[x265] split rate calculation functions to luma and chroma to simplify luma path
Steve Borho
steve at borho.org
Mon Jun 30 19:57:02 CEST 2014
On Sun, Jun 29, 2014 at 9:52 PM, Satoshi Nakagawa <nakagawa424 at oki.com> wrote:
> # HG changeset patch
> # User Satoshi Nakagawa <nakagawa424 at oki.com>
> # Date 1404096538 -32400
> # Mon Jun 30 11:48:58 2014 +0900
> # Node ID 4c30d66afc78ed99385c04645d7b1303d80dea2c
> # Parent 32aa6cc3cf4d108ac92f5d29258b2c38ca888d29
> split rate calculation functions to luma and chroma to simplify luma path
I'm afraid this patch doesn't apply on top of the current tip + your
previous "fix emms" patch
Can you rebase this?
> diff -r 32aa6cc3cf4d -r 4c30d66afc78 source/Lib/TLibEncoder/TEncEntropy.cpp
> --- a/source/Lib/TLibEncoder/TEncEntropy.cpp Thu Jun 26 17:19:08 2014 -0700
> +++ b/source/Lib/TLibEncoder/TEncEntropy.cpp Mon Jun 30 11:48:58 2014 +0900
> @@ -325,7 +325,7 @@
> }
> else
> {
> - m_entropyCoder->codeQtCbf(cu, absPartIdx, TEXT_LUMA, cu->getTransformIdx(absPartIdx), absPartIdxStep, tuSize, tuSize, (subdiv == 0));
> + m_entropyCoder->codeQtCbf(cu, absPartIdx, TEXT_LUMA, cu->getTransformIdx(absPartIdx));
> }
>
> if (cbfY || cbfU || cbfV)
> @@ -342,7 +342,7 @@
> }
> if (cbfY)
> {
> - m_entropyCoder->codeCoeffNxN(cu, (cu->getCoeffY() + offsetLuma), absPartIdx, tuSize, TEXT_LUMA);
> + m_entropyCoder->codeCoeffNxN(cu, (cu->getCoeffY() + offsetLuma), absPartIdx, log2TrSize, TEXT_LUMA);
> }
>
> int chFmt = cu->getChromaFormat();
> @@ -351,7 +351,7 @@
> uint32_t partNum = cu->getPic()->getNumPartInCU() >> ((depth - 1) << 1);
> if ((absPartIdx & (partNum - 1)) == (partNum - 1))
> {
> - uint32_t trSizeC = 1 << log2TrSize;
> + const uint32_t log2TrSizeC = 2;
> const bool splitIntoSubTUs = (chFmt == CHROMA_422);
>
> uint32_t curPartNum = cu->getPic()->getNumPartInCU() >> ((depth - 1) << 1);
> @@ -364,10 +364,10 @@
> do
> {
> uint32_t cbf = cu->getCbf(tuIterator.m_absPartIdxTURelCU, (TextType)chromaId, trIdx + splitIntoSubTUs);
> - uint32_t subTUIndex = tuIterator.m_section * trSizeC * trSizeC;
> if (cbf)
> {
> - m_entropyCoder->codeCoeffNxN(cu, (coeffChroma + m_bakChromaOffset + subTUIndex), tuIterator.m_absPartIdxTURelCU, trSizeC, (TextType)chromaId);
> + uint32_t subTUOffset = tuIterator.m_section << (log2TrSizeC * 2);
> + m_entropyCoder->codeCoeffNxN(cu, (coeffChroma + m_bakChromaOffset + subTUOffset), tuIterator.m_absPartIdxTURelCU, log2TrSizeC, (TextType)chromaId);
> }
> }
> while (isNextTUSection(&tuIterator));
> @@ -376,7 +376,7 @@
> }
> else
> {
> - uint32_t trSizeC = tuSize >> hChromaShift;
> + uint32_t log2TrSizeC = log2TrSize - hChromaShift;
> const bool splitIntoSubTUs = (chFmt == CHROMA_422);
> uint32_t curPartNum = cu->getPic()->getNumPartInCU() >> (depth << 1);
> for (uint32_t chromaId = TEXT_CHROMA_U; chromaId <= TEXT_CHROMA_V; chromaId++)
> @@ -387,10 +387,10 @@
> do
> {
> uint32_t cbf = cu->getCbf(tuIterator.m_absPartIdxTURelCU, (TextType)chromaId, trIdx + splitIntoSubTUs);
> - uint32_t subTUIndex = tuIterator.m_section * trSizeC * trSizeC;
> if (cbf)
> {
> - m_entropyCoder->codeCoeffNxN(cu, (coeffChroma + offsetChroma + subTUIndex), tuIterator.m_absPartIdxTURelCU, trSizeC, (TextType)chromaId);
> + uint32_t subTUOffset = tuIterator.m_section << (log2TrSizeC * 2);
> + m_entropyCoder->codeCoeffNxN(cu, (coeffChroma + offsetChroma + subTUOffset), tuIterator.m_absPartIdxTURelCU, log2TrSizeC, (TextType)chromaId);
> }
> }
> while (isNextTUSection(&tuIterator));
> @@ -540,11 +540,6 @@
> m_entropyCoder->codeQtRootCbf(cu, absPartIdx);
> }
>
> -void TEncEntropy::encodeQtCbfZero(TComDataCU* cu, TextType ttype, uint32_t trDepth)
> -{
> - m_entropyCoder->codeQtCbfZero(cu, ttype, trDepth);
> -}
> -
> void TEncEntropy::encodeQtRootCbfZero(TComDataCU* cu)
> {
> m_entropyCoder->codeQtRootCbfZero(cu);
> @@ -593,11 +588,6 @@
> xEncodeTransform(cu, lumaOffset, chromaOffset, absPartIdx, absPartIdxStep, depth, cuSize, 0, bCodeDQP);
> }
>
> -void TEncEntropy::encodeCoeffNxN(TComDataCU* cu, coeff_t* coeff, uint32_t absPartIdx, uint32_t trSize, TextType ttype)
> -{
> - m_entropyCoder->codeCoeffNxN(cu, coeff, absPartIdx, trSize, ttype);
> -}
> -
> void TEncEntropy::estimateBit(estBitsSbacStruct* estBitsSBac, int trSize, TextType ttype)
> {
> ttype = ttype == TEXT_LUMA ? TEXT_LUMA : TEXT_CHROMA;
> diff -r 32aa6cc3cf4d -r 4c30d66afc78 source/Lib/TLibEncoder/TEncEntropy.h
> --- a/source/Lib/TLibEncoder/TEncEntropy.h Thu Jun 26 17:19:08 2014 -0700
> +++ b/source/Lib/TLibEncoder/TEncEntropy.h Mon Jun 30 11:48:58 2014 +0900
> @@ -114,14 +114,24 @@
>
> void encodeTransformSubdivFlag(uint32_t symbol, uint32_t ctx);
> void encodeQtCbf(TComDataCU* cu, uint32_t absPartIdx, uint32_t absPartIdxStep, uint32_t width, uint32_t height, TextType ttype, uint32_t trDepth, bool lowestLevel);
> - void encodeQtCbfZero(TComDataCU* cu, TextType ttype, uint32_t trDepth);
> + void encodeQtCbf(TComDataCU* cu, uint32_t absPartIdx, TextType ttype, uint32_t trDepth)
> + {
> + m_entropyCoder->codeQtCbf(cu, absPartIdx, ttype, trDepth);
> + }
> + void encodeQtCbfZero(TComDataCU* cu, TextType ttype, uint32_t trDepth)
> + {
> + m_entropyCoder->codeQtCbfZero(cu, ttype, trDepth);
> + }
> +
> void encodeQtRootCbfZero(TComDataCU* cu);
> void encodeQtRootCbf(TComDataCU* cu, uint32_t absPartIdx);
> void encodeQP(TComDataCU* cu, uint32_t absPartIdx);
> void encodeScalingList(TComScalingList* scalingList);
> void encodeCoeff(TComDataCU* cu, uint32_t absPartIdx, uint32_t depth, uint32_t cuSize, bool& bCodeDQP);
> - void encodeCoeffNxN(TComDataCU* cu, coeff_t* pcCoeff, uint32_t absPartIdx, uint32_t trSize, TextType ttype);
> -
> + void encodeCoeffNxN(TComDataCU* cu, coeff_t* coeff, uint32_t absPartIdx, uint32_t log2TrSize, TextType ttype)
> + {
> + m_entropyCoder->codeCoeffNxN(cu, coeff, absPartIdx, log2TrSize, ttype);
> + }
>
> void estimateBit(estBitsSbacStruct* estBitsSbac, int trSize, TextType ttype);
> void encodeSaoOffset(SaoLcuParam* saoLcuParam, uint32_t compIdx);
> diff -r 32aa6cc3cf4d -r 4c30d66afc78 source/Lib/TLibEncoder/TEncSbac.cpp
> --- a/source/Lib/TLibEncoder/TEncSbac.cpp Thu Jun 26 17:19:08 2014 -0700
> +++ b/source/Lib/TLibEncoder/TEncSbac.cpp Mon Jun 30 11:48:58 2014 +0900
> @@ -1873,6 +1873,25 @@
> }
> }
>
> +void TEncSbac::codeQtCbf(TComDataCU* cu, uint32_t absPartIdx, TextType ttype, uint32_t trDepth)
> +{
> + uint32_t ctx = cu->getCtxQtCbf(ttype, trDepth);
> + uint32_t cbf = cu->getCbf(absPartIdx, ttype, trDepth);
> + m_cabac->encodeBin(cbf, m_contextModels[OFF_QT_CBF_CTX + ctx]);
> +
> + DTRACE_CABAC_VL(g_nSymbolCounter++)
> + DTRACE_CABAC_T("\tparseQtCbf()")
> + DTRACE_CABAC_T("\tsymbol=")
> + DTRACE_CABAC_V(cbf)
> + DTRACE_CABAC_T("\tctx=")
> + DTRACE_CABAC_V(ctx)
> + DTRACE_CABAC_T("\tetype=")
> + DTRACE_CABAC_V(ttype)
> + DTRACE_CABAC_T("\tuiAbsPartIdx=")
> + DTRACE_CABAC_V(absPartIdx)
> + DTRACE_CABAC_T("\n")
> +}
> +
> void TEncSbac::codeTransformSkipFlags(TComDataCU* cu, uint32_t absPartIdx, uint32_t trSize, TextType ttype)
> {
> if (cu->getCUTransquantBypass(absPartIdx))
> @@ -1999,8 +2018,9 @@
> }
> }
>
> -void TEncSbac::codeCoeffNxN(TComDataCU* cu, coeff_t* coeff, uint32_t absPartIdx, uint32_t trSize, TextType ttype)
> +void TEncSbac::codeCoeffNxN(TComDataCU* cu, coeff_t* coeff, uint32_t absPartIdx, uint32_t log2TrSize, TextType ttype)
> {
> + uint32_t trSize = 1 << log2TrSize;
> #if ENC_DEC_TRACE
> DTRACE_CABAC_VL(g_nSymbolCounter++)
> DTRACE_CABAC_T("\tparseCoeffNxN()\teType=")
> @@ -2028,8 +2048,6 @@
>
> X265_CHECK(trSize <= m_slice->getSPS()->getMaxTrSize(), "transform size out of range\n");
>
> - const uint32_t log2TrSize = g_convertToBit[trSize] + 2;
> -
> // compute number of significant coefficients
> uint32_t numSig = primitives.count_nonzero(coeff, (1 << (log2TrSize << 1)));
>
> diff -r 32aa6cc3cf4d -r 4c30d66afc78 source/Lib/TLibEncoder/TEncSbac.h
> --- a/source/Lib/TLibEncoder/TEncSbac.h Thu Jun 26 17:19:08 2014 -0700
> +++ b/source/Lib/TLibEncoder/TEncSbac.h Mon Jun 30 11:48:58 2014 +0900
> @@ -114,6 +114,7 @@
> void codePredMode(TComDataCU* cu, uint32_t absPartIdx);
> void codeTransformSubdivFlag(uint32_t symbol, uint32_t ctx);
> void codeQtCbf(TComDataCU* cu, uint32_t absPartIdx, TextType ttype, uint32_t trDepth, uint32_t absPartIdxStep, uint32_t width, uint32_t height, bool lowestLevel);
> + void codeQtCbf(TComDataCU* cu, uint32_t absPartIdx, TextType ttype, uint32_t trDepth);
> void codeQtRootCbf(TComDataCU* cu, uint32_t absPartIdx);
> void codeQtCbfZero(TComDataCU* cu, TextType ttype, uint32_t trDepth);
> void codeQtRootCbfZero(TComDataCU* cu);
> @@ -127,7 +128,7 @@
> void codeDeltaQP(TComDataCU* cu, uint32_t absPartIdx);
>
> void codeLastSignificantXY(uint32_t posx, uint32_t posy, uint32_t log2TrSize, TextType ttype, uint32_t scanIdx);
> - void codeCoeffNxN(TComDataCU* cu, coeff_t* coef, uint32_t absPartIdx, uint32_t trSize, TextType ttype);
> + void codeCoeffNxN(TComDataCU* cu, coeff_t* coef, uint32_t absPartIdx, uint32_t log2TrSize, TextType ttype);
> void codeTransformSkipFlags(TComDataCU* cu, uint32_t absPartIdx, uint32_t trSize, TextType ttype);
>
> // -------------------------------------------------------------------------------------------------------------------
> diff -r 32aa6cc3cf4d -r 4c30d66afc78 source/Lib/TLibEncoder/TEncSearch.cpp
> --- a/source/Lib/TLibEncoder/TEncSearch.cpp Thu Jun 26 17:19:08 2014 -0700
> +++ b/source/Lib/TLibEncoder/TEncSearch.cpp Mon Jun 30 11:48:58 2014 +0900
> @@ -151,7 +151,7 @@
> m_rdCost->setCrDistortionWeight(lambdaOffset);
> }
>
> -void TEncSearch::xEncSubdivCbfQT(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t absPartIdxStep, uint32_t width, uint32_t height, bool bLuma, bool bChroma)
> +void TEncSearch::xEncSubdivCbfQTLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx)
> {
> uint32_t fullDepth = cu->getDepth(0) + trDepth;
> uint32_t trMode = cu->getTransformIdx(absPartIdx);
> @@ -177,77 +177,116 @@
> else
> {
> X265_CHECK(log2TrSize > cu->getQuadtreeTULog2MinSizeInCU(absPartIdx), "transform size too small\n");
> - if (bLuma)
> - {
> - m_entropyCoder->encodeTransformSubdivFlag(subdiv, 5 - log2TrSize);
> - }
> + m_entropyCoder->encodeTransformSubdivFlag(subdiv, 5 - log2TrSize);
> }
>
> - if (bChroma)
> - {
> - int chFmt = cu->getChromaFormat();
> - if ((log2TrSize > 2) && !(chFmt == CHROMA_444))
> - {
> - if (trDepth == 0 || cu->getCbf(absPartIdx, TEXT_CHROMA_U, trDepth - 1))
> - m_entropyCoder->encodeQtCbf(cu, absPartIdx, absPartIdxStep, (width >> m_hChromaShift), (height >> m_vChromaShift), TEXT_CHROMA_U, trDepth, (subdiv == 0));
> -
> - if (trDepth == 0 || cu->getCbf(absPartIdx, TEXT_CHROMA_V, trDepth - 1))
> - m_entropyCoder->encodeQtCbf(cu, absPartIdx, absPartIdxStep, (width >> m_hChromaShift), (height >> m_vChromaShift), TEXT_CHROMA_V, trDepth, (subdiv == 0));
> - }
> - }
> -
> - if (subdiv)
> - {
> - TComTURecurse tuIterator;
> - initSection(&tuIterator, QUAD_SPLIT, absPartIdxStep);
> - width >>= 1;
> - height >>= 1;
> -
> - uint32_t qtPartNum = cu->getPic()->getNumPartInCU() >> ((fullDepth + 1) << 1);
> - for (uint32_t part = 0; part < 4; part++)
> - {
> - xEncSubdivCbfQT(cu, trDepth + 1, absPartIdx + part * qtPartNum, tuIterator.m_absPartIdxStep, width, height, bLuma, bChroma);
> - }
> -
> - return;
> - }
> -
> - //===== Cbfs =====
> - if (bLuma)
> - {
> - m_entropyCoder->encodeQtCbf(cu, absPartIdx, absPartIdxStep, width, height, TEXT_LUMA, trMode, (subdiv == 0));
> - }
> -}
> -
> -void TEncSearch::xEncCoeffQT(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, TextType ttype)
> -{
> - if (!cu->getCbf(absPartIdx, ttype, trDepth))
> - return;
> -
> - uint32_t fullDepth = cu->getDepth(0) + trDepth;
> - uint32_t trMode = cu->getTransformIdx(absPartIdx);
> - uint32_t subdiv = (trMode > trDepth ? 1 : 0);
> -
> if (subdiv)
> {
> uint32_t qtPartNum = cu->getPic()->getNumPartInCU() >> ((fullDepth + 1) << 1);
> for (uint32_t part = 0; part < 4; part++)
> {
> - xEncCoeffQT(cu, trDepth + 1, absPartIdx + part * qtPartNum, ttype);
> + xEncSubdivCbfQTLuma(cu, trDepth + 1, absPartIdx + part * qtPartNum);
> }
>
> return;
> }
>
> - uint32_t origTrDepth = trDepth;
> -
> + //===== Cbfs =====
> + m_entropyCoder->encodeQtCbf(cu, absPartIdx, TEXT_LUMA, trMode);
> +}
> +
> +void TEncSearch::xEncSubdivCbfQTChroma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t absPartIdxStep, uint32_t width, uint32_t height)
> +{
> + uint32_t fullDepth = cu->getDepth(0) + trDepth;
> + uint32_t trMode = cu->getTransformIdx(absPartIdx);
> + uint32_t subdiv = (trMode > trDepth ? 1 : 0);
> uint32_t log2TrSize = cu->getSlice()->getSPS()->getLog2MaxCodingBlockSize() - fullDepth;
> - int chFmt = cu->getChromaFormat();
> - if ((ttype != TEXT_LUMA) && (log2TrSize == 2) && !(chFmt == CHROMA_444))
> +
> + int chFmt = cu->getChromaFormat();
> + if ((log2TrSize > 2) && !(chFmt == CHROMA_444))
> + {
> + if (trDepth == 0 || cu->getCbf(absPartIdx, TEXT_CHROMA_U, trDepth - 1))
> + m_entropyCoder->encodeQtCbf(cu, absPartIdx, absPartIdxStep, (width >> m_hChromaShift), (height >> m_vChromaShift), TEXT_CHROMA_U, trDepth, (subdiv == 0));
> +
> + if (trDepth == 0 || cu->getCbf(absPartIdx, TEXT_CHROMA_V, trDepth - 1))
> + m_entropyCoder->encodeQtCbf(cu, absPartIdx, absPartIdxStep, (width >> m_hChromaShift), (height >> m_vChromaShift), TEXT_CHROMA_V, trDepth, (subdiv == 0));
> + }
> +
> + if (subdiv)
> + {
> + absPartIdxStep >>= 2;
> + width >>= 1;
> + height >>= 1;
> +
> + uint32_t qtPartNum = cu->getPic()->getNumPartInCU() >> ((fullDepth + 1) << 1);
> + for (uint32_t part = 0; part < 4; part++)
> + {
> + xEncSubdivCbfQTChroma(cu, trDepth + 1, absPartIdx + part * qtPartNum, absPartIdxStep, width, height);
> + }
> + }
> +}
> +
> +void TEncSearch::xEncCoeffQTLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx)
> +{
> + const TextType ttype = TEXT_LUMA;
> +
> + if (!cu->getCbf(absPartIdx, ttype, trDepth))
> + return;
> +
> + uint32_t fullDepth = cu->getDepth(0) + trDepth;
> + uint32_t trMode = cu->getTransformIdx(absPartIdx);
> +
> + if (trMode > trDepth)
> + {
> + uint32_t qtPartNum = cu->getPic()->getNumPartInCU() >> ((fullDepth + 1) << 1);
> + for (uint32_t part = 0; part < 4; part++)
> + {
> + xEncCoeffQTLuma(cu, trDepth + 1, absPartIdx + part * qtPartNum);
> + }
> +
> + return;
> + }
> +
> + uint32_t log2TrSize = cu->getSlice()->getSPS()->getLog2MaxCodingBlockSize() - fullDepth;
> + uint32_t qtLayer = cu->getSlice()->getSPS()->getQuadtreeTULog2MaxSize() - log2TrSize;
> + uint32_t log2UnitSize = cu->getPic()->getLog2UnitSize();
> + uint32_t coeffOffset = absPartIdx << (log2UnitSize * 2);
> + coeff_t* coeff = m_qtTempCoeff[ttype][qtLayer] + coeffOffset;
> + m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, log2TrSize, ttype);
> +}
> +
> +void TEncSearch::xEncCoeffQTChroma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, TextType ttype)
> +{
> + if (!cu->getCbf(absPartIdx, ttype, trDepth))
> + return;
> +
> + uint32_t fullDepth = cu->getDepth(0) + trDepth;
> + uint32_t trMode = cu->getTransformIdx(absPartIdx);
> +
> + if (trMode > trDepth)
> + {
> + uint32_t qtPartNum = cu->getPic()->getNumPartInCU() >> ((fullDepth + 1) << 1);
> + for (uint32_t part = 0; part < 4; part++)
> + {
> + xEncCoeffQTChroma(cu, trDepth + 1, absPartIdx + part * qtPartNum, ttype);
> + }
> +
> + return;
> + }
> +
> + uint32_t log2TrSize = cu->getSlice()->getSPS()->getLog2MaxCodingBlockSize() - fullDepth;
> + uint32_t qtLayer = cu->getSlice()->getSPS()->getQuadtreeTULog2MaxSize() - log2TrSize;
> + uint32_t log2UnitSize = cu->getPic()->getLog2UnitSize();
> +
> + uint32_t trDepthC = trDepth;
> + uint32_t log2TrSizeC = log2TrSize - m_hChromaShift;
> + int chFmt = cu->getChromaFormat();
> + if ((log2TrSize == 2) && !(chFmt == CHROMA_444))
> {
> X265_CHECK(trDepth > 0, "transform size too small\n");
> - trDepth--;
> - uint32_t qpdiv = cu->getPic()->getNumPartInCU() >> ((cu->getDepth(0) + trDepth) << 1);
> + trDepthC--;
> + log2TrSizeC++;
> + uint32_t qpdiv = cu->getPic()->getNumPartInCU() >> ((cu->getDepth(0) + trDepthC) << 1);
> bool bFirstQ = ((absPartIdx & (qpdiv - 1)) == 0);
> if (!bFirstQ)
> {
> @@ -255,132 +294,122 @@
> }
> }
>
> - //===== coefficients =====
> - uint32_t chroma = (ttype != TEXT_LUMA ? 1 : 0);
> - int cspx = chroma ? m_hChromaShift : 0;
> - int cspy = chroma ? m_vChromaShift : 0;
> - uint32_t width = cu->getCUSize(0) >> (trDepth + cspx);
> - uint32_t height = cu->getCUSize(0) >> (trDepth + cspy);
> - uint32_t coeffOffset = absPartIdx << (cu->getPic()->getLog2UnitSize() * 2 - (cspx + cspy));
> - uint32_t qtLayer = cu->getSlice()->getSPS()->getQuadtreeTULog2MaxSize() - log2TrSize;
> - coeff_t* coeff = m_qtTempCoeff[ttype][qtLayer] + coeffOffset;
> -
> - if (width == height)
> + if (chFmt != CHROMA_422)
> {
> - m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, width, ttype);
> + uint32_t shift = (chFmt == CHROMA_420) ? 2 : 0;
> + uint32_t coeffOffset = absPartIdx << (log2UnitSize * 2 - shift);
> + coeff_t* coeff = m_qtTempCoeff[ttype][qtLayer] + coeffOffset;
> + m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, log2TrSizeC, ttype);
> }
> else
> {
> - uint32_t subTUSize = width * width;
> - uint32_t partIdxesPerSubTU = cu->getPic()->getNumPartInCU() >> (((cu->getDepth(absPartIdx) + trDepth) << 1) + 1);
> -
> - if (cu->getCbf(absPartIdx, ttype, origTrDepth + 1))
> - m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, width, ttype);
> - if (cu->getCbf(absPartIdx + partIdxesPerSubTU, ttype, origTrDepth + 1))
> - m_entropyCoder->encodeCoeffNxN(cu, coeff + subTUSize, absPartIdx + partIdxesPerSubTU, width, ttype);
> + uint32_t coeffOffset = absPartIdx << (log2UnitSize * 2 - 1);
> + coeff_t* coeff = m_qtTempCoeff[ttype][qtLayer] + coeffOffset;
> + uint32_t subTUSize = 1 << (log2TrSizeC * 2);
> + uint32_t partIdxesPerSubTU = cu->getPic()->getNumPartInCU() >> (((cu->getDepth(absPartIdx) + trDepthC) << 1) + 1);
> + if (cu->getCbf(absPartIdx, ttype, trDepth + 1))
> + m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, log2TrSizeC, ttype);
> + if (cu->getCbf(absPartIdx + partIdxesPerSubTU, ttype, trDepth + 1))
> + m_entropyCoder->encodeCoeffNxN(cu, coeff + subTUSize, absPartIdx + partIdxesPerSubTU, log2TrSizeC, ttype);
> }
> }
>
> -void TEncSearch::xEncIntraHeader(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, bool bLuma, bool bChroma)
> +void TEncSearch::xEncIntraHeaderLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx)
> {
> - if (bLuma)
> + // CU header
> + if (absPartIdx == 0)
> {
> - // CU header
> + if (!cu->getSlice()->isIntra())
> + {
> + if (cu->getSlice()->getPPS()->getTransquantBypassEnableFlag())
> + {
> + m_entropyCoder->encodeCUTransquantBypassFlag(cu, 0);
> + }
> + m_entropyCoder->encodeSkipFlag(cu, 0);
> + m_entropyCoder->encodePredMode(cu, 0);
> + }
> +
> + m_entropyCoder->encodePartSize(cu, 0, cu->getDepth(0));
> + }
> + // luma prediction mode
> + if (cu->getPartitionSize(0) == SIZE_2Nx2N)
> + {
> if (absPartIdx == 0)
> {
> - if (!cu->getSlice()->isIntra())
> + m_entropyCoder->encodeIntraDirModeLuma(cu, 0);
> + }
> + }
> + else
> + {
> + uint32_t qtNumParts = cu->getTotalNumPart() >> 2;
> + if (trDepth == 0)
> + {
> + X265_CHECK(absPartIdx == 0, "unexpected absPartIdx %d\n", absPartIdx);
> + for (uint32_t part = 0; part < 4; part++)
> {
> - if (cu->getSlice()->getPPS()->getTransquantBypassEnableFlag())
> - {
> - m_entropyCoder->encodeCUTransquantBypassFlag(cu, 0);
> - }
> - m_entropyCoder->encodeSkipFlag(cu, 0);
> - m_entropyCoder->encodePredMode(cu, 0);
> - }
> -
> - m_entropyCoder->encodePartSize(cu, 0, cu->getDepth(0));
> - }
> - // luma prediction mode
> - if (cu->getPartitionSize(0) == SIZE_2Nx2N)
> - {
> - if (absPartIdx == 0)
> - {
> - m_entropyCoder->encodeIntraDirModeLuma(cu, 0);
> + m_entropyCoder->encodeIntraDirModeLuma(cu, part * qtNumParts);
> }
> }
> - else
> + else if ((absPartIdx & (qtNumParts - 1)) == 0)
> {
> - uint32_t qtNumParts = cu->getTotalNumPart() >> 2;
> - if (trDepth == 0)
> - {
> - X265_CHECK(absPartIdx == 0, "unexpected absPartIdx %d\n", absPartIdx);
> - for (uint32_t part = 0; part < 4; part++)
> - {
> - m_entropyCoder->encodeIntraDirModeLuma(cu, part * qtNumParts);
> - }
> - }
> - else if ((absPartIdx & (qtNumParts - 1)) == 0)
> - {
> - m_entropyCoder->encodeIntraDirModeLuma(cu, absPartIdx);
> - }
> - }
> - }
> - if (bChroma)
> - {
> - // chroma prediction mode
> - if ((cu->getPartitionSize(0) == SIZE_2Nx2N) || !(cu->getChromaFormat() == CHROMA_444))
> - {
> - if (absPartIdx == 0)
> - {
> - m_entropyCoder->encodeIntraDirModeChroma(cu, absPartIdx);
> - }
> - }
> - else
> - {
> - uint32_t qtNumParts = cu->getTotalNumPart() >> 2;
> - X265_CHECK(trDepth > 0, "unexpected trDepth %d\n", trDepth);
> - if ((absPartIdx & (qtNumParts - 1)) == 0)
> - m_entropyCoder->encodeIntraDirModeChroma(cu, absPartIdx);
> + m_entropyCoder->encodeIntraDirModeLuma(cu, absPartIdx);
> }
> }
> }
>
> -uint32_t TEncSearch::xGetIntraBitsQT(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t absPartIdxStep, bool bLuma, bool bChroma)
> +void TEncSearch::xEncIntraHeaderChroma(TComDataCU* cu, uint32_t absPartIdx)
> +{
> + // chroma prediction mode
> + if ((cu->getPartitionSize(0) == SIZE_2Nx2N) || !(cu->getChromaFormat() == CHROMA_444))
> + {
> + if (absPartIdx == 0)
> + {
> + m_entropyCoder->encodeIntraDirModeChroma(cu, absPartIdx);
> + }
> + }
> + else
> + {
> + uint32_t qtNumParts = cu->getTotalNumPart() >> 2;
> + if ((absPartIdx & (qtNumParts - 1)) == 0)
> + m_entropyCoder->encodeIntraDirModeChroma(cu, absPartIdx);
> + }
> +}
> +
> +uint32_t TEncSearch::xGetIntraBitsQTLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx)
> {
> m_entropyCoder->resetBits();
> - xEncIntraHeader(cu, trDepth, absPartIdx, bLuma, bChroma);
> - xEncSubdivCbfQT(cu, trDepth, absPartIdx, absPartIdxStep, cu->getCUSize(absPartIdx), cu->getCUSize(absPartIdx), bLuma, bChroma);
> -
> - if (bLuma)
> - {
> - xEncCoeffQT(cu, trDepth, absPartIdx, TEXT_LUMA);
> - }
> - if (bChroma)
> - {
> - xEncCoeffQT(cu, trDepth, absPartIdx, TEXT_CHROMA_U);
> - xEncCoeffQT(cu, trDepth, absPartIdx, TEXT_CHROMA_V);
> - }
> + xEncIntraHeaderLuma(cu, trDepth, absPartIdx);
> + xEncSubdivCbfQTLuma(cu, trDepth, absPartIdx);
> + xEncCoeffQTLuma(cu, trDepth, absPartIdx);
> return m_entropyCoder->getNumberOfWrittenBits();
> }
>
> -uint32_t TEncSearch::xGetIntraBitsQTLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t log2TrSize, coeff_t* coeff)
> +uint32_t TEncSearch::xGetIntraBitsQTChroma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t absPartIdxStep)
> {
> m_entropyCoder->resetBits();
> - xEncIntraHeader(cu, trDepth, absPartIdx, true, false);
> - xEncSubdivCbfQT(cu, trDepth, absPartIdx, 0, cu->getCUSize(absPartIdx), cu->getCUSize(absPartIdx), true, false);
> -
> - if (cu->getCbf(absPartIdx, TEXT_LUMA, trDepth))
> - {
> - m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, 1 << log2TrSize, TEXT_LUMA);
> - }
> -
> + xEncIntraHeaderChroma(cu, absPartIdx);
> + xEncSubdivCbfQTChroma(cu, trDepth, absPartIdx, absPartIdxStep, cu->getCUSize(absPartIdx), cu->getCUSize(absPartIdx));
> + xEncCoeffQTChroma(cu, trDepth, absPartIdx, TEXT_CHROMA_U);
> + xEncCoeffQTChroma(cu, trDepth, absPartIdx, TEXT_CHROMA_V);
> return m_entropyCoder->getNumberOfWrittenBits();
> }
>
> -uint32_t TEncSearch::xGetIntraBitsQTChroma(TComDataCU* cu, uint32_t absPartIdx, uint32_t log2TrSizeC, uint32_t chromaId, coeff_t* coeff)
> +uint32_t TEncSearch::xGetIntraBitsLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t log2TrSize, coeff_t* coeff)
> {
> m_entropyCoder->resetBits();
> - m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, 1 << log2TrSizeC, (TextType)chromaId);
> + xEncIntraHeaderLuma(cu, trDepth, absPartIdx);
> + xEncSubdivCbfQTLuma(cu, trDepth, absPartIdx);
> +
> + if (cu->getCbf(absPartIdx, TEXT_LUMA, trDepth))
> + m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, log2TrSize, TEXT_LUMA);
> +
> + return m_entropyCoder->getNumberOfWrittenBits();
> +}
> +
> +uint32_t TEncSearch::xGetIntraBitsChroma(TComDataCU* cu, uint32_t absPartIdx, uint32_t log2TrSizeC, uint32_t chromaId, coeff_t* coeff)
> +{
> + m_entropyCoder->resetBits();
> + m_entropyCoder->encodeCoeffNxN(cu, coeff, absPartIdx, log2TrSizeC, (TextType)chromaId);
> return m_entropyCoder->getNumberOfWrittenBits();
> }
>
> @@ -688,7 +717,7 @@
> }
> else
> {
> - uint32_t singleBits = xGetIntraBitsQTLuma(cu, trDepth, absPartIdx, log2TrSize, coeff);
> + uint32_t singleBits = xGetIntraBitsLuma(cu, trDepth, absPartIdx, log2TrSize, coeff);
> if (m_rdCost->psyRdEnabled())
> singleCostTmp = m_rdCost->calcPsyRdCost(singleDistYTmp, singleBits, singlePsyEnergyYTmp);
> else
> @@ -748,7 +777,7 @@
> }
> cu->setCbfSubParts(singleCbfY << trDepth, TEXT_LUMA, absPartIdx, fullDepth);
>
> - uint32_t singleBits = xGetIntraBitsQTLuma(cu, trDepth, absPartIdx, log2TrSize, coeffY);
> + uint32_t singleBits = xGetIntraBitsLuma(cu, trDepth, absPartIdx, log2TrSize, coeffY);
> if (m_param->rdPenalty && (log2TrSize == 5) && !isIntraSlice)
> singleBits *= 4;
>
> @@ -799,7 +828,7 @@
> m_rdGoOnSbacCoder->load(m_rdSbacCoders[fullDepth][CI_QT_TRAFO_ROOT]);
>
> //----- determine rate and r-d cost -----
> - uint32_t splitBits = xGetIntraBitsQT(cu, trDepth, absPartIdx, 0, true, false);
> + uint32_t splitBits = xGetIntraBitsQTLuma(cu, trDepth, absPartIdx);
> if (m_rdCost->psyRdEnabled())
> splitCost = m_rdCost->calcPsyRdCost(splitDistY, splitBits, splitPsyEnergyY);
> else
> @@ -1174,7 +1203,7 @@
> }
> else
> {
> - uint32_t bitsTmp = singleCbfCTmp ? xGetIntraBitsQTChroma(cu, absPartIdxC, log2TrSizeC, chromaId, coeff) : 0;
> + uint32_t bitsTmp = singleCbfCTmp ? xGetIntraBitsChroma(cu, absPartIdxC, log2TrSizeC, chromaId, coeff) : 0;
> if (m_rdCost->psyRdEnabled())
> {
> uint32_t zorder = cu->getZorderIdxInCU() + absPartIdxC;
> @@ -1889,7 +1918,7 @@
> m_rdGoOnSbacCoder->load(m_rdSbacCoders[depth][CI_CURR_BEST]);
> }
>
> - uint32_t bits = xGetIntraBitsQT(cu, initTrDepth, absPartIdxC, tuIterator.m_absPartIdxStep, false, true);
> + uint32_t bits = xGetIntraBitsQTChroma(cu, initTrDepth, absPartIdxC, tuIterator.m_absPartIdxStep);
> uint64_t cost = 0;
> if (m_rdCost->psyRdEnabled())
> cost = m_rdCost->calcPsyRdCost(dist, bits, cu->m_psyEnergy);
> @@ -2736,7 +2765,7 @@
> do
> {
> uint32_t absPartIdxC = tuIterator.m_absPartIdxTURelCU;
> - uint32_t subTUBufferOffset = trSizeC * trSizeC * tuIterator.m_section;
> + uint32_t subTUOffset = tuIterator.m_section << (log2TrSizeC * 2);
>
> int16_t *curResiU = resiYuv->getCbAddr(absPartIdxC);
> int16_t *curResiV = resiYuv->getCrAddr(absPartIdxC);
> @@ -2747,13 +2776,13 @@
> int curChromaQpOffset = cu->getSlice()->getPPS()->getChromaCbQpOffset() + cu->getSlice()->getSliceQpDeltaCb();
> m_trQuant->setQPforQuant(cu->getQP(0), TEXT_CHROMA, cu->getSlice()->getSPS()->getQpBDOffsetC(), curChromaQpOffset, chFmt);
> m_trQuant->selectLambda(TEXT_CHROMA_U);
> - absSumU = m_trQuant->transformNxN(cu, curResiU, strideResiC, coeffCurU + subTUBufferOffset,
> + absSumU = m_trQuant->transformNxN(cu, curResiU, strideResiC, coeffCurU + subTUOffset,
> trSizeC, TEXT_CHROMA_U, absPartIdxC, &lastPosU, false, curuseRDOQ);
>
> curChromaQpOffset = cu->getSlice()->getPPS()->getChromaCrQpOffset() + cu->getSlice()->getSliceQpDeltaCr();
> m_trQuant->setQPforQuant(cu->getQP(0), TEXT_CHROMA, cu->getSlice()->getSPS()->getQpBDOffsetC(), curChromaQpOffset, chFmt);
> m_trQuant->selectLambda(TEXT_CHROMA_V);
> - absSumV = m_trQuant->transformNxN(cu, curResiV, strideResiC, coeffCurV + subTUBufferOffset,
> + absSumV = m_trQuant->transformNxN(cu, curResiV, strideResiC, coeffCurV + subTUOffset,
> trSizeC, TEXT_CHROMA_V, absPartIdxC, &lastPosV, false, curuseRDOQ);
>
> cu->setCbfPartRange(absSumU ? setCbf : 0, TEXT_CHROMA_U, absPartIdxC, tuIterator.m_absPartIdxStep);
> @@ -2766,7 +2795,7 @@
>
> int scalingListType = 3 + TEXT_CHROMA_U;
> X265_CHECK(scalingListType < 6, "scalingListType too large %d\n", scalingListType);
> - m_trQuant->invtransformNxN(cu->getCUTransquantBypass(absPartIdxC), REG_DCT, curResiU, strideResiC, coeffCurU + subTUBufferOffset, trSizeC, scalingListType, false, lastPosU);
> + m_trQuant->invtransformNxN(cu->getCUTransquantBypass(absPartIdxC), REG_DCT, curResiU, strideResiC, coeffCurU + subTUOffset, trSizeC, scalingListType, false, lastPosU);
> }
> else
> {
> @@ -2779,7 +2808,7 @@
>
> int scalingListType = 3 + TEXT_CHROMA_V;
> X265_CHECK(scalingListType < 6, "scalingListType too large %d\n", scalingListType);
> - m_trQuant->invtransformNxN(cu->getCUTransquantBypass(absPartIdxC), REG_DCT, curResiV, strideResiC, coeffCurV + subTUBufferOffset, trSizeC, scalingListType, false, lastPosV);
> + m_trQuant->invtransformNxN(cu->getCUTransquantBypass(absPartIdxC), REG_DCT, curResiV, strideResiC, coeffCurV + subTUOffset, trSizeC, scalingListType, false, lastPosV);
> }
> else
> {
> @@ -2923,9 +2952,9 @@
> cu->setCbfSubParts(absSum[TEXT_LUMA][0] ? setCbf : 0, TEXT_LUMA, absPartIdx, depth);
>
> m_entropyCoder->resetBits();
> - m_entropyCoder->encodeQtCbf(cu, absPartIdx, 0, trSize, trSize, TEXT_LUMA, trMode, true);
> + m_entropyCoder->encodeQtCbf(cu, absPartIdx, TEXT_LUMA, trMode);
> if (absSum[TEXT_LUMA][0])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurY, absPartIdx, trSize, TEXT_LUMA);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurY, absPartIdx, log2TrSize, TEXT_LUMA);
> singleBitsComp[TEXT_LUMA][0] = m_entropyCoder->getNumberOfWrittenBits();
>
> uint32_t singleBitsPrev = singleBitsComp[TEXT_LUMA][0];
> @@ -2938,7 +2967,7 @@
> do
> {
> uint32_t absPartIdxC = tuIterator.m_absPartIdxTURelCU;
> - uint32_t subTUBufferOffset = trSizeC * trSizeC * tuIterator.m_section;
> + uint32_t subTUOffset = tuIterator.m_section << (log2TrSizeC * 2);
>
> cu->setTransformSkipPartRange(0, TEXT_CHROMA_U, absPartIdxC, tuIterator.m_absPartIdxStep);
> cu->setTransformSkipPartRange(0, TEXT_CHROMA_V, absPartIdxC, tuIterator.m_absPartIdxStep);
> @@ -2951,26 +2980,26 @@
> int curChromaQpOffset = cu->getSlice()->getPPS()->getChromaCbQpOffset() + cu->getSlice()->getSliceQpDeltaCb();
> m_trQuant->setQPforQuant(cu->getQP(0), TEXT_CHROMA, cu->getSlice()->getSPS()->getQpBDOffsetC(), curChromaQpOffset, chFmt);
> m_trQuant->selectLambda(TEXT_CHROMA_U);
> - absSum[TEXT_CHROMA_U][tuIterator.m_section] = m_trQuant->transformNxN(cu, resiYuv->getCbAddr(absPartIdxC), resiYuv->m_cwidth, coeffCurU + subTUBufferOffset,
> + absSum[TEXT_CHROMA_U][tuIterator.m_section] = m_trQuant->transformNxN(cu, resiYuv->getCbAddr(absPartIdxC), resiYuv->m_cwidth, coeffCurU + subTUOffset,
> trSizeC, TEXT_CHROMA_U, absPartIdxC, &lastPos[TEXT_CHROMA_U][tuIterator.m_section], false, curuseRDOQ);
> //Cr transform
> curChromaQpOffset = cu->getSlice()->getPPS()->getChromaCrQpOffset() + cu->getSlice()->getSliceQpDeltaCr();
> m_trQuant->setQPforQuant(cu->getQP(0), TEXT_CHROMA, cu->getSlice()->getSPS()->getQpBDOffsetC(), curChromaQpOffset, chFmt);
> m_trQuant->selectLambda(TEXT_CHROMA_V);
> - absSum[TEXT_CHROMA_V][tuIterator.m_section] = m_trQuant->transformNxN(cu, resiYuv->getCrAddr(absPartIdxC), resiYuv->m_cwidth, coeffCurV + subTUBufferOffset,
> + absSum[TEXT_CHROMA_V][tuIterator.m_section] = m_trQuant->transformNxN(cu, resiYuv->getCrAddr(absPartIdxC), resiYuv->m_cwidth, coeffCurV + subTUOffset,
> trSizeC, TEXT_CHROMA_V, absPartIdxC, &lastPos[TEXT_CHROMA_V][tuIterator.m_section], false, curuseRDOQ);
>
> cu->setCbfPartRange(absSum[TEXT_CHROMA_U][tuIterator.m_section] ? setCbf : 0, TEXT_CHROMA_U, absPartIdxC, tuIterator.m_absPartIdxStep);
> cu->setCbfPartRange(absSum[TEXT_CHROMA_V][tuIterator.m_section] ? setCbf : 0, TEXT_CHROMA_V, absPartIdxC, tuIterator.m_absPartIdxStep);
>
> - m_entropyCoder->encodeQtCbf(cu, absPartIdxC, tuIterator.m_absPartIdxStep, trSizeC, trSizeC, TEXT_CHROMA_U, trMode, true);
> + m_entropyCoder->encodeQtCbf(cu, absPartIdxC, TEXT_CHROMA_U, trMode);
> if (absSum[TEXT_CHROMA_U][tuIterator.m_section])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurU + subTUBufferOffset, absPartIdxC, trSizeC, TEXT_CHROMA_U);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurU + subTUOffset, absPartIdxC, log2TrSizeC, TEXT_CHROMA_U);
> singleBitsComp[TEXT_CHROMA_U][tuIterator.m_section] = m_entropyCoder->getNumberOfWrittenBits() - singleBitsPrev;
>
> - m_entropyCoder->encodeQtCbf(cu, absPartIdxC, tuIterator.m_absPartIdxStep, trSizeC, trSizeC, TEXT_CHROMA_V, trMode, true);
> + m_entropyCoder->encodeQtCbf(cu, absPartIdxC, TEXT_CHROMA_V, trMode);
> if (absSum[TEXT_CHROMA_V][tuIterator.m_section])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurV + subTUBufferOffset, absPartIdxC, trSizeC, TEXT_CHROMA_V);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurV + subTUOffset, absPartIdxC, log2TrSizeC, TEXT_CHROMA_V);
> uint32_t newBits = m_entropyCoder->getNumberOfWrittenBits();
> singleBitsComp[TEXT_CHROMA_V][tuIterator.m_section] = newBits - (singleBitsPrev + singleBitsComp[TEXT_CHROMA_U][tuIterator.m_section]);
>
> @@ -3105,7 +3134,7 @@
> do
> {
> uint32_t absPartIdxC = tuIterator.m_absPartIdxTURelCU;
> - uint32_t subTUBufferOffset = trSizeC * trSizeC * tuIterator.m_section;
> + uint32_t subTUOffset = tuIterator.m_section << (log2TrSizeC * 2);
>
> int16_t *curResiU = m_qtTempShortYuv[qtLayer].getCbAddr(absPartIdxC);
> int16_t *curResiV = m_qtTempShortYuv[qtLayer].getCrAddr(absPartIdxC);
> @@ -3123,7 +3152,7 @@
>
> int scalingListType = 3 + TEXT_CHROMA_U;
> X265_CHECK(scalingListType < 6, "scalingListType too large %d\n", scalingListType);
> - m_trQuant->invtransformNxN(cu->getCUTransquantBypass(absPartIdxC), REG_DCT, curResiU, strideResiC, coeffCurU + subTUBufferOffset,
> + m_trQuant->invtransformNxN(cu->getCUTransquantBypass(absPartIdxC), REG_DCT, curResiU, strideResiC, coeffCurU + subTUOffset,
> trSizeC, scalingListType, false, lastPos[TEXT_CHROMA_U][tuIterator.m_section]);
> uint32_t dist = primitives.sse_ss[partSizeC](resiYuv->getCbAddr(absPartIdxC), resiYuv->m_cwidth,
> curResiU, strideResiC);
> @@ -3166,7 +3195,7 @@
> {
> absSum[TEXT_CHROMA_U][tuIterator.m_section] = 0;
> #if CHECKED_BUILD || _DEBUG
> - ::memset(coeffCurU + subTUBufferOffset, 0, sizeof(coeff_t) * numCoeffC);
> + ::memset(coeffCurU + subTUOffset, 0, sizeof(coeff_t) * numCoeffC);
> #endif
> if (checkTransformSkipUV)
> {
> @@ -3215,7 +3244,7 @@
>
> int scalingListType = 3 + TEXT_CHROMA_V;
> X265_CHECK(scalingListType < 6, "scalingListType too large %d\n", scalingListType);
> - m_trQuant->invtransformNxN(cu->getCUTransquantBypass(absPartIdxC), REG_DCT, curResiV, strideResiC, coeffCurV + subTUBufferOffset,
> + m_trQuant->invtransformNxN(cu->getCUTransquantBypass(absPartIdxC), REG_DCT, curResiV, strideResiC, coeffCurV + subTUOffset,
> trSizeC, scalingListType, false, lastPos[TEXT_CHROMA_V][tuIterator.m_section]);
> uint32_t dist = primitives.sse_ss[partSizeC](resiYuv->getCrAddr(absPartIdxC), resiYuv->m_cwidth,
> curResiV, strideResiC);
> @@ -3259,7 +3288,7 @@
> {
> absSum[TEXT_CHROMA_V][tuIterator.m_section] = 0;
> #if CHECKED_BUILD || _DEBUG
> - ::memset(coeffCurV + subTUBufferOffset, 0, sizeof(coeff_t) * numCoeffC);
> + ::memset(coeffCurV + subTUOffset, 0, sizeof(coeff_t) * numCoeffC);
> #endif
> if (checkTransformSkipUV)
> {
> @@ -3330,8 +3359,8 @@
> if (absSumTransformSkipY)
> {
> m_entropyCoder->resetBits();
> - m_entropyCoder->encodeQtCbf(cu, absPartIdx, 0, trSize, trSize, TEXT_LUMA, trMode, true);
> - m_entropyCoder->encodeCoeffNxN(cu, tsCoeffY, absPartIdx, trSize, TEXT_LUMA);
> + m_entropyCoder->encodeQtCbf(cu, absPartIdx, TEXT_LUMA, trMode);
> + m_entropyCoder->encodeCoeffNxN(cu, tsCoeffY, absPartIdx, log2TrSize, TEXT_LUMA);
> const uint32_t skipSingleBitsY = m_entropyCoder->getNumberOfWrittenBits();
>
> m_trQuant->setQPforQuant(cu->getQP(0), TEXT_LUMA, QP_BD_OFFSET, 0, chFmt);
> @@ -3396,7 +3425,7 @@
> do
> {
> uint32_t absPartIdxC = tuIterator.m_absPartIdxTURelCU;
> - uint32_t subTUBufferOffset = trSizeC * trSizeC * tuIterator.m_section;
> + uint32_t subTUOffset = tuIterator.m_section << (log2TrSizeC * 2);
>
> int16_t *curResiU = m_qtTempShortYuv[qtLayer].getCbAddr(absPartIdxC);
> int16_t *curResiV = m_qtTempShortYuv[qtLayer].getCrAddr(absPartIdxC);
> @@ -3433,8 +3462,8 @@
>
> if (absSumTransformSkipU)
> {
> - m_entropyCoder->encodeQtCbf(cu, absPartIdxC, tuIterator.m_absPartIdxStep, trSizeC, trSizeC, TEXT_CHROMA_U, trMode, true);
> - m_entropyCoder->encodeCoeffNxN(cu, tsCoeffU, absPartIdxC, trSizeC, TEXT_CHROMA_U);
> + m_entropyCoder->encodeQtCbf(cu, absPartIdxC, TEXT_CHROMA_U, trMode);
> + m_entropyCoder->encodeCoeffNxN(cu, tsCoeffU, absPartIdxC, log2TrSizeC, TEXT_CHROMA_U);
> singleBitsComp[TEXT_CHROMA_U][tuIterator.m_section] = m_entropyCoder->getNumberOfWrittenBits();
>
> curChromaQpOffset = cu->getSlice()->getPPS()->getChromaCbQpOffset() + cu->getSlice()->getSliceQpDeltaCb();
> @@ -3475,14 +3504,14 @@
> singlePsyEnergyComp[TEXT_CHROMA_U][tuIterator.m_section] = nonZeroPsyEnergyU;
> absSum[TEXT_CHROMA_U][tuIterator.m_section] = absSumTransformSkipU;
> bestTransformMode[TEXT_CHROMA_U][tuIterator.m_section] = 1;
> - memcpy(coeffCurU + subTUBufferOffset, tsCoeffU, sizeof(coeff_t) * numCoeffC);
> + memcpy(coeffCurU + subTUOffset, tsCoeffU, sizeof(coeff_t) * numCoeffC);
> primitives.square_copy_ss[sizeIdxC](curResiU, strideResiC, tsResiU, trSizeC);
> }
>
> if (absSumTransformSkipV)
> {
> - m_entropyCoder->encodeQtCbf(cu, absPartIdxC, tuIterator.m_absPartIdxStep, trSizeC, trSizeC, TEXT_CHROMA_V, trMode, true);
> - m_entropyCoder->encodeCoeffNxN(cu, tsCoeffV, absPartIdxC, trSizeC, TEXT_CHROMA_V);
> + m_entropyCoder->encodeQtCbf(cu, absPartIdxC, TEXT_CHROMA_V, trMode);
> + m_entropyCoder->encodeCoeffNxN(cu, tsCoeffV, absPartIdxC, log2TrSizeC, TEXT_CHROMA_V);
> singleBitsComp[TEXT_CHROMA_V][tuIterator.m_section] = m_entropyCoder->getNumberOfWrittenBits() - singleBitsComp[TEXT_CHROMA_U][tuIterator.m_section];
>
> curChromaQpOffset = cu->getSlice()->getPPS()->getChromaCrQpOffset() + cu->getSlice()->getSliceQpDeltaCr();
> @@ -3523,7 +3552,7 @@
> singlePsyEnergyComp[TEXT_CHROMA_V][tuIterator.m_section] = nonZeroPsyEnergyV;
> absSum[TEXT_CHROMA_V][tuIterator.m_section] = absSumTransformSkipV;
> bestTransformMode[TEXT_CHROMA_V][tuIterator.m_section] = 1;
> - memcpy(coeffCurV + subTUBufferOffset, tsCoeffV, sizeof(coeff_t) * numCoeffC);
> + memcpy(coeffCurV + subTUOffset, tsCoeffV, sizeof(coeff_t) * numCoeffC);
> primitives.square_copy_ss[sizeIdxC](curResiV, strideResiC, tsResiV, trSizeC);
> }
>
> @@ -3556,32 +3585,32 @@
> m_entropyCoder->encodeQtCbf(cu, absPartIdx, absPartIdxStep, trSizeC, trHeightC, TEXT_CHROMA_V, trMode, true);
> }
>
> - m_entropyCoder->encodeQtCbf(cu, absPartIdx, 0, trSize, trSize, TEXT_LUMA, trMode, true);
> + m_entropyCoder->encodeQtCbf(cu, absPartIdx, TEXT_LUMA, trMode);
> if (absSum[TEXT_LUMA][0])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurY, absPartIdx, trSize, TEXT_LUMA);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurY, absPartIdx, log2TrSize, TEXT_LUMA);
>
> if (bCodeChroma)
> {
> if (!splitIntoSubTUs)
> {
> if (absSum[TEXT_CHROMA_U][0])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurU, absPartIdx, trSizeC, TEXT_CHROMA_U);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurU, absPartIdx, log2TrSizeC, TEXT_CHROMA_U);
> if (absSum[TEXT_CHROMA_V][0])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurV, absPartIdx, trSizeC, TEXT_CHROMA_V);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurV, absPartIdx, log2TrSizeC, TEXT_CHROMA_V);
> }
> else
> {
> - uint32_t subTUSize = trSizeC * trSizeC;
> + uint32_t subTUSize = 1 << (log2TrSizeC * 2);
> uint32_t partIdxesPerSubTU = absPartIdxStep >> 1;
>
> if (absSum[TEXT_CHROMA_U][0])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurU, absPartIdx, trSizeC, TEXT_CHROMA_U);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurU, absPartIdx, log2TrSizeC, TEXT_CHROMA_U);
> if (absSum[TEXT_CHROMA_U][1])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurU + subTUSize, absPartIdx + partIdxesPerSubTU, trSizeC, TEXT_CHROMA_U);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurU + subTUSize, absPartIdx + partIdxesPerSubTU, log2TrSizeC, TEXT_CHROMA_U);
> if (absSum[TEXT_CHROMA_V][0])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurV, absPartIdx, trSizeC, TEXT_CHROMA_V);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurV, absPartIdx, log2TrSizeC, TEXT_CHROMA_V);
> if (absSum[TEXT_CHROMA_V][1])
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurV + subTUSize, absPartIdx + partIdxesPerSubTU, trSizeC, TEXT_CHROMA_V);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurV + subTUSize, absPartIdx + partIdxesPerSubTU, log2TrSizeC, TEXT_CHROMA_V);
> }
> }
>
> @@ -3769,7 +3798,6 @@
> X265_CHECK(cu->getPredictionMode(absPartIdx) != MODE_INTRA, "xEncodeResidualQT() with intra block\n");
>
> bool mCodeAll = true;
> - uint32_t trSize = 1 << log2TrSize;
> uint32_t trWidthC = 1 << log2TrSizeC;
> uint32_t trHeightC = splitIntoSubTUs ? (trWidthC << 1) : trWidthC;
>
> @@ -3821,49 +3849,48 @@
>
> if (bSubdivAndCbf)
> {
> - m_entropyCoder->encodeQtCbf(cu, absPartIdx, 0, trSize, trSize, TEXT_LUMA, trMode, true);
> + m_entropyCoder->encodeQtCbf(cu, absPartIdx, TEXT_LUMA, trMode);
> }
> else
> {
> if (ttype == TEXT_LUMA && cu->getCbf(absPartIdx, TEXT_LUMA, trMode))
> {
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurY, absPartIdx, trSize, TEXT_LUMA);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurY, absPartIdx, log2TrSize, TEXT_LUMA);
> }
> if (bCodeChroma)
> {
> uint32_t coeffOffsetC = coeffOffsetY >> (m_hChromaShift + m_vChromaShift);
> coeff_t *coeffCurU = m_qtTempCoeff[1][qtLayer] + coeffOffsetC;
> coeff_t *coeffCurV = m_qtTempCoeff[2][qtLayer] + coeffOffsetC;
> - uint32_t trSizeC = 1 << log2TrSizeC;
>
> if (!splitIntoSubTUs)
> {
> if (ttype == TEXT_CHROMA_U && cu->getCbf(absPartIdx, TEXT_CHROMA_U, trMode))
> {
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurU, absPartIdx, trSizeC, TEXT_CHROMA_U);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurU, absPartIdx, log2TrSizeC, TEXT_CHROMA_U);
> }
> if (ttype == TEXT_CHROMA_V && cu->getCbf(absPartIdx, TEXT_CHROMA_V, trMode))
> {
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurV, absPartIdx, trSizeC, TEXT_CHROMA_V);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurV, absPartIdx, log2TrSizeC, TEXT_CHROMA_V);
> }
> }
> else
> {
> uint32_t partIdxesPerSubTU = cu->getPic()->getNumPartInCU() >> (((cu->getDepth(absPartIdx) + trModeC) << 1) + 1);
> - uint32_t subTUSize = trSizeC * trSizeC;
> + uint32_t subTUSize = 1 << (log2TrSizeC * 2);
> if (ttype == TEXT_CHROMA_U && cu->getCbf(absPartIdx, TEXT_CHROMA_U, trMode))
> {
> if (cu->getCbf(absPartIdx, ttype, trMode + 1))
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurU, absPartIdx, trSizeC, TEXT_CHROMA_U);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurU, absPartIdx, log2TrSizeC, TEXT_CHROMA_U);
> if (cu->getCbf(absPartIdx + partIdxesPerSubTU, ttype, trMode + 1))
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurU + subTUSize, absPartIdx + partIdxesPerSubTU, trSizeC, TEXT_CHROMA_U);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurU + subTUSize, absPartIdx + partIdxesPerSubTU, log2TrSizeC, TEXT_CHROMA_U);
> }
> if (ttype == TEXT_CHROMA_V && cu->getCbf(absPartIdx, TEXT_CHROMA_V, trMode))
> {
> if (cu->getCbf(absPartIdx, ttype, trMode + 1))
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurV, absPartIdx, trSizeC, TEXT_CHROMA_V);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurV, absPartIdx, log2TrSizeC, TEXT_CHROMA_V);
> if (cu->getCbf(absPartIdx + partIdxesPerSubTU, ttype, trMode + 1))
> - m_entropyCoder->encodeCoeffNxN(cu, coeffCurV + subTUSize, absPartIdx + partIdxesPerSubTU, trSizeC, TEXT_CHROMA_V);
> + m_entropyCoder->encodeCoeffNxN(cu, coeffCurV + subTUSize, absPartIdx + partIdxesPerSubTU, log2TrSizeC, TEXT_CHROMA_V);
> }
> }
> }
> diff -r 32aa6cc3cf4d -r 4c30d66afc78 source/Lib/TLibEncoder/TEncSearch.h
> --- a/source/Lib/TLibEncoder/TEncSearch.h Thu Jun 26 17:19:08 2014 -0700
> +++ b/source/Lib/TLibEncoder/TEncSearch.h Mon Jun 30 11:48:58 2014 +0900
> @@ -193,13 +193,17 @@
> // Intra search
> // --------------------------------------------------------------------------------------------
>
> - void xEncSubdivCbfQT(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t absPartIdxStep, uint32_t width, uint32_t height, bool bLuma, bool bChroma);
> + void xEncSubdivCbfQTLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx);
> + void xEncSubdivCbfQTChroma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t absPartIdxStep, uint32_t width, uint32_t height);
>
> - void xEncCoeffQT(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, TextType ttype);
> - void xEncIntraHeader(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, bool bLuma, bool bChroma);
> - uint32_t xGetIntraBitsQT(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t absPartIdxStep, bool bLuma, bool bChroma);
> - uint32_t xGetIntraBitsQTLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t log2TrSize, coeff_t* coeff);
> - uint32_t xGetIntraBitsQTChroma(TComDataCU* cu, uint32_t absPartIdx, uint32_t log2TrSizeC, uint32_t chromaId, coeff_t* coeff);
> + void xEncCoeffQTLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx);
> + void xEncCoeffQTChroma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, TextType ttype);
> + void xEncIntraHeaderLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx);
> + void xEncIntraHeaderChroma(TComDataCU* cu, uint32_t absPartIdx);
> + uint32_t xGetIntraBitsQTLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx);
> + uint32_t xGetIntraBitsQTChroma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t absPartIdxStep);
> + uint32_t xGetIntraBitsLuma(TComDataCU* cu, uint32_t trDepth, uint32_t absPartIdx, uint32_t log2TrSize, coeff_t* coeff);
> + uint32_t xGetIntraBitsChroma(TComDataCU* cu, uint32_t absPartIdx, uint32_t log2TrSizeC, uint32_t chromaId, coeff_t* coeff);
> void xIntraCodingLumaBlk(TComDataCU* cu, uint32_t absPartIdx, uint32_t log2TrSize, TComYuv* fencYuv, TComYuv* predYuv, ShortYuv* resiYuv,
> int16_t* reconQt, uint32_t reconQtStride, coeff_t* coeff,
> uint32_t& cbf, uint32_t& outDist);
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel
--
Steve Borho
More information about the x265-devel
mailing list