root/source/encoder/analysis.cpp

DEFINITIONS

1. create
2. destroy
3. compressCTU
4. collectPUStatistics
5. loadTUDepth
6. tryLossless
7. qprdRefine
8. compressIntraCU
9. processTasks
10. processPmode
11. compressInterCU_dist
12. compressInterCU_rd0_4
13. compressInterCU_rd5_6
14. recodeCU
15. checkMerge2Nx2N_rd0_4
16. checkMerge2Nx2N_rd5_6
17. checkInter_rd0_4
18. checkInter_rd5_6
19. checkBidir2Nx2N
20. encodeResidue
21. addSplitFlagCost
22. topSkipMinDepth
23. recursionDepthCheck
24. complexityCheckCU
25. calculateQpforCuSize
26. normFactor
27. calculateNormFactor
28. findSameContentRefCount

/*****************************************************************************
* Copyright (C) 2013-2017 MulticoreWare, Inc
*
* Authors: Deepthi Nandakumar <deepthi@multicorewareinc.com>
*          Steve Borho <steve@borho.org>
*          Min Chen <chenm003@163.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
*
* This program is also available under a commercial proprietary license.
* For more information, contact us at license @ x265.com.
*****************************************************************************/

#include "common.h"
#include "frame.h"
#include "framedata.h"
#include "picyuv.h"
#include "primitives.h"
#include "threading.h"

#include "analysis.h"
#include "rdcost.h"
#include "encoder.h"

using namespace X265_NS;

/* An explanation of rate distortion levels (--rd-level)
 * 
 * rd-level 0 generates no recon per CU (NO RDO or Quant)
 *
 *   sa8d selection between merge / skip / inter / intra and split
 *   no recon pixels generated until CTU analysis is complete, requiring
 *   intra predictions to use source pixels
 *
 * rd-level 1 uses RDO for merge and skip, sa8d for all else
 *
 *   RDO selection between merge and skip
 *   sa8d selection between (merge/skip) / inter modes / intra and split
 *   intra prediction uses reconstructed pixels
 *
 * rd-level 2 uses RDO for merge/skip and split
 *
 *   RDO selection between merge and skip
 *   sa8d selection between (merge/skip) / inter modes / intra
 *   RDO split decisions
 *
 * rd-level 3 uses RDO for merge/skip/best inter/intra
 *
 *   RDO selection between merge and skip
 *   sa8d selection of best inter mode
 *   sa8d decisions include chroma residual cost
 *   RDO selection between (merge/skip) / best inter mode / intra / split
 *
 * rd-level 4 enables RDOQuant
 *   chroma residual cost included in satd decisions, including subpel refine
 *    (as a result of --subme 3 being used by preset slow)
 *
 * rd-level 5,6 does RDO for each inter mode
 */

Analysis::Analysis()
{
    m_reuseInterDataCTU = NULL;
    m_reuseRef = NULL;
    m_bHD = false;
    m_evaluateInter = 0;
}

bool Analysis::create(ThreadLocalData *tld)
{
    m_tld = tld;
    m_bTryLossless = m_param->bCULossless && !m_param->bLossless && m_param->rdLevel >= 2;

    int costArrSize = 1;
    uint32_t maxDQPDepth = g_log2Size[m_param->maxCUSize] - g_log2Size[m_param->rc.qgSize];
    for (uint32_t i = 1; i <= maxDQPDepth; i++)
        costArrSize += (1 << (i * 2));
    cacheCost = X265_MALLOC(uint64_t, costArrSize);

    int csp = m_param->internalCsp;
    uint32_t cuSize = m_param->maxCUSize;

    bool ok = true;
    for (uint32_t depth = 0; depth <= m_param->maxCUDepth; depth++, cuSize >>= 1)
    {
        ModeDepth &md = m_modeDepth[depth];

        md.cuMemPool.create(depth, csp, MAX_PRED_TYPES, *m_param);
        ok &= md.fencYuv.create(cuSize, csp);

        for (int j = 0; j < MAX_PRED_TYPES; j++)
        {
            md.pred[j].cu.initialize(md.cuMemPool, depth, *m_param, j);
            ok &= md.pred[j].predYuv.create(cuSize, csp);
            ok &= md.pred[j].reconYuv.create(cuSize, csp);
            md.pred[j].fencYuv = &md.fencYuv;
        }
    }
    if (m_param->sourceHeight >= 1080)
        m_bHD = true;

    return ok;
}

void Analysis::destroy()
{
    for (uint32_t i = 0; i <= m_param->maxCUDepth; i++)
    {
        m_modeDepth[i].cuMemPool.destroy();
        m_modeDepth[i].fencYuv.destroy();

        for (int j = 0; j < MAX_PRED_TYPES; j++)
        {
            m_modeDepth[i].pred[j].predYuv.destroy();
            m_modeDepth[i].pred[j].reconYuv.destroy();
        }
    }
    X265_FREE(cacheCost);
}

Mode& Analysis::compressCTU(CUData& ctu, Frame& frame, const CUGeom& cuGeom, const Entropy& initialContext)
{
    m_slice = ctu.m_slice;
    m_frame = &frame;
    m_bChromaSa8d = m_param->rdLevel >= 3;

#if _DEBUG || CHECKED_BUILD
    invalidateContexts(0);
#endif

    int qp = setLambdaFromQP(ctu, m_slice->m_pps->bUseDQP ? calculateQpforCuSize(ctu, cuGeom) : m_slice->m_sliceQp);
    ctu.setQPSubParts((int8_t)qp, 0, 0);

    m_rqt[0].cur.load(initialContext);
    ctu.m_meanQP = initialContext.m_meanQP;
    m_modeDepth[0].fencYuv.copyFromPicYuv(*m_frame->m_fencPic, ctu.m_cuAddr, 0);

    if (m_param->bSsimRd)
        calculateNormFactor(ctu, qp);

    uint32_t numPartition = ctu.m_numPartitions;
    if (m_param->bCTUInfo && (*m_frame->m_ctuInfo + ctu.m_cuAddr))
    {
        x265_ctu_info_t* ctuTemp = *m_frame->m_ctuInfo + ctu.m_cuAddr;
        if (ctuTemp->ctuPartitions)
        {
            int32_t depthIdx = 0;
            uint32_t maxNum8x8Partitions = 64;
            uint8_t* depthInfoPtr = m_frame->m_addOnDepth[ctu.m_cuAddr];
            uint8_t* contentInfoPtr = m_frame->m_addOnCtuInfo[ctu.m_cuAddr];
            int* prevCtuInfoChangePtr = m_frame->m_addOnPrevChange[ctu.m_cuAddr];
            do
            {
                uint8_t depth = (uint8_t)ctuTemp->ctuPartitions[depthIdx];
                uint8_t content = (uint8_t)(*((int32_t *)ctuTemp->ctuInfo + depthIdx));
                int prevCtuInfoChange = m_frame->m_prevCtuInfoChange[ctu.m_cuAddr * maxNum8x8Partitions + depthIdx];
                memset(depthInfoPtr, depth, sizeof(uint8_t) * numPartition >> 2 * depth);
                memset(contentInfoPtr, content, sizeof(uint8_t) * numPartition >> 2 * depth);
                memset(prevCtuInfoChangePtr, 0, sizeof(int) * numPartition >> 2 * depth);
                for (uint32_t l = 0; l < numPartition >> 2 * depth; l++)
                    prevCtuInfoChangePtr[l] = prevCtuInfoChange;
                depthInfoPtr += ctu.m_numPartitions >> 2 * depth;
                contentInfoPtr += ctu.m_numPartitions >> 2 * depth;
                prevCtuInfoChangePtr += ctu.m_numPartitions >> 2 * depth;
                depthIdx++;
            } while (ctuTemp->ctuPartitions[depthIdx] != 0);

            m_additionalCtuInfo = m_frame->m_addOnCtuInfo[ctu.m_cuAddr];
            m_prevCtuInfoChange = m_frame->m_addOnPrevChange[ctu.m_cuAddr];
            memcpy(ctu.m_cuDepth, m_frame->m_addOnDepth[ctu.m_cuAddr], sizeof(uint8_t) * numPartition);
            //Calculate log2CUSize from depth
            for (uint32_t i = 0; i < cuGeom.numPartitions; i++)
                ctu.m_log2CUSize[i] = (uint8_t)m_param->maxLog2CUSize - ctu.m_cuDepth[i];
        }
    }

    if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead)
    {
        m_multipassAnalysis = (analysis2PassFrameData*)m_frame->m_analysis2Pass.analysisFramedata;
        m_multipassDepth = &m_multipassAnalysis->depth[ctu.m_cuAddr * ctu.m_numPartitions];
        if (m_slice->m_sliceType != I_SLICE)
        {
            int numPredDir = m_slice->isInterP() ? 1 : 2;
            for (int dir = 0; dir < numPredDir; dir++)
            {
                m_multipassMv[dir] = &m_multipassAnalysis->m_mv[dir][ctu.m_cuAddr * ctu.m_numPartitions];
                m_multipassMvpIdx[dir] = &m_multipassAnalysis->mvpIdx[dir][ctu.m_cuAddr * ctu.m_numPartitions];
                m_multipassRef[dir] = &m_multipassAnalysis->ref[dir][ctu.m_cuAddr * ctu.m_numPartitions];
            }
            m_multipassModes = &m_multipassAnalysis->modes[ctu.m_cuAddr * ctu.m_numPartitions];
        }
    }

    if (m_param->analysisReuseMode && m_slice->m_sliceType != I_SLICE && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel < 10)
    {
        int numPredDir = m_slice->isInterP() ? 1 : 2;
        m_reuseInterDataCTU = (analysis_inter_data*)m_frame->m_analysisData.interData;
        m_reuseRef = &m_reuseInterDataCTU->ref[ctu.m_cuAddr * X265_MAX_PRED_MODE_PER_CTU * numPredDir];
        m_reuseDepth = &m_reuseInterDataCTU->depth[ctu.m_cuAddr * ctu.m_numPartitions];
        m_reuseModes = &m_reuseInterDataCTU->modes[ctu.m_cuAddr * ctu.m_numPartitions];
        if (m_param->analysisReuseLevel > 4)
        {
            m_reusePartSize = &m_reuseInterDataCTU->partSize[ctu.m_cuAddr * ctu.m_numPartitions];
            m_reuseMergeFlag = &m_reuseInterDataCTU->mergeFlag[ctu.m_cuAddr * ctu.m_numPartitions];
        }
        if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE)
            for (int i = 0; i < X265_MAX_PRED_MODE_PER_CTU * numPredDir; i++)
                m_reuseRef[i] = -1;
    }
    ProfileCUScope(ctu, totalCTUTime, totalCTUs);

    if (m_slice->m_sliceType == I_SLICE)
    {
        analysis_intra_data* intraDataCTU = (analysis_intra_data*)m_frame->m_analysisData.intraData;
        if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1)
        {
            memcpy(ctu.m_cuDepth, &intraDataCTU->depth[ctu.m_cuAddr * numPartition], sizeof(uint8_t) * numPartition);
            memcpy(ctu.m_lumaIntraDir, &intraDataCTU->modes[ctu.m_cuAddr * numPartition], sizeof(uint8_t) * numPartition);
            memcpy(ctu.m_partSize, &intraDataCTU->partSizes[ctu.m_cuAddr * numPartition], sizeof(char) * numPartition);
            memcpy(ctu.m_chromaIntraDir, &intraDataCTU->chromaModes[ctu.m_cuAddr * numPartition], sizeof(uint8_t) * numPartition);
        }
        compressIntraCU(ctu, cuGeom, qp);
    }
    else
    {
        if (m_param->bIntraRefresh && m_slice->m_sliceType == P_SLICE &&
            ctu.m_cuPelX / m_param->maxCUSize >= frame.m_encData->m_pir.pirStartCol
            && ctu.m_cuPelX / m_param->maxCUSize < frame.m_encData->m_pir.pirEndCol)
            compressIntraCU(ctu, cuGeom, qp);
        else if (!m_param->rdLevel)
        {
            /* In RD Level 0/1, copy source pixels into the reconstructed block so
             * they are available for intra predictions */
            m_modeDepth[0].fencYuv.copyToPicYuv(*m_frame->m_reconPic, ctu.m_cuAddr, 0);

            compressInterCU_rd0_4(ctu, cuGeom, qp);

            /* generate residual for entire CTU at once and copy to reconPic */
            encodeResidue(ctu, cuGeom);
        }
        else if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel == 10)
        {
            analysis_inter_data* interDataCTU = (analysis_inter_data*)m_frame->m_analysisData.interData;
            int posCTU = ctu.m_cuAddr * numPartition;
            memcpy(ctu.m_cuDepth, &interDataCTU->depth[posCTU], sizeof(uint8_t) * numPartition);
            memcpy(ctu.m_predMode, &interDataCTU->modes[posCTU], sizeof(uint8_t) * numPartition);
            memcpy(ctu.m_partSize, &interDataCTU->partSize[posCTU], sizeof(uint8_t) * numPartition);
            if (m_slice->m_sliceType == P_SLICE || m_param->bIntraInBFrames)
            {
                analysis_intra_data* intraDataCTU = (analysis_intra_data*)m_frame->m_analysisData.intraData;
                memcpy(ctu.m_lumaIntraDir, &intraDataCTU->modes[posCTU], sizeof(uint8_t) * numPartition);
                memcpy(ctu.m_chromaIntraDir, &intraDataCTU->chromaModes[posCTU], sizeof(uint8_t) * numPartition);
            }
            //Calculate log2CUSize from depth
            for (uint32_t i = 0; i < cuGeom.numPartitions; i++)
                ctu.m_log2CUSize[i] = (uint8_t)m_param->maxLog2CUSize - ctu.m_cuDepth[i];

            qprdRefine (ctu, cuGeom, qp, qp);
            return *m_modeDepth[0].bestMode;
        }
        else if (m_param->bDistributeModeAnalysis && m_param->rdLevel >= 2)
            compressInterCU_dist(ctu, cuGeom, qp);
        else if (m_param->rdLevel <= 4)
            compressInterCU_rd0_4(ctu, cuGeom, qp);
        else
            compressInterCU_rd5_6(ctu, cuGeom, qp);
    }

    if (m_param->bEnableRdRefine || m_param->bOptCUDeltaQP)
        qprdRefine(ctu, cuGeom, qp, qp);

    if (m_param->csvLogLevel >= 2)
        collectPUStatistics(ctu, cuGeom);

    return *m_modeDepth[0].bestMode;
}

void Analysis::collectPUStatistics(const CUData& ctu, const CUGeom& cuGeom)
{
    uint8_t depth = 0;
    uint8_t partSize = 0;
    for (uint32_t absPartIdx = 0; absPartIdx < ctu.m_numPartitions; absPartIdx += ctu.m_numPartitions >> (depth * 2))
    {
        depth = ctu.m_cuDepth[absPartIdx];
        partSize = ctu.m_partSize[absPartIdx];
        uint32_t numPU = nbPartsTable[(int)partSize];
        int shift = 2 * (m_param->maxCUDepth + 1 - depth);
        for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
        {
            PredictionUnit pu(ctu, cuGeom, puIdx);
            int puabsPartIdx = ctu.getPUOffset(puIdx, absPartIdx);
            int mode = 1;
            if (ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_Nx2N || ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_2NxN)
                mode = 2;
            else if (ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_2NxnU || ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_2NxnD || ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_nLx2N || ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_nRx2N)
                 mode = 3;

            if (ctu.m_predMode[puabsPartIdx + absPartIdx] == MODE_SKIP)
            {
                ctu.m_encData->m_frameStats.cntSkipPu[depth] += (uint64_t)(1 << shift);
                ctu.m_encData->m_frameStats.totalPu[depth] += (uint64_t)(1 << shift);
            }
            else if (ctu.m_predMode[puabsPartIdx + absPartIdx] == MODE_INTRA)
            {
                if (ctu.m_partSize[puabsPartIdx + absPartIdx] == SIZE_NxN)
                {
                    ctu.m_encData->m_frameStats.cnt4x4++;
                    ctu.m_encData->m_frameStats.totalPu[4]++;
                }
                else
                {
                    ctu.m_encData->m_frameStats.cntIntraPu[depth] += (uint64_t)(1 << shift);
                    ctu.m_encData->m_frameStats.totalPu[depth] += (uint64_t)(1 << shift);
                }
            }
            else if (mode == 3)
            {
                ctu.m_encData->m_frameStats.cntAmp[depth] += (uint64_t)(1 << shift);
                ctu.m_encData->m_frameStats.totalPu[depth] += (uint64_t)(1 << shift);
                break;
            }
            else
            {
                if (ctu.m_mergeFlag[puabsPartIdx + absPartIdx])
                    ctu.m_encData->m_frameStats.cntMergePu[depth][ctu.m_partSize[puabsPartIdx + absPartIdx]] += (1 << shift) / mode;
                else
                    ctu.m_encData->m_frameStats.cntInterPu[depth][ctu.m_partSize[puabsPartIdx + absPartIdx]] += (1 << shift) / mode;

                ctu.m_encData->m_frameStats.totalPu[depth] += (1 << shift) / mode;
            }
        }
    }
}

int32_t Analysis::loadTUDepth(CUGeom cuGeom, CUData parentCTU)
{
    float predDepth = 0;
    CUData* neighbourCU;
    uint8_t count = 0;
    int32_t maxTUDepth = -1;
    neighbourCU = m_slice->m_refFrameList[0][0]->m_encData->m_picCTU;
    predDepth += neighbourCU->m_refTuDepth[cuGeom.geomRecurId];
    count++;
    if (m_slice->isInterB())
    {
        neighbourCU = m_slice->m_refFrameList[1][0]->m_encData->m_picCTU;
        predDepth += neighbourCU->m_refTuDepth[cuGeom.geomRecurId];
        count++;
    }
    if (parentCTU.m_cuAbove)
    {
        predDepth += parentCTU.m_cuAbove->m_refTuDepth[cuGeom.geomRecurId];
        count++;
        if (parentCTU.m_cuAboveLeft)
        {
            predDepth += parentCTU.m_cuAboveLeft->m_refTuDepth[cuGeom.geomRecurId];
            count++;
        }
        if (parentCTU.m_cuAboveRight)
        {
            predDepth += parentCTU.m_cuAboveRight->m_refTuDepth[cuGeom.geomRecurId];
            count++;
        }
    }
    if (parentCTU.m_cuLeft)
    {
        predDepth += parentCTU.m_cuLeft->m_refTuDepth[cuGeom.geomRecurId];
        count++;
    }
    predDepth /= count;

    if (predDepth == 0)
        maxTUDepth = 0;
    else if (predDepth < 1)
        maxTUDepth = 1;
    else if (predDepth >= 1 && predDepth <= 1.5)
        maxTUDepth = 2;
    else if (predDepth > 1.5 && predDepth <= 2.5)
        maxTUDepth = 3;
    else
        maxTUDepth = -1;

    return maxTUDepth;
}

void Analysis::tryLossless(const CUGeom& cuGeom)
{
    ModeDepth& md = m_modeDepth[cuGeom.depth];

    if (!md.bestMode->distortion)
        /* already lossless */
        return;
    else if (md.bestMode->cu.isIntra(0))
    {
        md.pred[PRED_LOSSLESS].initCosts();
        md.pred[PRED_LOSSLESS].cu.initLosslessCU(md.bestMode->cu, cuGeom);
        PartSize size = (PartSize)md.pred[PRED_LOSSLESS].cu.m_partSize[0];
        checkIntra(md.pred[PRED_LOSSLESS], cuGeom, size);
        checkBestMode(md.pred[PRED_LOSSLESS], cuGeom.depth);
    }
    else
    {
        md.pred[PRED_LOSSLESS].initCosts();
        md.pred[PRED_LOSSLESS].cu.initLosslessCU(md.bestMode->cu, cuGeom);
        md.pred[PRED_LOSSLESS].predYuv.copyFromYuv(md.bestMode->predYuv);
        encodeResAndCalcRdInterCU(md.pred[PRED_LOSSLESS], cuGeom);
        checkBestMode(md.pred[PRED_LOSSLESS], cuGeom.depth);
    }
}

void Analysis::qprdRefine(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp, int32_t lqp)
{
    uint32_t depth = cuGeom.depth;
    ModeDepth& md = m_modeDepth[depth];
    md.bestMode = NULL;

    bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;

    int bestCUQP = qp;
    int lambdaQP = lqp;

    bool doQPRefine = (bDecidedDepth && depth <= m_slice->m_pps->maxCuDQPDepth) || (!bDecidedDepth && depth == m_slice->m_pps->maxCuDQPDepth);
    if (m_param->analysisReuseLevel == 10)
        doQPRefine = false;

    if (doQPRefine)
    {
        uint64_t bestCUCost, origCUCost, cuCost, cuPrevCost;

        int cuIdx = (cuGeom.childOffset - 1) / 3;
        bestCUCost = origCUCost = cacheCost[cuIdx];

        int direction = m_param->bOptCUDeltaQP ? 1 : 2;

        for (int dir = direction; dir >= -direction; dir -= (direction * 2))
        {
            if (m_param->bOptCUDeltaQP && ((dir != 1) || ((qp + 3) >= (int32_t)parentCTU.m_meanQP)))
                break;

            int threshold = 1;
            int failure = 0;
            cuPrevCost = origCUCost;

            int modCUQP = qp + dir;
            while (modCUQP >= m_param->rc.qpMin && modCUQP <= QP_MAX_SPEC)
            {
                if (m_param->bOptCUDeltaQP && modCUQP > (int32_t)parentCTU.m_meanQP)
                    break;

                recodeCU(parentCTU, cuGeom, modCUQP, qp);
                cuCost = md.bestMode->rdCost;

                COPY2_IF_LT(bestCUCost, cuCost, bestCUQP, modCUQP);
                if (cuCost < cuPrevCost)
                    failure = 0;
                else
                    failure++;

                if (failure > threshold)
                    break;

                cuPrevCost = cuCost;
                modCUQP += dir;
            }
        }
        lambdaQP = bestCUQP;
    }

    recodeCU(parentCTU, cuGeom, bestCUQP, lambdaQP);

    /* Copy best data to encData CTU and recon */
    md.bestMode->cu.copyToPic(depth);
    md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, parentCTU.m_cuAddr, cuGeom.absPartIdx);
}

void Analysis::compressIntraCU(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp)
{
    uint32_t depth = cuGeom.depth;
    ModeDepth& md = m_modeDepth[depth];
    md.bestMode = NULL;

    bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);
    bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);

    bool bAlreadyDecided = parentCTU.m_lumaIntraDir[cuGeom.absPartIdx] != (uint8_t)ALL_IDX;
    bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;
    int split = 0;
    if (m_param->intraRefine)
    {
        split = ((cuGeom.log2CUSize == (uint32_t)(g_log2Size[m_param->minCUSize] + 1)) && bDecidedDepth);
        if (cuGeom.log2CUSize == (uint32_t)(g_log2Size[m_param->minCUSize]) && !bDecidedDepth)
            bAlreadyDecided = false;
    }

    if (bAlreadyDecided)
    {
        if (bDecidedDepth)
        {
            Mode& mode = md.pred[0];
            md.bestMode = &mode;
            mode.cu.initSubCU(parentCTU, cuGeom, qp);
            if (m_param->intraRefine != 2 || parentCTU.m_lumaIntraDir[cuGeom.absPartIdx] <= 1)
            {
                memcpy(mode.cu.m_lumaIntraDir, parentCTU.m_lumaIntraDir + cuGeom.absPartIdx, cuGeom.numPartitions);
                memcpy(mode.cu.m_chromaIntraDir, parentCTU.m_chromaIntraDir + cuGeom.absPartIdx, cuGeom.numPartitions);
            }
            checkIntra(mode, cuGeom, (PartSize)parentCTU.m_partSize[cuGeom.absPartIdx]);

            if (m_bTryLossless)
                tryLossless(cuGeom);

            if (mightSplit)
                addSplitFlagCost(*md.bestMode, cuGeom.depth);
        }
    }
    else if (cuGeom.log2CUSize != MAX_LOG2_CU_SIZE && mightNotSplit)
    {
        md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
        checkIntra(md.pred[PRED_INTRA], cuGeom, SIZE_2Nx2N);
        checkBestMode(md.pred[PRED_INTRA], depth);

        if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3)
        {
            md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom, qp);
            checkIntra(md.pred[PRED_INTRA_NxN], cuGeom, SIZE_NxN);
            checkBestMode(md.pred[PRED_INTRA_NxN], depth);
        }

        if (m_bTryLossless)
            tryLossless(cuGeom);

        if (mightSplit)
            addSplitFlagCost(*md.bestMode, cuGeom.depth);
    }

    // stop recursion if we reach the depth of previous analysis decision
    mightSplit &= !(bAlreadyDecided && bDecidedDepth) || split;

    if (mightSplit)
    {
        Mode* splitPred = &md.pred[PRED_SPLIT];
        splitPred->initCosts();
        CUData* splitCU = &splitPred->cu;
        splitCU->initSubCU(parentCTU, cuGeom, qp);

        uint32_t nextDepth = depth + 1;
        ModeDepth& nd = m_modeDepth[nextDepth];
        invalidateContexts(nextDepth);
        Entropy* nextContext = &m_rqt[depth].cur;
        int32_t nextQP = qp;

        for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
        {
            const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);
            if (childGeom.flags & CUGeom::PRESENT)
            {
                m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                m_rqt[nextDepth].cur.load(*nextContext);

                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));

                compressIntraCU(parentCTU, childGeom, nextQP);

                // Save best CU and pred data for this sub CU
                splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
                splitPred->addSubCosts(*nd.bestMode);
                nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);
                nextContext = &nd.bestMode->contexts;
            }
            else
            {
                /* record the depth of this non-present sub-CU */
                splitCU->setEmptyPart(childGeom, subPartIdx);

                /* Set depth of non-present CU to 0 to ensure that correct CU is fetched as reference to code deltaQP */
                if (bAlreadyDecided)
                    memset(parentCTU.m_cuDepth + childGeom.absPartIdx, 0, childGeom.numPartitions);
            }
        }
        nextContext->store(splitPred->contexts);
        if (mightNotSplit)
            addSplitFlagCost(*splitPred, cuGeom.depth);
        else
            updateModeCost(*splitPred);

        checkDQPForSplitPred(*splitPred, cuGeom);
        checkBestMode(*splitPred, depth);
    }

    if (m_param->bEnableRdRefine && depth <= m_slice->m_pps->maxCuDQPDepth)
    {
        int cuIdx = (cuGeom.childOffset - 1) / 3;
        cacheCost[cuIdx] = md.bestMode->rdCost;
    }

    /* Save Intra CUs TU depth only when analysis mode is OFF */
    if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4 && !m_param->analysisReuseMode)
    {
        CUData* ctu = md.bestMode->cu.m_encData->getPicCTU(parentCTU.m_cuAddr);
        int8_t maxTUDepth = -1;
        for (uint32_t i = 0; i < cuGeom.numPartitions; i++)
            maxTUDepth = X265_MAX(maxTUDepth, md.pred[PRED_INTRA].cu.m_tuDepth[i]);
        ctu->m_refTuDepth[cuGeom.geomRecurId] = maxTUDepth;
    }

    /* Copy best data to encData CTU and recon */
    md.bestMode->cu.copyToPic(depth);
    if (md.bestMode != &md.pred[PRED_SPLIT])
        md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, parentCTU.m_cuAddr, cuGeom.absPartIdx);
}

void Analysis::PMODE::processTasks(int workerThreadId)
{
#if DETAILED_CU_STATS
    int fe = master.m_modeDepth[cuGeom.depth].pred[PRED_2Nx2N].cu.m_encData->m_frameEncoderID;
    master.m_stats[fe].countPModeTasks++;
    ScopedElapsedTime pmodeTime(master.m_stats[fe].pmodeTime);
#endif
    ProfileScopeEvent(pmode);
    master.processPmode(*this, master.m_tld[workerThreadId].analysis);
}

/* process pmode jobs until none remain; may be called by the master thread or by
 * a bonded peer (slave) thread via pmodeTasks() */
void Analysis::processPmode(PMODE& pmode, Analysis& slave)
{
    /* acquire a mode task, else exit early */
    int task;
    pmode.m_lock.acquire();
    if (pmode.m_jobTotal > pmode.m_jobAcquired)
    {
        task = pmode.m_jobAcquired++;
        pmode.m_lock.release();
    }
    else
    {
        pmode.m_lock.release();
        return;
    }

    ModeDepth& md = m_modeDepth[pmode.cuGeom.depth];

    /* setup slave Analysis */
    if (&slave != this)
    {
        slave.m_slice = m_slice;
        slave.m_frame = m_frame;
        slave.m_param = m_param;
        slave.m_bChromaSa8d = m_param->rdLevel >= 3;
        slave.setLambdaFromQP(md.pred[PRED_2Nx2N].cu, m_rdCost.m_qp);
        slave.invalidateContexts(0);
        slave.m_rqt[pmode.cuGeom.depth].cur.load(m_rqt[pmode.cuGeom.depth].cur);
    }

    /* perform Mode task, repeat until no more work is available */
    do
    {
        uint32_t refMasks[2] = { 0, 0 };

        if (m_param->rdLevel <= 4)
        {
            switch (pmode.modes[task])
            {
            case PRED_INTRA:
                slave.checkIntraInInter(md.pred[PRED_INTRA], pmode.cuGeom);
                if (m_param->rdLevel > 2)
                    slave.encodeIntraInInter(md.pred[PRED_INTRA], pmode.cuGeom);
                break;

            case PRED_2Nx2N:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3];

                slave.checkInter_rd0_4(md.pred[PRED_2Nx2N], pmode.cuGeom, SIZE_2Nx2N, refMasks);
                if (m_slice->m_sliceType == B_SLICE)
                    slave.checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], pmode.cuGeom);
                break;

            case PRED_Nx2N:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[2]; /* left */
                refMasks[1] = m_splitRefIdx[1] | m_splitRefIdx[3]; /* right */

                slave.checkInter_rd0_4(md.pred[PRED_Nx2N], pmode.cuGeom, SIZE_Nx2N, refMasks);
                break;

            case PRED_2NxN:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1]; /* top */
                refMasks[1] = m_splitRefIdx[2] | m_splitRefIdx[3]; /* bot */

                slave.checkInter_rd0_4(md.pred[PRED_2NxN], pmode.cuGeom, SIZE_2NxN, refMasks);
                break;

            case PRED_2NxnU:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1]; /* 25% top */
                refMasks[1] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3]; /* 75% bot */

                slave.checkInter_rd0_4(md.pred[PRED_2NxnU], pmode.cuGeom, SIZE_2NxnU, refMasks);
                break;

            case PRED_2NxnD:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3]; /* 75% top */
                refMasks[1] = m_splitRefIdx[2] | m_splitRefIdx[3]; /* 25% bot */

                slave.checkInter_rd0_4(md.pred[PRED_2NxnD], pmode.cuGeom, SIZE_2NxnD, refMasks);
                break;

            case PRED_nLx2N:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[2]; /* 25% left */
                refMasks[1] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3]; /* 75% right */

                slave.checkInter_rd0_4(md.pred[PRED_nLx2N], pmode.cuGeom, SIZE_nLx2N, refMasks);
                break;

            case PRED_nRx2N:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3]; /* 75% left */
                refMasks[1] = m_splitRefIdx[1] | m_splitRefIdx[3]; /* 25% right */

                slave.checkInter_rd0_4(md.pred[PRED_nRx2N], pmode.cuGeom, SIZE_nRx2N, refMasks);
                break;

            default:
                X265_CHECK(0, "invalid job ID for parallel mode analysis\n");
                break;
            }
        }
        else
        {
            switch (pmode.modes[task])
            {
            case PRED_INTRA:
                slave.checkIntra(md.pred[PRED_INTRA], pmode.cuGeom, SIZE_2Nx2N);
                if (pmode.cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3)
                    slave.checkIntra(md.pred[PRED_INTRA_NxN], pmode.cuGeom, SIZE_NxN);
                break;

            case PRED_2Nx2N:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3];

                slave.checkInter_rd5_6(md.pred[PRED_2Nx2N], pmode.cuGeom, SIZE_2Nx2N, refMasks);
                md.pred[PRED_BIDIR].rdCost = MAX_INT64;
                if (m_slice->m_sliceType == B_SLICE)
                {
                    slave.checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], pmode.cuGeom);
                    if (md.pred[PRED_BIDIR].sa8dCost < MAX_INT64)
                        slave.encodeResAndCalcRdInterCU(md.pred[PRED_BIDIR], pmode.cuGeom);
                }
                break;

            case PRED_Nx2N:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[2]; /* left */
                refMasks[1] = m_splitRefIdx[1] | m_splitRefIdx[3]; /* right */

                slave.checkInter_rd5_6(md.pred[PRED_Nx2N], pmode.cuGeom, SIZE_Nx2N, refMasks);
                break;

            case PRED_2NxN:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1]; /* top */
                refMasks[1] = m_splitRefIdx[2] | m_splitRefIdx[3]; /* bot */

                slave.checkInter_rd5_6(md.pred[PRED_2NxN], pmode.cuGeom, SIZE_2NxN, refMasks);
                break;

            case PRED_2NxnU:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1]; /* 25% top */
                refMasks[1] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3]; /* 75% bot */

                slave.checkInter_rd5_6(md.pred[PRED_2NxnU], pmode.cuGeom, SIZE_2NxnU, refMasks);
                break;

            case PRED_2NxnD:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3]; /* 75% top */
                refMasks[1] = m_splitRefIdx[2] | m_splitRefIdx[3]; /* 25% bot */
                slave.checkInter_rd5_6(md.pred[PRED_2NxnD], pmode.cuGeom, SIZE_2NxnD, refMasks);
                break;

            case PRED_nLx2N:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[2]; /* 25% left */
                refMasks[1] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3]; /* 75% right */

                slave.checkInter_rd5_6(md.pred[PRED_nLx2N], pmode.cuGeom, SIZE_nLx2N, refMasks);
                break;

            case PRED_nRx2N:
                refMasks[0] = m_splitRefIdx[0] | m_splitRefIdx[1] | m_splitRefIdx[2] | m_splitRefIdx[3]; /* 75% left */
                refMasks[1] = m_splitRefIdx[1] | m_splitRefIdx[3]; /* 25% right */
                slave.checkInter_rd5_6(md.pred[PRED_nRx2N], pmode.cuGeom, SIZE_nRx2N, refMasks);
                break;

            default:
                X265_CHECK(0, "invalid job ID for parallel mode analysis\n");
                break;
            }
        }

        task = -1;
        pmode.m_lock.acquire();
        if (pmode.m_jobTotal > pmode.m_jobAcquired)
            task = pmode.m_jobAcquired++;
        pmode.m_lock.release();
    }
    while (task >= 0);
}

uint32_t Analysis::compressInterCU_dist(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp)
{
    uint32_t depth = cuGeom.depth;
    uint32_t cuAddr = parentCTU.m_cuAddr;
    ModeDepth& md = m_modeDepth[depth];
    md.bestMode = NULL;

    bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);
    bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);
    uint32_t minDepth = m_param->rdLevel <= 4 ? topSkipMinDepth(parentCTU, cuGeom) : 0;
    uint32_t splitRefs[4] = { 0, 0, 0, 0 };

    X265_CHECK(m_param->rdLevel >= 2, "compressInterCU_dist does not support RD 0 or 1\n");

    PMODE pmode(*this, cuGeom);

    if (mightNotSplit && depth >= minDepth)
    {
        /* Initialize all prediction CUs based on parentCTU */
        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);

        if (m_param->rdLevel <= 4)
            checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
        else
            checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
    }

    bool bNoSplit = false;
    bool splitIntra = true;
    if (md.bestMode)
    {
        bNoSplit = md.bestMode->cu.isSkipped(0);
        if (mightSplit && depth && depth >= minDepth && !bNoSplit && m_param->rdLevel <= 4)
            bNoSplit = recursionDepthCheck(parentCTU, cuGeom, *md.bestMode);
    }

    if (mightSplit && !bNoSplit)
    {
        Mode* splitPred = &md.pred[PRED_SPLIT];
        splitPred->initCosts();
        CUData* splitCU = &splitPred->cu;
        splitCU->initSubCU(parentCTU, cuGeom, qp);

        uint32_t nextDepth = depth + 1;
        ModeDepth& nd = m_modeDepth[nextDepth];
        invalidateContexts(nextDepth);
        Entropy* nextContext = &m_rqt[depth].cur;
        int nextQP = qp;
        splitIntra = false;

        for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
        {
            const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);
            if (childGeom.flags & CUGeom::PRESENT)
            {
                m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                m_rqt[nextDepth].cur.load(*nextContext);

                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));

                splitRefs[subPartIdx] = compressInterCU_dist(parentCTU, childGeom, nextQP);

                // Save best CU and pred data for this sub CU
                splitIntra |= nd.bestMode->cu.isIntra(0);
                splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
                splitPred->addSubCosts(*nd.bestMode);

                nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);
                nextContext = &nd.bestMode->contexts;
            }
            else
                splitCU->setEmptyPart(childGeom, subPartIdx);
        }
        nextContext->store(splitPred->contexts);

        if (mightNotSplit)
            addSplitFlagCost(*splitPred, cuGeom.depth);
        else
            updateModeCost(*splitPred);

        checkDQPForSplitPred(*splitPred, cuGeom);
    }

    if (mightNotSplit && depth >= minDepth)
    {
        int bTryAmp = m_slice->m_sps->maxAMPDepth > depth;
        int bTryIntra = (m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames) && (!m_param->limitReferences || splitIntra) && (cuGeom.log2CUSize != MAX_LOG2_CU_SIZE);

        if (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth && m_slice->m_pps->maxCuDQPDepth != 0)
            setLambdaFromQP(parentCTU, qp);

        if (bTryIntra)
        {
            md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
            if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3 && m_param->rdLevel >= 5)
                md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom, qp);
            pmode.modes[pmode.m_jobTotal++] = PRED_INTRA;
        }
        md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_2Nx2N;
        md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);
        if (m_param->bEnableRectInter)
        {
            md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_2NxN;
            md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_Nx2N;
        }
        if (bTryAmp)
        {
            md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_2NxnU;
            md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_2NxnD;
            md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_nLx2N;
            md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp); pmode.modes[pmode.m_jobTotal++] = PRED_nRx2N;
        }

        m_splitRefIdx[0] = splitRefs[0]; m_splitRefIdx[1] = splitRefs[1]; m_splitRefIdx[2] = splitRefs[2]; m_splitRefIdx[3] = splitRefs[3];

        pmode.tryBondPeers(*m_frame->m_encData->m_jobProvider, pmode.m_jobTotal);

        /* participate in processing jobs, until all are distributed */
        processPmode(pmode, *this);

        /* the master worker thread (this one) does merge analysis. By doing
         * merge after all the other jobs are at least started, we usually avoid
         * blocking on another thread */

        if (m_param->rdLevel <= 4)
        {
            {
                ProfileCUScope(parentCTU, pmodeBlockTime, countPModeMasters);
                pmode.waitForExit();
            }

            /* select best inter mode based on sa8d cost */
            Mode *bestInter = &md.pred[PRED_2Nx2N];

            if (m_param->bEnableRectInter)
            {
                if (md.pred[PRED_Nx2N].sa8dCost < bestInter->sa8dCost)
                    bestInter = &md.pred[PRED_Nx2N];
                if (md.pred[PRED_2NxN].sa8dCost < bestInter->sa8dCost)
                    bestInter = &md.pred[PRED_2NxN];
            }

            if (bTryAmp)
            {
                if (md.pred[PRED_2NxnU].sa8dCost < bestInter->sa8dCost)
                    bestInter = &md.pred[PRED_2NxnU];
                if (md.pred[PRED_2NxnD].sa8dCost < bestInter->sa8dCost)
                    bestInter = &md.pred[PRED_2NxnD];
                if (md.pred[PRED_nLx2N].sa8dCost < bestInter->sa8dCost)
                    bestInter = &md.pred[PRED_nLx2N];
                if (md.pred[PRED_nRx2N].sa8dCost < bestInter->sa8dCost)
                    bestInter = &md.pred[PRED_nRx2N];
            }

            if (m_param->rdLevel > 2)
            {
                /* RD selection between merge, inter, bidir and intra */
                if (!m_bChromaSa8d && (m_csp != X265_CSP_I400)) /* When m_bChromaSa8d is enabled, chroma MC has already been done */
                {
                    uint32_t numPU = bestInter->cu.getNumPartInter(0);
                    for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
                    {
                        PredictionUnit pu(bestInter->cu, cuGeom, puIdx);
                        motionCompensation(bestInter->cu, pu, bestInter->predYuv, false, true);
                    }
                }
                encodeResAndCalcRdInterCU(*bestInter, cuGeom);
                checkBestMode(*bestInter, depth);

                /* If BIDIR is available and within 17/16 of best inter option, choose by RDO */
                if (m_slice->m_sliceType == B_SLICE && md.pred[PRED_BIDIR].sa8dCost != MAX_INT64 &&
                    md.pred[PRED_BIDIR].sa8dCost * 16 <= bestInter->sa8dCost * 17)
                {
                    encodeResAndCalcRdInterCU(md.pred[PRED_BIDIR], cuGeom);
                    checkBestMode(md.pred[PRED_BIDIR], depth);
                }

                if (bTryIntra)
                    checkBestMode(md.pred[PRED_INTRA], depth);
            }
            else /* m_param->rdLevel == 2 */
            {
                if (!md.bestMode || bestInter->sa8dCost < md.bestMode->sa8dCost)
                    md.bestMode = bestInter;

                if (m_slice->m_sliceType == B_SLICE && md.pred[PRED_BIDIR].sa8dCost < md.bestMode->sa8dCost)
                    md.bestMode = &md.pred[PRED_BIDIR];

                if (bTryIntra && md.pred[PRED_INTRA].sa8dCost < md.bestMode->sa8dCost)
                {
                    md.bestMode = &md.pred[PRED_INTRA];
                    encodeIntraInInter(*md.bestMode, cuGeom);
                }
                else if (!md.bestMode->cu.m_mergeFlag[0])
                {
                    /* finally code the best mode selected from SA8D costs */
                    uint32_t numPU = md.bestMode->cu.getNumPartInter(0);
                    for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
                    {
                        PredictionUnit pu(md.bestMode->cu, cuGeom, puIdx);
                        motionCompensation(md.bestMode->cu, pu, md.bestMode->predYuv, false, true);
                    }
                    encodeResAndCalcRdInterCU(*md.bestMode, cuGeom);
                }
            }
        }
        else
        {
            {
                ProfileCUScope(parentCTU, pmodeBlockTime, countPModeMasters);
                pmode.waitForExit();
            }

            checkBestMode(md.pred[PRED_2Nx2N], depth);
            if (m_slice->m_sliceType == B_SLICE && md.pred[PRED_BIDIR].sa8dCost < MAX_INT64)
                checkBestMode(md.pred[PRED_BIDIR], depth);

            if (m_param->bEnableRectInter)
            {
                checkBestMode(md.pred[PRED_Nx2N], depth);
                checkBestMode(md.pred[PRED_2NxN], depth);
            }

            if (bTryAmp)
            {
                checkBestMode(md.pred[PRED_2NxnU], depth);
                checkBestMode(md.pred[PRED_2NxnD], depth);
                checkBestMode(md.pred[PRED_nLx2N], depth);
                checkBestMode(md.pred[PRED_nRx2N], depth);
            }

            if (bTryIntra)
            {
                checkBestMode(md.pred[PRED_INTRA], depth);
                if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3)
                    checkBestMode(md.pred[PRED_INTRA_NxN], depth);
            }
        }

        if (m_bTryLossless)
            tryLossless(cuGeom);

        if (mightSplit)
            addSplitFlagCost(*md.bestMode, cuGeom.depth);
    }

    /* compare split RD cost against best cost */
    if (mightSplit && !bNoSplit)
        checkBestMode(md.pred[PRED_SPLIT], depth);

    /* determine which motion references the parent CU should search */
    uint32_t refMask;
    if (!(m_param->limitReferences & X265_REF_LIMIT_DEPTH))
        refMask = 0;
    else if (md.bestMode == &md.pred[PRED_SPLIT])
        refMask = splitRefs[0] | splitRefs[1] | splitRefs[2] | splitRefs[3];
    else
    {
        /* use best merge/inter mode, in case of intra use 2Nx2N inter references */
        CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;
        uint32_t numPU = cu.getNumPartInter(0);
        refMask = 0;
        for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))
            refMask |= cu.getBestRefIdx(subPartIdx);
    }

    if (mightNotSplit)
    {
        /* early-out statistics */
        FrameData& curEncData = *m_frame->m_encData;
        FrameData::RCStatCU& cuStat = curEncData.m_cuStat[parentCTU.m_cuAddr];
        uint64_t temp = cuStat.avgCost[depth] * cuStat.count[depth];
        cuStat.count[depth] += 1;
        cuStat.avgCost[depth] = (temp + md.bestMode->rdCost) / cuStat.count[depth];
    }

    /* Copy best data to encData CTU and recon */
    md.bestMode->cu.copyToPic(depth);
    md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, cuAddr, cuGeom.absPartIdx);

    return refMask;
}

SplitData Analysis::compressInterCU_rd0_4(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp)
{
    if (parentCTU.m_vbvAffected && calculateQpforCuSize(parentCTU, cuGeom, 1))
        return compressInterCU_rd5_6(parentCTU, cuGeom, qp);

    uint32_t depth = cuGeom.depth;
    uint32_t cuAddr = parentCTU.m_cuAddr;
    ModeDepth& md = m_modeDepth[depth];
    md.bestMode = NULL;

    if (m_param->searchMethod == X265_SEA)
    {
        int numPredDir = m_slice->isInterP() ? 1 : 2;
        int offset = (int)(m_frame->m_reconPic->m_cuOffsetY[parentCTU.m_cuAddr] + m_frame->m_reconPic->m_buOffsetY[cuGeom.absPartIdx]);
        for (int list = 0; list < numPredDir; list++)
            for (int i = 0; i < m_frame->m_encData->m_slice->m_numRefIdx[list]; i++)
                for (int planes = 0; planes < INTEGRAL_PLANE_NUM; planes++)
                    m_modeDepth[depth].fencYuv.m_integral[list][i][planes] = m_frame->m_encData->m_slice->m_refFrameList[list][i]->m_encData->m_meIntegral[planes] + offset;
    }

    PicYuv& reconPic = *m_frame->m_reconPic;

    bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);
    bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);
    uint32_t minDepth = topSkipMinDepth(parentCTU, cuGeom);
    bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;
    bool skipModes = false; /* Skip any remaining mode analyses at current depth */
    bool skipRecursion = false; /* Skip recursion */
    bool splitIntra = true;
    bool skipRectAmp = false;
    bool chooseMerge = false;
    bool bCtuInfoCheck = false;
    int sameContentRef = 0;

    if (m_evaluateInter == 1)
    {
        skipRectAmp = !!md.bestMode;
        mightSplit &= false;
        minDepth = depth;
    }

    if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)
        m_maxTUDepth = loadTUDepth(cuGeom, parentCTU);

    SplitData splitData[4];
    splitData[0].initSplitCUData();
    splitData[1].initSplitCUData();
    splitData[2].initSplitCUData();
    splitData[3].initSplitCUData();

    // avoid uninitialize value in below reference
    if (m_param->limitModes)
    {
        md.pred[PRED_2Nx2N].bestME[0][0].mvCost = 0; // L0
        md.pred[PRED_2Nx2N].bestME[0][1].mvCost = 0; // L1
        md.pred[PRED_2Nx2N].sa8dCost = 0;
    }

    if (m_param->bCTUInfo && depth <= parentCTU.m_cuDepth[cuGeom.absPartIdx])
    {
        if (bDecidedDepth && m_additionalCtuInfo[cuGeom.absPartIdx])
            sameContentRef = findSameContentRefCount(parentCTU, cuGeom);
        if (depth < parentCTU.m_cuDepth[cuGeom.absPartIdx])
        {
            mightNotSplit &= bDecidedDepth;
            bCtuInfoCheck = skipRecursion = false;
            skipModes = true;
        }
        else if (mightNotSplit && bDecidedDepth)
        {
            if (m_additionalCtuInfo[cuGeom.absPartIdx])
            {
                bCtuInfoCheck = skipRecursion = true;
                md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
                md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
                checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
                if (!sameContentRef)
                {
                    if ((m_param->bCTUInfo & 2) && (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth))
                    {
                        qp -= int32_t(0.04 * qp);
                        setLambdaFromQP(parentCTU, qp);
                    }
                    if (m_param->bCTUInfo & 4)
                        skipModes = false;
                }
                if (sameContentRef || (!sameContentRef && !(m_param->bCTUInfo & 4)))
                {
                    if (m_param->rdLevel)
                        skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0);
                    if ((m_param->bCTUInfo & 4) && sameContentRef)
                        skipModes = md.bestMode && true;
                }
            }
            else
            {
                md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
                md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
                checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
                if (m_param->rdLevel)
                    skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0);
            }
            mightSplit &= !bDecidedDepth;
        }
    }
    if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
    {
        if (mightNotSplit && depth == m_reuseDepth[cuGeom.absPartIdx])
        {
            if (m_reuseModes[cuGeom.absPartIdx] == MODE_SKIP)
            {
                md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
                md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
                checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);

                skipRecursion = !!m_param->bEnableRecursionSkip && md.bestMode;
                if (m_param->rdLevel)
                    skipModes = m_param->bEnableEarlySkip && md.bestMode;
            }
            if (m_param->analysisReuseLevel > 4 && m_reusePartSize[cuGeom.absPartIdx] == SIZE_2Nx2N)
            {
                if (m_reuseModes[cuGeom.absPartIdx] != MODE_INTRA  && m_reuseModes[cuGeom.absPartIdx] != 4)
                {
                    skipRectAmp = true && !!md.bestMode;
                    chooseMerge = !!m_reuseMergeFlag[cuGeom.absPartIdx] && !!md.bestMode;
                }
            }
        }
    }
    if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead && m_multipassAnalysis)
    {
        if (mightNotSplit && depth == m_multipassDepth[cuGeom.absPartIdx])
        {
            if (m_multipassModes[cuGeom.absPartIdx] == MODE_SKIP)
            {
                md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
                md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
                checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);

                skipRecursion = !!m_param->bEnableRecursionSkip && md.bestMode;
                if (m_param->rdLevel)
                    skipModes = m_param->bEnableEarlySkip && md.bestMode;
            }
        }
    }

    /* Step 1. Evaluate Merge/Skip candidates for likely early-outs, if skip mode was not set above */
    if (mightNotSplit && depth >= minDepth && !md.bestMode && !bCtuInfoCheck) /* TODO: Re-evaluate if analysis load/save still works */
    {
        /* Compute Merge Cost */
        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
        checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
        if (m_param->rdLevel)
            skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0); // TODO: sa8d threshold per depth
    }

    if (md.bestMode && m_param->bEnableRecursionSkip && !bCtuInfoCheck)
    {
        skipRecursion = md.bestMode->cu.isSkipped(0);
        if (mightSplit && depth >= minDepth && !skipRecursion)
        {
            if (depth)
                skipRecursion = recursionDepthCheck(parentCTU, cuGeom, *md.bestMode);
            if (m_bHD && !skipRecursion && m_param->rdLevel == 2 && md.fencYuv.m_size != MAX_CU_SIZE)
                skipRecursion = complexityCheckCU(*md.bestMode);
        }
    }

    /* Step 2. Evaluate each of the 4 split sub-blocks in series */
    if (mightSplit && !skipRecursion)
    {
        if (bCtuInfoCheck && m_param->bCTUInfo & 2)
            qp = int((1 / 0.96) * qp + 0.5);
        Mode* splitPred = &md.pred[PRED_SPLIT];
        splitPred->initCosts();
        CUData* splitCU = &splitPred->cu;
        splitCU->initSubCU(parentCTU, cuGeom, qp);

        uint32_t nextDepth = depth + 1;
        ModeDepth& nd = m_modeDepth[nextDepth];
        invalidateContexts(nextDepth);
        Entropy* nextContext = &m_rqt[depth].cur;
        int nextQP = qp;
        splitIntra = false;

        for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
        {
            const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);
            if (childGeom.flags & CUGeom::PRESENT)
            {
                m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                m_rqt[nextDepth].cur.load(*nextContext);

                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));

                splitData[subPartIdx] = compressInterCU_rd0_4(parentCTU, childGeom, nextQP);

                // Save best CU and pred data for this sub CU
                splitIntra |= nd.bestMode->cu.isIntra(0);
                splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
                splitPred->addSubCosts(*nd.bestMode);

                if (m_param->rdLevel)
                    nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);
                else
                    nd.bestMode->predYuv.copyToPartYuv(splitPred->predYuv, childGeom.numPartitions * subPartIdx);
                if (m_param->rdLevel > 1)
                    nextContext = &nd.bestMode->contexts;
            }
            else
                splitCU->setEmptyPart(childGeom, subPartIdx);
        }
        nextContext->store(splitPred->contexts);

        if (mightNotSplit)
            addSplitFlagCost(*splitPred, cuGeom.depth);
        else if (m_param->rdLevel > 1)
            updateModeCost(*splitPred);
        else
            splitPred->sa8dCost = m_rdCost.calcRdSADCost((uint32_t)splitPred->distortion, splitPred->sa8dBits);
    }

    /* Split CUs
     *   0  1
     *   2  3 */
    uint32_t allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;
    /* Step 3. Evaluate ME (2Nx2N, rect, amp) and intra modes at current depth */
    if (mightNotSplit && (depth >= minDepth || (m_param->bCTUInfo && !md.bestMode)))
    {
        if (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth && m_slice->m_pps->maxCuDQPDepth != 0)
            setLambdaFromQP(parentCTU, qp);

        if (!skipModes)
        {
            uint32_t refMasks[2];
            refMasks[0] = allSplitRefs;
            md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
            checkInter_rd0_4(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);

            if (m_param->limitReferences & X265_REF_LIMIT_CU)
            {
                CUData& cu = md.pred[PRED_2Nx2N].cu;
                uint32_t refMask = cu.getBestRefIdx(0);
                allSplitRefs = splitData[0].splitRefs = splitData[1].splitRefs = splitData[2].splitRefs = splitData[3].splitRefs = refMask;
            }

            if (m_slice->m_sliceType == B_SLICE)
            {
                md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);
                checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], cuGeom);
            }

            Mode *bestInter = &md.pred[PRED_2Nx2N];
            if (!skipRectAmp)
            {
                if (m_param->bEnableRectInter)
                {
                    uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;
                    uint32_t threshold_2NxN, threshold_Nx2N;

                    if (m_slice->m_sliceType == P_SLICE)
                    {
                        threshold_2NxN = splitData[0].mvCost[0] + splitData[1].mvCost[0];
                        threshold_Nx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];
                    }
                    else
                    {
                        threshold_2NxN = (splitData[0].mvCost[0] + splitData[1].mvCost[0]
                                       + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;
                        threshold_Nx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]
                                       + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;
                    }

                    int try_2NxN_first = threshold_2NxN < threshold_Nx2N;
                    if (try_2NxN_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxN)
                    {
                        refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */
                        refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */
                        md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkInter_rd0_4(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);
                        if (md.pred[PRED_2NxN].sa8dCost < bestInter->sa8dCost)
                            bestInter = &md.pred[PRED_2NxN];
                    }

                    if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_Nx2N)
                    {
                        refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* left */
                        refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* right */
                        md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkInter_rd0_4(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N, refMasks);
                        if (md.pred[PRED_Nx2N].sa8dCost < bestInter->sa8dCost)
                            bestInter = &md.pred[PRED_Nx2N];
                    }

                    if (!try_2NxN_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxN)
                    {
                        refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */
                        refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */
                        md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkInter_rd0_4(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);
                        if (md.pred[PRED_2NxN].sa8dCost < bestInter->sa8dCost)
                            bestInter = &md.pred[PRED_2NxN];
                    }
                }

                if (m_slice->m_sps->maxAMPDepth > depth)
                {
                    uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;
                    uint32_t threshold_2NxnU, threshold_2NxnD, threshold_nLx2N, threshold_nRx2N;

                    if (m_slice->m_sliceType == P_SLICE)
                    {
                        threshold_2NxnU = splitData[0].mvCost[0] + splitData[1].mvCost[0];
                        threshold_2NxnD = splitData[2].mvCost[0] + splitData[3].mvCost[0];

                        threshold_nLx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];
                        threshold_nRx2N = splitData[1].mvCost[0] + splitData[3].mvCost[0];
                    }
                    else
                    {
                        threshold_2NxnU = (splitData[0].mvCost[0] + splitData[1].mvCost[0]
                                         + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;
                        threshold_2NxnD = (splitData[2].mvCost[0] + splitData[3].mvCost[0]
                                         + splitData[2].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;

                        threshold_nLx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]
                                        + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;
                        threshold_nRx2N = (splitData[1].mvCost[0] + splitData[3].mvCost[0]
                                        + splitData[1].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;
                    }

                    bool bHor = false, bVer = false;
                    if (bestInter->cu.m_partSize[0] == SIZE_2NxN)
                        bHor = true;
                    else if (bestInter->cu.m_partSize[0] == SIZE_Nx2N)
                        bVer = true;
                    else if (bestInter->cu.m_partSize[0] == SIZE_2Nx2N &&
                        md.bestMode && md.bestMode->cu.getQtRootCbf(0))
                    {
                        bHor = true;
                        bVer = true;
                    }

                    if (bHor)
                    {
                        int try_2NxnD_first = threshold_2NxnD < threshold_2NxnU;
                        if (try_2NxnD_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnD)
                        {
                            refMasks[0] = allSplitRefs;                                    /* 75% top */
                            refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */
                            md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd0_4(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);
                            if (md.pred[PRED_2NxnD].sa8dCost < bestInter->sa8dCost)
                                bestInter = &md.pred[PRED_2NxnD];
                        }

                        if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnU)
                        {
                            refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* 25% top */
                            refMasks[1] = allSplitRefs;                                    /* 75% bot */
                            md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd0_4(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU, refMasks);
                            if (md.pred[PRED_2NxnU].sa8dCost < bestInter->sa8dCost)
                                bestInter = &md.pred[PRED_2NxnU];
                        }

                        if (!try_2NxnD_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_2NxnD)
                        {
                            refMasks[0] = allSplitRefs;                                    /* 75% top */
                            refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */
                            md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd0_4(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);
                            if (md.pred[PRED_2NxnD].sa8dCost < bestInter->sa8dCost)
                                bestInter = &md.pred[PRED_2NxnD];
                        }
                    }
                    if (bVer)
                    {
                        int try_nRx2N_first = threshold_nRx2N < threshold_nLx2N;
                        if (try_nRx2N_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nRx2N)
                        {
                            refMasks[0] = allSplitRefs;                                    /* 75% left  */
                            refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */
                            md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd0_4(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);
                            if (md.pred[PRED_nRx2N].sa8dCost < bestInter->sa8dCost)
                                bestInter = &md.pred[PRED_nRx2N];
                        }

                        if (splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nLx2N)
                        {
                            refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* 25% left  */
                            refMasks[1] = allSplitRefs;                                    /* 75% right */
                            md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd0_4(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N, refMasks);
                            if (md.pred[PRED_nLx2N].sa8dCost < bestInter->sa8dCost)
                                bestInter = &md.pred[PRED_nLx2N];
                        }

                        if (!try_nRx2N_first && splitCost < md.pred[PRED_2Nx2N].sa8dCost + threshold_nRx2N)
                        {
                            refMasks[0] = allSplitRefs;                                    /* 75% left  */
                            refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */
                            md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd0_4(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);
                            if (md.pred[PRED_nRx2N].sa8dCost < bestInter->sa8dCost)
                                bestInter = &md.pred[PRED_nRx2N];
                        }
                    }
                }
            }
            bool bTryIntra = (m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames) && cuGeom.log2CUSize != MAX_LOG2_CU_SIZE && !((m_param->bCTUInfo & 4) && bCtuInfoCheck);
            if (m_param->rdLevel >= 3)
            {
                /* Calculate RD cost of best inter option */
                if ((!m_bChromaSa8d && (m_csp != X265_CSP_I400)) || (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)) /* When m_bChromaSa8d is enabled, chroma MC has already been done */
                {
                    uint32_t numPU = bestInter->cu.getNumPartInter(0);
                    for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
                    {
                        PredictionUnit pu(bestInter->cu, cuGeom, puIdx);
                        motionCompensation(bestInter->cu, pu, bestInter->predYuv, false, true);
                    }
                }

                if (!chooseMerge)
                {
                    encodeResAndCalcRdInterCU(*bestInter, cuGeom);
                    checkBestMode(*bestInter, depth);

                    /* If BIDIR is available and within 17/16 of best inter option, choose by RDO */
                    if (m_slice->m_sliceType == B_SLICE && md.pred[PRED_BIDIR].sa8dCost != MAX_INT64 &&
                        md.pred[PRED_BIDIR].sa8dCost * 16 <= bestInter->sa8dCost * 17)
                    {
                        uint32_t numPU = md.pred[PRED_BIDIR].cu.getNumPartInter(0);
                        if (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)
                            for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
                            {
                                PredictionUnit pu(md.pred[PRED_BIDIR].cu, cuGeom, puIdx);
                                motionCompensation(md.pred[PRED_BIDIR].cu, pu, md.pred[PRED_BIDIR].predYuv, true, true);
                            }
                        encodeResAndCalcRdInterCU(md.pred[PRED_BIDIR], cuGeom);
                        checkBestMode(md.pred[PRED_BIDIR], depth);
                    }
                }

                if ((bTryIntra && md.bestMode->cu.getQtRootCbf(0)) ||
                    md.bestMode->sa8dCost == MAX_INT64)
                {
                    if (!m_param->limitReferences || splitIntra)
                    {
                        ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);
                        md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkIntraInInter(md.pred[PRED_INTRA], cuGeom);
                        encodeIntraInInter(md.pred[PRED_INTRA], cuGeom);
                        checkBestMode(md.pred[PRED_INTRA], depth);
                    }
                    else
                    {
                        ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);
                    }
                }
            }
            else
            {
                /* SA8D choice between merge/skip, inter, bidir, and intra */
                if (!md.bestMode || bestInter->sa8dCost < md.bestMode->sa8dCost)
                    md.bestMode = bestInter;

                if (m_slice->m_sliceType == B_SLICE &&
                    md.pred[PRED_BIDIR].sa8dCost < md.bestMode->sa8dCost)
                    md.bestMode = &md.pred[PRED_BIDIR];

                if (bTryIntra || md.bestMode->sa8dCost == MAX_INT64)
                {
                    if (!m_param->limitReferences || splitIntra)
                    {
                        ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);
                        md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkIntraInInter(md.pred[PRED_INTRA], cuGeom);
                        if (md.pred[PRED_INTRA].sa8dCost < md.bestMode->sa8dCost)
                            md.bestMode = &md.pred[PRED_INTRA];
                    }
                    else
                    {
                        ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);
                    }
                }

                /* finally code the best mode selected by SA8D costs:
                 * RD level 2 - fully encode the best mode
                 * RD level 1 - generate recon pixels
                 * RD level 0 - generate chroma prediction */
                if (md.bestMode->cu.m_mergeFlag[0] && md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N)
                {
                    /* prediction already generated for this CU, and if rd level
                     * is not 0, it is already fully encoded */
                }
                else if (md.bestMode->cu.isInter(0))
                {
                    uint32_t numPU = md.bestMode->cu.getNumPartInter(0);
                    if (m_csp != X265_CSP_I400)
                    {
                        for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
                        {
                            PredictionUnit pu(md.bestMode->cu, cuGeom, puIdx);
                            motionCompensation(md.bestMode->cu, pu, md.bestMode->predYuv, false, true);
                        }
                    }
                    if (m_param->rdLevel == 2)
                        encodeResAndCalcRdInterCU(*md.bestMode, cuGeom);
                    else if (m_param->rdLevel == 1)
                    {
                        /* generate recon pixels with no rate distortion considerations */
                        CUData& cu = md.bestMode->cu;

                        uint32_t tuDepthRange[2];
                        cu.getInterTUQtDepthRange(tuDepthRange, 0);
                        m_rqt[cuGeom.depth].tmpResiYuv.subtract(*md.bestMode->fencYuv, md.bestMode->predYuv, cuGeom.log2CUSize, m_frame->m_fencPic->m_picCsp);
                        residualTransformQuantInter(*md.bestMode, cuGeom, 0, 0, tuDepthRange);
                        if (cu.getQtRootCbf(0))
                            md.bestMode->reconYuv.addClip(md.bestMode->predYuv, m_rqt[cuGeom.depth].tmpResiYuv, cu.m_log2CUSize[0], m_frame->m_fencPic->m_picCsp);
                        else
                        {
                            md.bestMode->reconYuv.copyFromYuv(md.bestMode->predYuv);
                            if (cu.m_mergeFlag[0] && cu.m_partSize[0] == SIZE_2Nx2N)
                                cu.setPredModeSubParts(MODE_SKIP);
                        }
                    }
                }
                else
                {
                    if (m_param->rdLevel == 2)
                        encodeIntraInInter(*md.bestMode, cuGeom);
                    else if (m_param->rdLevel == 1)
                    {
                        /* generate recon pixels with no rate distortion considerations */
                        CUData& cu = md.bestMode->cu;

                        uint32_t tuDepthRange[2];
                        cu.getIntraTUQtDepthRange(tuDepthRange, 0);

                        residualTransformQuantIntra(*md.bestMode, cuGeom, 0, 0, tuDepthRange);
                        if (m_csp != X265_CSP_I400)
                        {
                            getBestIntraModeChroma(*md.bestMode, cuGeom);
                            residualQTIntraChroma(*md.bestMode, cuGeom, 0, 0);
                        }
                        md.bestMode->reconYuv.copyFromPicYuv(reconPic, cu.m_cuAddr, cuGeom.absPartIdx); // TODO:
                    }
                }
            }
        } // !earlyskip

        if (m_bTryLossless)
            tryLossless(cuGeom);

        if (mightSplit)
            addSplitFlagCost(*md.bestMode, cuGeom.depth);
    }

    if (mightSplit && !skipRecursion)
    {
        Mode* splitPred = &md.pred[PRED_SPLIT];
        if (!md.bestMode)
            md.bestMode = splitPred;
        else if (m_param->rdLevel > 1)
            checkBestMode(*splitPred, cuGeom.depth);
        else if (splitPred->sa8dCost < md.bestMode->sa8dCost)
            md.bestMode = splitPred;

        checkDQPForSplitPred(*md.bestMode, cuGeom);
    }

    /* determine which motion references the parent CU should search */
    SplitData splitCUData;
    splitCUData.initSplitCUData();

    if (m_param->limitReferences & X265_REF_LIMIT_DEPTH)
    {
        if (md.bestMode == &md.pred[PRED_SPLIT])
            splitCUData.splitRefs = allSplitRefs;
        else 
        {
            /* use best merge/inter mode, in case of intra use 2Nx2N inter references */
            CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;
            uint32_t numPU = cu.getNumPartInter(0);
            for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))
                splitCUData.splitRefs |= cu.getBestRefIdx(subPartIdx);
        }
    }

    if (m_param->limitModes)
    {
        splitCUData.mvCost[0] = md.pred[PRED_2Nx2N].bestME[0][0].mvCost; // L0
        splitCUData.mvCost[1] = md.pred[PRED_2Nx2N].bestME[0][1].mvCost; // L1
        splitCUData.sa8dCost    = md.pred[PRED_2Nx2N].sa8dCost;
    }
    
    if (mightNotSplit && md.bestMode->cu.isSkipped(0))
    {
        FrameData& curEncData = *m_frame->m_encData;
        FrameData::RCStatCU& cuStat = curEncData.m_cuStat[parentCTU.m_cuAddr];
        uint64_t temp = cuStat.avgCost[depth] * cuStat.count[depth];
        cuStat.count[depth] += 1;
        cuStat.avgCost[depth] = (temp + md.bestMode->rdCost) / cuStat.count[depth];
    }

    /* Copy best data to encData CTU and recon */
    md.bestMode->cu.copyToPic(depth);
    if (m_param->rdLevel)
        md.bestMode->reconYuv.copyToPicYuv(reconPic, cuAddr, cuGeom.absPartIdx);

    if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)
    {
        if (mightNotSplit)
        {
            CUData* ctu = md.bestMode->cu.m_encData->getPicCTU(parentCTU.m_cuAddr);
            int8_t maxTUDepth = -1;
            for (uint32_t i = 0; i < cuGeom.numPartitions; i++)
                maxTUDepth = X265_MAX(maxTUDepth, md.bestMode->cu.m_tuDepth[i]);
            ctu->m_refTuDepth[cuGeom.geomRecurId] = maxTUDepth;
        }
    }

    return splitCUData;
}

SplitData Analysis::compressInterCU_rd5_6(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp)
{
    if (parentCTU.m_vbvAffected && !calculateQpforCuSize(parentCTU, cuGeom, 1))
        return compressInterCU_rd0_4(parentCTU, cuGeom, qp);

    uint32_t depth = cuGeom.depth;
    ModeDepth& md = m_modeDepth[depth];
    md.bestMode = NULL;

    if (m_param->searchMethod == X265_SEA)
    {
        int numPredDir = m_slice->isInterP() ? 1 : 2;
        int offset = (int)(m_frame->m_reconPic->m_cuOffsetY[parentCTU.m_cuAddr] + m_frame->m_reconPic->m_buOffsetY[cuGeom.absPartIdx]);
        for (int list = 0; list < numPredDir; list++)
            for (int i = 0; i < m_frame->m_encData->m_slice->m_numRefIdx[list]; i++)
                for (int planes = 0; planes < INTEGRAL_PLANE_NUM; planes++)
                    m_modeDepth[depth].fencYuv.m_integral[list][i][planes] = m_frame->m_encData->m_slice->m_refFrameList[list][i]->m_encData->m_meIntegral[planes] + offset;
    }

    bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);
    bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);
    bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;
    bool skipRecursion = false;
    bool skipModes = false;
    bool splitIntra = true;
    bool skipRectAmp = false;
    bool bCtuInfoCheck = false;
    int sameContentRef = 0;

    if (m_evaluateInter == 1)
    {
        skipRectAmp = !!md.bestMode;
        mightSplit &= false;
    }

    // avoid uninitialize value in below reference
    if (m_param->limitModes)
    {
        md.pred[PRED_2Nx2N].bestME[0][0].mvCost = 0; // L0
        md.pred[PRED_2Nx2N].bestME[0][1].mvCost = 0; // L1
        md.pred[PRED_2Nx2N].rdCost = 0;
    }

    if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)
        m_maxTUDepth = loadTUDepth(cuGeom, parentCTU);

    SplitData splitData[4];
    splitData[0].initSplitCUData();
    splitData[1].initSplitCUData();
    splitData[2].initSplitCUData();
    splitData[3].initSplitCUData();
    uint32_t allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;
    uint32_t refMasks[2];
    if (m_param->bCTUInfo && depth <= parentCTU.m_cuDepth[cuGeom.absPartIdx])
    {
        if (bDecidedDepth && m_additionalCtuInfo[cuGeom.absPartIdx])
            sameContentRef = findSameContentRefCount(parentCTU, cuGeom);
        if (depth < parentCTU.m_cuDepth[cuGeom.absPartIdx])
        {
            mightNotSplit &= bDecidedDepth;
            bCtuInfoCheck = skipRecursion = false;
            skipModes = true;
        }
        else if (mightNotSplit && bDecidedDepth)
        {
            if (m_additionalCtuInfo[cuGeom.absPartIdx])
            {
                bCtuInfoCheck = skipRecursion = true;
                refMasks[0] = allSplitRefs;
                md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);
                checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);
                if (!sameContentRef)
                {
                    if ((m_param->bCTUInfo & 2) && (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth))
                    {
                        qp -= int32_t(0.04 * qp);
                        setLambdaFromQP(parentCTU, qp);
                    }
                    if (m_param->bCTUInfo & 4)
                        skipModes = false;
                }
                if (sameContentRef || (!sameContentRef && !(m_param->bCTUInfo & 4)))
                {
                    if (m_param->rdLevel)
                        skipModes = m_param->bEnableEarlySkip && md.bestMode && md.bestMode->cu.isSkipped(0);
                    if ((m_param->bCTUInfo & 4) && sameContentRef)
                        skipModes = md.bestMode && true;
                }
            }
            else
            {
                md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
                md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
                checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
                skipModes = !!m_param->bEnableEarlySkip && md.bestMode;
                refMasks[0] = allSplitRefs;
                md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);
                checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);
            }
            mightSplit &= !bDecidedDepth;
        }
    }
    if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
    {
        if (mightNotSplit && depth == m_reuseDepth[cuGeom.absPartIdx])
        {
            if (m_reuseModes[cuGeom.absPartIdx] == MODE_SKIP)
            {
                md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
                md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
                checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
                skipModes = !!m_param->bEnableEarlySkip && md.bestMode;
                refMasks[0] = allSplitRefs;
                md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);
                checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);

                if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)
                    skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);
            }
            if (m_param->analysisReuseLevel > 4 && m_reusePartSize[cuGeom.absPartIdx] == SIZE_2Nx2N)
                skipRectAmp = true && !!md.bestMode;
        }
    }

    if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead && m_multipassAnalysis)
    {
        if (mightNotSplit && depth == m_multipassDepth[cuGeom.absPartIdx])
        {
            if (m_multipassModes[cuGeom.absPartIdx] == MODE_SKIP)
            {
                md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
                md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
                checkMerge2Nx2N_rd0_4(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);

                skipModes = !!m_param->bEnableEarlySkip && md.bestMode;
                refMasks[0] = allSplitRefs;
                md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);
                checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);

                if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)
                    skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);
            }
        }
    }

    /* Step 1. Evaluate Merge/Skip candidates for likely early-outs */
    if (mightNotSplit && !md.bestMode && !bCtuInfoCheck)
    {
        md.pred[PRED_SKIP].cu.initSubCU(parentCTU, cuGeom, qp);
        md.pred[PRED_MERGE].cu.initSubCU(parentCTU, cuGeom, qp);
        checkMerge2Nx2N_rd5_6(md.pred[PRED_SKIP], md.pred[PRED_MERGE], cuGeom);
        skipModes = m_param->bEnableEarlySkip && md.bestMode && !md.bestMode->cu.getQtRootCbf(0);
        refMasks[0] = allSplitRefs;
        md.pred[PRED_2Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
        checkInter_rd5_6(md.pred[PRED_2Nx2N], cuGeom, SIZE_2Nx2N, refMasks);
        checkBestMode(md.pred[PRED_2Nx2N], cuGeom.depth);

        if (m_param->bEnableRecursionSkip && depth && m_modeDepth[depth - 1].bestMode)
            skipRecursion = md.bestMode && !md.bestMode->cu.getQtRootCbf(0);
    }

    // estimate split cost
    /* Step 2. Evaluate each of the 4 split sub-blocks in series */
    if (mightSplit && !skipRecursion)
    {
        if (bCtuInfoCheck && m_param->bCTUInfo & 2)
            qp = int((1 / 0.96) * qp + 0.5);
        Mode* splitPred = &md.pred[PRED_SPLIT];
        splitPred->initCosts();
        CUData* splitCU = &splitPred->cu;
        splitCU->initSubCU(parentCTU, cuGeom, qp);

        uint32_t nextDepth = depth + 1;
        ModeDepth& nd = m_modeDepth[nextDepth];
        invalidateContexts(nextDepth);
        Entropy* nextContext = &m_rqt[depth].cur;
        int nextQP = qp;
        splitIntra = false;

        for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
        {
            const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);
            if (childGeom.flags & CUGeom::PRESENT)
            {
                m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                m_rqt[nextDepth].cur.load(*nextContext);

                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));

                splitData[subPartIdx] = compressInterCU_rd5_6(parentCTU, childGeom, nextQP);

                // Save best CU and pred data for this sub CU
                splitIntra |= nd.bestMode->cu.isIntra(0);
                splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
                splitPred->addSubCosts(*nd.bestMode);
                nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);
                nextContext = &nd.bestMode->contexts;
            }
            else
            {
                splitCU->setEmptyPart(childGeom, subPartIdx);
            }
        }
        nextContext->store(splitPred->contexts);
        if (mightNotSplit)
            addSplitFlagCost(*splitPred, cuGeom.depth);
        else
            updateModeCost(*splitPred);

        checkDQPForSplitPred(*splitPred, cuGeom);
    }

    /* Split CUs
     *   0  1
     *   2  3 */
    allSplitRefs = splitData[0].splitRefs | splitData[1].splitRefs | splitData[2].splitRefs | splitData[3].splitRefs;
    /* Step 3. Evaluate ME (2Nx2N, rect, amp) and intra modes at current depth */
    if (mightNotSplit)
    {
        if (m_slice->m_pps->bUseDQP && depth <= m_slice->m_pps->maxCuDQPDepth && m_slice->m_pps->maxCuDQPDepth != 0)
            setLambdaFromQP(parentCTU, qp);

        if (!skipModes)
        {
            refMasks[0] = allSplitRefs;

            if (m_param->limitReferences & X265_REF_LIMIT_CU)
            {
                CUData& cu = md.pred[PRED_2Nx2N].cu;
                uint32_t refMask = cu.getBestRefIdx(0);
                allSplitRefs = splitData[0].splitRefs = splitData[1].splitRefs = splitData[2].splitRefs = splitData[3].splitRefs = refMask;
            }

            if (m_slice->m_sliceType == B_SLICE)
            {
                md.pred[PRED_BIDIR].cu.initSubCU(parentCTU, cuGeom, qp);
                checkBidir2Nx2N(md.pred[PRED_2Nx2N], md.pred[PRED_BIDIR], cuGeom);
                if (md.pred[PRED_BIDIR].sa8dCost < MAX_INT64)
                {
                    uint32_t numPU = md.pred[PRED_BIDIR].cu.getNumPartInter(0);
                    if (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)
                        for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
                        {
                            PredictionUnit pu(md.pred[PRED_BIDIR].cu, cuGeom, puIdx);
                            motionCompensation(md.pred[PRED_BIDIR].cu, pu, md.pred[PRED_BIDIR].predYuv, true, true);
                        }
                    encodeResAndCalcRdInterCU(md.pred[PRED_BIDIR], cuGeom);
                    checkBestMode(md.pred[PRED_BIDIR], cuGeom.depth);
                }
            }

            if (!skipRectAmp)
            {
                if (m_param->bEnableRectInter)
                {
                    uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;
                    uint32_t threshold_2NxN, threshold_Nx2N;

                    if (m_slice->m_sliceType == P_SLICE)
                    {
                        threshold_2NxN = splitData[0].mvCost[0] + splitData[1].mvCost[0];
                        threshold_Nx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];
                    }
                    else
                    {
                        threshold_2NxN = (splitData[0].mvCost[0] + splitData[1].mvCost[0]
                                       + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;
                        threshold_Nx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]
                                       + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;
                    }

                    int try_2NxN_first = threshold_2NxN < threshold_Nx2N;
                    if (try_2NxN_first && splitCost < md.bestMode->rdCost + threshold_2NxN)
                    {
                        refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */
                        refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */
                        md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkInter_rd5_6(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);
                        checkBestMode(md.pred[PRED_2NxN], cuGeom.depth);
                    }

                    if (splitCost < md.bestMode->rdCost + threshold_Nx2N)
                    {
                        refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* left */
                        refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* right */
                        md.pred[PRED_Nx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkInter_rd5_6(md.pred[PRED_Nx2N], cuGeom, SIZE_Nx2N, refMasks);
                        checkBestMode(md.pred[PRED_Nx2N], cuGeom.depth);
                    }

                    if (!try_2NxN_first && splitCost < md.bestMode->rdCost + threshold_2NxN)
                    {
                        refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* top */
                        refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* bot */
                        md.pred[PRED_2NxN].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkInter_rd5_6(md.pred[PRED_2NxN], cuGeom, SIZE_2NxN, refMasks);
                        checkBestMode(md.pred[PRED_2NxN], cuGeom.depth);
                    }
                }

                // Try AMP (SIZE_2NxnU, SIZE_2NxnD, SIZE_nLx2N, SIZE_nRx2N)
                if (m_slice->m_sps->maxAMPDepth > depth)
                {
                    uint64_t splitCost = splitData[0].sa8dCost + splitData[1].sa8dCost + splitData[2].sa8dCost + splitData[3].sa8dCost;
                    uint32_t threshold_2NxnU, threshold_2NxnD, threshold_nLx2N, threshold_nRx2N;

                    if (m_slice->m_sliceType == P_SLICE)
                    {
                        threshold_2NxnU = splitData[0].mvCost[0] + splitData[1].mvCost[0];
                        threshold_2NxnD = splitData[2].mvCost[0] + splitData[3].mvCost[0];

                        threshold_nLx2N = splitData[0].mvCost[0] + splitData[2].mvCost[0];
                        threshold_nRx2N = splitData[1].mvCost[0] + splitData[3].mvCost[0];
                    }
                    else
                    {
                        threshold_2NxnU = (splitData[0].mvCost[0] + splitData[1].mvCost[0]
                                        + splitData[0].mvCost[1] + splitData[1].mvCost[1] + 1) >> 1;
                        threshold_2NxnD = (splitData[2].mvCost[0] + splitData[3].mvCost[0]
                                        + splitData[2].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;

                        threshold_nLx2N = (splitData[0].mvCost[0] + splitData[2].mvCost[0]
                                        + splitData[0].mvCost[1] + splitData[2].mvCost[1] + 1) >> 1;
                        threshold_nRx2N = (splitData[1].mvCost[0] + splitData[3].mvCost[0]
                                        + splitData[1].mvCost[1] + splitData[3].mvCost[1] + 1) >> 1;
                    }

                    bool bHor = false, bVer = false;
                    if (md.bestMode->cu.m_partSize[0] == SIZE_2NxN)
                        bHor = true;
                    else if (md.bestMode->cu.m_partSize[0] == SIZE_Nx2N)
                        bVer = true;
                    else if (md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N && !md.bestMode->cu.m_mergeFlag[0])
                    {
                        bHor = true;
                        bVer = true;
                    }

                    if (bHor)
                    {
                        int try_2NxnD_first = threshold_2NxnD < threshold_2NxnU;
                        if (try_2NxnD_first && splitCost < md.bestMode->rdCost + threshold_2NxnD)
                        {
                            refMasks[0] = allSplitRefs;                                    /* 75% top */
                            refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */
                            md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd5_6(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);
                            checkBestMode(md.pred[PRED_2NxnD], cuGeom.depth);
                        }

                        if (splitCost < md.bestMode->rdCost + threshold_2NxnU)
                        {
                            refMasks[0] = splitData[0].splitRefs | splitData[1].splitRefs; /* 25% top */
                            refMasks[1] = allSplitRefs;                                    /* 75% bot */
                            md.pred[PRED_2NxnU].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd5_6(md.pred[PRED_2NxnU], cuGeom, SIZE_2NxnU, refMasks);
                            checkBestMode(md.pred[PRED_2NxnU], cuGeom.depth);
                        }

                        if (!try_2NxnD_first && splitCost < md.bestMode->rdCost + threshold_2NxnD)
                        {
                            refMasks[0] = allSplitRefs;                                    /* 75% top */
                            refMasks[1] = splitData[2].splitRefs | splitData[3].splitRefs; /* 25% bot */
                            md.pred[PRED_2NxnD].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd5_6(md.pred[PRED_2NxnD], cuGeom, SIZE_2NxnD, refMasks);
                            checkBestMode(md.pred[PRED_2NxnD], cuGeom.depth);
                        }
                    }

                    if (bVer)
                    {
                        int try_nRx2N_first = threshold_nRx2N < threshold_nLx2N;
                        if (try_nRx2N_first && splitCost < md.bestMode->rdCost + threshold_nRx2N)
                        {
                            refMasks[0] = allSplitRefs;                                    /* 75% left  */
                            refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */
                            md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd5_6(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);
                            checkBestMode(md.pred[PRED_nRx2N], cuGeom.depth);
                        }

                        if (splitCost < md.bestMode->rdCost + threshold_nLx2N)
                        {
                            refMasks[0] = splitData[0].splitRefs | splitData[2].splitRefs; /* 25% left  */
                            refMasks[1] = allSplitRefs;                                    /* 75% right */
                            md.pred[PRED_nLx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd5_6(md.pred[PRED_nLx2N], cuGeom, SIZE_nLx2N, refMasks);
                            checkBestMode(md.pred[PRED_nLx2N], cuGeom.depth);
                        }

                        if (!try_nRx2N_first && splitCost < md.bestMode->rdCost + threshold_nRx2N)
                        {
                            refMasks[0] = allSplitRefs;                                    /* 75% left  */
                            refMasks[1] = splitData[1].splitRefs | splitData[3].splitRefs; /* 25% right */
                            md.pred[PRED_nRx2N].cu.initSubCU(parentCTU, cuGeom, qp);
                            checkInter_rd5_6(md.pred[PRED_nRx2N], cuGeom, SIZE_nRx2N, refMasks);
                            checkBestMode(md.pred[PRED_nRx2N], cuGeom.depth);
                        }
                    }
                }
            }

            if ((m_slice->m_sliceType != B_SLICE || m_param->bIntraInBFrames) && (cuGeom.log2CUSize != MAX_LOG2_CU_SIZE) && !((m_param->bCTUInfo & 4) && bCtuInfoCheck))
            {
                if (!m_param->limitReferences || splitIntra)
                {
                    ProfileCounter(parentCTU, totalIntraCU[cuGeom.depth]);
                    md.pred[PRED_INTRA].cu.initSubCU(parentCTU, cuGeom, qp);
                    checkIntra(md.pred[PRED_INTRA], cuGeom, SIZE_2Nx2N);
                    checkBestMode(md.pred[PRED_INTRA], depth);

                    if (cuGeom.log2CUSize == 3 && m_slice->m_sps->quadtreeTULog2MinSize < 3)
                    {
                        md.pred[PRED_INTRA_NxN].cu.initSubCU(parentCTU, cuGeom, qp);
                        checkIntra(md.pred[PRED_INTRA_NxN], cuGeom, SIZE_NxN);
                        checkBestMode(md.pred[PRED_INTRA_NxN], depth);
                    }
                }
                else
                {
                    ProfileCounter(parentCTU, skippedIntraCU[cuGeom.depth]);
                }
            }
        }

        if ((md.bestMode->cu.isInter(0) && !(md.bestMode->cu.m_mergeFlag[0] && md.bestMode->cu.m_partSize[0] == SIZE_2Nx2N)) && (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400))
        {
            uint32_t numPU = md.bestMode->cu.getNumPartInter(0);

            for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
            {
                PredictionUnit pu(md.bestMode->cu, cuGeom, puIdx);
                motionCompensation(md.bestMode->cu, pu, md.bestMode->predYuv, false, m_csp != X265_CSP_I400);
            }
            encodeResAndCalcRdInterCU(*md.bestMode, cuGeom);
        }
        if (m_bTryLossless)
            tryLossless(cuGeom);

        if (mightSplit)
            addSplitFlagCost(*md.bestMode, cuGeom.depth);
    }

    if ((m_limitTU & X265_TU_LIMIT_NEIGH) && cuGeom.log2CUSize >= 4)
    {
        if (mightNotSplit)
        {
            CUData* ctu = md.bestMode->cu.m_encData->getPicCTU(parentCTU.m_cuAddr);
            int8_t maxTUDepth = -1;
            for (uint32_t i = 0; i < cuGeom.numPartitions; i++)
                maxTUDepth = X265_MAX(maxTUDepth, md.bestMode->cu.m_tuDepth[i]);
            ctu->m_refTuDepth[cuGeom.geomRecurId] = maxTUDepth;
        }
    }

    /* compare split RD cost against best cost */
    if (mightSplit && !skipRecursion)
        checkBestMode(md.pred[PRED_SPLIT], depth);

    if (m_param->bEnableRdRefine && depth <= m_slice->m_pps->maxCuDQPDepth)
    {
        int cuIdx = (cuGeom.childOffset - 1) / 3;
        cacheCost[cuIdx] = md.bestMode->rdCost;
    }

       /* determine which motion references the parent CU should search */
    SplitData splitCUData;
    splitCUData.initSplitCUData();
    if (m_param->limitReferences & X265_REF_LIMIT_DEPTH)
    {
        if (md.bestMode == &md.pred[PRED_SPLIT])
            splitCUData.splitRefs = allSplitRefs;
        else
        {
            /* use best merge/inter mode, in case of intra use 2Nx2N inter references */
            CUData& cu = md.bestMode->cu.isIntra(0) ? md.pred[PRED_2Nx2N].cu : md.bestMode->cu;
            uint32_t numPU = cu.getNumPartInter(0);
            for (uint32_t puIdx = 0, subPartIdx = 0; puIdx < numPU; puIdx++, subPartIdx += cu.getPUOffset(puIdx, 0))
                splitCUData.splitRefs |= cu.getBestRefIdx(subPartIdx);
        }
    }

    if (m_param->limitModes)
    {
        splitCUData.mvCost[0] = md.pred[PRED_2Nx2N].bestME[0][0].mvCost; // L0
        splitCUData.mvCost[1] = md.pred[PRED_2Nx2N].bestME[0][1].mvCost; // L1
        splitCUData.sa8dCost    = md.pred[PRED_2Nx2N].rdCost;
    }

    /* Copy best data to encData CTU and recon */
    md.bestMode->cu.copyToPic(depth);
    md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, parentCTU.m_cuAddr, cuGeom.absPartIdx);

    return splitCUData;
}

void Analysis::recodeCU(const CUData& parentCTU, const CUGeom& cuGeom, int32_t qp, int32_t lqp)
{
    uint32_t depth = cuGeom.depth;
    ModeDepth& md = m_modeDepth[depth];
    md.bestMode = NULL;

    m_evaluateInter = 0;
    bool mightSplit = !(cuGeom.flags & CUGeom::LEAF);
    bool mightNotSplit = !(cuGeom.flags & CUGeom::SPLIT_MANDATORY);
    bool bDecidedDepth = parentCTU.m_cuDepth[cuGeom.absPartIdx] == depth;

    int split = (m_param->interRefine && cuGeom.log2CUSize == (uint32_t)(g_log2Size[m_param->minCUSize] + 1)
                && bDecidedDepth && parentCTU.m_predMode[cuGeom.absPartIdx] == MODE_SKIP);

    if (bDecidedDepth)
    {
        setLambdaFromQP(parentCTU, qp, lqp);

        Mode& mode = md.pred[0];
        md.bestMode = &mode;
        mode.cu.initSubCU(parentCTU, cuGeom, qp);
        PartSize size = (PartSize)parentCTU.m_partSize[cuGeom.absPartIdx];
        if (parentCTU.isIntra(cuGeom.absPartIdx))
        {
            if (m_param->intraRefine != 2 || parentCTU.m_lumaIntraDir[cuGeom.absPartIdx] <= 1)
            {
                memcpy(mode.cu.m_lumaIntraDir, parentCTU.m_lumaIntraDir + cuGeom.absPartIdx, cuGeom.numPartitions);
                memcpy(mode.cu.m_chromaIntraDir, parentCTU.m_chromaIntraDir + cuGeom.absPartIdx, cuGeom.numPartitions);
            }
            checkIntra(mode, cuGeom, size);
        }
        else
        {
            mode.cu.copyFromPic(parentCTU, cuGeom, m_csp, false);
            uint32_t numPU = parentCTU.getNumPartInter(cuGeom.absPartIdx);
            for (uint32_t part = 0; part < numPU; part++)
            {
                PredictionUnit pu(mode.cu, cuGeom, part);
                if (m_param->analysisReuseLevel == 10)
                {
                    analysis_inter_data* interDataCTU = (analysis_inter_data*)m_frame->m_analysisData.interData;
                    int cuIdx = (mode.cu.m_cuAddr * parentCTU.m_numPartitions) + cuGeom.absPartIdx;
                    mode.cu.m_mergeFlag[pu.puAbsPartIdx] = interDataCTU->mergeFlag[cuIdx + part];
                    mode.cu.setPUInterDir(interDataCTU->interDir[cuIdx + part], pu.puAbsPartIdx, part);
                    for (int list = 0; list < m_slice->isInterB() + 1; list++)
                    {
                        mode.cu.setPUMv(list, interDataCTU->mv[list][cuIdx + part], pu.puAbsPartIdx, part);
                        mode.cu.setPURefIdx(list, interDataCTU->refIdx[list][cuIdx + part], pu.puAbsPartIdx, part);
                        mode.cu.m_mvpIdx[list][pu.puAbsPartIdx] = interDataCTU->mvpIdx[list][cuIdx + part];
                    }
                    if (!mode.cu.m_mergeFlag[pu.puAbsPartIdx])
                    {
                        if (m_param->mvRefine)
                            m_me.setSourcePU(*mode.fencYuv, pu.ctuAddr, pu.cuAbsPartIdx, pu.puAbsPartIdx, pu.width, pu.height, m_param->searchMethod, m_param->subpelRefine, false);
                        //AMVP
                        MV mvc[(MD_ABOVE_LEFT + 1) * 2 + 2];
                        mode.cu.getNeighbourMV(part, pu.puAbsPartIdx, mode.interNeighbours);
                        for (int list = 0; list < m_slice->isInterB() + 1; list++)
                        {
                            int ref = mode.cu.m_refIdx[list][pu.puAbsPartIdx];
                            if (ref == -1)
                                continue;
                            mode.cu.getPMV(mode.interNeighbours, list, ref, mode.amvpCand[list][ref], mvc);
                            MV mvp = mode.amvpCand[list][ref][mode.cu.m_mvpIdx[list][pu.puAbsPartIdx]];
                            if (m_param->mvRefine)
                            {
                                MV outmv;
                                searchMV(mode, pu, list, ref, outmv);
                                mode.cu.setPUMv(list, outmv, pu.puAbsPartIdx, part);
                            }
                            mode.cu.m_mvd[list][pu.puAbsPartIdx] = mode.cu.m_mv[list][pu.puAbsPartIdx] - mvp;
                        }
                    }
                    else if(m_param->scaleFactor)
                    {
                        MVField candMvField[MRG_MAX_NUM_CANDS][2]; // double length for mv of both lists
                        uint8_t candDir[MRG_MAX_NUM_CANDS];
                        mode.cu.getInterMergeCandidates(pu.puAbsPartIdx, part, candMvField, candDir);
                        uint8_t mvpIdx = mode.cu.m_mvpIdx[0][pu.puAbsPartIdx];
                        mode.cu.setPUInterDir(candDir[mvpIdx], pu.puAbsPartIdx, part);
                        mode.cu.setPUMv(0, candMvField[mvpIdx][0].mv, pu.puAbsPartIdx, part);
                        mode.cu.setPUMv(1, candMvField[mvpIdx][1].mv, pu.puAbsPartIdx, part);
                        mode.cu.setPURefIdx(0, (int8_t)candMvField[mvpIdx][0].refIdx, pu.puAbsPartIdx, part);
                        mode.cu.setPURefIdx(1, (int8_t)candMvField[mvpIdx][1].refIdx, pu.puAbsPartIdx, part);
                    }
                }
                motionCompensation(mode.cu, pu, mode.predYuv, true, (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400));
            }
            if (!m_param->interRefine && parentCTU.isSkipped(cuGeom.absPartIdx))
                encodeResAndCalcRdSkipCU(mode);
            else
                encodeResAndCalcRdInterCU(mode, cuGeom);

            /* checkMerge2Nx2N function performs checkDQP after encoding residual, do the same */
            bool mergeInter2Nx2N = size == SIZE_2Nx2N && mode.cu.m_mergeFlag[0];
            if (parentCTU.isSkipped(cuGeom.absPartIdx) || mergeInter2Nx2N)
                checkDQP(mode, cuGeom);
        }

        if (m_bTryLossless)
            tryLossless(cuGeom);

        if (mightSplit)
            addSplitFlagCost(*md.bestMode, cuGeom.depth);

        if (mightSplit && m_param->rdLevel < 5)
            checkDQPForSplitPred(*md.bestMode, cuGeom);

        if (m_param->interRefine && parentCTU.m_predMode[cuGeom.absPartIdx] == MODE_SKIP  && !mode.cu.isSkipped(0))
        {
            m_evaluateInter = 1;
            m_param->rdLevel > 4 ? compressInterCU_rd5_6(parentCTU, cuGeom, qp) : compressInterCU_rd0_4(parentCTU, cuGeom, qp);
        }
    }
    if (!bDecidedDepth || split)
    {
        Mode* splitPred = &md.pred[PRED_SPLIT];
        if (!split)
            md.bestMode = splitPred;
        splitPred->initCosts();
        CUData* splitCU = &splitPred->cu;
        splitCU->initSubCU(parentCTU, cuGeom, qp);

        uint32_t nextDepth = depth + 1;
        ModeDepth& nd = m_modeDepth[nextDepth];
        invalidateContexts(nextDepth);
        Entropy* nextContext = &m_rqt[depth].cur;
        int nextQP = qp;

        for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
        {
            const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);
            if (childGeom.flags & CUGeom::PRESENT)
            {
                m_modeDepth[0].fencYuv.copyPartToYuv(nd.fencYuv, childGeom.absPartIdx);
                m_rqt[nextDepth].cur.load(*nextContext);

                if (m_slice->m_pps->bUseDQP && nextDepth <= m_slice->m_pps->maxCuDQPDepth)
                    nextQP = setLambdaFromQP(parentCTU, calculateQpforCuSize(parentCTU, childGeom));

                int lamdaQP = m_param->analysisReuseLevel == 10 ? nextQP : lqp;

                if (split)
                    m_param->rdLevel > 4 ? compressInterCU_rd5_6(parentCTU, childGeom, nextQP) : compressInterCU_rd0_4(parentCTU, childGeom, nextQP);
                else
                    qprdRefine(parentCTU, childGeom, nextQP, lamdaQP);

                // Save best CU and pred data for this sub CU
                splitCU->copyPartFrom(nd.bestMode->cu, childGeom, subPartIdx);
                splitPred->addSubCosts(*nd.bestMode);
                nd.bestMode->reconYuv.copyToPartYuv(splitPred->reconYuv, childGeom.numPartitions * subPartIdx);
                nextContext = &nd.bestMode->contexts;
            }
            else
            {
                splitCU->setEmptyPart(childGeom, subPartIdx);
                // Set depth of non-present CU to 0 to ensure that correct CU is fetched as reference to code deltaQP
                memset(parentCTU.m_cuDepth + childGeom.absPartIdx, 0, childGeom.numPartitions);
            }
        }
        nextContext->store(splitPred->contexts);
        if (mightNotSplit)
            addSplitFlagCost(*splitPred, cuGeom.depth);
        else
            updateModeCost(*splitPred);

        if (m_param->interRefine)
        {
            if (m_param->rdLevel > 1)
                checkBestMode(*splitPred, cuGeom.depth);
            else if (splitPred->sa8dCost < md.bestMode->sa8dCost)
                md.bestMode = splitPred;
        }

        checkDQPForSplitPred(*splitPred, cuGeom);

        /* Copy best data to encData CTU and recon */
        md.bestMode->cu.copyToPic(depth);
        md.bestMode->reconYuv.copyToPicYuv(*m_frame->m_reconPic, parentCTU.m_cuAddr, cuGeom.absPartIdx);
    }
}

/* sets md.bestMode if a valid merge candidate is found, else leaves it NULL */
void Analysis::checkMerge2Nx2N_rd0_4(Mode& skip, Mode& merge, const CUGeom& cuGeom)
{
    uint32_t depth = cuGeom.depth;
    ModeDepth& md = m_modeDepth[depth];
    Yuv *fencYuv = &md.fencYuv;

    /* Note that these two Mode instances are named MERGE and SKIP but they may
     * hold the reverse when the function returns. We toggle between the two modes */
    Mode* tempPred = &merge;
    Mode* bestPred = &skip;

    X265_CHECK(m_slice->m_sliceType != I_SLICE, "Evaluating merge in I slice\n");

    tempPred->initCosts();
    tempPred->cu.setPartSizeSubParts(SIZE_2Nx2N);
    tempPred->cu.setPredModeSubParts(MODE_INTER);
    tempPred->cu.m_mergeFlag[0] = true;

    bestPred->initCosts();
    bestPred->cu.setPartSizeSubParts(SIZE_2Nx2N);
    bestPred->cu.setPredModeSubParts(MODE_INTER);
    bestPred->cu.m_mergeFlag[0] = true;

    MVField candMvField[MRG_MAX_NUM_CANDS][2]; // double length for mv of both lists
    uint8_t candDir[MRG_MAX_NUM_CANDS];
    uint32_t numMergeCand = tempPred->cu.getInterMergeCandidates(0, 0, candMvField, candDir);
    PredictionUnit pu(merge.cu, cuGeom, 0);

    bestPred->sa8dCost = MAX_INT64;
    int bestSadCand = -1;
    int sizeIdx = cuGeom.log2CUSize - 2;
    int safeX, maxSafeMv;
    if (m_param->bIntraRefresh && m_slice->m_sliceType == P_SLICE)
    {
        safeX = m_slice->m_refFrameList[0][0]->m_encData->m_pir.pirEndCol * m_param->maxCUSize - 3;
        maxSafeMv = (safeX - tempPred->cu.m_cuPelX) * 4;
    }
    for (uint32_t i = 0; i < numMergeCand; ++i)
    {
        if (m_bFrameParallel)
        {
            // Parallel slices bound check
            if (m_param->maxSlices > 1)
            {
                // NOTE: First row in slice can't negative
                if (X265_MIN(candMvField[i][0].mv.y, candMvField[i][1].mv.y) < m_sliceMinY)
                    continue;

                // Last row in slice can't reference beyond bound since it is another slice area
                // TODO: we may beyond bound in future since these area have a chance to finish because we use parallel slices. Necessary prepare research on load balance
                if (X265_MAX(candMvField[i][0].mv.y, candMvField[i][1].mv.y) > m_sliceMaxY)
                    continue;
            }

            if (candMvField[i][0].mv.y >= (m_param->searchRange + 1) * 4 ||
                candMvField[i][1].mv.y >= (m_param->searchRange + 1) * 4)
                continue;
        }

        if (m_param->bIntraRefresh && m_slice->m_sliceType == P_SLICE &&
            tempPred->cu.m_cuPelX / m_param->maxCUSize < m_frame->m_encData->m_pir.pirEndCol &&
            candMvField[i][0].mv.x > maxSafeMv)
            // skip merge candidates which reference beyond safe reference area
            continue;

        tempPred->cu.m_mvpIdx[0][0] = (uint8_t)i; // merge candidate ID is stored in L0 MVP idx
        X265_CHECK(m_slice->m_sliceType == B_SLICE || !(candDir[i] & 0x10), " invalid merge for P slice\n");
        tempPred->cu.m_interDir[0] = candDir[i];
        tempPred->cu.m_mv[0][0] = candMvField[i][0].mv;
        tempPred->cu.m_mv[1][0] = candMvField[i][1].mv;
        tempPred->cu.m_refIdx[0][0] = (int8_t)candMvField[i][0].refIdx;
        tempPred->cu.m_refIdx[1][0] = (int8_t)candMvField[i][1].refIdx;
        motionCompensation(tempPred->cu, pu, tempPred->predYuv, true, m_bChromaSa8d && (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400));

        tempPred->sa8dBits = getTUBits(i, numMergeCand);
        tempPred->distortion = primitives.cu[sizeIdx].sa8d(fencYuv->m_buf[0], fencYuv->m_size, tempPred->predYuv.m_buf[0], tempPred->predYuv.m_size);
        if (m_bChromaSa8d && (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400))
        {
            tempPred->distortion += primitives.chroma[m_csp].cu[sizeIdx].sa8d(fencYuv->m_buf[1], fencYuv->m_csize, tempPred->predYuv.m_buf[1], tempPred->predYuv.m_csize);
            tempPred->distortion += primitives.chroma[m_csp].cu[sizeIdx].sa8d(fencYuv->m_buf[2], fencYuv->m_csize, tempPred->predYuv.m_buf[2], tempPred->predYuv.m_csize);
        }
        tempPred->sa8dCost = m_rdCost.calcRdSADCost((uint32_t)tempPred->distortion, tempPred->sa8dBits);

        if (tempPred->sa8dCost < bestPred->sa8dCost)
        {
            bestSadCand = i;
            std::swap(tempPred, bestPred);
        }
    }

    /* force mode decision to take inter or intra */
    if (bestSadCand < 0)
        return;

    /* calculate the motion compensation for chroma for the best mode selected */
    if ((!m_bChromaSa8d && (m_csp != X265_CSP_I400)) || (m_frame->m_fencPic->m_picCsp == X265_CSP_I400 && m_csp != X265_CSP_I400)) /* Chroma MC was done above */
        motionCompensation(bestPred->cu, pu, bestPred->predYuv, false, true);

    if (m_param->rdLevel)
    {
        if (m_param->bLossless)
            bestPred->rdCost = MAX_INT64;
        else
            encodeResAndCalcRdSkipCU(*bestPred);

        /* Encode with residual */
        tempPred->cu.m_mvpIdx[0][0] = (uint8_t)bestSadCand;
        tempPred->cu.setPUInterDir(candDir[bestSadCand], 0, 0);
        tempPred->cu.setPUMv(0, candMvField[bestSadCand][0].mv, 0, 0);
        tempPred->cu.setPUMv(1, candMvField[bestSadCand][1].mv, 0, 0);
        tempPred->cu.setPURefIdx(0, (int8_t)candMvField[bestSadCand][0].refIdx, 0, 0);
        tempPred->cu.setPURefIdx(1, (int8_t)candMvField[bestSadCand][1].refIdx, 0, 0);
        tempPred->sa8dCost = bestPred->sa8dCost;
        tempPred->sa8dBits = bestPred->sa8dBits;
        tempPred->predYuv.copyFromYuv(bestPred->predYuv);

        encodeResAndCalcRdInterCU(*tempPred, cuGeom);

        md.bestMode = tempPred->rdCost < bestPred->rdCost ? tempPred : bestPred;
    }
    else
        md.bestMode = bestPred;

    /* broadcast sets of MV field data */
    md.bestMode->cu.setPUInterDir(candDir[bestSadCand], 0, 0);
    md.bestMode->cu.setPUMv(0, candMvField[bestSadCand][0].mv, 0, 0);
    md.bestMode->cu.setPUMv(1, candMvField[bestSadCand][1].mv, 0, 0);
    md.bestMode->cu.setPURefIdx(0, (int8_t)candMvField[bestSadCand][0].refIdx, 0, 0);
    md.bestMode->cu.setPURefIdx(1, (int8_t)candMvField[bestSadCand][1].refIdx, 0, 0);
    checkDQP(*md.bestMode, cuGeom);
}

/* sets md.bestMode if a valid merge candidate is found, else leaves it NULL */
void Analysis::checkMerge2Nx2N_rd5_6(Mode& skip, Mode& merge, const CUGeom& cuGeom)
{
    uint32_t depth = cuGeom.depth;

    /* Note that these two Mode instances are named MERGE and SKIP but they may
     * hold the reverse when the function returns. We toggle between the two modes */
    Mode* tempPred = &merge;
    Mode* bestPred = &skip;

    merge.initCosts();
    merge.cu.setPredModeSubParts(MODE_INTER);
    merge.cu.setPartSizeSubParts(SIZE_2Nx2N);
    merge.cu.m_mergeFlag[0] = true;

    skip.initCosts();
    skip.cu.setPredModeSubParts(MODE_INTER);
    skip.cu.setPartSizeSubParts(SIZE_2Nx2N);
    skip.cu.m_mergeFlag[0] = true;

    MVField candMvField[MRG_MAX_NUM_CANDS][2]; // double length for mv of both lists
    uint8_t candDir[MRG_MAX_NUM_CANDS];
    uint32_t numMergeCand = merge.cu.getInterMergeCandidates(0, 0, candMvField, candDir);
    PredictionUnit pu(merge.cu, cuGeom, 0);

    bool foundCbf0Merge = false;
    bool triedPZero = false, triedBZero = false;
    bestPred->rdCost = MAX_INT64;

    int safeX, maxSafeMv;
    if (m_param->bIntraRefresh && m_slice->m_sliceType == P_SLICE)
    {
        safeX = m_slice->m_refFrameList[0][0]->m_encData->m_pir.pirEndCol * m_param->maxCUSize - 3;
        maxSafeMv = (safeX - tempPred->cu.m_cuPelX) * 4;
    }
    for (uint32_t i = 0; i < numMergeCand; i++)
    {
        if (m_bFrameParallel)
        {
            // Parallel slices bound check
            if (m_param->maxSlices > 1)
            {
                // NOTE: First row in slice can't negative
                if (X265_MIN(candMvField[i][0].mv.y, candMvField[i][1].mv.y) < m_sliceMinY)
                    continue;

                // Last row in slice can't reference beyond bound since it is another slice area
                // TODO: we may beyond bound in future since these area have a chance to finish because we use parallel slices. Necessary prepare research on load balance
                if (X265_MAX(candMvField[i][0].mv.y, candMvField[i][1].mv.y) > m_sliceMaxY)
                    continue;
            }

            if (candMvField[i][0].mv.y >= (m_param->searchRange + 1) * 4 ||
                candMvField[i][1].mv.y >= (m_param->searchRange + 1) * 4)
                continue;
        }

        /* the merge candidate list is packed with MV(0,0) ref 0 when it is not full */
        if (candDir[i] == 1 && !candMvField[i][0].mv.word && !candMvField[i][0].refIdx)
        {
            if (triedPZero)
                continue;
            triedPZero = true;
        }
        else if (candDir[i] == 3 &&
            !candMvField[i][0].mv.word && !candMvField[i][0].refIdx &&
            !candMvField[i][1].mv.word && !candMvField[i][1].refIdx)
        {
            if (triedBZero)
                continue;
            triedBZero = true;
        }
        if (m_param->bIntraRefresh && m_slice->m_sliceType == P_SLICE &&
            tempPred->cu.m_cuPelX / m_param->maxCUSize < m_frame->m_encData->m_pir.pirEndCol &&
            candMvField[i][0].mv.x > maxSafeMv)
            // skip merge candidates which reference beyond safe reference area
            continue;
        tempPred->cu.m_mvpIdx[0][0] = (uint8_t)i;    /* merge candidate ID is stored in L0 MVP idx */
        tempPred->cu.m_interDir[0] = candDir[i];
        tempPred->cu.m_mv[0][0] = candMvField[i][0].mv;
        tempPred->cu.m_mv[1][0] = candMvField[i][1].mv;
        tempPred->cu.m_refIdx[0][0] = (int8_t)candMvField[i][0].refIdx;
        tempPred->cu.m_refIdx[1][0] = (int8_t)candMvField[i][1].refIdx;
        tempPred->cu.setPredModeSubParts(MODE_INTER); /* must be cleared between encode iterations */

        motionCompensation(tempPred->cu, pu, tempPred->predYuv, true, m_csp != X265_CSP_I400);

        uint8_t hasCbf = true;
        bool swapped = false;
        if (!foundCbf0Merge)
        {
            /* if the best prediction has CBF (not a skip) then try merge with residual */

            encodeResAndCalcRdInterCU(*tempPred, cuGeom);
            hasCbf = tempPred->cu.getQtRootCbf(0);
            foundCbf0Merge = !hasCbf;

            if (tempPred->rdCost < bestPred->rdCost)
            {
                std::swap(tempPred, bestPred);
                swapped = true;
            }
        }
        if (!m_param->bLossless && hasCbf)
        {
            /* try merge without residual (skip), if not lossless coding */

            if (swapped)
            {
                tempPred->cu.m_mvpIdx[0][0] = (uint8_t)i;
                tempPred->cu.m_interDir[0] = candDir[i];
                tempPred->cu.m_mv[0][0] = candMvField[i][0].mv;
                tempPred->cu.m_mv[1][0] = candMvField[i][1].mv;
                tempPred->cu.m_refIdx[0][0] = (int8_t)candMvField[i][0].refIdx;
                tempPred->cu.m_refIdx[1][0] = (int8_t)candMvField[i][1].refIdx;
                tempPred->cu.setPredModeSubParts(MODE_INTER);
                tempPred->predYuv.copyFromYuv(bestPred->predYuv);
            }

            encodeResAndCalcRdSkipCU(*tempPred);

            if (tempPred->rdCost < bestPred->rdCost)
                std::swap(tempPred, bestPred);
        }
    }

    if (bestPred->rdCost < MAX_INT64)
    {
        m_modeDepth[depth].bestMode = bestPred;

        /* broadcast sets of MV field data */
        uint32_t bestCand = bestPred->cu.m_mvpIdx[0][0];
        bestPred->cu.setPUInterDir(candDir[bestCand], 0, 0);
        bestPred->cu.setPUMv(0, candMvField[bestCand][0].mv, 0, 0);
        bestPred->cu.setPUMv(1, candMvField[bestCand][1].mv, 0, 0);
        bestPred->cu.setPURefIdx(0, (int8_t)candMvField[bestCand][0].refIdx, 0, 0);
        bestPred->cu.setPURefIdx(1, (int8_t)candMvField[bestCand][1].refIdx, 0, 0);
        checkDQP(*bestPred, cuGeom);
    }
}

void Analysis::checkInter_rd0_4(Mode& interMode, const CUGeom& cuGeom, PartSize partSize, uint32_t refMask[2])
{
    interMode.initCosts();
    interMode.cu.setPartSizeSubParts(partSize);
    interMode.cu.setPredModeSubParts(MODE_INTER);
    int numPredDir = m_slice->isInterP() ? 1 : 2;

    if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
    {
        int refOffset = cuGeom.geomRecurId * 16 * numPredDir + partSize * numPredDir * 2;
        int index = 0;

        uint32_t numPU = interMode.cu.getNumPartInter(0);
        for (uint32_t part = 0; part < numPU; part++)
        {
            MotionData* bestME = interMode.bestME[part];
            for (int32_t i = 0; i < numPredDir; i++)
                bestME[i].ref = m_reuseRef[refOffset + index++];
        }
    }

    if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead && m_multipassAnalysis)
    {
        uint32_t numPU = interMode.cu.getNumPartInter(0);
        for (uint32_t part = 0; part < numPU; part++)
        {
            MotionData* bestME = interMode.bestME[part];
            for (int32_t i = 0; i < numPredDir; i++)
            {
                bestME[i].ref = m_multipassRef[i][cuGeom.absPartIdx];
                bestME[i].mv = m_multipassMv[i][cuGeom.absPartIdx];
                bestME[i].mvpIdx = m_multipassMvpIdx[i][cuGeom.absPartIdx];
            }
        }
    }
    predInterSearch(interMode, cuGeom, m_bChromaSa8d && (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400), refMask);

    /* predInterSearch sets interMode.sa8dBits */
    const Yuv& fencYuv = *interMode.fencYuv;
    Yuv& predYuv = interMode.predYuv;
    int part = partitionFromLog2Size(cuGeom.log2CUSize);
    interMode.distortion = primitives.cu[part].sa8d(fencYuv.m_buf[0], fencYuv.m_size, predYuv.m_buf[0], predYuv.m_size);
    if (m_bChromaSa8d && (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400))
    {
        interMode.distortion += primitives.chroma[m_csp].cu[part].sa8d(fencYuv.m_buf[1], fencYuv.m_csize, predYuv.m_buf[1], predYuv.m_csize);
        interMode.distortion += primitives.chroma[m_csp].cu[part].sa8d(fencYuv.m_buf[2], fencYuv.m_csize, predYuv.m_buf[2], predYuv.m_csize);
    }
    interMode.sa8dCost = m_rdCost.calcRdSADCost((uint32_t)interMode.distortion, interMode.sa8dBits);

    if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1)
    {
        int refOffset = cuGeom.geomRecurId * 16 * numPredDir + partSize * numPredDir * 2;
        int index = 0;

        uint32_t numPU = interMode.cu.getNumPartInter(0);
        for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
        {
            MotionData* bestME = interMode.bestME[puIdx];
            for (int32_t i = 0; i < numPredDir; i++)
                m_reuseRef[refOffset + index++] = bestME[i].ref;
        }
    }
}

void Analysis::checkInter_rd5_6(Mode& interMode, const CUGeom& cuGeom, PartSize partSize, uint32_t refMask[2])
{
    interMode.initCosts();
    interMode.cu.setPartSizeSubParts(partSize);
    interMode.cu.setPredModeSubParts(MODE_INTER);
    int numPredDir = m_slice->isInterP() ? 1 : 2;

    if (m_param->analysisReuseMode == X265_ANALYSIS_LOAD && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1 && m_param->analysisReuseLevel != 10)
    {
        int refOffset = cuGeom.geomRecurId * 16 * numPredDir + partSize * numPredDir * 2;
        int index = 0;

        uint32_t numPU = interMode.cu.getNumPartInter(0);
        for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
        {
            MotionData* bestME = interMode.bestME[puIdx];
            for (int32_t i = 0; i < numPredDir; i++)
                bestME[i].ref = m_reuseRef[refOffset + index++];
        }
    }

    if (m_param->analysisMultiPassRefine && m_param->rc.bStatRead && m_multipassAnalysis)
    {
        uint32_t numPU = interMode.cu.getNumPartInter(0);
        for (uint32_t part = 0; part < numPU; part++)
        {
            MotionData* bestME = interMode.bestME[part];
            for (int32_t i = 0; i < numPredDir; i++)
            {
                bestME[i].ref = m_multipassRef[i][cuGeom.absPartIdx];
                bestME[i].mv = m_multipassMv[i][cuGeom.absPartIdx];
                bestME[i].mvpIdx = m_multipassMvpIdx[i][cuGeom.absPartIdx];
            }
        }
    }

    predInterSearch(interMode, cuGeom, m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400, refMask);

    /* predInterSearch sets interMode.sa8dBits, but this is ignored */
    encodeResAndCalcRdInterCU(interMode, cuGeom);

    if (m_param->analysisReuseMode == X265_ANALYSIS_SAVE && m_reuseInterDataCTU && m_param->analysisReuseLevel > 1)
    {
        int refOffset = cuGeom.geomRecurId * 16 * numPredDir + partSize * numPredDir * 2;
        int index = 0;

        uint32_t numPU = interMode.cu.getNumPartInter(0);
        for (uint32_t puIdx = 0; puIdx < numPU; puIdx++)
        {
            MotionData* bestME = interMode.bestME[puIdx];
            for (int32_t i = 0; i < numPredDir; i++)
                m_reuseRef[refOffset + index++] = bestME[i].ref;
        }
    }
}

void Analysis::checkBidir2Nx2N(Mode& inter2Nx2N, Mode& bidir2Nx2N, const CUGeom& cuGeom)
{
    CUData& cu = bidir2Nx2N.cu;

    if (cu.isBipredRestriction() || inter2Nx2N.bestME[0][0].cost == MAX_UINT || inter2Nx2N.bestME[0][1].cost == MAX_UINT)
    {
        bidir2Nx2N.sa8dCost = MAX_INT64;
        bidir2Nx2N.rdCost = MAX_INT64;
        return;
    }

    const Yuv& fencYuv = *bidir2Nx2N.fencYuv;
    MV   mvzero(0, 0);
    int  partEnum = cuGeom.log2CUSize - 2;

    bidir2Nx2N.bestME[0][0] = inter2Nx2N.bestME[0][0];
    bidir2Nx2N.bestME[0][1] = inter2Nx2N.bestME[0][1];
    MotionData* bestME = bidir2Nx2N.bestME[0];
    int ref0    = bestME[0].ref;
    MV  mvp0    = bestME[0].mvp;
    int mvpIdx0 = bestME[0].mvpIdx;
    int ref1    = bestME[1].ref;
    MV  mvp1    = bestME[1].mvp;
    int mvpIdx1 = bestME[1].mvpIdx;

    bidir2Nx2N.initCosts();
    cu.setPartSizeSubParts(SIZE_2Nx2N);
    cu.setPredModeSubParts(MODE_INTER);
    cu.setPUInterDir(3, 0, 0);
    cu.setPURefIdx(0, (int8_t)ref0, 0, 0);
    cu.setPURefIdx(1, (int8_t)ref1, 0, 0);
    cu.m_mvpIdx[0][0] = (uint8_t)mvpIdx0;
    cu.m_mvpIdx[1][0] = (uint8_t)mvpIdx1;
    cu.m_mergeFlag[0] = 0;

    /* Estimate cost of BIDIR using best 2Nx2N L0 and L1 motion vectors */
    cu.setPUMv(0, bestME[0].mv, 0, 0);
    cu.m_mvd[0][0] = bestME[0].mv - mvp0;

    cu.setPUMv(1, bestME[1].mv, 0, 0);
    cu.m_mvd[1][0] = bestME[1].mv - mvp1;

    PredictionUnit pu(cu, cuGeom, 0);
    motionCompensation(cu, pu, bidir2Nx2N.predYuv, true, m_bChromaSa8d && (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400));

    int sa8d = primitives.cu[partEnum].sa8d(fencYuv.m_buf[0], fencYuv.m_size, bidir2Nx2N.predYuv.m_buf[0], bidir2Nx2N.predYuv.m_size);
    if (m_bChromaSa8d && (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400))
    {
        /* Add in chroma distortion */
        sa8d += primitives.chroma[m_csp].cu[partEnum].sa8d(fencYuv.m_buf[1], fencYuv.m_csize, bidir2Nx2N.predYuv.m_buf[1], bidir2Nx2N.predYuv.m_csize);
        sa8d += primitives.chroma[m_csp].cu[partEnum].sa8d(fencYuv.m_buf[2], fencYuv.m_csize, bidir2Nx2N.predYuv.m_buf[2], bidir2Nx2N.predYuv.m_csize);
    }
    bidir2Nx2N.sa8dBits = bestME[0].bits + bestME[1].bits + m_listSelBits[2] - (m_listSelBits[0] + m_listSelBits[1]);
    bidir2Nx2N.sa8dCost = sa8d + m_rdCost.getCost(bidir2Nx2N.sa8dBits);

    bool bTryZero = bestME[0].mv.notZero() || bestME[1].mv.notZero();
    if (bTryZero)
    {
        /* Do not try zero MV if unidir motion predictors are beyond
         * valid search area */
        MV mvmin, mvmax;
        int merange = X265_MAX(m_param->sourceWidth, m_param->sourceHeight);
        setSearchRange(cu, mvzero, merange, mvmin, mvmax);
        mvmax.y += 2; // there is some pad for subpel refine
        mvmin <<= 2;
        mvmax <<= 2;

        bTryZero &= bestME[0].mvp.checkRange(mvmin, mvmax);
        bTryZero &= bestME[1].mvp.checkRange(mvmin, mvmax);
    }
    if (bTryZero)
    {
        /* Estimate cost of BIDIR using coincident blocks */
        Yuv& tmpPredYuv = m_rqt[cuGeom.depth].tmpPredYuv;

        int zsa8d;

        if (m_bChromaSa8d && (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400))
        {
            cu.m_mv[0][0] = mvzero;
            cu.m_mv[1][0] = mvzero;

            motionCompensation(cu, pu, tmpPredYuv, true, true);
            zsa8d  = primitives.cu[partEnum].sa8d(fencYuv.m_buf[0], fencYuv.m_size, tmpPredYuv.m_buf[0], tmpPredYuv.m_size);
            zsa8d += primitives.chroma[m_csp].cu[partEnum].sa8d(fencYuv.m_buf[1], fencYuv.m_csize, tmpPredYuv.m_buf[1], tmpPredYuv.m_csize);
            zsa8d += primitives.chroma[m_csp].cu[partEnum].sa8d(fencYuv.m_buf[2], fencYuv.m_csize, tmpPredYuv.m_buf[2], tmpPredYuv.m_csize);

        }
        else
        {
            pixel *fref0 = m_slice->m_mref[0][ref0].getLumaAddr(pu.ctuAddr, pu.cuAbsPartIdx);
            pixel *fref1 = m_slice->m_mref[1][ref1].getLumaAddr(pu.ctuAddr, pu.cuAbsPartIdx);
            intptr_t refStride = m_slice->m_mref[0][0].lumaStride;

            primitives.pu[partEnum].pixelavg_pp(tmpPredYuv.m_buf[0], tmpPredYuv.m_size, fref0, refStride, fref1, refStride, 32);
            zsa8d = primitives.cu[partEnum].sa8d(fencYuv.m_buf[0], fencYuv.m_size, tmpPredYuv.m_buf[0], tmpPredYuv.m_size);
        }

        uint32_t bits0 = bestME[0].bits - m_me.bitcost(bestME[0].mv, mvp0) + m_me.bitcost(mvzero, mvp0);
        uint32_t bits1 = bestME[1].bits - m_me.bitcost(bestME[1].mv, mvp1) + m_me.bitcost(mvzero, mvp1);
        uint32_t zcost = zsa8d + m_rdCost.getCost(bits0) + m_rdCost.getCost(bits1);

        /* refine MVP selection for zero mv, updates: mvp, mvpidx, bits, cost */
        mvp0 = checkBestMVP(inter2Nx2N.amvpCand[0][ref0], mvzero, mvpIdx0, bits0, zcost);
        mvp1 = checkBestMVP(inter2Nx2N.amvpCand[1][ref1], mvzero, mvpIdx1, bits1, zcost);

        uint32_t zbits = bits0 + bits1 + m_listSelBits[2] - (m_listSelBits[0] + m_listSelBits[1]);
        zcost = zsa8d + m_rdCost.getCost(zbits);

        if (zcost < bidir2Nx2N.sa8dCost)
        {
            bidir2Nx2N.sa8dBits = zbits;
            bidir2Nx2N.sa8dCost = zcost;

            cu.setPUMv(0, mvzero, 0, 0);
            cu.m_mvd[0][0] = mvzero - mvp0;
            cu.m_mvpIdx[0][0] = (uint8_t)mvpIdx0;

            cu.setPUMv(1, mvzero, 0, 0);
            cu.m_mvd[1][0] = mvzero - mvp1;
            cu.m_mvpIdx[1][0] = (uint8_t)mvpIdx1;

            if (m_bChromaSa8d) /* real MC was already performed */
                bidir2Nx2N.predYuv.copyFromYuv(tmpPredYuv);
            else
                motionCompensation(cu, pu, bidir2Nx2N.predYuv, true, m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400);
        }
        else if (m_bChromaSa8d && (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400))
        {
            /* recover overwritten motion vectors */
            cu.m_mv[0][0] = bestME[0].mv;
            cu.m_mv[1][0] = bestME[1].mv;
        }
    }
}

void Analysis::encodeResidue(const CUData& ctu, const CUGeom& cuGeom)
{
    if (cuGeom.depth < ctu.m_cuDepth[cuGeom.absPartIdx] && cuGeom.depth < ctu.m_encData->m_param->maxCUDepth)
    {
        for (uint32_t subPartIdx = 0; subPartIdx < 4; subPartIdx++)
        {
            const CUGeom& childGeom = *(&cuGeom + cuGeom.childOffset + subPartIdx);
            if (childGeom.flags & CUGeom::PRESENT)
                encodeResidue(ctu, childGeom);
        }
        return;
    }

    uint32_t absPartIdx = cuGeom.absPartIdx;
    int sizeIdx = cuGeom.log2CUSize - 2;

    /* reuse the bestMode data structures at the current depth */
    Mode *bestMode = m_modeDepth[cuGeom.depth].bestMode;
    CUData& cu = bestMode->cu;

    cu.copyFromPic(ctu, cuGeom, m_csp);

    PicYuv& reconPic = *m_frame->m_reconPic;

    Yuv& fencYuv = m_modeDepth[cuGeom.depth].fencYuv;
    if (cuGeom.depth)
        m_modeDepth[0].fencYuv.copyPartToYuv(fencYuv, absPartIdx);
    X265_CHECK(bestMode->fencYuv == &fencYuv, "invalid fencYuv\n");

    if (cu.isIntra(0))
    {
        ProfileCUScope(ctu, intraRDOElapsedTime[cuGeom.depth], countIntraRDO[cuGeom.depth]); // not really RDO, but close enough
        
        uint32_t tuDepthRange[2];
        cu.getIntraTUQtDepthRange(tuDepthRange, 0);

        residualTransformQuantIntra(*bestMode, cuGeom, 0, 0, tuDepthRange);
        if (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400)
        {
            getBestIntraModeChroma(*bestMode, cuGeom);
            residualQTIntraChroma(*bestMode, cuGeom, 0, 0);
        }
    }
    else // if (cu.isInter(0))
    {
        ProfileCUScope(ctu, interRDOElapsedTime[cuGeom.depth], countInterRDO[cuGeom.depth]); // not really RDO, but close enough

        X265_CHECK(!ctu.isSkipped(absPartIdx), "skip not expected prior to transform\n");

        /* Calculate residual for current CU part into depth sized resiYuv */

        ShortYuv& resiYuv = m_rqt[cuGeom.depth].tmpResiYuv;

        /* at RD 0, the prediction pixels are accumulated into the top depth predYuv */
        Yuv& predYuv = m_modeDepth[0].bestMode->predYuv;
        pixel* predY = predYuv.getLumaAddr(absPartIdx);

        primitives.cu[sizeIdx].sub_ps(resiYuv.m_buf[0], resiYuv.m_size,
                                      fencYuv.m_buf[0], predY,
                                      fencYuv.m_size, predYuv.m_size);

        if (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400)
        {
            pixel* predU = predYuv.getCbAddr(absPartIdx);
            pixel* predV = predYuv.getCrAddr(absPartIdx);
            primitives.chroma[m_csp].cu[sizeIdx].sub_ps(resiYuv.m_buf[1], resiYuv.m_csize,
                                                 fencYuv.m_buf[1], predU,
                                                 fencYuv.m_csize, predYuv.m_csize);

            primitives.chroma[m_csp].cu[sizeIdx].sub_ps(resiYuv.m_buf[2], resiYuv.m_csize,
                                                 fencYuv.m_buf[2], predV,
                                                 fencYuv.m_csize, predYuv.m_csize);
        }

        uint32_t tuDepthRange[2];
        cu.getInterTUQtDepthRange(tuDepthRange, 0);

        residualTransformQuantInter(*bestMode, cuGeom, 0, 0, tuDepthRange);

        if (cu.m_mergeFlag[0] && cu.m_partSize[0] == SIZE_2Nx2N && !cu.getQtRootCbf(0))
            cu.setPredModeSubParts(MODE_SKIP);

        /* residualTransformQuantInter() wrote transformed residual back into
         * resiYuv. Generate the recon pixels by adding it to the prediction */

        if (cu.m_cbf[0][0])
            primitives.cu[sizeIdx].add_ps(reconPic.getLumaAddr(cu.m_cuAddr, absPartIdx), reconPic.m_stride,
                                          predY, resiYuv.m_buf[0], predYuv.m_size, resiYuv.m_size);
        else
            primitives.cu[sizeIdx].copy_pp(reconPic.getLumaAddr(cu.m_cuAddr, absPartIdx), reconPic.m_stride,
                                           predY, predYuv.m_size);
        if (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400)
        {
             pixel* predU = predYuv.getCbAddr(absPartIdx);
             pixel* predV = predYuv.getCrAddr(absPartIdx);
            if (cu.m_cbf[1][0])
                primitives.chroma[m_csp].cu[sizeIdx].add_ps(reconPic.getCbAddr(cu.m_cuAddr, absPartIdx), reconPic.m_strideC,
                                                        predU, resiYuv.m_buf[1], predYuv.m_csize, resiYuv.m_csize);
            else
                primitives.chroma[m_csp].cu[sizeIdx].copy_pp(reconPic.getCbAddr(cu.m_cuAddr, absPartIdx), reconPic.m_strideC,
                                                         predU, predYuv.m_csize);

            if (cu.m_cbf[2][0])
                primitives.chroma[m_csp].cu[sizeIdx].add_ps(reconPic.getCrAddr(cu.m_cuAddr, absPartIdx), reconPic.m_strideC,
                                                        predV, resiYuv.m_buf[2], predYuv.m_csize, resiYuv.m_csize);
            else
                primitives.chroma[m_csp].cu[sizeIdx].copy_pp(reconPic.getCrAddr(cu.m_cuAddr, absPartIdx), reconPic.m_strideC,
                                                         predV, predYuv.m_csize);
        }
    }

    cu.updatePic(cuGeom.depth, m_frame->m_fencPic->m_picCsp);
}

void Analysis::addSplitFlagCost(Mode& mode, uint32_t depth)
{
    if (m_param->rdLevel >= 3)
    {
        /* code the split flag (0 or 1) and update bit costs */
        mode.contexts.resetBits();
        mode.contexts.codeSplitFlag(mode.cu, 0, depth);
        uint32_t bits = mode.contexts.getNumberOfWrittenBits();
        mode.totalBits += bits;
        updateModeCost(mode);
    }
    else if (m_param->rdLevel <= 1)
    {
        mode.sa8dBits++;
        mode.sa8dCost = m_rdCost.calcRdSADCost((uint32_t)mode.distortion, mode.sa8dBits);
    }
    else
    {
        mode.totalBits++;
        updateModeCost(mode);
    }
}

uint32_t Analysis::topSkipMinDepth(const CUData& parentCTU, const CUGeom& cuGeom)
{
    /* Do not attempt to code a block larger than the largest block in the
     * co-located CTUs in L0 and L1 */
    int currentQP = parentCTU.m_qp[0];
    int previousQP = currentQP;
    uint32_t minDepth0 = 4, minDepth1 = 4;
    uint32_t sum = 0;
    int numRefs = 0;
    if (m_slice->m_numRefIdx[0])
    {
        numRefs++;
        const CUData& cu = *m_slice->m_refFrameList[0][0]->m_encData->getPicCTU(parentCTU.m_cuAddr);
        previousQP = cu.m_qp[0];
        if (!cu.m_cuDepth[cuGeom.absPartIdx])
            return 0;
        for (uint32_t i = 0; i < cuGeom.numPartitions; i += 4)
        {
            uint32_t d = cu.m_cuDepth[cuGeom.absPartIdx + i];
            minDepth0 = X265_MIN(d, minDepth0);
            sum += d;
        }
    }
    if (m_slice->m_numRefIdx[1])
    {
        numRefs++;
        const CUData& cu = *m_slice->m_refFrameList[1][0]->m_encData->getPicCTU(parentCTU.m_cuAddr);
        if (!cu.m_cuDepth[cuGeom.absPartIdx])
            return 0;
        for (uint32_t i = 0; i < cuGeom.numPartitions; i += 4)
        {
            uint32_t d = cu.m_cuDepth[cuGeom.absPartIdx + i];
            minDepth1 = X265_MIN(d, minDepth1);
            sum += d;
        }
    }
    if (!numRefs)
        return 0;

    uint32_t minDepth = X265_MIN(minDepth0, minDepth1);
    uint32_t thresh = minDepth * numRefs * (cuGeom.numPartitions >> 2);

    /* allow block size growth if QP is raising or avg depth is
     * less than 1.5 of min depth */
    if (minDepth && currentQP >= previousQP && (sum <= thresh + (thresh >> 1)))
        minDepth -= 1;

    return minDepth;
}

/* returns true if recursion should be stopped */
bool Analysis::recursionDepthCheck(const CUData& parentCTU, const CUGeom& cuGeom, const Mode& bestMode)
{
    /* early exit when the RD cost of best mode at depth n is less than the sum
     * of average of RD cost of the neighbor CU's(above, aboveleft, aboveright,
     * left, colocated) and avg cost of that CU at depth "n" with weightage for
     * each quantity */

    uint32_t depth = cuGeom.depth;
    FrameData& curEncData = *m_frame->m_encData;
    FrameData::RCStatCU& cuStat = curEncData.m_cuStat[parentCTU.m_cuAddr];
    uint64_t cuCost = cuStat.avgCost[depth] * cuStat.count[depth];
    uint64_t cuCount = cuStat.count[depth];

    uint64_t neighCost = 0, neighCount = 0;
    const CUData* above = parentCTU.m_cuAbove;
    if (above)
    {
        FrameData::RCStatCU& astat = curEncData.m_cuStat[above->m_cuAddr];
        neighCost += astat.avgCost[depth] * astat.count[depth];
        neighCount += astat.count[depth];

        const CUData* aboveLeft = parentCTU.m_cuAboveLeft;
        if (aboveLeft)
        {
            FrameData::RCStatCU& lstat = curEncData.m_cuStat[aboveLeft->m_cuAddr];
            neighCost += lstat.avgCost[depth] * lstat.count[depth];
            neighCount += lstat.count[depth];
        }

        const CUData* aboveRight = parentCTU.m_cuAboveRight;
        if (aboveRight)
        {
            FrameData::RCStatCU& rstat = curEncData.m_cuStat[aboveRight->m_cuAddr];
            neighCost += rstat.avgCost[depth] * rstat.count[depth];
            neighCount += rstat.count[depth];
        }
    }
    const CUData* left = parentCTU.m_cuLeft;
    if (left)
    {
        FrameData::RCStatCU& nstat = curEncData.m_cuStat[left->m_cuAddr];
        neighCost += nstat.avgCost[depth] * nstat.count[depth];
        neighCount += nstat.count[depth];
    }

    // give 60% weight to all CU's and 40% weight to neighbour CU's
    if (neighCount + cuCount)
    {
        uint64_t avgCost = ((3 * cuCost) + (2 * neighCost)) / ((3 * cuCount) + (2 * neighCount));
        uint64_t curCost = m_param->rdLevel > 1 ? bestMode.rdCost : bestMode.sa8dCost;
        if (curCost < avgCost && avgCost)
            return true;
    }

    return false;
}

bool Analysis::complexityCheckCU(const Mode& bestMode)
{
    uint32_t mean = 0;
    uint32_t homo = 0;
    uint32_t cuSize = bestMode.fencYuv->m_size;
    for (uint32_t y = 0; y < cuSize; y++) {
        for (uint32_t x = 0; x < cuSize; x++) {
            mean += (bestMode.fencYuv->m_buf[0][y * cuSize + x]);
        }
    }
    mean = mean / (cuSize * cuSize);
    for (uint32_t y = 0 ; y < cuSize; y++){
        for (uint32_t x = 0 ; x < cuSize; x++){
            homo += abs(int(bestMode.fencYuv->m_buf[0][y * cuSize + x] - mean));
        }
    }
    homo = homo / (cuSize * cuSize);

    if (homo < (.1 * mean))
        return true;

    return false;
}

int Analysis::calculateQpforCuSize(const CUData& ctu, const CUGeom& cuGeom, int32_t complexCheck, double baseQp)
{
    FrameData& curEncData = *m_frame->m_encData;
    double qp = baseQp >= 0 ? baseQp : curEncData.m_cuStat[ctu.m_cuAddr].baseQp;

    if (m_param->analysisMultiPassDistortion && m_param->rc.bStatRead)
    {
        m_multipassAnalysis = (analysis2PassFrameData*)m_frame->m_analysis2Pass.analysisFramedata;
        if ((m_multipassAnalysis->threshold[ctu.m_cuAddr] < 0.9 || m_multipassAnalysis->threshold[ctu.m_cuAddr] > 1.1)
            && m_multipassAnalysis->highDistortionCtuCount && m_multipassAnalysis->lowDistortionCtuCount)
            qp += m_multipassAnalysis->offset[ctu.m_cuAddr];
    }

    int loopIncr;
    if (m_param->rc.qgSize == 8)
        loopIncr = 8;
    else
        loopIncr = 16;
    /* Use cuTree offsets if cuTree enabled and frame is referenced, else use AQ offsets */
    bool isReferenced = IS_REFERENCED(m_frame);
    double *qpoffs;
    if (complexCheck)
        qpoffs = m_frame->m_lowres.qpAqOffset;
    else
        qpoffs = (isReferenced && m_param->rc.cuTree) ? m_frame->m_lowres.qpCuTreeOffset : m_frame->m_lowres.qpAqOffset;
    if (qpoffs)
    {
        uint32_t width = m_frame->m_fencPic->m_picWidth;
        uint32_t height = m_frame->m_fencPic->m_picHeight;
        uint32_t block_x = ctu.m_cuPelX + g_zscanToPelX[cuGeom.absPartIdx];
        uint32_t block_y = ctu.m_cuPelY + g_zscanToPelY[cuGeom.absPartIdx];
        uint32_t maxCols = (m_frame->m_fencPic->m_picWidth + (loopIncr - 1)) / loopIncr;
        uint32_t blockSize = m_param->maxCUSize >> cuGeom.depth;
        double qp_offset = 0;
        uint32_t cnt = 0;
        uint32_t idx;

        for (uint32_t block_yy = block_y; block_yy < block_y + blockSize && block_yy < height; block_yy += loopIncr)
        {
            for (uint32_t block_xx = block_x; block_xx < block_x + blockSize && block_xx < width; block_xx += loopIncr)
            {
                idx = ((block_yy / loopIncr) * (maxCols)) + (block_xx / loopIncr);
                qp_offset += qpoffs[idx];
                cnt++;
            }
        }

        qp_offset /= cnt;
        qp += qp_offset;
        if (complexCheck)
        {
            int32_t offset = (int32_t)(qp_offset * 100 + .5);
            double threshold = (1 - ((x265_ADAPT_RD_STRENGTH - m_param->dynamicRd) * 0.5));
            int32_t max_threshold = (int32_t)(threshold * 100 + .5);
            if (offset < max_threshold)
                return 1;
            else
                return 0;
        }
    }

    return x265_clip3(m_param->rc.qpMin, m_param->rc.qpMax, (int)(qp + 0.5));
}

void Analysis::normFactor(const pixel* src, uint32_t blockSize, CUData& ctu, int qp, TextType ttype)
{
    static const int ssim_c1 = (int)(.01 * .01 * PIXEL_MAX * PIXEL_MAX * 64 + .5); // 416
    static const int ssim_c2 = (int)(.03 * .03 * PIXEL_MAX * PIXEL_MAX * 64 * 63 + .5); // 235963
    int shift = (X265_DEPTH - 8);

    double s = 1 + 0.005 * qp;

    // Calculate denominator of normalization factor
    uint64_t fDc_den = 0, fAc_den = 0;

    // 1. Calculate dc component
    uint64_t z_o = 0;
    for (uint32_t block_yy = 0; block_yy < blockSize; block_yy += 4)
    {
        for (uint32_t block_xx = 0; block_xx < blockSize; block_xx += 4)
        {
            uint32_t temp = src[block_yy * blockSize + block_xx] >> shift;
            z_o += temp * temp; // 2 * (Z(0)) pow(2)
        }
    }
    fDc_den = (2 * z_o)  + (blockSize * blockSize * ssim_c1); // 2 * (Z(0)) pow(2) + N * C1
    fDc_den /= ((blockSize >> 2) * (blockSize >> 2));

    // 2. Calculate ac component
    uint64_t z_k = 0;
    for (uint32_t block_yy = 0; block_yy < blockSize; block_yy += 1)
    {
        for (uint32_t block_xx = 0; block_xx < blockSize; block_xx += 1)
        {
            uint32_t temp = src[block_yy * blockSize + block_xx] >> shift;
            z_k += temp * temp;
        }
    }

    // Remove the DC part
    z_k -= z_o;

    fAc_den = z_k + int(s * z_k) + ssim_c2;
    fAc_den /= ((blockSize >> 2) * (blockSize >> 2));

    ctu.m_fAc_den[ttype] = fAc_den;
    ctu.m_fDc_den[ttype] = fDc_den;
}

void Analysis::calculateNormFactor(CUData& ctu, int qp)
{
    const pixel* srcY = m_modeDepth[0].fencYuv.m_buf[0];
    uint32_t blockSize = m_modeDepth[0].fencYuv.m_size;

    normFactor(srcY, blockSize, ctu, qp, TEXT_LUMA);

    if (m_csp != X265_CSP_I400 && m_frame->m_fencPic->m_picCsp != X265_CSP_I400)
    {
        const pixel* srcU = m_modeDepth[0].fencYuv.m_buf[1];
        const pixel* srcV = m_modeDepth[0].fencYuv.m_buf[2];
        uint32_t blockSizeC = m_modeDepth[0].fencYuv.m_csize;

        normFactor(srcU, blockSizeC, ctu, qp, TEXT_CHROMA_U);
        normFactor(srcV, blockSizeC, ctu, qp, TEXT_CHROMA_V);
    }
}

int Analysis::findSameContentRefCount(const CUData& parentCTU, const CUGeom& cuGeom)
{
    int sameContentRef = 0;
    int m_curPoc = parentCTU.m_slice->m_poc;
    int prevChange = m_prevCtuInfoChange[cuGeom.absPartIdx];
    int numPredDir = m_slice->isInterP() ? 1 : 2;
    for (int list = 0; list < numPredDir; list++)
    {
        for (int i = 0; i < m_frame->m_encData->m_slice->m_numRefIdx[list]; i++)
        {
            int refPoc = m_frame->m_encData->m_slice->m_refFrameList[list][i]->m_poc;
            int refPrevChange = m_frame->m_encData->m_slice->m_refFrameList[list][i]->m_addOnPrevChange[parentCTU.m_cuAddr][cuGeom.absPartIdx];
            if ((refPoc < prevChange && refPoc < m_curPoc) || (refPoc > m_curPoc && prevChange < m_curPoc && refPrevChange > m_curPoc) || ((refPoc == prevChange) && (m_additionalCtuInfo[cuGeom.absPartIdx] == CTU_INFO_CHANGE)))
                sameContentRef++;    /* Content changed */
        }
    }
    return sameContentRef;
}