/* The copyright in this software is being made available under the BSD
* License, included below. This software may be subject to other third party
* and contributor rights, including patent rights, and no such rights are
* granted under this license.
*
* Copyright (c) 2010-2019, ITU/ISO/IEC
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* * Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
* * Neither the name of the ITU/ISO/IEC nor the names of its contributors may
* be used to endorse or promote products derived from this software without
* specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
/** \file TrQuant.cpp
\brief transform and quantization class
*/
#include "TrQuant.h"
#include "TrQuant_EMT.h"
#include "UnitTools.h"
#include "ContextModelling.h"
#include "CodingStructure.h"
#include "CrossCompPrediction.h"
#include "dtrace_buffer.h"
#include <stdlib.h>
#include <limits>
#include <memory.h>
#include "QuantRDOQ.h"
#include "DepQuant.h"
#if RExt__DECODER_DEBUG_TOOL_STATISTICS
#include "CommonLib/CodingStatistics.h"
#endif
struct coeffGroupRDStats
{
int iNNZbeforePos0;
double d64CodedLevelandDist; // distortion and level cost only
double d64UncodedDist; // all zero coded block distortion
double d64SigCost;
double d64SigCost_0;
};
FwdTrans *fastFwdTrans[NUM_TRANS_TYPE][g_numTransformMatrixSizes] =
{
{ fastForwardDCT2_B2, fastForwardDCT2_B4, fastForwardDCT2_B8, fastForwardDCT2_B16, fastForwardDCT2_B32, fastForwardDCT2_B64 },
{ nullptr, fastForwardDCT8_B4, fastForwardDCT8_B8, fastForwardDCT8_B16, fastForwardDCT8_B32, nullptr },
{ nullptr, fastForwardDST7_B4, fastForwardDST7_B8, fastForwardDST7_B16, fastForwardDST7_B32, nullptr },
};
InvTrans *fastInvTrans[NUM_TRANS_TYPE][g_numTransformMatrixSizes] =
{
{ fastInverseDCT2_B2, fastInverseDCT2_B4, fastInverseDCT2_B8, fastInverseDCT2_B16, fastInverseDCT2_B32, fastInverseDCT2_B64 },
{ nullptr, fastInverseDCT8_B4, fastInverseDCT8_B8, fastInverseDCT8_B16, fastInverseDCT8_B32, nullptr },
{ nullptr, fastInverseDST7_B4, fastInverseDST7_B8, fastInverseDST7_B16, fastInverseDST7_B32, nullptr },
};
//! \ingroup CommonLib
//! \{
// ====================================================================================================================
// TrQuant class member functions
// ====================================================================================================================
TrQuant::TrQuant() : m_quant( nullptr )
{
// allocate temporary buffers
m_plTempCoeff = (TCoeff*) xMalloc( TCoeff, MAX_CU_SIZE * MAX_CU_SIZE );
m_mtsCoeffs = new TCoeff*[ NUM_TRAFO_MODES_MTS ];
for( int i = 0; i < NUM_TRAFO_MODES_MTS; i++ )
{
m_mtsCoeffs[i] = (TCoeff*) xMalloc( TCoeff, MAX_CU_SIZE * MAX_CU_SIZE );
}
}
TrQuant::~TrQuant()
{
if( m_quant )
{
delete m_quant;
m_quant = nullptr;
}
// delete temporary buffers
if ( m_plTempCoeff )
{
xFree( m_plTempCoeff );
m_plTempCoeff = nullptr;
}
if( m_mtsCoeffs )
{
for( int i = 0; i < NUM_TRAFO_MODES_MTS; i++ )
{
xFree( m_mtsCoeffs[i] );
m_mtsCoeffs[i] = nullptr;
}
delete[] m_mtsCoeffs;
m_mtsCoeffs = nullptr;
}
}
#if ENABLE_SPLIT_PARALLELISM
void TrQuant::copyState( const TrQuant& other )
{
m_quant->copyState( *other.m_quant );
}
#endif
void TrQuant::xDeQuant(const TransformUnit &tu,
CoeffBuf &dstCoeff,
const ComponentID &compID,
const QpParam &cQP)
{
m_quant->dequant( tu, dstCoeff, compID, cQP );
}
void TrQuant::init( const Quant* otherQuant,
const uint32_t uiMaxTrSize,
const bool bUseRDOQ,
const bool bUseRDOQTS,
#if T0196_SELECTIVE_RDOQ
const bool useSelectiveRDOQ,
#endif
const bool bEnc,
const bool useTransformSkipFast
)
{
m_uiMaxTrSize = uiMaxTrSize;
m_bEnc = bEnc;
m_useTransformSkipFast = useTransformSkipFast;
delete m_quant;
m_quant = nullptr;
if( bUseRDOQ || !bEnc )
{
m_quant = new DepQuant( otherQuant, bEnc );
}
else
m_quant = new Quant( otherQuant );
if( m_quant )
{
m_quant->init( uiMaxTrSize, bUseRDOQ, bUseRDOQTS, useSelectiveRDOQ );
}
}
void TrQuant::invTransformNxN( TransformUnit &tu, const ComponentID &compID, PelBuf &pResi, const QpParam &cQP )
{
const CompArea &area = tu.blocks[compID];
const uint32_t uiWidth = area.width;
const uint32_t uiHeight = area.height;
#if MAX_TB_SIZE_SIGNALLING
CHECK( uiWidth > tu.cs->sps->getMaxTbSize() || uiHeight > tu.cs->sps->getMaxTbSize(), "Maximal allowed transformation size exceeded!" );
#else
CHECK( uiWidth > MAX_TB_SIZEY || uiHeight > MAX_TB_SIZEY, "Maximal allowed transformation size exceeded!" );
#endif
if (tu.cu->transQuantBypass)
{
// where should this logic go?
const bool rotateResidual = TU::isNonTransformedResidualRotated(tu, compID);
const CCoeffBuf pCoeff = tu.getCoeffs(compID);
for (uint32_t y = 0, coefficientIndex = 0; y < uiHeight; y++)
{
for (uint32_t x = 0; x < uiWidth; x++, coefficientIndex++)
{
pResi.at(x, y) = rotateResidual ? pCoeff.at(pCoeff.width - x - 1, pCoeff.height - y - 1) : pCoeff.at(x, y);
}
}
}
else
{
CoeffBuf tempCoeff = CoeffBuf( m_plTempCoeff, area );
xDeQuant( tu, tempCoeff, compID, cQP );
DTRACE_COEFF_BUF( D_TCOEFF, tempCoeff, tu, tu.cu->predMode, compID );
if( isLuma(compID) && tu.mtsIdx == 1 )
{
xITransformSkip( tempCoeff, pResi, tu, compID );
}
else
{
xIT( tu, compID, tempCoeff, pResi );
}
}
//DTRACE_BLOCK_COEFF(tu.getCoeffs(compID), tu, tu.cu->predMode, compID);
DTRACE_PEL_BUF( D_RESIDUALS, pResi, tu, tu.cu->predMode, compID);
invRdpcmNxN(tu, compID, pResi);
}
void TrQuant::invRdpcmNxN(TransformUnit& tu, const ComponentID &compID, PelBuf &pcResidual)
{
const CompArea &area = tu.blocks[compID];
if (CU::isRDPCMEnabled(*tu.cu) && (tu.mtsIdx==1 || tu.cu->transQuantBypass))
{
const uint32_t uiWidth = area.width;
const uint32_t uiHeight = area.height;
RDPCMMode rdpcmMode = RDPCM_OFF;
if (tu.cu->predMode == MODE_INTRA)
{
const ChannelType chType = toChannelType(compID);
const uint32_t uiChFinalMode = PU::getFinalIntraMode(*tu.cs->getPU(area.pos(), chType), chType);
if (uiChFinalMode == VER_IDX || uiChFinalMode == HOR_IDX)
{
rdpcmMode = (uiChFinalMode == VER_IDX) ? RDPCM_VER : RDPCM_HOR;
}
}
else // not intra case
{
rdpcmMode = RDPCMMode(tu.rdpcm[compID]);
}
const TCoeff pelMin = (TCoeff) std::numeric_limits<Pel>::min();
const TCoeff pelMax = (TCoeff) std::numeric_limits<Pel>::max();
if (rdpcmMode == RDPCM_VER)
{
for (uint32_t uiX = 0; uiX < uiWidth; uiX++)
{
TCoeff accumulator = pcResidual.at(uiX, 0); // 32-bit accumulator
for (uint32_t uiY = 1; uiY < uiHeight; uiY++)
{
accumulator += pcResidual.at(uiX, uiY);
pcResidual.at(uiX, uiY) = (Pel) Clip3<TCoeff>(pelMin, pelMax, accumulator);
}
}
}
else if (rdpcmMode == RDPCM_HOR)
{
for (uint32_t uiY = 0; uiY < uiHeight; uiY++)
{
TCoeff accumulator = pcResidual.at(0, uiY);
for (uint32_t uiX = 1; uiX < uiWidth; uiX++)
{
accumulator += pcResidual.at(uiX, uiY);
pcResidual.at(uiX, uiY) = (Pel) Clip3<TCoeff>(pelMin, pelMax, accumulator);
}
}
}
}
}
// ------------------------------------------------------------------------------------------------
// Logical transform
// ------------------------------------------------------------------------------------------------
void TrQuant::getTrTypes ( TransformUnit tu, const ComponentID compID, int &trTypeHor, int &trTypeVer )
{
bool mtsActivated = CU::isIntra( *tu.cu ) ? tu.cs->sps->getUseIntraMTS() : tu.cs->sps->getUseInterMTS() && CU::isInter( *tu.cu );
bool mtsImplicit = CU::isIntra( *tu.cu ) && tu.cs->sps->getUseImplicitMTS() && compID == COMPONENT_Y;
trTypeHor = DCT2;
trTypeVer = DCT2;
if (tu.cu->ispMode && isLuma(compID))
{
TU::getTransformTypeISP(tu, compID, trTypeHor, trTypeVer);
return;
}
if( tu.cu->sbtInfo && compID == COMPONENT_Y )
{
uint8_t sbtIdx = tu.cu->getSbtIdx();
uint8_t sbtPos = tu.cu->getSbtPos();
if( sbtIdx == SBT_VER_HALF || sbtIdx == SBT_VER_QUAD )
{
assert( tu.lwidth() <= MTS_INTER_MAX_CU_SIZE );
if( tu.lheight() > MTS_INTER_MAX_CU_SIZE )
{
trTypeHor = trTypeVer = DCT2;
}
else
{
if( sbtPos == SBT_POS0 ) { trTypeHor = DCT8; trTypeVer = DST7; }
else { trTypeHor = DST7; trTypeVer = DST7; }
}
}
else
{
assert( tu.lheight() <= MTS_INTER_MAX_CU_SIZE );
if( tu.lwidth() > MTS_INTER_MAX_CU_SIZE )
{
trTypeHor = trTypeVer = DCT2;
}
else
{
if( sbtPos == SBT_POS0 ) { trTypeHor = DST7; trTypeVer = DCT8; }
else { trTypeHor = DST7; trTypeVer = DST7; }
}
}
return;
}
if ( mtsActivated )
{
if( compID == COMPONENT_Y )
{
if ( tu.mtsIdx > 1 )
{
int indHor = ( tu.mtsIdx - 2 ) & 1;
int indVer = ( tu.mtsIdx - 2 ) >> 1;
trTypeHor = indHor ? DCT8 : DST7;
trTypeVer = indVer ? DCT8 : DST7;
}
}
}
else if ( mtsImplicit )
{
int width = tu.blocks[compID].width;
int height = tu.blocks[compID].height;
bool widthDstOk = width >= 4 && width <= 16;
bool heightDstOk = height >= 4 && height <= 16;
if ( width < height && widthDstOk )
trTypeHor = DST7;
else if ( height < width && heightDstOk )
trTypeVer = DST7;
else if ( width == height && widthDstOk )
trTypeHor = trTypeVer = DST7;
}
}
void TrQuant::xT( const TransformUnit &tu, const ComponentID &compID, const CPelBuf &resi, CoeffBuf &dstCoeff, const int width, const int height )
{
const unsigned maxLog2TrDynamicRange = tu.cs->sps->getMaxLog2TrDynamicRange( toChannelType( compID ) );
const unsigned bitDepth = tu.cs->sps->getBitDepth( toChannelType( compID ) );
const int TRANSFORM_MATRIX_SHIFT = g_transformMatrixShift[TRANSFORM_FORWARD];
const uint32_t transformWidthIndex = g_aucLog2[width ] - 1; // nLog2WidthMinus1, since transform start from 2-point
const uint32_t transformHeightIndex = g_aucLog2[height] - 1; // nLog2HeightMinus1, since transform start from 2-point
int trTypeHor = DCT2;
int trTypeVer = DCT2;
getTrTypes ( tu, compID, trTypeHor, trTypeVer );
const int skipWidth = ( trTypeHor != DCT2 && width == 32 ) ? 16 : width > JVET_C0024_ZERO_OUT_TH ? width - JVET_C0024_ZERO_OUT_TH : 0;
const int skipHeight = ( trTypeVer != DCT2 && height == 32 ) ? 16 : height > JVET_C0024_ZERO_OUT_TH ? height - JVET_C0024_ZERO_OUT_TH : 0;
#if RExt__DECODER_DEBUG_TOOL_STATISTICS
if ( trTypeHor != DCT2 )
{
CodingStatistics::IncrementStatisticTool( CodingStatisticsClassType{ STATS__TOOL_EMT, uint32_t( width ), uint32_t( height ), compID } );
}
#endif
ALIGN_DATA( MEMORY_ALIGN_DEF_SIZE, TCoeff block[MAX_TB_SIZEY * MAX_TB_SIZEY] );
const Pel *resiBuf = resi.buf;
const int resiStride = resi.stride;
for( int y = 0; y < height; y++ )
{
for( int x = 0; x < width; x++ )
{
block[( y * width ) + x] = resiBuf[( y * resiStride ) + x];
}
}
if( width > 1 && height > 1 ) // 2-D transform
{
const int shift_1st = ((g_aucLog2[width ]) + bitDepth + TRANSFORM_MATRIX_SHIFT) - maxLog2TrDynamicRange + COM16_C806_TRANS_PREC;
const int shift_2nd = (g_aucLog2[height]) + TRANSFORM_MATRIX_SHIFT + COM16_C806_TRANS_PREC;
CHECK( shift_1st < 0, "Negative shift" );
CHECK( shift_2nd < 0, "Negative shift" );
TCoeff *tmp = ( TCoeff * ) alloca( width * height * sizeof( TCoeff ) );
fastFwdTrans[trTypeHor][transformWidthIndex ](block, tmp, shift_1st, height, 0, skipWidth);
fastFwdTrans[trTypeVer][transformHeightIndex](tmp, dstCoeff.buf, shift_2nd, width, skipWidth, skipHeight);
}
else if( height == 1 ) //1-D horizontal transform
{
const int shift = ((g_aucLog2[width ]) + bitDepth + TRANSFORM_MATRIX_SHIFT) - maxLog2TrDynamicRange + COM16_C806_TRANS_PREC;
CHECK( shift < 0, "Negative shift" );
CHECKD( ( transformWidthIndex < 0 ), "There is a problem with the width." );
fastFwdTrans[trTypeHor][transformWidthIndex]( block, dstCoeff.buf, shift, 1, 0, skipWidth );
}
else //if (iWidth == 1) //1-D vertical transform
{
int shift = ( ( g_aucLog2[height] ) + bitDepth + TRANSFORM_MATRIX_SHIFT ) - maxLog2TrDynamicRange + COM16_C806_TRANS_PREC;
CHECK( shift < 0, "Negative shift" );
CHECKD( ( transformHeightIndex < 0 ), "There is a problem with the height." );
fastFwdTrans[trTypeVer][transformHeightIndex]( block, dstCoeff.buf, shift, 1, 0, skipHeight );
}
}
void TrQuant::xIT( const TransformUnit &tu, const ComponentID &compID, const CCoeffBuf &pCoeff, PelBuf &pResidual )
{
const int width = pCoeff.width;
const int height = pCoeff.height;
const unsigned maxLog2TrDynamicRange = tu.cs->sps->getMaxLog2TrDynamicRange( toChannelType( compID ) );
const unsigned bitDepth = tu.cs->sps->getBitDepth( toChannelType( compID ) );
const int TRANSFORM_MATRIX_SHIFT = g_transformMatrixShift[TRANSFORM_INVERSE];
const TCoeff clipMinimum = -( 1 << maxLog2TrDynamicRange );
const TCoeff clipMaximum = ( 1 << maxLog2TrDynamicRange ) - 1;
const uint32_t transformWidthIndex = g_aucLog2[width ] - 1; // nLog2WidthMinus1, since transform start from 2-point
const uint32_t transformHeightIndex = g_aucLog2[height] - 1; // nLog2HeightMinus1, since transform start from 2-point
int trTypeHor = DCT2;
int trTypeVer = DCT2;
getTrTypes ( tu, compID, trTypeHor, trTypeVer );
const int skipWidth = ( trTypeHor != DCT2 && width == 32 ) ? 16 : width > JVET_C0024_ZERO_OUT_TH ? width - JVET_C0024_ZERO_OUT_TH : 0;
const int skipHeight = ( trTypeVer != DCT2 && height == 32 ) ? 16 : height > JVET_C0024_ZERO_OUT_TH ? height - JVET_C0024_ZERO_OUT_TH : 0;
TCoeff *block = ( TCoeff * ) alloca( width * height * sizeof( TCoeff ) );
if( width > 1 && height > 1 ) //2-D transform
{
const int shift_1st = TRANSFORM_MATRIX_SHIFT + 1 + COM16_C806_TRANS_PREC; // 1 has been added to shift_1st at the expense of shift_2nd
const int shift_2nd = ( TRANSFORM_MATRIX_SHIFT + maxLog2TrDynamicRange - 1 ) - bitDepth + COM16_C806_TRANS_PREC;
CHECK( shift_1st < 0, "Negative shift" );
CHECK( shift_2nd < 0, "Negative shift" );
TCoeff *tmp = ( TCoeff * ) alloca( width * height * sizeof( TCoeff ) );
fastInvTrans[trTypeVer][transformHeightIndex](pCoeff.buf, tmp, shift_1st, width, skipWidth, skipHeight, clipMinimum, clipMaximum);
fastInvTrans[trTypeHor][transformWidthIndex] (tmp, block, shift_2nd, height, 0, skipWidth, clipMinimum, clipMaximum);
}
else if( width == 1 ) //1-D vertical transform
{
int shift = ( TRANSFORM_MATRIX_SHIFT + maxLog2TrDynamicRange - 1 ) - bitDepth + COM16_C806_TRANS_PREC;
CHECK( shift < 0, "Negative shift" );
CHECK( ( transformHeightIndex < 0 ), "There is a problem with the height." );
fastInvTrans[trTypeVer][transformHeightIndex]( pCoeff.buf, block, shift + 1, 1, 0, skipHeight, clipMinimum, clipMaximum );
}
else //if(iHeight == 1) //1-D horizontal transform
{
const int shift = ( TRANSFORM_MATRIX_SHIFT + maxLog2TrDynamicRange - 1 ) - bitDepth + COM16_C806_TRANS_PREC;
CHECK( shift < 0, "Negative shift" );
CHECK( ( transformWidthIndex < 0 ), "There is a problem with the width." );
fastInvTrans[trTypeHor][transformWidthIndex]( pCoeff.buf, block, shift + 1, 1, 0, skipWidth, clipMinimum, clipMaximum );
}
Pel *resiBuf = pResidual.buf;
int resiStride = pResidual.stride;
for( int y = 0; y < height; y++ )
{
for( int x = 0; x < width; x++ )
{
resiBuf[( y * resiStride ) + x] = Pel( block[( y * width ) + x] );
}
}
}
/** Wrapper function between HM interface and core NxN transform skipping
*/
void TrQuant::xITransformSkip(const CCoeffBuf &pCoeff,
PelBuf &pResidual,
const TransformUnit &tu,
const ComponentID &compID)
{
const CompArea &area = tu.blocks[compID];
const int width = area.width;
const int height = area.height;
const int maxLog2TrDynamicRange = tu.cs->sps->getMaxLog2TrDynamicRange(toChannelType(compID));
const int channelBitDepth = tu.cs->sps->getBitDepth(toChannelType(compID));
int iTransformShift = getTransformShift(channelBitDepth, area.size(), maxLog2TrDynamicRange);
if( tu.cs->sps->getSpsRangeExtension().getExtendedPrecisionProcessingFlag() )
{
iTransformShift = std::max<int>( 0, iTransformShift );
}
int iWHScale = 1;
const bool rotateResidual = TU::isNonTransformedResidualRotated( tu, compID );
if( iTransformShift >= 0 )
{
const TCoeff offset = iTransformShift == 0 ? 0 : ( 1 << ( iTransformShift - 1 ) );
for( uint32_t y = 0; y < height; y++ )
{
for( uint32_t x = 0; x < width; x++ )
{
pResidual.at( x, y ) = Pel( ( ( rotateResidual ? pCoeff.at( pCoeff.width - x - 1, pCoeff.height - y - 1 ) : pCoeff.at( x, y ) ) * iWHScale + offset ) >> iTransformShift );
}
}
}
else //for very high bit depths
{
iTransformShift = -iTransformShift;
for( uint32_t y = 0; y < height; y++ )
{
for( uint32_t x = 0; x < width; x++ )
{
pResidual.at( x, y ) = Pel( ( rotateResidual ? pCoeff.at( pCoeff.width - x - 1, pCoeff.height - y - 1 ) : pCoeff.at( x, y ) ) * iWHScale << iTransformShift );
}
}
}
}
void TrQuant::xQuant(TransformUnit &tu, const ComponentID &compID, const CCoeffBuf &pSrc, TCoeff &uiAbsSum, const QpParam &cQP, const Ctx& ctx)
{
m_quant->quant( tu, compID, pSrc, uiAbsSum, cQP, ctx );
}
void TrQuant::transformNxN( TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, std::vector<TrMode>* trModes, const int maxCand, double* diagRatio, double* horVerRatio )
{
CodingStructure &cs = *tu.cs;
const CompArea &rect = tu.blocks[compID];
const uint32_t width = rect.width;
const uint32_t height = rect.height;
const CPelBuf resiBuf = cs.getResiBuf(rect);
#if MAX_TB_SIZE_SIGNALLING
CHECK( cs.sps->getMaxTbSize() < width, "Unsupported transformation size" );
#else
CHECK( MAX_TB_SIZEY < width, "Unsupported transformation size" );
#endif
int pos = 0;
std::vector<TrCost> trCosts;
std::vector<TrMode>::iterator it = trModes->begin();
const double facBB[] = { 1.2, 1.3, 1.3, 1.4, 1.5 };
while( it != trModes->end() )
{
tu.mtsIdx = it->first;
CoeffBuf tempCoeff( m_mtsCoeffs[tu.mtsIdx], rect );
if( tu.noResidual )
{
int sumAbs = 0;
trCosts.push_back( TrCost( sumAbs, pos++ ) );
it++;
continue;
}
if( isLuma(compID) && tu.mtsIdx == 1 )
{
xTransformSkip( tu, compID, resiBuf, tempCoeff.buf );
}
else
{
xT( tu, compID, resiBuf, tempCoeff, width, height );
}
int sumAbs = 0;
for( int pos = 0; pos < width*height; pos++ )
{
sumAbs += abs( tempCoeff.buf[pos] );
}
double scaleSAD=1.0;
if (isLuma(compID) && tu.mtsIdx==1 && ((g_aucLog2[width] + g_aucLog2[height]) & 1) == 1 )
{
scaleSAD=1.0/1.414213562; // compensate for not scaling transform skip coefficients by 1/sqrt(2)
}
trCosts.push_back( TrCost( int(sumAbs*scaleSAD), pos++ ) );
it++;
}
// it gets the distribution of the DCT-II coefficients energy, which will be useful to discard ISP tests
CoeffBuf coeffsDCT( m_mtsCoeffs[0], rect );
xGetCoeffEnergy( tu, compID, coeffsDCT, diagRatio, horVerRatio );
int numTests = 0;
std::vector<TrCost>::iterator itC = trCosts.begin();
const double fac = facBB[g_aucLog2[std::max(width, height)]-2];
const double thr = fac * trCosts.begin()->first;
const double thrTS = trCosts.begin()->first;
while( itC != trCosts.end() )
{
const bool testTr = itC->first <= ( itC->second == 1 ? thrTS : thr ) && numTests <= maxCand;
trModes->at( itC->second ).second = testTr;
numTests += testTr;
itC++;
}
}
void TrQuant::transformNxN( TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, const Ctx &ctx, const bool loadTr, double* diagRatio, double* horVerRatio )
{
CodingStructure &cs = *tu.cs;
const CompArea &rect = tu.blocks[compID];
const uint32_t uiWidth = rect.width;
const uint32_t uiHeight = rect.height;
const CPelBuf resiBuf = cs.getResiBuf(rect);
CoeffBuf rpcCoeff = tu.getCoeffs(compID);
if( tu.noResidual )
{
uiAbsSum = 0;
TU::setCbfAtDepth( tu, compID, tu.depth, uiAbsSum > 0 );
return;
}
RDPCMMode rdpcmMode = RDPCM_OFF;
rdpcmNxN(tu, compID, cQP, uiAbsSum, rdpcmMode);
if (rdpcmMode == RDPCM_OFF)
{
uiAbsSum = 0;
// transform and quantize
if (CU::isLosslessCoded(*tu.cu))
{
const bool rotateResidual = TU::isNonTransformedResidualRotated( tu, compID );
for( uint32_t y = 0; y < uiHeight; y++ )
{
for( uint32_t x = 0; x < uiWidth; x++ )
{
const Pel currentSample = resiBuf.at( x, y );
if( rotateResidual )
{
rpcCoeff.at( uiWidth - x - 1, uiHeight - y - 1 ) = currentSample;
}
else
{
rpcCoeff.at( x, y ) = currentSample;
}
uiAbsSum += TCoeff( abs( currentSample ) );
}
}
}
else
{
#if MAX_TB_SIZE_SIGNALLING
CHECK( cs.sps->getMaxTbSize() < uiWidth, "Unsupported transformation size" );
#else
CHECK( MAX_TB_SIZEY < uiWidth, "Unsupported transformation size" );
#endif
CoeffBuf tempCoeff( loadTr ? m_mtsCoeffs[tu.mtsIdx] : m_plTempCoeff, rect );
DTRACE_PEL_BUF( D_RESIDUALS, resiBuf, tu, tu.cu->predMode, compID );
if( !loadTr )
{
if( isLuma(compID) && tu.mtsIdx == 1 )
{
xTransformSkip( tu, compID, resiBuf, tempCoeff.buf );
}
else
{
xT( tu, compID, resiBuf, tempCoeff, uiWidth, uiHeight );
}
}
//we do this only with the DCT-II coefficients
if( isLuma(compID) &&
!loadTr && tu.mtsIdx == 0
)
{
//it gets the distribution of the coefficients energy, which will be useful to discard ISP tests
xGetCoeffEnergy( tu, compID, tempCoeff, diagRatio, horVerRatio );
}
DTRACE_COEFF_BUF( D_TCOEFF, tempCoeff, tu, tu.cu->predMode, compID );
xQuant( tu, compID, tempCoeff, uiAbsSum, cQP, ctx );
DTRACE_COEFF_BUF( D_TCOEFF, tu.getCoeffs( compID ), tu, tu.cu->predMode, compID );
}
}
// set coded block flag (CBF)
TU::setCbfAtDepth (tu, compID, tu.depth, uiAbsSum > 0);
}
void TrQuant::xGetCoeffEnergy( TransformUnit &tu, const ComponentID &compID, const CoeffBuf& coeffs, double* diagRatio, double* horVerRatio )
{
if( nullptr == diagRatio || nullptr == horVerRatio ) return;
if( tu.cu->predMode == MODE_INTRA && !tu.cu->ispMode && isLuma( compID ) && CU::canUseISPSplit( *tu.cu, compID ) != NOT_INTRA_SUBPARTITIONS )
{
const int width = tu.cu->blocks[compID].width;
const int height = tu.cu->blocks[compID].height;
const int log2Sl = width <= height ? g_aucLog2[height >> g_aucLog2[width]] : g_aucLog2[width >> g_aucLog2[height]];
const int diPos1 = width <= height ? width : height;
const int diPos2 = width <= height ? height : width;
const int ofsPos1 = width <= height ? 1 : coeffs.stride;
const int ofsPos2 = width <= height ? coeffs.stride : 1;
int wdtE = 0, hgtE = 0, diaE = 0;
int* gtE = width <= height ? &wdtE : &hgtE;
int* stE = width <= height ? &hgtE : &wdtE;
for( int pos1 = 0; pos1 < diPos1; pos1++ )
{
const int posN = pos1 << log2Sl;
for( int pos2 = 0; pos2 < diPos2; pos2++ )
{
const int blkP = pos1 * ofsPos1 + pos2 * ofsPos2;
if( posN > pos2 ) *gtE += abs( coeffs.buf[ blkP ] );
if( posN < pos2 ) *stE += abs( coeffs.buf[ blkP ] );
if( posN == pos2 ) diaE += abs( coeffs.buf[ blkP ] );
}
}
*horVerRatio = 0 == wdtE && 0 == hgtE ? 1 : double( wdtE ) / double( hgtE );
*diagRatio = 0 == wdtE && 0 == hgtE && 0 == diaE ? 1 : double( diaE ) / double( wdtE + hgtE );
}
}
void TrQuant::applyForwardRDPCM(TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, const RDPCMMode &mode)
{
const bool bLossless = tu.cu->transQuantBypass;
const uint32_t uiWidth = tu.blocks[compID].width;
const uint32_t uiHeight = tu.blocks[compID].height;
const bool rotateResidual = TU::isNonTransformedResidualRotated(tu, compID);
const uint32_t uiSizeMinus1 = (uiWidth * uiHeight) - 1;
const CPelBuf pcResidual = tu.cs->getResiBuf(tu.blocks[compID]);
const CoeffBuf pcCoeff = tu.getCoeffs(compID);
uint32_t uiX = 0;
uint32_t uiY = 0;
uint32_t &majorAxis = (mode == RDPCM_VER) ? uiX : uiY;
uint32_t &minorAxis = (mode == RDPCM_VER) ? uiY : uiX;
const uint32_t majorAxisLimit = (mode == RDPCM_VER) ? uiWidth : uiHeight;
const uint32_t minorAxisLimit = (mode == RDPCM_VER) ? uiHeight : uiWidth;
const bool bUseHalfRoundingPoint = (mode != RDPCM_OFF);
uiAbsSum = 0;
for (majorAxis = 0; majorAxis < majorAxisLimit; majorAxis++)
{
TCoeff accumulatorValue = 0; // 32-bit accumulator
for (minorAxis = 0; minorAxis < minorAxisLimit; minorAxis++)
{
const uint32_t sampleIndex = (uiY * uiWidth) + uiX;
const uint32_t coefficientIndex = (rotateResidual ? (uiSizeMinus1-sampleIndex) : sampleIndex);
const Pel currentSample = pcResidual.at(uiX, uiY);
const TCoeff encoderSideDelta = TCoeff(currentSample) - accumulatorValue;
Pel reconstructedDelta;
if (bLossless)
{
pcCoeff.buf[coefficientIndex] = encoderSideDelta;
reconstructedDelta = (Pel) encoderSideDelta;
}
else
{
m_quant->transformSkipQuantOneSample(tu, compID, encoderSideDelta, pcCoeff.buf[coefficientIndex], coefficientIndex, cQP, bUseHalfRoundingPoint);
m_quant->invTrSkipDeQuantOneSample (tu, compID, pcCoeff.buf[coefficientIndex], reconstructedDelta, coefficientIndex, cQP);
}
uiAbsSum += abs(pcCoeff.buf[coefficientIndex]);
if (mode != RDPCM_OFF)
{
accumulatorValue += reconstructedDelta;
}
}
}
}
void TrQuant::rdpcmNxN(TransformUnit &tu, const ComponentID &compID, const QpParam &cQP, TCoeff &uiAbsSum, RDPCMMode &rdpcmMode)
{
if (!CU::isRDPCMEnabled(*tu.cu) || (tu.mtsIdx!=1 && !tu.cu->transQuantBypass))
{
rdpcmMode = RDPCM_OFF;
}
else if (CU::isIntra(*tu.cu))
{
const ChannelType chType = toChannelType(compID);
const uint32_t uiChFinalMode = PU::getFinalIntraMode(*tu.cs->getPU(tu.blocks[compID].pos(), chType), chType);
if (uiChFinalMode == VER_IDX || uiChFinalMode == HOR_IDX)
{
rdpcmMode = (uiChFinalMode == VER_IDX) ? RDPCM_VER : RDPCM_HOR;
applyForwardRDPCM(tu, compID, cQP, uiAbsSum, rdpcmMode);
}
else
{
rdpcmMode = RDPCM_OFF;
}
}
else // not intra, need to select the best mode
{
const CompArea &area = tu.blocks[compID];
const uint32_t uiWidth = area.width;
const uint32_t uiHeight = area.height;
RDPCMMode bestMode = NUMBER_OF_RDPCM_MODES;
TCoeff bestAbsSum = std::numeric_limits<TCoeff>::max();
TCoeff bestCoefficients[MAX_TB_SIZEY * MAX_TB_SIZEY];
for (uint32_t modeIndex = 0; modeIndex < NUMBER_OF_RDPCM_MODES; modeIndex++)
{
const RDPCMMode mode = RDPCMMode(modeIndex);
TCoeff currAbsSum = 0;
applyForwardRDPCM(tu, compID, cQP, uiAbsSum, rdpcmMode);
if (currAbsSum < bestAbsSum)
{
bestMode = mode;
bestAbsSum = currAbsSum;
if (mode != RDPCM_OFF)
{
CoeffBuf(bestCoefficients, uiWidth, uiHeight).copyFrom(tu.getCoeffs(compID));
}
}
}
rdpcmMode = bestMode;
uiAbsSum = bestAbsSum;
if (rdpcmMode != RDPCM_OFF) //the TU is re-transformed and quantized if DPCM_OFF is returned, so there is no need to preserve it here
{
tu.getCoeffs(compID).copyFrom(CoeffBuf(bestCoefficients, uiWidth, uiHeight));
}
}
tu.rdpcm[compID] = rdpcmMode;
}
void TrQuant::xTransformSkip(const TransformUnit &tu, const ComponentID &compID, const CPelBuf &resi, TCoeff* psCoeff)
{
const SPS &sps = *tu.cs->sps;
const CompArea &rect = tu.blocks[compID];
const uint32_t width = rect.width;
const uint32_t height = rect.height;
const ChannelType chType = toChannelType(compID);
const int channelBitDepth = sps.getBitDepth(chType);
const int maxLog2TrDynamicRange = sps.getMaxLog2TrDynamicRange(chType);
int iTransformShift = getTransformShift(channelBitDepth, rect.size(), maxLog2TrDynamicRange);
if( sps.getSpsRangeExtension().getExtendedPrecisionProcessingFlag() )
{
iTransformShift = std::max<int>( 0, iTransformShift );
}
int iWHScale = 1;
const bool rotateResidual = TU::isNonTransformedResidualRotated( tu, compID );
const uint32_t uiSizeMinus1 = ( width * height ) - 1;
if( iTransformShift >= 0 )
{
for( uint32_t y = 0, coefficientIndex = 0; y < height; y++ )
{
for( uint32_t x = 0; x < width; x++, coefficientIndex++ )
{
psCoeff[rotateResidual ? uiSizeMinus1 - coefficientIndex : coefficientIndex] = ( TCoeff( resi.at( x, y ) ) * iWHScale ) << iTransformShift;
}
}
}
else //for very high bit depths
{
iTransformShift = -iTransformShift;
const TCoeff offset = 1 << ( iTransformShift - 1 );
for( uint32_t y = 0, coefficientIndex = 0; y < height; y++ )
{
for( uint32_t x = 0; x < width; x++, coefficientIndex++ )
{
psCoeff[rotateResidual ? uiSizeMinus1 - coefficientIndex : coefficientIndex] = ( TCoeff( resi.at( x, y ) ) * iWHScale + offset ) >> iTransformShift;
}
}
}
}
//! \}