Pel>(-clip, +clip, val0-ref) + Clip3<Pel>(-clip, +clip, val1-ref);
+ }
+#else
static inline int clipALF(const int clip, const short ref, const short val0, const short val1)
{
return Clip3<int>(-clip, +clip, val0-ref) + Clip3<int>(-clip, +clip, val1-ref);
}
+#endif
static constexpr int AlfNumClippingValues[MAX_NUM_CHANNEL_TYPE] = { 4, 4 };
static constexpr int MaxAlfNumClippingValues = 4;
@@ -99,8 +106,12 @@ public:
template<AlfFilterType filtType>
static void filterBlk(AlfClassifier **classifier, const PelUnitBuf &recDst, const CPelUnitBuf &recSrc,
const Area &blkDst, const Area &blk, const ComponentID compId, const short *filterSet,
- const short *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
- int vbPos);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#else
+ const short *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#endif
+ int vbPos);
void (*m_deriveClassificationBlk)(AlfClassifier **classifier, int **laplacian[NUM_DIRECTIONS], const CPelBuf &srcLuma,
const Area &blkDst, const Area &blk, const int shift, const int vbCTUHeight,
int vbPos);
@@ -116,11 +127,19 @@ public:
uint8_t* getCcAlfControlIdc(const ComponentID compID) { return m_ccAlfFilterControl[compID-1]; }
void (*m_filter5x5Blk)(AlfClassifier **classifier, const PelUnitBuf &recDst, const CPelUnitBuf &recSrc,
const Area &blkDst, const Area &blk, const ComponentID compId, const short *filterSet,
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#else
const short *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#endif
int vbPos);
void (*m_filter7x7Blk)(AlfClassifier **classifier, const PelUnitBuf &recDst, const CPelUnitBuf &recSrc,
const Area &blkDst, const Area &blk, const ComponentID compId, const short *filterSet,
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#else
const short *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#endif
int vbPos);
#ifdef TARGET_SIMD_X86
@@ -138,8 +157,13 @@ protected:
static const int m_fixedFilterSetCoeff[ALF_FIXED_FILTER_NUM][MAX_NUM_ALF_LUMA_COEFF];
short m_fixedFilterSetCoeffDec[NUM_FIXED_FILTER_SETS][MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF];
short m_coeffApsLuma[ALF_CTB_MAX_NUM_APS][MAX_NUM_ALF_LUMA_COEFF * MAX_NUM_ALF_CLASSES];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel m_clippApsLuma[ALF_CTB_MAX_NUM_APS][MAX_NUM_ALF_LUMA_COEFF * MAX_NUM_ALF_CLASSES];
+ Pel m_clipDefault[MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF];
+#else
short m_clippApsLuma[ALF_CTB_MAX_NUM_APS][MAX_NUM_ALF_LUMA_COEFF * MAX_NUM_ALF_CLASSES];
short m_clipDefault[MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF];
+#endif
bool m_created = false;
short m_chromaCoeffFinal[MAX_NUM_ALF_ALTERNATIVES_CHROMA][MAX_NUM_ALF_CHROMA_COEFF];
AlfParam* m_alfParamChroma;
@@ -148,8 +172,13 @@ protected:
std::vector<AlfFilterShape> m_filterShapes[MAX_NUM_CHANNEL_TYPE];
AlfClassifier** m_classifier;
short m_coeffFinal[MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel m_clippFinal[MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF];
+ Pel m_chromaClippFinal[MAX_NUM_ALF_ALTERNATIVES_CHROMA][MAX_NUM_ALF_CHROMA_COEFF];
+#else
short m_clippFinal[MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF];
short m_chromaClippFinal[MAX_NUM_ALF_ALTERNATIVES_CHROMA][MAX_NUM_ALF_CHROMA_COEFF];
+#endif
int** m_laplacian[NUM_DIRECTIONS];
int * m_laplacianPtr[NUM_DIRECTIONS][m_CLASSIFICATION_BLK_SIZE + 5];
int m_laplacianData[NUM_DIRECTIONS][m_CLASSIFICATION_BLK_SIZE + 5][m_CLASSIFICATION_BLK_SIZE + 5];
diff --git a/source/Lib/CommonLib/AlfParameters.h b/source/Lib/CommonLib/AlfParameters.h
index 989952d81..fe7f23ae8 100644
--- a/source/Lib/CommonLib/AlfParameters.h
+++ b/source/Lib/CommonLib/AlfParameters.h
@@ -129,10 +129,18 @@ struct AlfParam
bool enabledFlag[MAX_NUM_COMPONENT]; // alf_slice_enable_flag, alf_chroma_idc
bool nonLinearFlag[MAX_NUM_CHANNEL_TYPE]; // alf_[luma/chroma]_clip_flag
short lumaCoeff[MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF]; // alf_coeff_luma_delta[i][j]
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel lumaClipp[MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF]; // alf_clipp_luma_[i][j]
+#else
short lumaClipp[MAX_NUM_ALF_CLASSES * MAX_NUM_ALF_LUMA_COEFF]; // alf_clipp_luma_[i][j]
+#endif
int numAlternativesChroma; // alf_chroma_num_alts_minus_one + 1
short chromaCoeff[MAX_NUM_ALF_ALTERNATIVES_CHROMA][MAX_NUM_ALF_CHROMA_COEFF]; // alf_coeff_chroma[i]
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel chromaClipp[MAX_NUM_ALF_ALTERNATIVES_CHROMA][MAX_NUM_ALF_CHROMA_COEFF]; // alf_clipp_chroma[i]
+#else
short chromaClipp[MAX_NUM_ALF_ALTERNATIVES_CHROMA][MAX_NUM_ALF_CHROMA_COEFF]; // alf_clipp_chroma[i]
+#endif
short filterCoeffDeltaIdx[MAX_NUM_ALF_CLASSES]; // filter_coeff_delta[i]
bool alfLumaCoeffFlag[MAX_NUM_ALF_CLASSES]; // alf_luma_coeff_flag[i]
int numLumaFilters; // number_of_filters_minus1 + 1
diff --git a/source/Lib/CommonLib/Buffer.cpp b/source/Lib/CommonLib/Buffer.cpp
index b1ed883e4..3db48cdb6 100644
--- a/source/Lib/CommonLib/Buffer.cpp
+++ b/source/Lib/CommonLib/Buffer.cpp
@@ -94,16 +94,32 @@ void addBIOAvgCore(const Pel* src0, int src0Stride, const Pel* src1, int src1Str
for (int x = 0; x < width; x += 4)
{
b = tmpx * (gradX0[x] - gradX1[x]) + tmpy * (gradY0[x] - gradY1[x]);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[x] = ClipPel(rightShift((src0[x] + src1[x] + b + offset), shift), clpRng);
+#else
dst[x] = ClipPel((int16_t)rightShift((src0[x] + src1[x] + b + offset), shift), clpRng);
+#endif
b = tmpx * (gradX0[x + 1] - gradX1[x + 1]) + tmpy * (gradY0[x + 1] - gradY1[x + 1]);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[x + 1] = ClipPel(rightShift((src0[x + 1] + src1[x + 1] + b + offset), shift), clpRng);
+#else
dst[x + 1] = ClipPel((int16_t)rightShift((src0[x + 1] + src1[x + 1] + b + offset), shift), clpRng);
+#endif
b = tmpx * (gradX0[x + 2] - gradX1[x + 2]) + tmpy * (gradY0[x + 2] - gradY1[x + 2]);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[x + 2] = ClipPel(rightShift((src0[x + 2] + src1[x + 2] + b + offset), shift), clpRng);
+#else
dst[x + 2] = ClipPel((int16_t)rightShift((src0[x + 2] + src1[x + 2] + b + offset), shift), clpRng);
+#endif
b = tmpx * (gradX0[x + 3] - gradX1[x + 3]) + tmpy * (gradY0[x + 3] - gradY1[x + 3]);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[x + 3] = ClipPel(rightShift((src0[x + 3] + src1[x + 3] + b + offset), shift), clpRng);
+#else
dst[x + 3] = ClipPel((int16_t)rightShift((src0[x + 3] + src1[x + 3] + b + offset), shift), clpRng);
+#endif
}
dst += dstStride; src0 += src0Stride; src1 += src1Stride;
gradX0 += gradStride; gradX1 += gradStride; gradY0 += gradStride; gradY1 += gradStride;
@@ -361,7 +377,11 @@ void AreaBuf<Pel>::addWeightedAvg(const AreaBuf<const Pel> &other1, const AreaBu
const unsigned src2Stride = other2.stride;
const unsigned destStride = stride;
const int clipbd = clpRng.bd;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shiftNum = IF_INTERNAL_FRAC_BITS(clipbd) + log2WeightBase;
+#else
const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clipbd)) + log2WeightBase;
+#endif
const int offset = (1 << (shiftNum - 1)) + (IF_INTERNAL_OFFS << log2WeightBase);
#define ADD_AVG_OP( ADDR ) dest[ADDR] = ClipPel( rightShift( ( src0[ADDR]*w0 + src2[ADDR]*w1 + offset ), shiftNum ), clpRng )
@@ -454,7 +474,11 @@ void AreaBuf<Pel>::addAvg( const AreaBuf<const Pel> &other1, const AreaBuf<const
const unsigned src2Stride = other2.stride;
const unsigned destStride = stride;
const int clipbd = clpRng.bd;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shiftNum = IF_INTERNAL_FRAC_BITS(clipbd) + 1;
+#else
const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clipbd)) + 1;
+#endif
const int offset = (1 << (shiftNum - 1)) + 2 * IF_INTERNAL_OFFS;
#if ENABLE_SIMD_OPT_BUFFER && defined(TARGET_SIMD_X86)
@@ -489,7 +513,11 @@ void AreaBuf<Pel>::toLast( const ClpRng& clpRng )
const uint32_t srcStride = stride;
const int clipbd = clpRng.bd;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shiftNum = IF_INTERNAL_FRAC_BITS(clipbd);
+#else
const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clipbd));
+#endif
const int offset = ( 1 << ( shiftNum - 1 ) ) + IF_INTERNAL_OFFS;
if (width == 1)
@@ -562,7 +590,11 @@ void AreaBuf<Pel>::roundToOutputBitdepth( const AreaBuf<const Pel> &src, const C
const unsigned destStride = stride;
const int32_t clipbd = clpRng.bd;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int32_t shiftDefault = IF_INTERNAL_FRAC_BITS(clipbd);
+#else
const int32_t shiftDefault = std::max<int>(2, (IF_INTERNAL_PREC - clipbd));
+#endif
const int32_t offsetDefault = (1<<(shiftDefault-1)) + IF_INTERNAL_OFFS;
if( width == 1 )
diff --git a/source/Lib/CommonLib/Buffer.h b/source/Lib/CommonLib/Buffer.h
index 5719f521c..4dfe5a27c 100644
--- a/source/Lib/CommonLib/Buffer.h
+++ b/source/Lib/CommonLib/Buffer.h
@@ -440,8 +440,13 @@ void AreaBuf<T>::removeWeightHighFreq(const AreaBuf<T>& other, const bool bClip,
{
#endif
int normalizer = ((1 << 16) + (bcwWeight > 0 ? (bcwWeight >> 1) : -(bcwWeight >> 1))) / bcwWeight;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Intermediate_Int weight0 = normalizer << log2WeightBase;
+ Intermediate_Int weight1 = bcwWeightOther * normalizer;
+#else
int weight0 = normalizer << log2WeightBase;
int weight1 = bcwWeightOther * normalizer;
+#endif
#define REM_HF_INC \
src += srcStride; \
dst += dstStride; \
diff --git a/source/Lib/CommonLib/CommonDef.h b/source/Lib/CommonLib/CommonDef.h
index 2fa1f4a5e..40bf83cdf 100644
--- a/source/Lib/CommonLib/CommonDef.h
+++ b/source/Lib/CommonLib/CommonDef.h
@@ -203,8 +203,10 @@ static const int MAX_BDOF_APPLICATION_REGION = 16;
static const int MAX_CPB_CNT = 32; ///< Upper bound of (cpb_cnt_minus1 + 1)
static const int MAX_NUM_LAYER_IDS = 64;
static const int COEF_REMAIN_BIN_REDUCTION = 5; ///< indicates the level at which the VLC transitions from Golomb-Rice to TU+EG(k)
+#if !JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
static const int COEFF_MIN = -32768;
static const int COEFF_MAX = 32767;
+#endif
static const int CU_DQP_TU_CMAX = 5; ///< max number bins for truncated unary
static const int CU_DQP_EG_k = 0; ///< expgolomb order
@@ -629,7 +631,11 @@ const char* read_x86_extension(const std::string &extStrId);
template <typename ValueType> inline ValueType leftShift (const ValueType value, const int shift) { return (shift >= 0) ? ( value << shift) : ( value >> -shift); }
template <typename ValueType> inline ValueType rightShift (const ValueType value, const int shift) { return (shift >= 0) ? ( value >> shift) : ( value << -shift); }
template <typename ValueType> inline ValueType leftShift_round (const ValueType value, const int shift) { return (shift >= 0) ? ( value << shift) : ((value + (ValueType(1) << (-shift - 1))) >> -shift); }
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+template <typename ValueType> inline ValueType rightShift_round(const ValueType value, const int shift) { return (shift > 0) ? ((value + (ValueType(1) << (shift - 1))) >> shift) : ( value << -shift); }
+#else
template <typename ValueType> inline ValueType rightShift_round(const ValueType value, const int shift) { return (shift >= 0) ? ((value + (ValueType(1) << (shift - 1))) >> shift) : ( value << -shift); }
+#endif
static inline int floorLog2(uint32_t x)
{
diff --git a/source/Lib/CommonLib/ContextModelling.cpp b/source/Lib/CommonLib/ContextModelling.cpp
index 56562db08..34d20b05a 100644
--- a/source/Lib/CommonLib/ContextModelling.cpp
+++ b/source/Lib/CommonLib/ContextModelling.cpp
@@ -69,6 +69,10 @@ CoeffCodingContext::CoeffCodingContext( const TransformUnit& tu, ComponentID com
, m_lastShiftX (0)
, m_lastShiftY (0)
, m_TrafoBypass (tu.cs->sps->getSpsRangeExtension().getTransformSkipContextEnabledFlag() && (tu.mtsIdx[m_compID] == MTS_SKIP))
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ , m_minCoeff (-(1 << tu.cs->sps->getMaxLog2TrDynamicRange(m_chType)))
+ , m_maxCoeff ((1 << tu.cs->sps->getMaxLog2TrDynamicRange(m_chType)) - 1)
+#endif
, m_scanPosLast (-1)
, m_subSetId (-1)
, m_subSetPos (-1)
diff --git a/source/Lib/CommonLib/ContextModelling.h b/source/Lib/CommonLib/ContextModelling.h
index b06839a3b..10160254f 100644
--- a/source/Lib/CommonLib/ContextModelling.h
+++ b/source/Lib/CommonLib/ContextModelling.h
@@ -102,6 +102,11 @@ public:
void decimateNumCtxBins(int n) { m_remainingContextBins -= n; }
void increaseNumCtxBins(int n) { m_remainingContextBins += n; }
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff minCoeff() const { return m_minCoeff; }
+ TCoeff maxCoeff() const { return m_maxCoeff; }
+#endif
+
unsigned sigCtxIdAbs( int scanPos, const TCoeff* coeff, const int state )
{
const uint32_t posY = m_scan[scanPos].y;
@@ -241,6 +246,13 @@ public:
return m_tsLrg1FlagCtxSet(numPos);
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ template <typename T> int sgn(T val)
+ {
+ return (T(0) < val) - (val < T(0));
+ }
+
+#endif
unsigned signCtxIdAbsTS(int scanPos, const TCoeff* coeff, int bdpcm)
{
const uint32_t posY = m_scan[scanPos].y;
@@ -252,11 +264,19 @@ public:
if (posX > 0)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ rightSign = sgn(pData[-1]);
+#else
rightSign = pData[-1];
+#endif
}
if (posY > 0)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ belowSign = sgn(pData[-(int)m_width]);
+#else
belowSign = pData[-(int)m_width];
+#endif
}
if ((rightSign == 0 && belowSign == 0) || ((rightSign*belowSign) < 0))
@@ -380,6 +400,10 @@ private:
const int m_lastShiftX;
const int m_lastShiftY;
const bool m_TrafoBypass;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const TCoeff m_minCoeff;
+ const TCoeff m_maxCoeff;
+#endif
// modified
int m_scanPosLast;
int m_subSetId;
diff --git a/source/Lib/CommonLib/DepQuant.cpp b/source/Lib/CommonLib/DepQuant.cpp
index 2bb92ba14..72f65178e 100644
--- a/source/Lib/CommonLib/DepQuant.cpp
+++ b/source/Lib/CommonLib/DepQuant.cpp
@@ -1084,7 +1084,11 @@ namespace DQIntern
}
#undef UPDATE
TCoeff sumGt1 = sumAbs1 - sumNum;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ m_sigFracBits = m_sigFracBitsArray[scanInfo.sigCtxOffsetNext + std::min<TCoeff>( (sumAbs1+1)>>1, 3 )];
+#else
m_sigFracBits = m_sigFracBitsArray[scanInfo.sigCtxOffsetNext + std::min( (sumAbs1+1)>>1, 3 )];
+#endif
m_coeffFracBits = m_gtxFracBitsArray[scanInfo.gtxCtxOffsetNext + (sumGt1 < 4 ? sumGt1 : 4)];
TCoeff sumAbs = m_absLevelsAndCtxInit[8 + scanInfo.nextInsidePos] >> 8;
@@ -1198,7 +1202,11 @@ namespace DQIntern
TCoeff sumNum = tinit & 7;
TCoeff sumAbs1 = ( tinit >> 3 ) & 31;
TCoeff sumGt1 = sumAbs1 - sumNum;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ m_sigFracBits = m_sigFracBitsArray[ scanInfo.sigCtxOffsetNext + std::min<TCoeff>( (sumAbs1+1)>>1, 3 ) ];
+#else
m_sigFracBits = m_sigFracBitsArray[ scanInfo.sigCtxOffsetNext + std::min( (sumAbs1+1)>>1, 3 ) ];
+#endif
m_coeffFracBits = m_gtxFracBitsArray[ scanInfo.gtxCtxOffsetNext + ( sumGt1 < 4 ? sumGt1 : 4 ) ];
}
}
diff --git a/source/Lib/CommonLib/InterPrediction.cpp b/source/Lib/CommonLib/InterPrediction.cpp
index 4adf3ac3c..ce9c6c1d5 100644
--- a/source/Lib/CommonLib/InterPrediction.cpp
+++ b/source/Lib/CommonLib/InterPrediction.cpp
@@ -779,7 +779,11 @@ void InterPrediction::xPredInterBlk ( const ComponentID& compID, const Predictio
JVET_J0090_SET_CACHE_ENABLE((srcPadStride == 0) && (bioApplied == false)); // Enabled only in non-DMVR-non-BDOF process, In DMVR process, srcPadStride is always non-zero
if (bioApplied && compID == COMPONENT_Y)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shift = IF_INTERNAL_FRAC_BITS(clpRng.bd);
+#else
const int shift = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
+#endif
int xOffset = (xFrac < 8) ? 1 : 0;
int yOffset = (yFrac < 8) ? 1 : 0;
const Pel* refPel = refBuf.buf - yOffset * refBuf.stride - xOffset;
@@ -1192,7 +1196,11 @@ void InterPrediction::xPredAffineBlk(const ComponentID &compID, const Prediction
}
if (enablePROF)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shift = IF_INTERNAL_FRAC_BITS(clpRng.bd);
+#else
const int shift = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
+#endif
const int xOffset = xFrac >> 3;
const int yOffset = yFrac >> 3;
@@ -1219,15 +1227,23 @@ void InterPrediction::xPredAffineBlk(const ComponentID &compID, const Prediction
PelBuf gradYBuf = gradYExt.subBuf(0, 0, blockWidth + 2, blockHeight + 2);
g_pelBufOP.profGradFilter(dstExtBuf.buf, dstExtBuf.stride, blockWidth + 2, blockHeight + 2, gradXBuf.stride, gradXBuf.buf, gradYBuf.buf, clpRng.bd);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel offset = (1 << (shift - 1)) + IF_INTERNAL_OFFS;
+#else
const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
const Pel offset = (1 << (shiftNum - 1)) + IF_INTERNAL_OFFS;
+#endif
Pel* src = dstExtBuf.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
Pel* gX = gradXBuf.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
Pel* gY = gradYBuf.bufAt(PROF_BORDER_EXT_W, PROF_BORDER_EXT_H);
Pel * dstY = dstBuf.bufAt(w, h);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ g_pelBufOP.applyPROF(dstY, dstBuf.stride, src, dstExtBuf.stride, blockWidth, blockHeight, gX, gY, gradXBuf.stride, dMvScaleHor, dMvScaleVer, blockWidth, bi, shift, offset, clpRng);
+#else
g_pelBufOP.applyPROF(dstY, dstBuf.stride, src, dstExtBuf.stride, blockWidth, blockHeight, gX, gY, gradXBuf.stride, dMvScaleHor, dMvScaleVer, blockWidth, bi, shiftNum, offset, clpRng);
+#endif
}
}
}
@@ -1280,7 +1296,11 @@ void InterPrediction::applyBiOptFlow(const PredictionUnit &pu, const CPelUnitBuf
const ClpRng& clpRng = pu.cu->cs->slice->clpRng(COMPONENT_Y);
const int bitDepth = clipBitDepths.recon[toChannelType(COMPONENT_Y)];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shiftNum = IF_INTERNAL_FRAC_BITS(bitDepth) + 1;
+#else
const int shiftNum = IF_INTERNAL_PREC + 1 - bitDepth;
+#endif
const int offset = (1 << (shiftNum - 1)) + 2 * IF_INTERNAL_OFFS;
const int limit = ( 1 << 4 ) - 1;
diff --git a/source/Lib/CommonLib/InterpolationFilter.cpp b/source/Lib/CommonLib/InterpolationFilter.cpp
index ba76ca68f..22148635c 100644
--- a/source/Lib/CommonLib/InterpolationFilter.cpp
+++ b/source/Lib/CommonLib/InterpolationFilter.cpp
@@ -415,7 +415,11 @@ void InterpolationFilter::filterCopy( const ClpRng& clpRng, const Pel *src, int
}
else if ( isFirst )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shift = IF_INTERNAL_FRAC_BITS(clpRng.bd);
+#else
const int shift = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
+#endif
if (biMCForDMVR)
{
@@ -464,7 +468,11 @@ void InterpolationFilter::filterCopy( const ClpRng& clpRng, const Pel *src, int
}
else
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shift = IF_INTERNAL_FRAC_BITS(clpRng.bd);
+#else
const int shift = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
+#endif
if (biMCForDMVR)
{
@@ -567,7 +575,11 @@ void InterpolationFilter::filter(const ClpRng& clpRng, Pel const *src, int srcSt
src -= ( N/2 - 1 ) * cStride;
int offset;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ int headRoom = IF_INTERNAL_FRAC_BITS(clpRng.bd);
+#else
int headRoom = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd));
+#endif
int shift = IF_FILTER_PREC;
// with the current settings (IF_INTERNAL_PREC = 14 and IF_FILTER_PREC = 6), though headroom can be
// negative for bit depths greater than 14, shift will remain non-negative for bit depths of 8->20
@@ -899,7 +911,11 @@ void InterpolationFilter::xWeightedGeoBlk(const PredictionUnit &pu, const uint32
const char log2WeightBase = 3;
const ClpRng clipRng = pu.cu->slice->clpRngs().comp[compIdx];
const int32_t clipbd = clipRng.bd;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int32_t shiftWeighted = IF_INTERNAL_FRAC_BITS(clipbd) + log2WeightBase;
+#else
const int32_t shiftWeighted = std::max<int>(2, (IF_INTERNAL_PREC - clipbd)) + log2WeightBase;
+#endif
const int32_t offsetWeighted = (1 << (shiftWeighted - 1)) + (IF_INTERNAL_OFFS << log2WeightBase);
const uint32_t scaleX = getComponentScaleX(compIdx, pu.chromaFormat);
const uint32_t scaleY = getComponentScaleY(compIdx, pu.chromaFormat);
diff --git a/source/Lib/CommonLib/InterpolationFilter.h b/source/Lib/CommonLib/InterpolationFilter.h
index 58a811c38..bae5ddda6 100644
--- a/source/Lib/CommonLib/InterpolationFilter.h
+++ b/source/Lib/CommonLib/InterpolationFilter.h
@@ -50,6 +50,9 @@
#define IF_INTERNAL_OFFS (1<<(IF_INTERNAL_PREC-1)) ///< Offset used internally
#define IF_INTERNAL_PREC_BILINEAR 10 ///< Number of bits for internal precision
#define IF_FILTER_PREC_BILINEAR 4 ///< Bilinear filter coeff precision so that intermediate value will not exceed 16 bit for SIMD - bit exact
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+#define IF_INTERNAL_FRAC_BITS(bd) std::max(2, IF_INTERNAL_PREC - int(bd))
+#endif
/**
* \brief Interpolation filter class
*/
diff --git a/source/Lib/CommonLib/Rom.cpp b/source/Lib/CommonLib/Rom.cpp
index 61814be66..d00d1da9e 100644
--- a/source/Lib/CommonLib/Rom.cpp
+++ b/source/Lib/CommonLib/Rom.cpp
@@ -723,7 +723,11 @@ void initGeoTemplate()
if (g_angle2mask[angleIdx] == -1)
continue;
g_globalGeoWeights[g_angle2mask[angleIdx]] = new int16_t[GEO_WEIGHT_MASK_SIZE * GEO_WEIGHT_MASK_SIZE];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ g_globalGeoEncSADmask[g_angle2mask[angleIdx]] = new Pel[GEO_WEIGHT_MASK_SIZE * GEO_WEIGHT_MASK_SIZE];
+#else
g_globalGeoEncSADmask[g_angle2mask[angleIdx]] = new int16_t[GEO_WEIGHT_MASK_SIZE * GEO_WEIGHT_MASK_SIZE];
+#endif
int distanceX = angleIdx;
int distanceY = (distanceX + (GEO_NUM_ANGLES >> 2)) % GEO_NUM_ANGLES;
@@ -775,7 +779,11 @@ void initGeoTemplate()
}
int16_t** g_GeoParams;
int16_t* g_globalGeoWeights [GEO_NUM_PRESTORED_MASK];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+Pel* g_globalGeoEncSADmask[GEO_NUM_PRESTORED_MASK];
+#else
int16_t* g_globalGeoEncSADmask[GEO_NUM_PRESTORED_MASK];
+#endif
int16_t g_weightOffset [GEO_NUM_PARTITION_MODE][GEO_NUM_CU_SIZE][GEO_NUM_CU_SIZE][2];
int8_t g_angle2mask[GEO_NUM_ANGLES] = { 0, -1, 1, 2, 3, 4, -1, -1, 5, -1, -1, 4, 3, 2, 1, -1, 0, -1, 1, 2, 3, 4, -1, -1, 5, -1, -1, 4, 3, 2, 1, -1 };
int8_t g_Dis[GEO_NUM_ANGLES] = { 8, 8, 8, 8, 4, 4, 2, 1, 0, -1, -2, -4, -4, -8, -8, -8, -8, -8, -8, -8, -4, -4, -2, -1, 0, 1, 2, 4, 4, 8, 8, 8 };
diff --git a/source/Lib/CommonLib/Rom.h b/source/Lib/CommonLib/Rom.h
index 54e7a3b1c..42d43aa10 100644
--- a/source/Lib/CommonLib/Rom.h
+++ b/source/Lib/CommonLib/Rom.h
@@ -77,7 +77,11 @@ extern const int g_quantScales [2/*0=4^n blocks, 1=2*4^n blocks*/][SCALING_LIS
extern const int g_invQuantScales[2/*0=4^n blocks, 1=2*4^n blocks*/][SCALING_LIST_REM_NUM]; // IQ(QP%6)
static const int g_numTransformMatrixSizes = 6;
+#if RExt__HIGH_PRECISION_FORWARD_TRANSFORM
+static const int g_transformMatrixShift[TRANSFORM_NUMBER_OF_DIRECTIONS] = { 14, 6 };
+#else
static const int g_transformMatrixShift[TRANSFORM_NUMBER_OF_DIRECTIONS] = { 6, 6 };
+#endif
// ====================================================================================================================
@@ -224,7 +228,11 @@ const int g_IBCBufferSize = 256 * 128;
void initGeoTemplate();
extern int16_t** g_GeoParams;
extern int16_t* g_globalGeoWeights [GEO_NUM_PRESTORED_MASK];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+extern Pel* g_globalGeoEncSADmask[GEO_NUM_PRESTORED_MASK];
+#else
extern int16_t* g_globalGeoEncSADmask[GEO_NUM_PRESTORED_MASK];
+#endif
extern int16_t g_weightOffset [GEO_NUM_PARTITION_MODE][GEO_NUM_CU_SIZE][GEO_NUM_CU_SIZE][2];
extern int8_t g_angle2mask [GEO_NUM_ANGLES];
extern int8_t g_Dis[GEO_NUM_ANGLES];
diff --git a/source/Lib/CommonLib/RomTr.cpp b/source/Lib/CommonLib/RomTr.cpp
index a60611a63..722df0e3c 100644
--- a/source/Lib/CommonLib/RomTr.cpp
+++ b/source/Lib/CommonLib/RomTr.cpp
@@ -345,6 +345,90 @@
{ b, -d, f, -h, j, -l, n, -p, r, -t, v, -x, z, -B, D, -F, E, -C, A, -y, w, -u, s, -q, o, -m, k, -i, g, -e, c, -a,}, \
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT && RExt__HIGH_PRECISION_FORWARD_TRANSFORM
+//--------------------------------------------------------------------------------------------------
+// DCT-2
+const TMatrixCoeff g_trCoreDCT2P2[TRANSFORM_NUMBER_OF_DIRECTIONS][2][2] =
+{
+ DEFINE_DCT2_P2_MATRIX(16384),
+ DEFINE_DCT2_P2_MATRIX(64)
+};
+
+const TMatrixCoeff g_trCoreDCT2P4 [TRANSFORM_NUMBER_OF_DIRECTIONS][4][4] =
+{
+ DEFINE_DCT2_P4_MATRIX (16384, 21266, 9224),
+ DEFINE_DCT2_P4_MATRIX ( 64, 83, 36)
+};
+
+const TMatrixCoeff g_trCoreDCT2P8[TRANSFORM_NUMBER_OF_DIRECTIONS][8][8] =
+{
+ DEFINE_DCT2_P8_MATRIX(16384, 21266, 9224, 22813, 19244, 12769, 4563),
+ DEFINE_DCT2_P8_MATRIX(64, 83, 36, 89, 75, 50, 18)
+};
+
+const TMatrixCoeff g_trCoreDCT2P16[TRANSFORM_NUMBER_OF_DIRECTIONS][16][16] =
+{
+ DEFINE_DCT2_P16_MATRIX(16384, 21266, 9224, 22813, 19244, 12769, 4563, 23120, 22063, 20450, 17972, 14642, 11109, 6446, 2316),
+ DEFINE_DCT2_P16_MATRIX( 64, 83, 36, 89, 75, 50, 18, 90, 87, 80, 70, 57, 43, 25, 9)
+};
+
+const TMatrixCoeff g_trCoreDCT2P32[TRANSFORM_NUMBER_OF_DIRECTIONS][32][32] =
+{
+ DEFINE_DCT2_P32_MATRIX(16384, 21266, 9224, 22813, 19244, 12769, 4563, 23120, 22063, 20450, 17972, 14642, 11109, 6446 , 2316, 23106, 22852, 22445, 21848, 20995, 19810, 18601, 17143, 15718, 13853, 11749, 9846, 7908, 5573, 3281, 946),
+ DEFINE_DCT2_P32_MATRIX( 64, 83, 36, 89, 75, 50, 18, 90, 87, 80, 70, 57, 43, 25, 9, 90, 90, 88, 85, 82, 78, 73, 67, 61, 54, 46, 38, 31, 22, 13, 4)
+};
+
+const TMatrixCoeff g_trCoreDCT2P64[TRANSFORM_NUMBER_OF_DIRECTIONS][64][64] =
+{
+ DEFINE_DCT2_P64_MATRIX(16384, 21266, 9224, 22813, 19244, 12769, 4563, 23129, 22063, 20450, 17972, 14642, 11109, 6446, 2316, 23106, 22852, 22445, 21848, 20995, 19810, 18601, 17143, 15718, 13853, 11749, 9846, 7908, 5573, 3281, 946, 23360, 23053, 23048, 23023, 22610, 22339, 21936, 21502, 21266, 20730, 20251, 19726, 18731, 18201, 17638, 16604, 15881, 15084, 14322, 13340, 12238, 11330, 10493, 9428, 8426, 7100, 6151, 5101, 3848, 2734, 1754, 574),
+ DEFINE_DCT2_P64_MATRIX( 64, 83, 36, 89, 75, 50, 18, 90, 87, 80, 70, 57, 43, 25, 9, 90, 90, 88, 85, 82, 78, 73, 67, 61, 54, 46, 38, 31, 22, 13, 4, 91, 90, 90, 90, 88, 87, 86, 84, 83, 81, 79, 77, 73, 71, 69, 65, 62, 59, 56, 52, 48, 44, 41, 37, 33, 28, 24, 20, 15, 11, 7, 2)
+};
+
+// DCT-8
+const TMatrixCoeff g_trCoreDCT8P4[TRANSFORM_NUMBER_OF_DIRECTIONS][4][4] =
+{
+ DEFINE_DCT8_P4_MATRIX(21505, 18893, 14081, 7425),
+ DEFINE_DCT8_P4_MATRIX(84, 74, 55, 29)
+};
+const TMatrixCoeff g_trCoreDCT8P8[TRANSFORM_NUMBER_OF_DIRECTIONS][8][8] =
+{
+ DEFINE_DCT8_P8_MATRIX(22018, 21790, 19958, 18154, 15363, 11754, 8148, 4350),
+ DEFINE_DCT8_P8_MATRIX( 86, 85, 78, 71, 60, 46, 32, 17)
+};
+const TMatrixCoeff g_trCoreDCT8P16[TRANSFORM_NUMBER_OF_DIRECTIONS][16][16] =
+{
+ DEFINE_DCT8_P16_MATRIX(22569, 22542, 22202, 21664, 20754, 19738, 18787, 17369, 15781, 14044, 12209, 10360, 8498, 6421, 4295, 1967),
+ DEFINE_DCT8_P16_MATRIX( 88, 88, 87, 85, 81, 77, 73, 68, 62, 55, 48, 40, 33, 25, 17, 8)
+};
+const TMatrixCoeff g_trCoreDCT8P32[TRANSFORM_NUMBER_OF_DIRECTIONS][32][32] =
+{
+ DEFINE_DCT8_P32_MATRIX(23065, 23136, 22715, 22533, 22544, 22053, 21901, 21463, 21131, 20385, 20019, 19708, 19007, 18415, 17448, 16894, 16143, 15230, 14312, 13616, 12679, 11526, 10770, 9720, 8606, 7734, 6623, 5414, 4478, 3225, 2291, 1043),
+ 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)
+};
+
+// DST-7
+const TMatrixCoeff g_trCoreDST7P4[TRANSFORM_NUMBER_OF_DIRECTIONS][4][4] =
+{
+ DEFINE_DST7_P4_MATRIX( 7425, 14081, 18893, 21505),
+ DEFINE_DST7_P4_MATRIX( 29, 55, 74, 84)
+};
+const TMatrixCoeff g_trCoreDST7P8[TRANSFORM_NUMBER_OF_DIRECTIONS][8][8] =
+{
+ DEFINE_DST7_P8_MATRIX( 4350, 8148, 11754, 15363, 18154, 19958, 21790, 22018),
+ DEFINE_DST7_P8_MATRIX( 17, 32, 46, 60, 71, 78, 85, 86)
+};
+const TMatrixCoeff g_trCoreDST7P16[TRANSFORM_NUMBER_OF_DIRECTIONS][16][16] =
+{
+ DEFINE_DST7_P16_MATRIX(1967, 4295, 6421, 8498, 10360, 12209, 14044, 15781, 17369, 18787, 19738, 20754, 21664, 22202, 22542, 22569),
+ DEFINE_DST7_P16_MATRIX( 8, 17, 25, 33, 40, 48, 55, 62, 68, 73, 77, 81, 85, 87, 88, 88)
+};
+const TMatrixCoeff g_trCoreDST7P32[TRANSFORM_NUMBER_OF_DIRECTIONS][32][32] =
+{
+ DEFINE_DST7_P32_MATRIX( 1043, 2291, 3225, 4478, 5414, 6623, 7734, 8606, 9720, 10770, 11526, 12679, 13616, 14312, 15230, 16143, 16894, 17448, 18415, 19007, 19708, 20019, 20385, 21131, 21463, 21901, 22053, 22544, 22533, 22715, 23136, 23065),
+ 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
//--------------------------------------------------------------------------------------------------
// DCT-2
const TMatrixCoeff g_trCoreDCT2P2[TRANSFORM_NUMBER_OF_DIRECTIONS][2][2] =
@@ -426,5 +510,5 @@ const TMatrixCoeff g_trCoreDST7P32[TRANSFORM_NUMBER_OF_DIRECTIONS][32][32] =
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)
};
-
+#endif
//--------------------------------------------------------------------------------------------------
diff --git a/source/Lib/CommonLib/TrQuant.cpp b/source/Lib/CommonLib/TrQuant.cpp
index 7532aea6f..69908d9af 100644
--- a/source/Lib/CommonLib/TrQuant.cpp
+++ b/source/Lib/CommonLib/TrQuant.cpp
@@ -229,18 +229,30 @@ void TrQuant::init( const Quant* otherQuant,
}
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+void TrQuant::fwdLfnstNxN( TCoeff* src, TCoeff* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize )
+#else
void TrQuant::fwdLfnstNxN( int* src, int* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize )
+#endif
{
const int8_t* trMat = ( size > 4 ) ? g_lfnst8x8[ mode ][ index ][ 0 ] : g_lfnst4x4[ mode ][ index ][ 0 ];
const int trSize = ( size > 4 ) ? 48 : 16;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff coef;
+ TCoeff* out = dst;
+#else
int coef;
int* out = dst;
-
+#endif
assert( index < 3 );
for( int j = 0; j < zeroOutSize; j++ )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff* srcPtr = src;
+#else
int* srcPtr = src;
+#endif
const int8_t* trMatTmp = trMat;
coef = 0;
for( int i = 0; i < trSize; i++ )
@@ -254,29 +266,46 @@ void TrQuant::fwdLfnstNxN( int* src, int* dst, const uint32_t mode, const uint32
::memset( out, 0, ( trSize - zeroOutSize ) * sizeof( int ) );
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+void TrQuant::invLfnstNxN( TCoeff* src, TCoeff* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize, const int maxLog2TrDynamicRange )
+{
+#else
void TrQuant::invLfnstNxN( int* src, int* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize )
{
int maxLog2TrDynamicRange = 15;
+#endif
const TCoeff outputMinimum = -( 1 << maxLog2TrDynamicRange );
const TCoeff outputMaximum = ( 1 << maxLog2TrDynamicRange ) - 1;
const int8_t* trMat = ( size > 4 ) ? g_lfnst8x8[ mode ][ index ][ 0 ] : g_lfnst4x4[ mode ][ index ][ 0 ];
const int trSize = ( size > 4 ) ? 48 : 16;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff resi;
+ TCoeff* out = dst;
+#else
int resi;
int* out = dst;
-
+#endif
assert( index < 3 );
for( int j = 0; j < trSize; j++ )
{
resi = 0;
const int8_t* trMatTmp = trMat;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff* srcPtr = src;
+#else
int* srcPtr = src;
+#endif
for( int i = 0; i < zeroOutSize; i++ )
{
resi += *srcPtr++ * *trMatTmp;
trMatTmp += trSize;
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ *out++ = Clip3<TCoeff>( outputMinimum, outputMaximum, ( resi + 64 ) >> 7 );
+#else
*out++ = Clip3( outputMinimum, outputMaximum, ( int ) ( resi + 64 ) >> 7 );
+#endif
trMat++;
}
}
@@ -309,6 +338,9 @@ bool TrQuant::getTransposeFlag( uint32_t intraMode )
void TrQuant::xInvLfnst( const TransformUnit &tu, const ComponentID compID )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int maxLog2TrDynamicRange = tu.cs->sps->getMaxLog2TrDynamicRange(toChannelType(compID));
+#endif
const CompArea& area = tu.blocks[ compID ];
const uint32_t width = area.width;
const uint32_t height = area.height;
@@ -352,8 +384,11 @@ void TrQuant::xInvLfnst( const TransformUnit &tu, const ComponentID compID )
scanPtr++;
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ invLfnstNxN( m_tempInMatrix, m_tempOutMatrix, g_lfnstLut[ intraMode ], lfnstIdx - 1, sbSize, ( tu4x4Flag || tu8x8Flag ) ? 8 : 16, maxLog2TrDynamicRange );
+#else
invLfnstNxN( m_tempInMatrix, m_tempOutMatrix, g_lfnstLut[ intraMode ], lfnstIdx - 1, sbSize, ( tu4x4Flag || tu8x8Flag ) ? 8 : 16 );
-
+#endif
lfnstTemp = m_tempOutMatrix; // inverse spectral rearrangement
if( transposeFlag )
diff --git a/source/Lib/CommonLib/TrQuant.h b/source/Lib/CommonLib/TrQuant.h
index 50f893da8..619f743f9 100644
--- a/source/Lib/CommonLib/TrQuant.h
+++ b/source/Lib/CommonLib/TrQuant.h
@@ -79,8 +79,13 @@ public:
);
void getTrTypes(const TransformUnit tu, const ComponentID compID, int &trTypeHor, int &trTypeVer);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ void fwdLfnstNxN( TCoeff* src, TCoeff* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize );
+ void invLfnstNxN( TCoeff* src, TCoeff* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize, const int maxLog2TrDynamicRange );
+#else
void fwdLfnstNxN( int* src, int* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize );
void invLfnstNxN( int* src, int* dst, const uint32_t mode, const uint32_t index, const uint32_t size, int zeroOutSize );
+#endif
uint32_t getLFNSTIntraMode( int wideAngPredMode );
bool getTransposeFlag ( uint32_t intraMode );
diff --git a/source/Lib/CommonLib/TrQuant_EMT.cpp b/source/Lib/CommonLib/TrQuant_EMT.cpp
index b21ede257..82e34f176 100644
--- a/source/Lib/CommonLib/TrQuant_EMT.cpp
+++ b/source/Lib/CommonLib/TrQuant_EMT.cpp
@@ -51,7 +51,11 @@
void fastForwardDCT2_B2(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2)
{
int j;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff E, O;
+#else
int E, O;
+#endif
TCoeff add = (shift > 0) ? (1 << (shift - 1)) : 0;
const TMatrixCoeff *iT = g_trCoreDCT2P2[TRANSFORM_FORWARD][0];
@@ -85,8 +89,13 @@ void fastForwardDCT2_B2(const TCoeff *src, TCoeff *dst, int shift, int line, int
void fastInverseDCT2_B2(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
{
int j;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff E, O;
+ TCoeff add = 1 << (shift - 1);
+#else
int E, O;
int add = 1 << (shift - 1);
+#endif
const TMatrixCoeff *iT = g_trCoreDCT2P2[TRANSFORM_INVERSE][0];
@@ -98,8 +107,13 @@ void fastInverseDCT2_B2(const TCoeff *src, TCoeff *dst, int shift, int line, int
O = iT[2] * (src[0] - src[line]);
/* Combining even and odd terms at each hierarchy levels to calculate the final spatial domain vector */
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[0] = Clip3<TCoeff>(outputMinimum, outputMaximum, (E + add) >> shift);
+ dst[1] = Clip3<TCoeff>(outputMinimum, outputMaximum, (O + add) >> shift);
+#else
dst[0] = Clip3(outputMinimum, outputMaximum, (E + add) >> shift);
dst[1] = Clip3(outputMinimum, outputMaximum, (O + add) >> shift);
+#endif
src++;
dst += 2;
@@ -176,8 +190,13 @@ void fastForwardDCT2_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int
void fastInverseDCT2_B4( const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum )
{
int j;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff E[2], O[2];
+ TCoeff add = 1 << ( shift - 1 );
+#else
int E[2], O[2];
int add = 1 << ( shift - 1 );
+#endif
const TMatrixCoeff *iT = g_trCoreDCT2P4[TRANSFORM_INVERSE][0];
@@ -191,10 +210,17 @@ void fastInverseDCT2_B4( const TCoeff *src, TCoeff *dst, int shift, int line, in
E[1] = iT[0 * 4 + 1] * src[ 0] + iT[2 * 4 + 1] * src[2 * line];
/* Combining even and odd terms at each hierarchy levels to calculate the final spatial domain vector */
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[0] = Clip3<TCoeff>( outputMinimum, outputMaximum, ( E[0] + O[0] + add ) >> shift );
+ dst[1] = Clip3<TCoeff>( outputMinimum, outputMaximum, ( E[1] + O[1] + add ) >> shift );
+ dst[2] = Clip3<TCoeff>( outputMinimum, outputMaximum, ( E[1] - O[1] + add ) >> shift );
+ dst[3] = Clip3<TCoeff>( outputMinimum, outputMaximum, ( E[0] - O[0] + add ) >> shift );
+#else
dst[0] = Clip3( outputMinimum, outputMaximum, ( E[0] + O[0] + add ) >> shift );
dst[1] = Clip3( outputMinimum, outputMaximum, ( E[1] + O[1] + add ) >> shift );
dst[2] = Clip3( outputMinimum, outputMaximum, ( E[1] - O[1] + add ) >> shift );
dst[3] = Clip3( outputMinimum, outputMaximum, ( E[0] - O[0] + add ) >> shift );
+#endif
src++;
dst += 4;
@@ -209,7 +235,11 @@ void fastInverseDCT2_B4( const TCoeff *src, TCoeff *dst, int shift, int line, in
template< int uiTrSize >
inline void _fastInverseMM( const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum, const TMatrixCoeff* iT )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const TCoeff rnd_factor = 1 << (shift - 1);
+#else
const int rnd_factor = 1 << (shift - 1);
+#endif
const int reducedLine = line - iSkipLine;
const int cutoff = uiTrSize - iSkipLine2;
@@ -217,12 +247,20 @@ inline void _fastInverseMM( const TCoeff *src, TCoeff *dst, int shift, int line,
{
for( int j = 0; j<uiTrSize; j++ )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff iSum = 0;
+#else
int iSum = 0;
+#endif
for( int k = 0; k<cutoff; k++)
{
iSum += src[k*line + i] * iT[k*uiTrSize + j];
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[i*uiTrSize + j] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iSum + rnd_factor) >> shift);
+#else
dst[i*uiTrSize + j] = Clip3(outputMinimum, outputMaximum, (int)(iSum + rnd_factor) >> shift);
+#endif
}
}
@@ -236,7 +274,11 @@ inline void _fastInverseMM( const TCoeff *src, TCoeff *dst, int shift, int line,
template< int uiTrSize >
inline void _fastForwardMM( const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TMatrixCoeff* tc )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const TCoeff rnd_factor = 1 << (shift - 1);
+#else
const int rnd_factor = 1 << (shift - 1);
+#endif
const int reducedLine = line - iSkipLine;
const int cutoff = uiTrSize - iSkipLine2;
TCoeff *pCoef;
@@ -247,7 +289,11 @@ inline void _fastForwardMM( const TCoeff *src, TCoeff *dst, int shift, int line,
const TMatrixCoeff* iT = tc;
for( int j = 0; j<cutoff; j++ )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff iSum = 0;
+#else
int iSum = 0;
+#endif
for( int k = 0; k<uiTrSize; k++ )
{
iSum += src[k] * iT[k];
@@ -344,9 +390,15 @@ void fastForwardDCT2_B8( const TCoeff *src, TCoeff *dst, int shift, int line, in
void fastInverseDCT2_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
{
int j, k;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff E[4], O[4];
+ TCoeff EE[2], EO[2];
+ TCoeff add = 1 << (shift - 1);
+#else
int E[4], O[4];
int EE[2], EO[2];
int add = 1 << (shift - 1);
+#endif
const TMatrixCoeff *iT = g_trCoreDCT2P8[TRANSFORM_INVERSE][0];
@@ -372,8 +424,13 @@ void fastInverseDCT2_B8(const TCoeff *src, TCoeff *dst, int shift, int line, int
for( k = 0; k < 4; k++ )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[k ] = Clip3<TCoeff>( outputMinimum, outputMaximum, ( E[ k] + O[ k] + add ) >> shift );
+ dst[k + 4] = Clip3<TCoeff>( outputMinimum, outputMaximum, ( E[3 - k] - O[3 - k] + add ) >> shift );
+#else
dst[k ] = Clip3( outputMinimum, outputMaximum, ( E[ k] + O[ k] + add ) >> shift );
dst[k + 4] = Clip3( outputMinimum, outputMaximum, ( E[3 - k] - O[3 - k] + add ) >> shift );
+#endif
}
src++;
dst += 8;
@@ -465,10 +522,17 @@ void fastForwardDCT2_B16(const TCoeff *src, TCoeff *dst, int shift, int line, in
void fastInverseDCT2_B16( const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum )
{
int j, k;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff E [8], O [8];
+ TCoeff EE [4], EO [4];
+ TCoeff EEE[2], EEO[2];
+ TCoeff add = 1 << ( shift - 1 );
+#else
int E [8], O [8];
int EE [4], EO [4];
int EEE[2], EEO[2];
int add = 1 << ( shift - 1 );
+#endif
const TMatrixCoeff *iT = g_trCoreDCT2P16[TRANSFORM_INVERSE][0];
@@ -504,8 +568,13 @@ void fastInverseDCT2_B16( const TCoeff *src, TCoeff *dst, int shift, int line, i
}
for( k = 0; k < 8; k++ )
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[k ] = Clip3<TCoeff>( outputMinimum, outputMaximum, ( E[ k] + O[ k] + add ) >> shift );
+ dst[k + 8] = Clip3<TCoeff>( outputMinimum, outputMaximum, ( E[7 - k] - O[7 - k] + add ) >> shift );
+#else
dst[k ] = Clip3( outputMinimum, outputMaximum, ( E[ k] + O[ k] + add ) >> shift );
dst[k + 8] = Clip3( outputMinimum, outputMaximum, ( E[7 - k] - O[7 - k] + add ) >> shift );
+#endif
}
src++;
dst += 16;
@@ -607,13 +676,21 @@ void fastForwardDCT2_B32( const TCoeff *src, TCoeff *dst, int shift, int line, i
*/
void fastInverseDCT2_B32(const TCoeff *src, TCoeff *dst, int shift, int line, int iSkipLine, int iSkipLine2, const TCoeff outputMinimum, const TCoeff outputMaximum)
{
-
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff j, k;
+ TCoeff E[16], O[16];
+ TCoeff EE[8], EO[8];
+ TCoeff EEE[4], EEO[4];
+ TCoeff EEEE[2], EEEO[2];
+ TCoeff add = 1 << (shift - 1);
+#else
int j, k;
int E[16], O[16];
int EE[8], EO[8];
int EEE[4], EEO[4];
int EEEE[2], EEEO[2];
int add = 1 << (shift - 1);
+#endif
const TMatrixCoeff *iT = g_trCoreDCT2P32[TRANSFORM_INVERSE][0];
@@ -659,8 +736,13 @@ void fastInverseDCT2_B32(const TCoeff *src, TCoeff *dst, int shift, int line, in
}
for (k = 0;k<16;k++)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[k] = Clip3<TCoeff>(outputMinimum, outputMaximum, (E[k] + O[k] + add) >> shift);
+ dst[k + 16] = Clip3<TCoeff>(outputMinimum, outputMaximum, (E[15 - k] - O[15 - k] + add) >> shift);
+#else
dst[k] = Clip3(outputMinimum, outputMaximum, (E[k] + O[k] + add) >> shift);
dst[k + 16] = Clip3(outputMinimum, outputMaximum, (E[15 - k] - O[15 - k] + add) >> shift);
+#endif
}
src++;
dst += 32;
@@ -851,8 +933,13 @@ void fastInverseDCT2_B64(const TCoeff *src, TCoeff *dst, int shift, int line, in
}
for (k = 0;k<32;k++)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[k] = Clip3<TCoeff>(outputMinimum, outputMaximum, (E[k] + O[k] + rnd_factor) >> shift);
+ dst[k + 32] = Clip3<TCoeff>(outputMinimum, outputMaximum, (E[31 - k] - O[31 - k] + rnd_factor) >> shift);
+#else
dst[k] = Clip3(outputMinimum, outputMaximum, (E[k] + O[k] + rnd_factor) >> shift);
dst[k + 32] = Clip3(outputMinimum, outputMaximum, (E[31 - k] - O[31 - k] + rnd_factor) >> shift);
+#endif
}
src++;
dst += uiTrSize;
@@ -871,7 +958,11 @@ void fastForwardDST7_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int
const TMatrixCoeff *iT = g_trCoreDST7P4[TRANSFORM_FORWARD][0];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff c[4];
+#else
int c[4];
+#endif
TCoeff *pCoeff = dst;
const int reducedLine = line - iSkipLine;
for (i = 0; i<reducedLine; i++)
@@ -918,10 +1009,17 @@ void fastInverseDST7_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int
c[2] = src[0 * line] - src[3 * line];
c[3] = iT[2] * src[1 * line];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[0] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[0] * c[0] + iT[1] * c[1] + c[3] + rnd_factor) >> shift);
+ dst[1] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[1] * c[2] - iT[0] * c[1] + c[3] + rnd_factor) >> shift);
+ dst[2] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[2] * (src[0 * line] - src[2 * line] + src[3 * line]) + rnd_factor) >> shift);
+ dst[3] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[1] * c[0] + iT[0] * c[2] - c[3] + rnd_factor) >> shift);
+#else
dst[0] = Clip3(outputMinimum, outputMaximum, (iT[0] * c[0] + iT[1] * c[1] + c[3] + rnd_factor) >> shift);
dst[1] = Clip3(outputMinimum, outputMaximum, (iT[1] * c[2] - iT[0] * c[1] + c[3] + rnd_factor) >> shift);
dst[2] = Clip3(outputMinimum, outputMaximum, (iT[2] * (src[0 * line] - src[2 * line] + src[3 * line]) + rnd_factor) >> shift);
dst[3] = Clip3(outputMinimum, outputMaximum, (iT[1] * c[0] + iT[0] * c[2] - c[3] + rnd_factor) >> shift);
+#endif
dst += 4;
src++;
@@ -1037,6 +1135,28 @@ void fastInverseDST7_B16(const TCoeff *src, TCoeff *dst, int shift, int line, in
t = iT[10] * src[5 * line];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[ 2] = Clip3<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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);
+#else
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);
@@ -1057,7 +1177,7 @@ void fastInverseDST7_B16(const TCoeff *src, TCoeff *dst, int shift, int line, in
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);
-
+#endif
src++;
dst += 16;
}
@@ -1209,6 +1329,42 @@ void fastInverseDST7_B32(const TCoeff *src, TCoeff *dst, int shift, int line, in
t[0] = iT[12] * src[6*line] + iT[25] * src[19*line];
t[1] = iT[25] * src[6*line] - iT[12] * src[19*line];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[ 0] = Clip3<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, (-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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, (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<TCoeff>(outputMinimum, outputMaximum, (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<TCoeff>(outputMinimum, outputMaximum, (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<TCoeff>(outputMinimum, outputMaximum, (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<TCoeff>(outputMinimum, outputMaximum, (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<TCoeff>(outputMinimum, outputMaximum, (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<TCoeff>(outputMinimum, outputMaximum, (iT[12]*c[0] + iT[25]*c[1] + add) >> shift);
+ dst[25] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[25]*c[0] - iT[12]*c[1] + add) >> shift);
+#else
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);
@@ -1243,7 +1399,7 @@ void fastInverseDST7_B32(const TCoeff *src, TCoeff *dst, int shift, int line, in
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);
-
+#endif
src++;
dst += 32;
}
@@ -1265,7 +1421,11 @@ void fastForwardDCT8_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int
int rnd_factor = 1 << (shift - 1);
const TMatrixCoeff *iT = g_trCoreDCT8P4[TRANSFORM_FORWARD][0];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff c[4];
+#else
int c[4];
+#endif
TCoeff *pCoeff = dst;
const int reducedLine = line - iSkipLine;
for (i = 0; i<reducedLine; i++)
@@ -1302,7 +1462,11 @@ void fastInverseDCT8_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int
const TMatrixCoeff *iT = g_trCoreDCT8P4[TRANSFORM_INVERSE][0];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ TCoeff c[4];
+#else
int c[4];
+#endif
const int reducedLine = line - iSkipLine;
for (i = 0; i<reducedLine; i++)
{
@@ -1312,11 +1476,17 @@ void fastInverseDCT8_B4(const TCoeff *src, TCoeff *dst, int shift, int line, int
c[2] = src[3 * line] - src[2 * line];
c[3] = iT[1] * src[1 * line];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[0] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[3] * c[0] + iT[2] * c[1] + c[3] + rnd_factor) >> shift);
+ dst[1] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[1] * (src[0 * line] - src[2 * line] - src[3 * line]) + rnd_factor) >> shift);
+ dst[2] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[3] * c[2] + iT[2] * c[0] - c[3] + rnd_factor) >> shift);
+ dst[3] = Clip3<TCoeff>(outputMinimum, outputMaximum, (iT[3] * c[1] - iT[2] * c[2] - c[3] + rnd_factor) >> shift);
+#else
dst[0] = Clip3(outputMinimum, outputMaximum, (iT[3] * c[0] + iT[2] * c[1] + c[3] + rnd_factor) >> shift);
dst[1] = Clip3(outputMinimum, outputMaximum, (iT[1] * (src[0 * line] - src[2 * line] - src[3 * line]) + rnd_factor) >> shift);
dst[2] = Clip3(outputMinimum, outputMaximum, (iT[3] * c[2] + iT[2] * c[0] - c[3] + rnd_factor) >> shift);
dst[3] = Clip3(outputMinimum, outputMaximum, (iT[3] * c[1] - iT[2] * c[2] - c[3] + rnd_factor) >> shift);
-
+#endif
dst += 4;
src++;
}
@@ -1430,6 +1600,26 @@ void fastInverseDCT8_B16(const TCoeff *src, TCoeff *dst, int shift, int line, in
t = iT[10] * src[5*line];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[ 1] = Clip3<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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 );
+#else
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);
@@ -1448,7 +1638,7 @@ void fastInverseDCT8_B16(const TCoeff *src, TCoeff *dst, int shift, int line, in
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 );
-
+#endif
src++;
dst += 16;
}
@@ -1603,6 +1793,42 @@ void fastInverseDCT8_B32(const TCoeff *src, TCoeff *dst, int shift, int line, in
t[0] = iT[12] * src[19 * line] + iT[25] * src[ 6 * line];
t[1] = iT[12] * src[ 6 * line] - iT[25] * src[19 * line];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ dst[ 0] = Clip3<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( 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<TCoeff>(outputMinimum, outputMaximum, ( - 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<TCoeff>(outputMinimum, outputMaximum, ( iT[12] * c[0] + iT[25] * c[1] + add) >> shift);
+ dst[19] = Clip3<TCoeff>(outputMinimum, outputMaximum, ( - iT[25] * c[0] + iT[12] * c[1] + add) >> shift);
+#else
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);
@@ -1638,6 +1864,7 @@ void fastInverseDCT8_B32(const TCoeff *src, TCoeff *dst, int shift, int line, in
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);
+#endif
src++;
dst += 32;
}
diff --git a/source/Lib/CommonLib/TypeDef.h b/source/Lib/CommonLib/TypeDef.h
index c82f1006d..c455aec1a 100644
--- a/source/Lib/CommonLib/TypeDef.h
+++ b/source/Lib/CommonLib/TypeDef.h
@@ -56,6 +56,9 @@
#define RETRAIN_CABAC 1 // CABAC initial values retrained on VTM-9.0rc1
+#define JVET_R0351_HIGH_BIT_DEPTH_SUPPORT 1 // JVET-R0351: high bit depth coding support (syntax changes, no mathematical differences for CTCs)
+#define JVET_R0351_HIGH_BIT_DEPTH_ENABLED 0 // JVET-R0351: high bit depth coding enabled (increases accuracies of some calculations, e.g. transforms)
+
#define JVET_R0058 1 // JVET-R0058: the combination of RPR, subpictures, and scalability
#define JVET_R0185_OLS_DPB_CLEANUP 1 // JVET-R0185: Replace if( !vps_all_independent_layers_flag ) condition on vps_num_dpb_params syntax element with if(!each_layer_is_an_ols_flag)
@@ -327,8 +330,12 @@ typedef std::pair<int, int> TrCost;
// This can be enabled by the makefile
#ifndef RExt__HIGH_BIT_DEPTH_SUPPORT
+#if JVET_R0351_HIGH_BIT_DEPTH_ENABLED
+#define RExt__HIGH_BIT_DEPTH_SUPPORT 1 ///< 0 (default) use data type definitions for 8-10 bit video, 1 = use larger data types to allow for up to 16-bit video (originally developed as part of N0188)
+#else
#define RExt__HIGH_BIT_DEPTH_SUPPORT 0 ///< 0 (default) use data type definitions for 8-10 bit video, 1 = use larger data types to allow for up to 16-bit video (originally developed as part of N0188)
#endif
+#endif
// SIMD optimizations
#define SIMD_ENABLE 1
diff --git a/source/Lib/CommonLib/WeightPrediction.cpp b/source/Lib/CommonLib/WeightPrediction.cpp
index cf20eb209..cab07bc6d 100644
--- a/source/Lib/CommonLib/WeightPrediction.cpp
+++ b/source/Lib/CommonLib/WeightPrediction.cpp
@@ -186,7 +186,11 @@ void WeightPrediction::addWeightBi(const CPelUnitBuf &pcYuvSrc0,
const int w0 = wp0[compID].w;
const int offset = wp0[compID].offset;
const int clipBD = clpRng.bd;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shiftNum = IF_INTERNAL_FRAC_BITS(clipBD);
+#else
const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clipBD));
+#endif
const int shift = wp0[compID].shift + shiftNum;
const int round = (enableRounding[compID] && (shift > 0)) ? (1 << (shift - 1)) : 0;
const int w1 = wp1[compID].w;
@@ -243,7 +247,11 @@ void WeightPrediction::addWeightBiComponent(const CPelUnitBuf &pcYuvSrc
const int w0 = wp0[compID].w;
const int offset = wp0[compID].offset;
const int clipBD = clpRng.bd;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shiftNum = IF_INTERNAL_FRAC_BITS(clipBD);
+#else
const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clipBD));
+#endif
const int shift = wp0[compID].shift + shiftNum;
const int round = (enableRounding[compID] && (shift > 0)) ? (1 << (shift - 1)) : 0;
const int w1 = wp1[compID].w;
@@ -304,7 +312,11 @@ void WeightPrediction::addWeightUni(const CPelUnitBuf &pcYuvSrc0,
const int w0 = wp0[compID].w;
const int offset = wp0[compID].offset;
const int clipBD = clpRng.bd;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int shiftNum = IF_INTERNAL_FRAC_BITS(clipBD);
+#else
const int shiftNum = std::max<int>(2, (IF_INTERNAL_PREC - clipBD));
+#endif
const int shift = wp0[compID].shift + shiftNum;
const uint32_t iSrc0Stride = pcYuvSrc0.bufs[compID].stride;
const uint32_t iDstStride = rpcYuvDst.bufs[compID].stride;
diff --git a/source/Lib/CommonLib/x86/AdaptiveLoopFilterX86.h b/source/Lib/CommonLib/x86/AdaptiveLoopFilterX86.h
index 12c44a2f8..447d02b5c 100644
--- a/source/Lib/CommonLib/x86/AdaptiveLoopFilterX86.h
+++ b/source/Lib/CommonLib/x86/AdaptiveLoopFilterX86.h
@@ -297,7 +297,11 @@ static void simdDeriveClassificationBlk(AlfClassifier **classifier, int **laplac
template<X86_VEXT vext>
static void simdFilter5x5Blk(AlfClassifier **classifier, const PelUnitBuf &recDst, const CPelUnitBuf &recSrc,
const Area &blkDst, const Area &blk, const ComponentID compId, const short *filterSet,
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#else
const short *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#endif
int vbPos)
{
@@ -484,7 +488,11 @@ static const uint16_t shuffleTab[4][2][8] = {
template<X86_VEXT vext>
static void simdFilter7x7Blk(AlfClassifier **classifier, const PelUnitBuf &recDst, const CPelUnitBuf &recSrc,
const Area &blkDst, const Area &blk, const ComponentID compId, const short *filterSet,
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#else
const short *fClipSet, const ClpRng &clpRng, CodingStructure &cs, const int vbCTUHeight,
+#endif
int vbPos)
{
CHECK((vbCTUHeight & (vbCTUHeight - 1)) != 0, "vbCTUHeight must be a power of 2");
diff --git a/source/Lib/CommonLib/x86/InterpolationFilterX86.h b/source/Lib/CommonLib/x86/InterpolationFilterX86.h
index c4e093f85..6969ebcea 100644
--- a/source/Lib/CommonLib/x86/InterpolationFilterX86.h
+++ b/source/Lib/CommonLib/x86/InterpolationFilterX86.h
@@ -68,7 +68,11 @@ static void fullPelCopySSE( const ClpRng& clpRng, const void*_src, int srcStride
{
Tsrc* src = (Tsrc*)_src;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ int headroom = IF_INTERNAL_FRAC_BITS(clpRng.bd);
+#else
int headroom = IF_INTERNAL_PREC - clpRng.bd;
+#endif
int headroom_offset = 1 << ( headroom - 1 );
int offset = IF_INTERNAL_OFFS;
__m128i voffset = _mm_set1_epi16( offset );
@@ -131,7 +135,11 @@ static void fullPelCopyAVX2( const ClpRng& clpRng, const void*_src, int srcStrid
#ifdef USE_AVX2
Tsrc* src = (Tsrc*)_src;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ int headroom = IF_INTERNAL_FRAC_BITS(clpRng.bd);
+#else
int headroom = IF_INTERNAL_PREC - clpRng.bd;
+#endif
int offset = 1 << ( headroom - 1 );
int internal_offset = IF_INTERNAL_OFFS;
@@ -1184,7 +1192,11 @@ static void simdFilter( const ClpRng& clpRng, Pel const *src, int srcStride, Pel
src -= ( N/2 - 1 ) * cStride;
int offset;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ int headRoom = IF_INTERNAL_FRAC_BITS(clpRng.bd);
+#else
int headRoom = std::max<int>( 2, ( IF_INTERNAL_PREC - clpRng.bd ) );
+#endif
int shift = IF_FILTER_PREC;
// with the current settings (IF_INTERNAL_PREC = 14 and IF_FILTER_PREC = 6), though headroom can be
// negative for bit depths greater than 14, shift will remain non-negative for bit depths of 8->20
@@ -1339,7 +1351,11 @@ void xWeightedGeoBlk_SSE(const PredictionUnit &pu, const uint32_t width, const u
const char log2WeightBase = 3;
const ClpRng clpRng = pu.cu->slice->clpRngs().comp[compIdx];
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const int32_t shiftWeighted = IF_INTERNAL_FRAC_BITS(clpRng.bd) + log2WeightBase;
+#else
const int32_t shiftWeighted = std::max<int>(2, (IF_INTERNAL_PREC - clpRng.bd)) + log2WeightBase;
+#endif
const int32_t offsetWeighted = (1 << (shiftWeighted - 1)) + (IF_INTERNAL_OFFS << log2WeightBase);
int16_t wIdx = floorLog2(pu.lwidth()) - GEO_MIN_CU_LOG2;
diff --git a/source/Lib/DecoderLib/CABACReader.cpp b/source/Lib/DecoderLib/CABACReader.cpp
index 248ca9e3d..c972b82d2 100644
--- a/source/Lib/DecoderLib/CABACReader.cpp
+++ b/source/Lib/DecoderLib/CABACReader.cpp
@@ -3185,10 +3185,19 @@ int CABACReader::last_sig_coeff( CoeffCodingContext& cctx, TransformUnit& tu, Co
return scanPos;
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+static void check_coeff_conformance(const CoeffCodingContext& cctx, const TCoeff coeff)
+#else
static void check_coeff_conformance(TCoeff coeff)
+#endif
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ CHECK( coeff < cctx.minCoeff() || coeff > cctx.maxCoeff(),
+ "TransCoeffLevel outside allowable range" );
+#else
CHECK( coeff < COEFF_MIN || coeff > COEFF_MAX,
"TransCoeffLevel should be in the range [-32768, 32767]" );
+#endif
}
void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* coeff, const int stateTransTable, int& state )
@@ -3339,7 +3348,11 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
sumAbs += AbsCoeff;
coeff[ sigBlkPos[k] ] = ( signPattern & ( 1u << 31 ) ? -AbsCoeff : AbsCoeff );
signPattern <<= 1;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ check_coeff_conformance( cctx, coeff[ sigBlkPos[k] ] );
+#else
check_coeff_conformance( coeff[ sigBlkPos[k] ] );
+#endif
}
if( numNonZero > numSigns )
{
@@ -3347,7 +3360,11 @@ void CABACReader::residual_coding_subblock( CoeffCodingContext& cctx, TCoeff* co
int AbsCoeff = coeff[ sigBlkPos[ k ] ];
sumAbs += AbsCoeff;
coeff[ sigBlkPos[k] ] = ( sumAbs & 1 ? -AbsCoeff : AbsCoeff );
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ check_coeff_conformance( cctx, coeff[ sigBlkPos[k] ] );
+#else
check_coeff_conformance( coeff[ sigBlkPos[k] ] );
+#endif
}
}
@@ -3464,7 +3481,11 @@ void CABACReader::residual_coding_subblockTS( CoeffCodingContext& cctx, TCoeff*
DTRACE( g_trace_ctx, D_SYNTAX_RESI, "ts_par_flag() bin=%d ctx=%d\n", parFlag, cctx.parityCtxIdAbsTS() );
cctx.decimateNumCtxBins(1);
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ coeff[ blkPos ] = (sign ? -1 : 1 ) * (TCoeff)(1 + parFlag + gt1Flag);
+#else
coeff[ blkPos ] = (sign ? -1 : 1 ) * (1 + parFlag + gt1Flag);
+#endif
}
lastScanPosPass1 = nextSigPos;
}
@@ -3538,7 +3559,11 @@ void CABACReader::residual_coding_subblockTS( CoeffCodingContext& cctx, TCoeff*
int AbsCoeff = coeff[ sigBlkPos[ k ] ];
coeff[ sigBlkPos[k] ] = ( signPattern & 1 ? -AbsCoeff : AbsCoeff );
signPattern >>= 1;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ check_coeff_conformance( cctx, coeff[ sigBlkPos[k] ] );
+#else
check_coeff_conformance( coeff[ sigBlkPos[k] ] );
+#endif
}
}
diff --git a/source/Lib/DecoderLib/VLCReader.cpp b/source/Lib/DecoderLib/VLCReader.cpp
index 518bbb73d..d7d34965b 100644
--- a/source/Lib/DecoderLib/VLCReader.cpp
+++ b/source/Lib/DecoderLib/VLCReader.cpp
@@ -5614,7 +5614,11 @@ void HLSyntaxReader::alfFilter( AlfParam& alfParam, const bool isChroma, const i
AlfFilterShape alfShape( isChroma ? 5 : 7 );
const int numFilters = isChroma ? 1 : alfParam.numLumaFilters;
short* coeff = isChroma ? alfParam.chromaCoeff[altIdx] : alfParam.lumaCoeff;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel* clipp = isChroma ? alfParam.chromaClipp[altIdx] : alfParam.lumaClipp;
+#else
short* clipp = isChroma ? alfParam.chromaClipp[altIdx] : alfParam.lumaClipp;
+#endif
// Filter coefficients
diff --git a/source/Lib/EncoderLib/EncAdaptiveLoopFilter.cpp b/source/Lib/EncoderLib/EncAdaptiveLoopFilter.cpp
index f8021943d..8fadc3b33 100644
--- a/source/Lib/EncoderLib/EncAdaptiveLoopFilter.cpp
+++ b/source/Lib/EncoderLib/EncAdaptiveLoopFilter.cpp
@@ -2255,7 +2255,11 @@ void EncAdaptiveLoopFilter::getBlkStats(AlfCovariance* alfCovariance, const AlfF
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel ELocal[MAX_NUM_ALF_LUMA_COEFF][MaxAlfNumClippingValues];
+#else
int ELocal[MAX_NUM_ALF_LUMA_COEFF][MaxAlfNumClippingValues];
+#endif
const int numBins = AlfNumClippingValues[channel];
int transposeIdx = 0;
@@ -2283,7 +2287,11 @@ void EncAdaptiveLoopFilter::getBlkStats(AlfCovariance* alfCovariance, const AlfF
{
weight = m_lumaLevelToWeightPLUT[org[j]];
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Intermediate_Int yLocal = org[j] - rec[j];
+#else
int yLocal = org[j] - rec[j];
+#endif
calcCovariance(ELocal, rec + j, recStride, shape, transposeIdx, channel, vbDistance);
for( int k = 0; k < shape.numCoeff; k++ )
{
@@ -2295,11 +2303,19 @@ void EncAdaptiveLoopFilter::getBlkStats(AlfCovariance* alfCovariance, const AlfF
{
if (m_alfWSSD)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance[classIdx].E[b0][b1][k][l] += weight * (ELocal[k][b0] * (double)ELocal[l][b1]);
+#else
alfCovariance[classIdx].E[b0][b1][k][l] += weight * (double)(ELocal[k][b0] * ELocal[l][b1]);
+#endif
}
else
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance[classIdx].E[b0][b1][k][l] += ELocal[k][b0] * (double)ELocal[l][b1];
+#else
alfCovariance[classIdx].E[b0][b1][k][l] += ELocal[k][b0] * ELocal[l][b1];
+#endif
}
}
}
@@ -2308,21 +2324,37 @@ void EncAdaptiveLoopFilter::getBlkStats(AlfCovariance* alfCovariance, const AlfF
{
if (m_alfWSSD)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance[classIdx].y[b][k] += weight * (ELocal[k][b] * (double)yLocal);
+#else
alfCovariance[classIdx].y[b][k] += weight * (double)(ELocal[k][b] * yLocal);
+#endif
}
else
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance[classIdx].y[b][k] += ELocal[k][b] * (double)yLocal;
+#else
alfCovariance[classIdx].y[b][k] += ELocal[k][b] * yLocal;
+#endif
}
}
}
if (m_alfWSSD)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance[classIdx].pixAcc += weight * (yLocal * (double)yLocal);
+#else
alfCovariance[classIdx].pixAcc += weight * (double)(yLocal * yLocal);
+#endif
}
else
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance[classIdx].pixAcc += yLocal * (double)yLocal;
+#else
alfCovariance[classIdx].pixAcc += yLocal * yLocal;
+#endif
}
}
org += orgStride;
@@ -2348,7 +2380,11 @@ void EncAdaptiveLoopFilter::getBlkStats(AlfCovariance* alfCovariance, const AlfF
}
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+void EncAdaptiveLoopFilter::calcCovariance(Pel ELocal[MAX_NUM_ALF_LUMA_COEFF][MaxAlfNumClippingValues], const Pel *rec, const int stride, const AlfFilterShape& shape, const int transposeIdx, const ChannelType channel, int vbDistance)
+#else
void EncAdaptiveLoopFilter::calcCovariance(int ELocal[MAX_NUM_ALF_LUMA_COEFF][MaxAlfNumClippingValues], const Pel *rec, const int stride, const AlfFilterShape& shape, const int transposeIdx, const ChannelType channel, int vbDistance)
+#endif
{
int clipTopRow = -4;
int clipBotRow = 4;
@@ -2369,7 +2405,11 @@ void EncAdaptiveLoopFilter::calcCovariance(int ELocal[MAX_NUM_ALF_LUMA_COEFF][Ma
int k = 0;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel curr = rec[0];
+#else
const short curr = rec[0];
+#endif
if( transposeIdx == 0 )
{
@@ -2665,7 +2705,11 @@ void EncAdaptiveLoopFilter::alfEncoderCtb(CodingStructure& cs, AlfParam& alfPar
for (int classIdx = 0; classIdx < MAX_NUM_ALF_CLASSES; classIdx++)
{
short* pCoeff = m_coeffFinal;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel* pClipp = m_clippFinal;
+#else
short* pClipp = m_clippFinal;
+#endif
for (int i = 0; i < MAX_NUM_ALF_LUMA_COEFF; i++)
{
m_filterTmp[i] = pCoeff[classIdx * MAX_NUM_ALF_LUMA_COEFF + i];
@@ -2743,7 +2787,11 @@ void EncAdaptiveLoopFilter::alfEncoderCtb(CodingStructure& cs, AlfParam& alfPar
else
{
short *pCoeff;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel *pClipp;
+#else
short *pClipp;
+#endif
if (useNewFilter && filterSetIdx == NUM_FIXED_FILTER_SETS)
{
pCoeff = m_coeffFinal;
@@ -3147,7 +3195,11 @@ void EncAdaptiveLoopFilter::alfReconstructor(CodingStructure& cs, const PelUnitB
const Area blkDst(xStart, yStart, w, h);
short filterSetIndex = alfCtuFilterIndex[ctuIdx];
short *coeff;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel *clip;
+#else
short *clip;
+#endif
if (filterSetIndex >= NUM_FIXED_FILTER_SETS)
{
coeff = m_coeffApsLuma[filterSetIndex - NUM_FIXED_FILTER_SETS];
@@ -3196,7 +3248,11 @@ void EncAdaptiveLoopFilter::alfReconstructor(CodingStructure& cs, const PelUnitB
Area blk(xPos, yPos, width, height);
short filterSetIndex = alfCtuFilterIndex[ctuIdx];
short *coeff;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel *clip;
+#else
short *clip;
+#endif
if (filterSetIndex >= NUM_FIXED_FILTER_SETS)
{
coeff = m_coeffApsLuma[filterSetIndex - NUM_FIXED_FILTER_SETS];
@@ -4147,7 +4203,11 @@ void EncAdaptiveLoopFilter::getBlkStatsCcAlf(AlfCovariance &alfCovariance, const
vbPos = m_picHeight;
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Pel ELocal[MAX_NUM_CC_ALF_CHROMA_COEFF][1];
+#else
int ELocal[MAX_NUM_CC_ALF_CHROMA_COEFF][1];
+#endif
for (int i = 0; i < compArea.height; i++)
{
@@ -4167,7 +4227,11 @@ void EncAdaptiveLoopFilter::getBlkStatsCcAlf(AlfCovariance &alfCovariance, const
weight = m_lumaLevelToWeightPLUT[org[j]];
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ Intermediate_Int yLocal = org[j] - rec[compID][j];
+#else
int yLocal = org[j] - rec[compID][j];
+#endif
calcCovarianceCcAlf( ELocal, rec[COMPONENT_Y] + ( j << getComponentScaleX(compID, m_chromaFormat)), recStride[COMPONENT_Y], shape, vbDistance );
@@ -4181,11 +4245,19 @@ void EncAdaptiveLoopFilter::getBlkStatsCcAlf(AlfCovariance &alfCovariance, const
{
if (m_alfWSSD)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance.E[b0][b1][k][l] += weight * (ELocal[k][b0] * (double)ELocal[l][b1]);
+#else
alfCovariance.E[b0][b1][k][l] += weight * (double) (ELocal[k][b0] * ELocal[l][b1]);
+#endif
}
else
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance.E[b0][b1][k][l] += ELocal[k][b0] * (double)ELocal[l][b1];
+#else
alfCovariance.E[b0][b1][k][l] += ELocal[k][b0] * ELocal[l][b1];
+#endif
}
}
}
@@ -4194,21 +4266,37 @@ void EncAdaptiveLoopFilter::getBlkStatsCcAlf(AlfCovariance &alfCovariance, const
{
if (m_alfWSSD)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance.y[b][k] += weight * (ELocal[k][b] * (double)yLocal);
+#else
alfCovariance.y[b][k] += weight * (double) (ELocal[k][b] * yLocal);
+#endif
}
else
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance.y[b][k] += ELocal[k][b] * (double)yLocal;
+#else
alfCovariance.y[b][k] += ELocal[k][b] * yLocal;
+#endif
}
}
}
if (m_alfWSSD)
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance.pixAcc += weight * (yLocal * (double)yLocal);
+#else
alfCovariance.pixAcc += weight * (double) (yLocal * yLocal);
+#endif
}
else
{
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ alfCovariance.pixAcc += yLocal * (double)yLocal;
+#else
alfCovariance.pixAcc += yLocal * yLocal;
+#endif
}
}
org += orgStride;
@@ -4248,7 +4336,11 @@ void EncAdaptiveLoopFilter::getBlkStatsCcAlf(AlfCovariance &alfCovariance, const
}
}
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+void EncAdaptiveLoopFilter::calcCovarianceCcAlf(Pel ELocal[MAX_NUM_CC_ALF_CHROMA_COEFF][1], const Pel *rec, const int stride, const AlfFilterShape& shape, int vbDistance)
+#else
void EncAdaptiveLoopFilter::calcCovarianceCcAlf(int ELocal[MAX_NUM_CC_ALF_CHROMA_COEFF][1], const Pel *rec, const int stride, const AlfFilterShape& shape, int vbDistance)
+#endif
{
CHECK(shape.filterType != CC_ALF, "Bad CC ALF shape");
diff --git a/source/Lib/EncoderLib/EncAdaptiveLoopFilter.h b/source/Lib/EncoderLib/EncAdaptiveLoopFilter.h
index 0aa905998..525aa16c1 100644
--- a/source/Lib/EncoderLib/EncAdaptiveLoopFilter.h
+++ b/source/Lib/EncoderLib/EncAdaptiveLoopFilter.h
@@ -318,13 +318,21 @@ private:
void getFrameStat( AlfCovariance* frameCov, AlfCovariance** ctbCov, uint8_t* ctbEnableFlags, uint8_t* ctbAltIdx, const int numClasses, int altIdx );
void deriveStatsForFiltering( PelUnitBuf& orgYuv, PelUnitBuf& recYuv, CodingStructure& cs );
void getBlkStats(AlfCovariance* alfCovariace, const AlfFilterShape& shape, AlfClassifier** classifier, Pel* org, const int orgStride, Pel* rec, const int recStride, const CompArea& areaDst, const CompArea& area, const ChannelType channel, int vbCTUHeight, int vbPos);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ void calcCovariance(Pel ELocal[MAX_NUM_ALF_LUMA_COEFF][MaxAlfNumClippingValues], const Pel *rec, const int stride, const AlfFilterShape& shape, const int transposeIdx, const ChannelType channel, int vbDistance);
+#else
void calcCovariance(int ELocal[MAX_NUM_ALF_LUMA_COEFF][MaxAlfNumClippingValues], const Pel *rec, const int stride, const AlfFilterShape& shape, const int transposeIdx, const ChannelType channel, int vbDistance);
+#endif
void deriveStatsForCcAlfFiltering(const PelUnitBuf &orgYuv, const PelUnitBuf &recYuv, const int compIdx,
const int maskStride, const uint8_t filterIdc, CodingStructure &cs);
void getBlkStatsCcAlf(AlfCovariance &alfCovariance, const AlfFilterShape &shape, const PelUnitBuf &orgYuv,
const PelUnitBuf &recYuv, const UnitArea &areaDst, const UnitArea &area,
const ComponentID compID, const int yPos);
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ void calcCovarianceCcAlf(Pel ELocal[MAX_NUM_CC_ALF_CHROMA_COEFF][1], const Pel* rec, const int stride, const AlfFilterShape& shape, int vbDistance);
+#else
void calcCovarianceCcAlf(int ELocal[MAX_NUM_CC_ALF_CHROMA_COEFF][1], const Pel* rec, const int stride, const AlfFilterShape& shape, int vbDistance);
+#endif
void mergeClasses(const AlfFilterShape& alfShape, AlfCovariance* cov, AlfCovariance* covMerged, int clipMerged[MAX_NUM_ALF_CLASSES][MAX_NUM_ALF_CLASSES][MAX_NUM_ALF_LUMA_COEFF], const int numClasses, short filterIndices[MAX_NUM_ALF_CLASSES][MAX_NUM_ALF_CLASSES]);
diff --git a/source/Lib/EncoderLib/VLCWriter.cpp b/source/Lib/EncoderLib/VLCWriter.cpp
index ec0a1bd15..4db625ea0 100644
--- a/source/Lib/EncoderLib/VLCWriter.cpp
+++ b/source/Lib/EncoderLib/VLCWriter.cpp
@@ -3291,7 +3291,11 @@ void HLSWriter::alfFilter( const AlfParam& alfParam, const bool isChroma, const
{
AlfFilterShape alfShape(isChroma ? 5 : 7);
const short* coeff = isChroma ? alfParam.chromaCoeff[altIdx] : alfParam.lumaCoeff;
+#if JVET_R0351_HIGH_BIT_DEPTH_SUPPORT
+ const Pel* clipp = isChroma ? alfParam.chromaClipp[altIdx] : alfParam.lumaClipp;
+#else
const short* clipp = isChroma ? alfParam.chromaClipp[altIdx] : alfParam.lumaClipp;
+#endif
const int numFilters = isChroma ? 1 : alfParam.numLumaFilters;
// vlc for all
--
GitLab