IntraSearch.cpp 66.3 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
/* The copyright in this software is being made available under the BSD
 * License, included below. This software may be subject to other third party
 * and contributor rights, including patent rights, and no such rights are
 * granted under this license.
 *
 * Copyright (c) 2010-2018, ITU/ISO/IEC
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 *  * Redistributions of source code must retain the above copyright notice,
 *    this list of conditions and the following disclaimer.
 *  * Redistributions in binary form must reproduce the above copyright notice,
 *    this list of conditions and the following disclaimer in the documentation
 *    and/or other materials provided with the distribution.
 *  * Neither the name of the ITU/ISO/IEC nor the names of its contributors may
 *    be used to endorse or promote products derived from this software without
 *    specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS
 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
 * THE POSSIBILITY OF SUCH DAMAGE.
 */

/** \file     EncSearch.cpp
 *  \brief    encoder intra search class
 */

#include "IntraSearch.h"

#include "EncModeCtrl.h"

#include "CommonLib/CommonDef.h"
#include "CommonLib/Rom.h"
#include "CommonLib/Picture.h"
#include "CommonLib/UnitTools.h"

#include "CommonLib/dtrace_next.h"
#include "CommonLib/dtrace_buffer.h"

#include <math.h>
#include <limits>

 //! \ingroup EncoderLib
 //! \{

IntraSearch::IntraSearch()
  : m_pSplitCS      (nullptr)
  , m_pFullCS       (nullptr)
  , m_pBestCS       (nullptr)
  , m_pcEncCfg      (nullptr)
  , m_pcTrQuant     (nullptr)
  , m_pcRdCost      (nullptr)
  , m_CABACEstimator(nullptr)
  , m_CtxCache      (nullptr)
  , m_isInitialized (false)
{
  for( uint32_t ch = 0; ch < MAX_NUM_TBLOCKS; ch++ )
  {
    m_pSharedPredTransformSkip[ch] = nullptr;
  }
}


void IntraSearch::destroy()
{
  CHECK( !m_isInitialized, "Not initialized" );

  if( m_pcEncCfg )
  {
    bool BTnoRQT = m_pcEncCfg->getQTBT();


    const uint32_t uiNumLayersToAllocateSplit = BTnoRQT ? 1 : m_pcEncCfg->getQuadtreeTULog2MaxSize() - m_pcEncCfg->getQuadtreeTULog2MinSize() + 1;
    const uint32_t uiNumLayersToAllocateFull  = BTnoRQT ? 1 : m_pcEncCfg->getQuadtreeTULog2MaxSize() - m_pcEncCfg->getQuadtreeTULog2MinSize() + 1;
    const int uiNumSaveLayersToAllocate = 2;

    for( uint32_t layer = 0; layer < uiNumSaveLayersToAllocate; layer++ )
    {
      m_pSaveCS[layer]->destroy();
      delete m_pSaveCS[layer];
    }

    uint32_t numWidths  = gp_sizeIdxInfo->numWidths();
    uint32_t numHeights = gp_sizeIdxInfo->numHeights();

    for( uint32_t width = 0; width < numWidths; width++ )
    {
      for( uint32_t height = 0; height < numHeights; height++ )
      {
        if( ( BTnoRQT || width == height ) && gp_sizeIdxInfo->isCuSize( gp_sizeIdxInfo->sizeFrom( width ) ) && gp_sizeIdxInfo->isCuSize( gp_sizeIdxInfo->sizeFrom( height ) ) )
        {
          for( uint32_t layer = 0; layer < uiNumLayersToAllocateSplit; layer++ )
          {
            m_pSplitCS[width][height][layer]->destroy();

            delete m_pSplitCS[width][height][layer];
          }

          for( uint32_t layer = 0; layer < uiNumLayersToAllocateFull; layer++ )
          {
            m_pFullCS[width][height][layer]->destroy();

            delete m_pFullCS[width][height][layer];
          }

          delete[] m_pSplitCS[width][height];
          delete[] m_pFullCS [width][height];

          m_pBestCS[width][height]->destroy();
          m_pTempCS[width][height]->destroy();

          delete m_pTempCS[width][height];
          delete m_pBestCS[width][height];
        }
      }

      delete[] m_pSplitCS[width];
      delete[] m_pFullCS [width];

      delete[] m_pTempCS[width];
      delete[] m_pBestCS[width];
    }

    delete[] m_pSplitCS;
    delete[] m_pFullCS;

    delete[] m_pBestCS;
    delete[] m_pTempCS;

    delete[] m_pSaveCS;
  }

  m_pSplitCS = m_pFullCS = nullptr;

  m_pBestCS = m_pTempCS = nullptr;

  m_pSaveCS = nullptr;

  for( uint32_t ch = 0; ch < MAX_NUM_TBLOCKS; ch++ )
  {
    delete[] m_pSharedPredTransformSkip[ch];
    m_pSharedPredTransformSkip[ch] = nullptr;
  }

  m_isInitialized = false;
}

IntraSearch::~IntraSearch()
{
  if( m_isInitialized )
  {
    destroy();
  }
}

void IntraSearch::init( EncCfg*        pcEncCfg,
                        TrQuant*       pcTrQuant,
                        RdCost*        pcRdCost,
                        CABACWriter*   CABACEstimator,
                        CtxCache*      ctxCache,
                        const uint32_t     maxCUWidth,
                        const uint32_t     maxCUHeight,
                        const uint32_t     maxTotalCUDepth
)
{
  CHECK(m_isInitialized, "Already initialized");
  m_pcEncCfg                     = pcEncCfg;
  m_pcTrQuant                    = pcTrQuant;
  m_pcRdCost                     = pcRdCost;
  m_CABACEstimator               = CABACEstimator;
  m_CtxCache                     = ctxCache;

  const ChromaFormat cform = pcEncCfg->getChromaFormatIdc();

  IntraPrediction::init( cform, pcEncCfg->getBitDepth( CHANNEL_TYPE_LUMA ) );

  for( uint32_t ch = 0; ch < MAX_NUM_TBLOCKS; ch++ )
  {
    m_pSharedPredTransformSkip[ch] = new Pel[MAX_CU_SIZE * MAX_CU_SIZE];
  }

  uint32_t numWidths  = gp_sizeIdxInfo->numWidths();
  uint32_t numHeights = gp_sizeIdxInfo->numHeights();

  bool BTnoRQT = m_pcEncCfg->getQTBT();

  const uint32_t uiNumLayersToAllocateSplit = BTnoRQT ? 1 : pcEncCfg->getQuadtreeTULog2MaxSize() - pcEncCfg->getQuadtreeTULog2MinSize() + 1;
  const uint32_t uiNumLayersToAllocateFull  = BTnoRQT ? 1 : pcEncCfg->getQuadtreeTULog2MaxSize() - pcEncCfg->getQuadtreeTULog2MinSize() + 1;

  m_pBestCS = new CodingStructure**[numWidths];
  m_pTempCS = new CodingStructure**[numWidths];

  m_pFullCS  = new CodingStructure***[numWidths];
  m_pSplitCS = new CodingStructure***[numWidths];

  for( uint32_t width = 0; width < numWidths; width++ )
  {
    m_pBestCS[width] = new CodingStructure*[numHeights];
    m_pTempCS[width] = new CodingStructure*[numHeights];

    m_pFullCS [width] = new CodingStructure**[numHeights];
    m_pSplitCS[width] = new CodingStructure**[numHeights];

    for( uint32_t height = 0; height < numHeights; height++ )
    {
      if( ( BTnoRQT || width == height ) && gp_sizeIdxInfo->isCuSize( gp_sizeIdxInfo->sizeFrom( width ) ) && gp_sizeIdxInfo->isCuSize( gp_sizeIdxInfo->sizeFrom( height ) ) )
      {
        m_pBestCS[width][height] = new CodingStructure( m_unitCache.cuCache, m_unitCache.puCache, m_unitCache.tuCache );
        m_pTempCS[width][height] = new CodingStructure( m_unitCache.cuCache, m_unitCache.puCache, m_unitCache.tuCache );

        m_pBestCS[width][height]->create( m_pcEncCfg->getChromaFormatIdc(), Area( 0, 0, gp_sizeIdxInfo->sizeFrom( width ), gp_sizeIdxInfo->sizeFrom( height ) ), false );
        m_pTempCS[width][height]->create( m_pcEncCfg->getChromaFormatIdc(), Area( 0, 0, gp_sizeIdxInfo->sizeFrom( width ), gp_sizeIdxInfo->sizeFrom( height ) ), false );
        m_pFullCS [width][height] = new CodingStructure*[uiNumLayersToAllocateFull];
        m_pSplitCS[width][height] = new CodingStructure*[uiNumLayersToAllocateSplit];

        for( uint32_t layer = 0; layer < uiNumLayersToAllocateFull; layer++ )
        {
          m_pFullCS [width][height][layer] = new CodingStructure( m_unitCache.cuCache, m_unitCache.puCache, m_unitCache.tuCache );

          m_pFullCS [width][height][layer]->create( m_pcEncCfg->getChromaFormatIdc(), Area( 0, 0, gp_sizeIdxInfo->sizeFrom( width ), gp_sizeIdxInfo->sizeFrom( height ) ), false );
        }

        for( uint32_t layer = 0; layer < uiNumLayersToAllocateSplit; layer++ )
        {
          m_pSplitCS[width][height][layer] = new CodingStructure( m_unitCache.cuCache, m_unitCache.puCache, m_unitCache.tuCache );

          m_pSplitCS[width][height][layer]->create( m_pcEncCfg->getChromaFormatIdc(), Area( 0, 0, gp_sizeIdxInfo->sizeFrom( width ), gp_sizeIdxInfo->sizeFrom( height ) ), false );
        }
      }
      else
      {
        m_pBestCS[width][height] = nullptr;
        m_pTempCS[width][height] = nullptr;

        m_pFullCS [width][height] = nullptr;
        m_pSplitCS[width][height] = nullptr;
      }
    }
  }

  const int uiNumSaveLayersToAllocate = 2;

  m_pSaveCS = new CodingStructure*[uiNumSaveLayersToAllocate];

  for( uint32_t depth = 0; depth < uiNumSaveLayersToAllocate; depth++ )
  {
    m_pSaveCS[depth] = new CodingStructure( m_unitCache.cuCache, m_unitCache.puCache, m_unitCache.tuCache );
    m_pSaveCS[depth]->create( UnitArea( cform, Area( 0, 0, maxCUWidth, maxCUHeight ) ), false );
  }

  m_isInitialized = true;
}


//////////////////////////////////////////////////////////////////////////
// INTRA PREDICTION
//////////////////////////////////////////////////////////////////////////

void IntraSearch::estIntraPredLumaQT( CodingUnit &cu, Partitioner &partitioner )
{
  CodingStructure       &cs            = *cu.cs;
  const SPS             &sps           = *cs.sps;
  const uint32_t             uiWidthBit    = cs.pcv->rectCUs ? g_aucLog2[partitioner.currArea().lwidth() ] : CU::getIntraSizeIdx(cu);
  const uint32_t             uiHeightBit   =                   g_aucLog2[partitioner.currArea().lheight()];

  // Lambda calculation at equivalent Qp of 4 is recommended because at that Qp, the quantization divisor is 1.
  const double sqrtLambdaForFirstPass = m_pcRdCost->getMotionLambda(cu.transQuantBypass) / double(1 << SCALE_BITS);


  //===== loop over partitions =====

  const TempCtx ctxStart          ( m_CtxCache, m_CABACEstimator->getCtx() );
  const TempCtx ctxStartIntraMode ( m_CtxCache, SubCtx( Ctx::IPredMode[CHANNEL_TYPE_LUMA],        m_CABACEstimator->getCtx() ) );

  CHECK( !cu.firstPU, "CU has no PUs" );
  const bool keepResi   = cs.pps->getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() || KEEP_PRED_AND_RESI_SIGNALS;


  uint32_t extraModes = 0; // add two extra modes, which would be used after uiMode <= DC_IDX is removed for cu.nsstIdx == 3


  const int width   = partitioner.currArea().lwidth();
  const int height  = partitioner.currArea().lheight();

  // Marking EMT usage for faster EMT
  // 0: EMT is either not applicable for current CU (cuWidth > EMT_INTRA_MAX_CU or cuHeight > EMT_INTRA_MAX_CU), not active in the config file or the fast decision algorithm is not used in this case
  // 1: EMT fast algorithm can be applied for the current CU, and the DCT2 is being checked
  // 2: EMT is being checked for current CU. Stored results of DCT2 can be utilized for speedup
  uint8_t emtUsageFlag = 0;
  const int maxSizeEMT = cs.pcv->noRQT ? EMT_INTRA_MAX_CU_WITH_QTBT : EMT_INTRA_MAX_CU;
  if( width <= maxSizeEMT && height <= maxSizeEMT && sps.getSpsNext().getUseIntraEMT() )
  {
    emtUsageFlag = cu.emtFlag == 1 ? 2 : 1;
  }

  bool isAllIntra = m_pcEncCfg->getIntraPeriod() == 1;

  if( cs.pcv->rectCUs )
  {
    if( width * height < 64 && !isAllIntra )
    {
      emtUsageFlag = 0; //this forces the recalculation of the candidates list. Why is this necessary? (to be checked)
    }
  }

  static_vector<uint32_t,   FAST_UDI_MAX_RDMODE_NUM> uiHadModeList;
  static_vector<double, FAST_UDI_MAX_RDMODE_NUM> CandCostList;
  static_vector<double, FAST_UDI_MAX_RDMODE_NUM> CandHadList;

320
321
322
323
324
#if JVET_L0283_MULTI_REF_LINE
  static_vector<int, FAST_UDI_MAX_RDMODE_NUM> extendRefList;
  static_vector<int, FAST_UDI_MAX_RDMODE_NUM>* nullList = NULL;
#endif

325
326
327
328
329
330
  auto &pu = *cu.firstPU;
  int puIndex = 0;
  {
    CandHadList.clear();
    CandCostList.clear();
    uiHadModeList.clear();
331
332
333
#if JVET_L0283_MULTI_REF_LINE
    extendRefList.clear();
#endif
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360

    CHECK(pu.cu != &cu, "PU is not contained in the CU");

    //===== determine set of modes to be tested (using prediction signal only) =====
    int numModesAvailable = NUM_LUMA_MODE; // total number of Intra modes
    static_vector< uint32_t, FAST_UDI_MAX_RDMODE_NUM > uiRdModeList;

    int numModesForFullRD = 3;
    if( cs.pcv->rectCUs )
    {
      numModesForFullRD = g_aucIntraModeNumFast_UseMPM_2D[uiWidthBit - MIN_CU_LOG2][uiHeightBit - MIN_CU_LOG2];
    }
    else
    {
      numModesForFullRD = m_pcEncCfg->getFastUDIUseMPMEnabled() ? g_aucIntraModeNumFast_UseMPM[uiWidthBit] : g_aucIntraModeNumFast_NotUseMPM[uiWidthBit];
      numModesForFullRD -= 1;
    }

#if INTRA_FULL_SEARCH
    numModesForFullRD = numModesAvailable;
#endif


    if( emtUsageFlag != 2 )
    {
      // this should always be true
      CHECK( !pu.Y().valid(), "PU is not valid" );
361
362
363
364
365
#if JVET_L0283_MULTI_REF_LINE
      bool isFirstLineOfCtu = (((pu.block(COMPONENT_Y).y)&((pu.cs->sps)->getMaxCUWidth() - 1)) == 0);
      int numOfPassesExtendRef = (isFirstLineOfCtu ? 1 : MRL_NUM_REF_LINES);
      pu.multiRefIdx = 0;
#endif
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424

      //===== init pattern for luma prediction =====
      initIntraPatternChType( cu, pu.Y(), IntraPrediction::useFilteredIntraRefSamples( COMPONENT_Y, pu, false, pu ) );
      if( numModesForFullRD != numModesAvailable )
      {
        CHECK( numModesForFullRD >= numModesAvailable, "Too many modes for full RD search" );

        const CompArea &area = pu.Y();

        PelBuf piOrg         = cs.getOrgBuf(area);
        PelBuf piPred        = cs.getPredBuf(area);

        DistParam distParam;

        const bool bUseHadamard = cu.transQuantBypass == 0;

        m_pcRdCost->setDistParam(distParam, piOrg, piPred, sps.getBitDepth(CHANNEL_TYPE_LUMA), COMPONENT_Y, bUseHadamard);

        distParam.applyWeight = false;

        bool bSatdChecked[NUM_INTRA_MODE];
        memset( bSatdChecked, 0, sizeof( bSatdChecked ) );

        {
          for( int modeIdx = 0; modeIdx < numModesAvailable; modeIdx++ )
          {
            uint32_t       uiMode = modeIdx;
            Distortion uiSad  = 0;

            // Skip checking extended Angular modes in the first round of SATD
            if( uiMode > DC_IDX && ( uiMode & 1 ) )
            {
              continue;
            }

            bSatdChecked[uiMode] = true;

            pu.intraDir[0] = modeIdx;

            if( useDPCMForFirstPassIntraEstimation( pu, uiMode ) )
            {
              encPredIntraDPCM( COMPONENT_Y, piOrg, piPred, uiMode );
            }
            else
            {
              predIntraAng( COMPONENT_Y, piPred, pu, IntraPrediction::useFilteredIntraRefSamples( COMPONENT_Y, pu, true, pu ) );
            }
            // use Hadamard transform here
            uiSad += distParam.distFunc(distParam);

            // NB xFracModeBitsIntra will not affect the mode for chroma that may have already been pre-estimated.
            m_CABACEstimator->getCtx() = SubCtx( Ctx::IPredMode[CHANNEL_TYPE_LUMA], ctxStartIntraMode );

            uint64_t fracModeBits = xFracModeBitsIntra(pu, uiMode, CHANNEL_TYPE_LUMA);

            double cost = ( double ) uiSad + ( double ) fracModeBits * sqrtLambdaForFirstPass;

            DTRACE( g_trace_ctx, D_INTRA_COST, "IntraHAD: %u, %llu, %f (%d)\n", uiSad, fracModeBits, cost, uiMode );

425
426
427
428
429
430
431
432
433
434
            updateCandList( uiMode, cost,  uiRdModeList, CandCostList
#if JVET_L0283_MULTI_REF_LINE
              , extendRefList, 0
#endif              
              , numModesForFullRD + extraModes );
            updateCandList(uiMode, (double) uiSad, uiHadModeList, CandHadList
#if JVET_L0283_MULTI_REF_LINE
              , *nullList, -1
#endif              
              , 3 + extraModes);
435
436
437
438
439
          }
        } // NSSTFlag

        // forget the extra modes
        uiRdModeList.resize( numModesForFullRD );
440
441
442
443
#if JVET_L0283_MULTI_REF_LINE
        CandCostList.resize(numModesForFullRD);
        extendRefList.resize(numModesForFullRD);
#endif
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
        static_vector<unsigned, FAST_UDI_MAX_RDMODE_NUM> parentCandList(FAST_UDI_MAX_RDMODE_NUM);
        std::copy_n(uiRdModeList.begin(), numModesForFullRD, parentCandList.begin());

        // Second round of SATD for extended Angular modes
        for (int modeIdx = 0; modeIdx < numModesForFullRD; modeIdx++)
        {
          unsigned parentMode = parentCandList[modeIdx];
          if (parentMode > (DC_IDX + 1) && parentMode < (NUM_LUMA_MODE - 1))
          {
            for (int subModeIdx = -1; subModeIdx <= 1; subModeIdx += 2)
            {
              unsigned mode = parentMode + subModeIdx;


              if (!bSatdChecked[mode])
              {
                pu.intraDir[0] = mode;

                if (useDPCMForFirstPassIntraEstimation(pu, mode))
                {
                  encPredIntraDPCM(COMPONENT_Y, piOrg, piPred, mode);
                }
                else
                {
                  predIntraAng(COMPONENT_Y, piPred, pu,
                               IntraPrediction::useFilteredIntraRefSamples(COMPONENT_Y, pu, true, pu));
                }
                // use Hadamard transform here
                Distortion sad = distParam.distFunc(distParam);

                // NB xFracModeBitsIntra will not affect the mode for chroma that may have already been pre-estimated.
                m_CABACEstimator->getCtx() = SubCtx(Ctx::IPredMode[CHANNEL_TYPE_LUMA], ctxStartIntraMode);

                uint64_t fracModeBits = xFracModeBitsIntra(pu, mode, CHANNEL_TYPE_LUMA);

                double cost = (double) sad + (double) fracModeBits * sqrtLambdaForFirstPass;

481
482
483
484
485
486
487
488
489
490
                updateCandList(mode, cost, uiRdModeList, CandCostList
#if JVET_L0283_MULTI_REF_LINE
                  , extendRefList, 0
#endif
                  , numModesForFullRD);
                updateCandList(mode, (double)sad, uiHadModeList, CandHadList
#if JVET_L0283_MULTI_REF_LINE
                  , *nullList, -1
#endif                 
                  , 3);
491
492
493
494
495
496

                bSatdChecked[mode] = true;
              }
            }
          }
        }
497
498
#if JVET_L0283_MULTI_REF_LINE
        pu.multiRefIdx = 1;
499
500
        const int  numMPMs = NUM_MOST_PROBABLE_MODES;
        unsigned  multiRefMPM [numMPMs];
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
        PU::getIntraMPMs(pu, multiRefMPM);
        for (int mRefNum = 1; mRefNum < numOfPassesExtendRef; mRefNum++)
        {
          int multiRefIdx = MULTI_REF_LINE_IDX[mRefNum];

          pu.multiRefIdx = multiRefIdx;
          {
            initIntraPatternChType(cu, pu.Y(), IntraPrediction::useFilteredIntraRefSamples(COMPONENT_Y, pu, false, pu));
          }
          for (int x = 0; x < numMPMs; x++)
          {
            uint32_t mode = multiRefMPM[x];
            {
              pu.intraDir[0] = mode;

              if (useDPCMForFirstPassIntraEstimation(pu, mode))
              {
                encPredIntraDPCM(COMPONENT_Y, piOrg, piPred, mode);
              }
              else
              {
                predIntraAng(COMPONENT_Y, piPred, pu, IntraPrediction::useFilteredIntraRefSamples(COMPONENT_Y, pu, true, pu));
              }

              // use Hadamard transform here
              Distortion sad = distParam.distFunc(distParam);

              // NB xFracModeBitsIntra will not affect the mode for chroma that may have already been pre-estimated.
              m_CABACEstimator->getCtx() = SubCtx(Ctx::IPredMode[CHANNEL_TYPE_LUMA], ctxStartIntraMode);

              uint64_t fracModeBits = xFracModeBitsIntra(pu, mode, CHANNEL_TYPE_LUMA);

              double cost = (double)sad + (double)fracModeBits * sqrtLambdaForFirstPass;
              updateCandList(mode, cost, uiRdModeList, CandCostList, extendRefList, multiRefIdx, numModesForFullRD);
            }
          }
        }
        CandCostList.resize(numModesForFullRD);
        extendRefList.resize(numModesForFullRD);
#endif
541
542
        if( m_pcEncCfg->getFastUDIUseMPMEnabled() )
        {
543
544
          const int numMPMs = NUM_MOST_PROBABLE_MODES;
          unsigned  uiPreds[numMPMs];
545

546
547
548
549
#if JVET_L0283_MULTI_REF_LINE
          pu.multiRefIdx = 0;
#endif

550
551
552
553
554
555
556
557
558
559
          const int numCand = PU::getIntraMPMs( pu, uiPreds );

          for( int j = 0; j < numCand; j++ )
          {
            bool mostProbableModeIncluded = false;
            int  mostProbableMode         = uiPreds[j];


            for( int i = 0; i < numModesForFullRD; i++ )
            {
560
561
562
#if JVET_L0283_MULTI_REF_LINE
              mostProbableModeIncluded |= (mostProbableMode == uiRdModeList[i] && extendRefList[i] == 0);
#else
563
              mostProbableModeIncluded |= ( mostProbableMode == uiRdModeList[i] );
564
#endif
565
566
567
            }
            if( !mostProbableModeIncluded )
            {
568
569
570
#if JVET_L0283_MULTI_REF_LINE
              extendRefList.push_back(0);
#endif
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
              numModesForFullRD++;
              uiRdModeList.push_back( mostProbableMode );
            }
          }
        }
      }
      else
      {
        for( int i = 0; i < numModesForFullRD; i++ )
        {
          uiRdModeList.push_back( i );
        }
      }
      if( emtUsageFlag == 1 )
      {
        // Store the modes to be checked with RD
        m_savedNumRdModes[puIndex] = numModesForFullRD;
        std::copy_n( uiRdModeList.begin(), numModesForFullRD, m_savedRdModeList[puIndex] );
589
590
591
#if JVET_L0283_MULTI_REF_LINE
        std::copy_n(extendRefList.begin(), numModesForFullRD, m_savedExtendRefList[puIndex]);
#endif
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
      }
    }
    else //emtUsage = 2 (here we potentially reduce the number of modes that will be full-RD checked)
    {
      if( isAllIntra && m_pcEncCfg->getFastIntraEMT() )
      {
        double thresholdSkipMode;
        if( cs.pcv->noRQT )
        {
          thresholdSkipMode = 1.0 + 1.4 / sqrt( ( double ) ( width*height ) );
        }
        else
        {
          switch( width )
          {
          case  4: thresholdSkipMode = 1.47; break; // Skip checking   4x4 Intra modes using the R-D cost in the DCT2-pass
          case  8: thresholdSkipMode = 1.28; break; // Skip checking   8x8 Intra modes using the R-D cost in the DCT2-pass
          case 16: thresholdSkipMode = 1.12; break; // Skip checking 16x16 Intra modes using the R-D cost in the DCT2-pass
          case 32: thresholdSkipMode = 1.06; break; // Skip checking 32x32 Intra modes using the R-D cost in the DCT2-pass
          default: thresholdSkipMode = 1.06; break; // Skip checking 32x32 Intra modes using the R-D cost in the DCT2-pass
          }
        }

        numModesForFullRD = 0;

        // Skip checking the modes with much larger R-D cost than the best mode
        for( int i = 0; i < m_savedNumRdModes[puIndex]; i++ )
        {
          if( m_modeCostStore[puIndex][i] <= thresholdSkipMode * m_bestModeCostStore[puIndex] )
          {
            uiRdModeList.push_back( m_savedRdModeList[puIndex][i] );
623
624
625
#if JVET_L0283_MULTI_REF_LINE
            extendRefList.push_back(m_savedExtendRefList[puIndex][i]);
#endif
626
627
628
629
630
631
632
633
634
635
            numModesForFullRD++;
          }
        }
      }
      else //this is necessary because we skip the candidates list calculation, since it was already obtained for the DCT-II. Now we load it
      {
        // Restore the modes to be checked with RD
        numModesForFullRD = m_savedNumRdModes[puIndex];
        uiRdModeList.resize( numModesForFullRD );
        std::copy_n( m_savedRdModeList[puIndex], m_savedNumRdModes[puIndex], uiRdModeList.begin() );
636
637
638
639
640
#if JVET_L0283_MULTI_REF_LINE
        CandCostList.resize(numModesForFullRD);
        extendRefList.resize(numModesForFullRD);
        std::copy_n(m_savedExtendRefList[puIndex], m_savedNumRdModes[puIndex], extendRefList.begin());
#endif
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
      }
    }


    CHECK( numModesForFullRD != uiRdModeList.size(), "Inconsistent state!" );

    // after this point, don't use numModesForFullRD

    // PBINTRA fast
    if( m_pcEncCfg->getUsePbIntraFast() && !cs.slice->isIntra() && cu.partSize == SIZE_2Nx2N && uiRdModeList.size() < numModesAvailable && emtUsageFlag != 2 )
    {
      if( CandHadList.size() < 3 || CandHadList[2] > cs.interHad * PBINTRA_RATIO )
      {
        uiRdModeList.resize( std::min<size_t>( uiRdModeList.size(), 2 ) );
      }
      if( CandHadList.size() < 2 || CandHadList[1] > cs.interHad * PBINTRA_RATIO )
      {
        uiRdModeList.resize( std::min<size_t>( uiRdModeList.size(), 1 ) );
      }
      if( CandHadList.size() < 1 || CandHadList[0] > cs.interHad * PBINTRA_RATIO )
      {
        cs.dist = std::numeric_limits<Distortion>::max();
        cs.interHad = 0;

        //===== reset context models =====
        m_CABACEstimator->getCtx() = SubCtx( Ctx::IPredMode       [CHANNEL_TYPE_LUMA], ctxStartIntraMode );

        return;
      }
    }

    //===== check modes (using r-d costs) =====
    uint32_t       uiBestPUMode  = 0;
674
675
676
#if JVET_L0283_MULTI_REF_LINE
    int            bestExtendRef = 0;
#endif
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693

    CodingStructure *csTemp = m_pTempCS[gp_sizeIdxInfo->idxFrom( cu.lwidth() )][gp_sizeIdxInfo->idxFrom( cu.lheight() )];
    CodingStructure *csBest = m_pBestCS[gp_sizeIdxInfo->idxFrom( cu.lwidth() )][gp_sizeIdxInfo->idxFrom( cu.lheight() )];

    csTemp->slice = cs.slice;
    csBest->slice = cs.slice;
    csTemp->initStructData();
    csBest->initStructData();

    // just to be sure
    numModesForFullRD = ( int ) uiRdModeList.size();
    for (uint32_t uiMode = 0; uiMode < numModesForFullRD; uiMode++)
    {
      // set luma prediction mode
      uint32_t uiOrgMode = uiRdModeList[uiMode];

      pu.intraDir[0] = uiOrgMode;
694
695
696
697
698
#if JVET_L0283_MULTI_REF_LINE
      int multiRefIdx = extendRefList[uiMode];
      pu.multiRefIdx  = multiRefIdx;
      CHECK(pu.multiRefIdx && (pu.intraDir[0] == DC_IDX || pu.intraDir[0] == PLANAR_IDX), "ERL");
#endif
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722


      // set context models
      m_CABACEstimator->getCtx() = ctxStart;

      // determine residual for partition
      cs.initSubStructure( *csTemp, partitioner.chType, cs.area, true );

      xRecurIntraCodingLumaQT( *csTemp, partitioner );

      if( emtUsageFlag == 1 && m_pcEncCfg->getFastIntraEMT() )
      {
        m_modeCostStore[puIndex][uiMode] = csTemp->cost; //cs.cost;
      }


      DTRACE( g_trace_ctx, D_INTRA_COST, "IntraCost T %f (%d) \n", csTemp->cost, uiOrgMode );

      // check r-d cost
      if( csTemp->cost < csBest->cost )
      {
        std::swap( csTemp, csBest );

        uiBestPUMode  = uiOrgMode;
723
724
725
#if JVET_L0283_MULTI_REF_LINE
        bestExtendRef = multiRefIdx;
#endif
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740

        if( ( emtUsageFlag == 1 ) && m_pcEncCfg->getFastIntraEMT() )
        {
          m_bestModeCostStore[puIndex] = csBest->cost; //cs.cost;
        }
      }

      csTemp->releaseIntermediateData();
    } // Mode loop

    cs.useSubStructure( *csBest, partitioner.chType, pu.singleChan( CHANNEL_TYPE_LUMA ), KEEP_PRED_AND_RESI_SIGNALS, true, keepResi, keepResi );

    csBest->releaseIntermediateData();
    //=== update PU data ====
    pu.intraDir[0] = uiBestPUMode;
741
742
743
#if JVET_L0283_MULTI_REF_LINE
    pu.multiRefIdx = bestExtendRef;
#endif
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
  }

  //===== reset context models =====
  m_CABACEstimator->getCtx() = ctxStart;
}

void IntraSearch::estIntraPredChromaQT(CodingUnit &cu, Partitioner &partitioner)
{
  const ChromaFormat format   = cu.chromaFormat;
  const uint32_t    numberValidComponents = getNumberValidComponents(format);
  CodingStructure &cs = *cu.cs;
  const TempCtx ctxStart  ( m_CtxCache, m_CABACEstimator->getCtx() );

  cs.setDecomp( cs.area.Cb(), false );

  auto &pu = *cu.firstPU;

  {
    uint32_t       uiBestMode = 0;
    Distortion uiBestDist = 0;
    double     dBestCost = MAX_DOUBLE;

    //----- init mode list ----
    {
      uint32_t  uiMinMode = 0;
      uint32_t  uiMaxMode = NUM_CHROMA_MODE;

      //----- check chroma modes -----
      uint32_t chromaCandModes[ NUM_CHROMA_MODE ];
      PU::getIntraChromaCandModes( pu, chromaCandModes );

      // create a temporary CS
      CodingStructure &saveCS = *m_pSaveCS[0];
      saveCS.pcv      = cs.pcv;
      saveCS.picture  = cs.picture;
      saveCS.area.repositionTo( cs.area );
      saveCS.clearTUs();

      if( CS::isDualITree( cs ) )
      {
        if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
        {
          partitioner.splitCurrArea( TU_MAX_TR_SPLIT, cs );

          do
          {
            cs.addTU( CS::getArea( cs, partitioner.currArea(), partitioner.chType ), partitioner.chType ).depth = partitioner.currTrDepth;
          } while( partitioner.nextPart( cs ) );

          partitioner.exitCurrSplit();
        }
        else
        cs.addTU( CS::getArea( cs, partitioner.currArea(), partitioner.chType ), partitioner.chType );
      }

      std::vector<TransformUnit*> orgTUs;


      // create a store for the TUs
      for( const auto &ptu : cs.tus )
      {
        // for split TUs in HEVC, add the TUs without Chroma parts for correct setting of Cbfs
        if( pu.contains( *ptu, CHANNEL_TYPE_CHROMA ) || ( !cs.pcv->noRQT && !ptu->Cb().valid() && !ptu->Cr().valid() ) )
        {
          saveCS.addTU( *ptu, partitioner.chType );
          orgTUs.push_back( ptu );
        }
      }
Xiang Ma's avatar
Xiang Ma committed
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
#if JVET_L0338_MDLM
      // SATD pre-selecting.
      int satdModeList[NUM_CHROMA_MODE];
      int64_t satdSortedCost[NUM_CHROMA_MODE];
      for (int i = 0; i < NUM_CHROMA_MODE; i++)
      {
        satdSortedCost[i] = 0; // for the mode not pre-select by SATD, do RDO by default, so set the initial value 0.
        satdModeList[i] = 0;
      }
      bool modeIsEnable[NUM_INTRA_MODE + 1]; // use intra mode idx to check whether enable
      for (int i = 0; i < NUM_INTRA_MODE + 1; i++)
      {
        modeIsEnable[i] = 1;
      }

      DistParam distParam;
      const bool useHadamard = true;
      pu.intraDir[1] = MDLM_L_IDX; // temporary assigned, just to indicate this is a MDLM mode. for luma down-sampling operation.

      initIntraPatternChType(cu, pu.Cb());
      initIntraPatternChType(cu, pu.Cr());
      xGetLumaRecPixels(pu, pu.Cb());

      for (int idx = uiMinMode; idx <= uiMaxMode - 1; idx++)
      {
        int mode = chromaCandModes[idx];
        satdModeList[idx] = mode;
        if (PU::isLMCMode(mode) && !PU::isLMCModeEnabled(pu, mode))
        {
          continue;
        }
843
        if ((mode == LM_CHROMA_IDX) || (mode == PLANAR_IDX) || (mode == DM_CHROMA_IDX)) // only pre-check regular modes and MDLM modes, not including DM ,Planar, and LM
Xiang Ma's avatar
Xiang Ma committed
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
        {
          continue;
        }
        pu.intraDir[1] = mode; // temporary assigned, for SATD checking.

        int64_t sad = 0;
        CodingStructure& cs = *(pu.cs);

        CompArea areaCb = pu.Cb();
        PelBuf orgCb = cs.getOrgBuf(areaCb);
        PelBuf predCb = cs.getPredBuf(areaCb);

        m_pcRdCost->setDistParam(distParam, orgCb, predCb, pu.cs->sps->getBitDepth(CHANNEL_TYPE_CHROMA), COMPONENT_Cb, useHadamard);
        distParam.applyWeight = false;

        if (PU::isLMCMode(mode))
        {
          predIntraChromaLM(COMPONENT_Cb, predCb, pu, areaCb, mode);
        }
        else
        {
          predIntraAng(COMPONENT_Cb, predCb, pu, false);
        }

        sad += distParam.distFunc(distParam);
869

Xiang Ma's avatar
Xiang Ma committed
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
        CompArea areaCr = pu.Cr();
        PelBuf orgCr = cs.getOrgBuf(areaCr);
        PelBuf predCr = cs.getPredBuf(areaCr);

        m_pcRdCost->setDistParam(distParam, orgCr, predCr, pu.cs->sps->getBitDepth(CHANNEL_TYPE_CHROMA), COMPONENT_Cr, useHadamard);
        distParam.applyWeight = false;

        if (PU::isLMCMode(mode))
        {
          predIntraChromaLM(COMPONENT_Cr, predCr, pu, areaCr, mode);
        }
        else
        {
          predIntraAng(COMPONENT_Cr, predCr, pu, false);
        }
        sad += distParam.distFunc(distParam);
        satdSortedCost[idx] = sad;
      }
      // sort the mode based on the cost from small to large.
      int tempIdx = 0;
      int64_t tempCost = 0;
      for (int i = uiMinMode; i <= uiMaxMode - 1; i++)
      {
        for (int j = i + 1; j <= uiMaxMode - 1; j++)
        {
          if (satdSortedCost[j] < satdSortedCost[i])
          {
            tempIdx = satdModeList[i];
            satdModeList[i] = satdModeList[j];
            satdModeList[j] = tempIdx;

            tempCost = satdSortedCost[i];
            satdSortedCost[i] = satdSortedCost[j];
            satdSortedCost[j] = tempCost;

          }
        }
      }
      int reducedModeNumber = 2; // reduce the number of chroma modes
      for (int i = 0; i < reducedModeNumber; i++)
      {
        modeIsEnable[satdModeList[uiMaxMode - 1 - i]] = 0; // disable the last reducedModeNumber modes
      }
#endif
914
915
916
917
918
919
920
921
922
923
924

      // save the dist
      Distortion baseDist = cs.dist;

      for (uint32_t uiMode = uiMinMode; uiMode < uiMaxMode; uiMode++)
      {
        const int chromaIntraMode = chromaCandModes[uiMode];
        if( PU::isLMCMode( chromaIntraMode ) && ! PU::isLMCModeEnabled( pu, chromaIntraMode ) )
        {
          continue;
        }
Xiang Ma's avatar
Xiang Ma committed
925
926
927
928
929
930
#if JVET_L0338_MDLM
        if (!modeIsEnable[chromaIntraMode] && PU::isLMCModeEnabled(pu, chromaIntraMode)) // when CCLM is disable, then MDLM is disable. not use satd checking
        {
          continue;
        }
#endif
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
        cs.setDecomp( pu.Cb(), false );
        cs.dist = baseDist;
        //----- restore context models -----
        m_CABACEstimator->getCtx() = ctxStart;

        //----- chroma coding -----
        pu.intraDir[1] = chromaIntraMode;

        xRecurIntraChromaCodingQT( cs, partitioner );

        if (cs.pps->getUseTransformSkip())
        {
          m_CABACEstimator->getCtx() = ctxStart;
        }

        uint64_t fracBits   = xGetIntraFracBitsQT( cs, partitioner, false, true );
        Distortion uiDist = cs.dist;
        double    dCost   = m_pcRdCost->calcRdCost( fracBits, uiDist - baseDist );

        //----- compare -----
        if( dCost < dBestCost )
        {
          for( uint32_t i = getFirstComponentOfChannel( CHANNEL_TYPE_CHROMA ); i < numberValidComponents; i++ )
          {
            const CompArea &area = pu.blocks[i];

            saveCS.getRecoBuf     ( area ).copyFrom( cs.getRecoBuf   ( area ) );
#if KEEP_PRED_AND_RESI_SIGNALS
            saveCS.getPredBuf     ( area ).copyFrom( cs.getPredBuf   ( area ) );
            saveCS.getResiBuf     ( area ).copyFrom( cs.getResiBuf   ( area ) );
#endif
            cs.picture->getRecoBuf( area ).copyFrom( cs.getRecoBuf( area ) );

            for( uint32_t j = 0; j < saveCS.tus.size(); j++ )
            {
              saveCS.tus[j]->copyComponentFrom( *orgTUs[j], area.compID );
            }
          }

          dBestCost  = dCost;
          uiBestDist = uiDist;
          uiBestMode = chromaIntraMode;
        }
      }

      for( uint32_t i = getFirstComponentOfChannel( CHANNEL_TYPE_CHROMA ); i < numberValidComponents; i++ )
      {
        const CompArea &area = pu.blocks[i];

        cs.getRecoBuf         ( area ).copyFrom( saveCS.getRecoBuf( area ) );
#if KEEP_PRED_AND_RESI_SIGNALS
        cs.getPredBuf         ( area ).copyFrom( saveCS.getPredBuf( area ) );
        cs.getResiBuf         ( area ).copyFrom( saveCS.getResiBuf( area ) );
#endif
        cs.picture->getRecoBuf( area ).copyFrom( cs.    getRecoBuf( area ) );

        for( uint32_t j = 0; j < saveCS.tus.size(); j++ )
        {
          orgTUs[ j ]->copyComponentFrom( *saveCS.tus[ j ], area.compID );
        }
      }
    }

    pu.intraDir[1] = uiBestMode;
    cs.dist        = uiBestDist;
  }

  //----- restore context models -----
  m_CABACEstimator->getCtx() = ctxStart;
}

void IntraSearch::IPCMSearch(CodingStructure &cs, Partitioner& partitioner)
{
  for (uint32_t ch = 0; ch < getNumberValidTBlocks( *cs.pcv ); ch++)
  {
    const ComponentID compID = ComponentID(ch);

    xEncPCM(cs, partitioner, compID);
  }

  cs.getPredBuf().fill(0);
  cs.getResiBuf().fill(0);
  cs.getOrgResiBuf().fill(0);

  cs.dist     = 0;
  cs.fracBits = 0;
  cs.cost     = 0;

  cs.setDecomp(cs.area);
  cs.picture->getRecoBuf(cs.area).copyFrom(cs.getRecoBuf());
}

void IntraSearch::xEncPCM(CodingStructure &cs, Partitioner& partitioner, const ComponentID &compID)
{
  TransformUnit &tu = *cs.getTU( partitioner.chType );

  const int  channelBitDepth = cs.sps->getBitDepth(toChannelType(compID));
  const uint32_t uiPCMBitDepth = cs.sps->getPCMBitDepth(toChannelType(compID));

  const int pcmShiftRight = (channelBitDepth - int(uiPCMBitDepth));

  CompArea  area    = tu.blocks[compID];
  PelBuf    pcmBuf  = tu.getPcmbuf  (compID);
  PelBuf    recBuf  = cs.getRecoBuf ( area );
  CPelBuf   orgBuf  = cs.getOrgBuf  ( area );

  CHECK(pcmShiftRight < 0, "Negative shift");

  for (uint32_t uiY = 0; uiY < pcmBuf.height; uiY++)
  {
    for (uint32_t uiX = 0; uiX < pcmBuf.width; uiX++)
    {
      // Encode
      pcmBuf.at(uiX, uiY) = orgBuf.at(uiX, uiY) >> pcmShiftRight;
      // Reconstruction
      recBuf.at(uiX, uiY) = pcmBuf.at(uiX, uiY) << pcmShiftRight;
    }
  }
}

// -------------------------------------------------------------------------------------------------------------------
// Intra search
// -------------------------------------------------------------------------------------------------------------------

void IntraSearch::xEncIntraHeader(CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma)
{
  CodingUnit &cu = *cs.getCU( partitioner.chType );

  if (bLuma)
  {
    bool isFirst = partitioner.currArea().lumaPos() == cs.area.lumaPos();

    // CU header
    if( isFirst )
    {
1066
      if( !cs.slice->isIntra() 
Xiaozhong Xu's avatar
Xiaozhong Xu committed
1067
1068
1069
#if JVET_L0293_CPR
        && cu.Y().valid()
#endif
1070
        )
1071
1072
1073
1074
1075
1076
1077
1078
      {
        if( cs.pps->getTransquantBypassEnabledFlag() )
        {
          m_CABACEstimator->cu_transquant_bypass_flag( cu );
        }
        m_CABACEstimator->cu_skip_flag( cu );
        m_CABACEstimator->pred_mode   ( cu );
      }
1079
1080
1081
#if JVET_L0283_MULTI_REF_LINE
      m_CABACEstimator->extend_ref_line(cu);
#endif
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
      if( CU::isIntra(cu) && cu.partSize == SIZE_2Nx2N )
      {
        m_CABACEstimator->pcm_data( cu );
        if( cu.ipcm )
        {
          return;
        }
      }
    }

    PredictionUnit &pu = *cs.getPU(partitioner.currArea().lumaPos(), partitioner.chType);

    // luma prediction mode
    if (cu.partSize == SIZE_2Nx2N)
    {
      if (isFirst)
      {
Xiaozhong Xu's avatar
Xiaozhong Xu committed
1099
1100
1101
1102
#if JVET_L0293_CPR
        if ( !cu.Y().valid())
          m_CABACEstimator->pred_mode( cu );
#endif
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
        m_CABACEstimator->intra_luma_pred_mode( pu );
      }
    }
  }

  if (bChroma)
  {
    bool isFirst = partitioner.currArea().Cb().valid() && partitioner.currArea().chromaPos() == cs.area.chromaPos();

    PredictionUnit &pu = *cs.getPU( partitioner.currArea().chromaPos(), CHANNEL_TYPE_CHROMA );

    if( cu.partSize == SIZE_2Nx2N )
    {
      if( isFirst )
      {
        m_CABACEstimator->intra_chroma_pred_mode( pu );
      }
    }
  }
}

void IntraSearch::xEncSubdivCbfQT(CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma)
{
  const UnitArea &currArea = partitioner.currArea();
  TransformUnit &currTU    = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType );
1128
#if HM_EMT_NSST_AS_IN_JEM
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
  CodingUnit &currCU       = *currTU.cu;
#endif
  uint32_t currDepth           = partitioner.currTrDepth;

  const bool subdiv        = currTU.depth > currDepth;

  if( cs.pcv->noRQT )
  {
    if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
    {
      CHECK( !subdiv, "TU split implied" );
    }
    else
      CHECK( subdiv, "No TU subdivision is allowed with QTBT" );
  }

  if (bChroma)
  {
    const uint32_t numberValidComponents = getNumberValidComponents(currArea.chromaFormat);

    for (uint32_t ch = COMPONENT_Cb; ch < numberValidComponents; ch++)
    {
      const ComponentID compID = ComponentID(ch);

      if( currDepth == 0 || TU::getCbfAtDepth( currTU, compID, currDepth - 1 ) )
      {
        const bool prevCbf = ( compID == COMPONENT_Cr ? TU::getCbfAtDepth( currTU, COMPONENT_Cb, currDepth ) : false );
        m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, compID, currDepth ), currArea.blocks[compID], currDepth, prevCbf );

      }
    }
  }

  if (subdiv)
  {
1164
#if HM_EMT_NSST_AS_IN_JEM
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
    if( currDepth == 0 && bLuma ) m_CABACEstimator->emt_cu_flag( currCU );
#endif

    if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
    {
      partitioner.splitCurrArea( TU_MAX_TR_SPLIT, cs );
    }
    else
    THROW( "Cannot perform an implicit split!" );

    do
    {
      xEncSubdivCbfQT( cs, partitioner, bLuma, bChroma );
    } while( partitioner.nextPart( cs ) );

    partitioner.exitCurrSplit();
  }
  else
  {
1184
#if HM_EMT_NSST_AS_IN_JEM
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
    if( currDepth == 0 && bLuma && TU::getCbfAtDepth( currTU, COMPONENT_Y, 0 ) ) m_CABACEstimator->emt_cu_flag( currCU );

#endif
    //===== Cbfs =====
    if (bLuma)
    {
      m_CABACEstimator->cbf_comp( cs, TU::getCbfAtDepth( currTU, COMPONENT_Y, currDepth ), currTU.Y(), currTU.depth );
    }
  }
}

void IntraSearch::xEncCoeffQT(CodingStructure &cs, Partitioner &partitioner, const ComponentID &compID)
{
  const UnitArea &currArea  = partitioner.currArea();
  TransformUnit &currTU     = *cs.getTU( currArea.blocks[partitioner.chType], partitioner.chType );
  uint32_t      currDepth       = partitioner.currTrDepth;
  const bool subdiv         = currTU.depth > currDepth;

  if (subdiv)
  {
    if (partitioner.canSplit(TU_MAX_TR_SPLIT, cs))
    {
      partitioner.splitCurrArea(TU_MAX_TR_SPLIT, cs);
    }
    else
      THROW("Implicit TU split not available!");

    do
    {
      xEncCoeffQT( cs, partitioner, compID );
    } while( partitioner.nextPart( cs ) );

    partitioner.exitCurrSplit();
  }
  else

  if( currArea.blocks[compID].valid() )
  {
    if( TU::hasCrossCompPredInfo( currTU, compID ) )
    {
      m_CABACEstimator->cross_comp_pred( currTU, compID );
    }
    if( TU::getCbf( currTU, compID ) )
    {
      m_CABACEstimator->residual_coding( currTU, compID );
    }
  }
}

uint64_t IntraSearch::xGetIntraFracBitsQT( CodingStructure &cs, Partitioner &partitioner, const bool &bLuma, const bool &bChroma )
{
  m_CABACEstimator->resetBits();

  xEncIntraHeader( cs, partitioner, bLuma, bChroma );
  xEncSubdivCbfQT( cs, partitioner, bLuma, bChroma );

  if( bLuma )
  {
    xEncCoeffQT( cs, partitioner, COMPONENT_Y );
  }
  if( bChroma )
  {
    xEncCoeffQT( cs, partitioner, COMPONENT_Cb );
    xEncCoeffQT( cs, partitioner, COMPONENT_Cr );
  }

  uint64_t fracBits = m_CABACEstimator->getEstFracBits();
  return fracBits;
}

uint64_t IntraSearch::xGetIntraFracBitsQTChroma(TransformUnit& currTU, const ComponentID &compID)
{
  m_CABACEstimator->resetBits();

  if( TU::hasCrossCompPredInfo( currTU, compID ) )
  {
    m_CABACEstimator->cross_comp_pred( currTU, compID );
  }
  if( TU::getCbf( currTU, compID ) )
  {
    m_CABACEstimator->residual_coding( currTU, compID );
  }

  uint64_t fracBits = m_CABACEstimator->getEstFracBits();
  return fracBits;
}

void IntraSearch::xIntraCodingTUBlock(TransformUnit &tu, const ComponentID &compID, const bool &checkCrossCPrediction, Distortion& ruiDist, const int &default0Save1Load2, uint32_t* numSig )
{
  if (!tu.blocks[compID].valid())
  {
    return;
  }

  CodingStructure &cs                       = *tu.cs;

  const CompArea      &area                 = tu.blocks[compID];
  const SPS           &sps                  = *cs.sps;
  const PPS           &pps                  = *cs.pps;

  const ChannelType    chType               = toChannelType(compID);
  const int            bitDepth             = sps.getBitDepth(chType);

  PelBuf         piOrg                      = cs.getOrgBuf    (area);
  PelBuf         piPred                     = cs.getPredBuf   (area);
  PelBuf         piResi                     = cs.getResiBuf   (area);
  PelBuf         piOrgResi                  = cs.getOrgResiBuf(area);
  PelBuf         piReco                     = cs.getRecoBuf   (area);

  const PredictionUnit &pu                  = *cs.getPU(area.pos(), chType);
  const uint32_t           uiChFinalMode        = PU::getFinalIntraMode(pu, chType);

  const bool           bUseCrossCPrediction = pps.getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() && isChroma( compID ) && PU::isChromaIntraModeCrossCheckMode( pu ) && checkCrossCPrediction;
  const bool           ccUseRecoResi        = m_pcEncCfg->getUseReconBasedCrossCPredictionEstimate();

1300
#if !JVET_L0059_MTS_SIMP
1301
  const uint8_t          transformIndex       = tu.cu->emtFlag && compID == COMPONENT_Y ? tu.emtIdx : DCT2_EMT ;
1302
#endif
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372

  //===== init availability pattern =====
  PelBuf sharedPredTS( m_pSharedPredTransformSkip[compID], area );
  if( default0Save1Load2 != 2 )
  {
    const bool bUseFilteredPredictions = IntraPrediction::useFilteredIntraRefSamples( compID, pu, true, tu );
    initIntraPatternChType( *tu.cu, area, bUseFilteredPredictions );

    //===== get prediction signal =====
    if( compID != COMPONENT_Y && PU::isLMCMode( uiChFinalMode ) )
    {
      {
        xGetLumaRecPixels( pu, area );
      }
      predIntraChromaLM( compID, piPred, pu, area, uiChFinalMode );
    }
    else
    {
      predIntraAng( compID, piPred, pu, bUseFilteredPredictions );
    }


    // save prediction
    if( default0Save1Load2 == 1 )
    {
      sharedPredTS.copyFrom( piPred );
    }
  }
  else
  {
    // load prediction
    piPred.copyFrom( sharedPredTS );
  }


  DTRACE( g_trace_ctx, D_PRED, "@(%4d,%4d) [%2dx%2d] IMode=%d\n", tu.lx(), tu.ly(), tu.lwidth(), tu.lheight(), uiChFinalMode );
  //DTRACE_PEL_BUF( D_PRED, piPred, tu, tu.cu->predMode, COMPONENT_Y );

  //===== get residual signal =====
  piResi.copyFrom( piOrg  );
  piResi.subtract( piPred );

  if (pps.getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() && isLuma(compID))
  {
    piOrgResi.copyFrom (piResi);
  }

  if (bUseCrossCPrediction)
  {
    if (xCalcCrossComponentPredictionAlpha(tu, compID, ccUseRecoResi) == 0)
    {
      return;
    }
    CrossComponentPrediction::crossComponentPrediction(tu, compID, cs.getResiBuf(tu.Y()), piResi, piResi, false);
  }

  //===== transform and quantization =====
  //--- init rate estimation arrays for RDOQ ---
  //--- transform and quantization           ---
  TCoeff uiAbsSum = 0;

  const QpParam cQP(tu, compID);

#if RDOQ_CHROMA_LAMBDA
  m_pcTrQuant->selectLambda(compID);
#endif


  m_pcTrQuant->transformNxN(tu, compID, cQP, uiAbsSum, m_CABACEstimator->getCtx());

1373
#if !JVET_L0059_MTS_SIMP
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
  if( transformIndex != DCT2_EMT && ( !tu.transformSkip[COMPONENT_Y] ) ) //this can only be true if compID is luma
  {
    *numSig = 0;
    TCoeff* coeffBuffer = tu.getCoeffs(compID).buf;
    for( uint32_t uiX = 0; uiX < tu.Y().area(); uiX++ )
    {
      if( coeffBuffer[uiX] )
      {
        ( *numSig )++;
        if( *numSig > g_EmtSigNumThr )
        {
          break;
        }
      }
    }
    //if the number of significant coeffs is less than the threshold, then only the default transform (which has a 0 index, but it is the DST7) is allowed
    if( transformIndex != 0 && *numSig <= g_EmtSigNumThr && !tu.transformSkip[compID] )
    {
      return;
    }
  }
1395
#endif
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577

  DTRACE( g_trace_ctx, D_TU_ABS_SUM, "%d: comp=%d, abssum=%d\n", DTRACE_GET_COUNTER( g_trace_ctx, D_TU_ABS_SUM ), compID, uiAbsSum );


  //--- inverse transform ---
  if (uiAbsSum > 0)
  {
    m_pcTrQuant->invTransformNxN(tu, compID, piResi, cQP);
  }
  else
  {
    piResi.fill(0);
  }

  //===== reconstruction =====
  if (bUseCrossCPrediction)
  {
    CrossComponentPrediction::crossComponentPrediction(tu, compID, cs.getResiBuf(tu.Y()), piResi, piResi, true);
  }

  piReco.reconstruct(piPred, piResi, cs.slice->clpRng( compID ));

  //===== update distortion =====
#if WCG_EXT
  if( m_pcEncCfg->getLumaLevelToDeltaQPMapping().isEnabled() )
  {
    const CPelBuf orgLuma = cs.getOrgBuf( cs.area.blocks[COMPONENT_Y] );
    ruiDist += m_pcRdCost->getDistPart( piOrg, piReco, bitDepth, compID, DF_SSE_WTD, &orgLuma );
  }
  else
#endif
  {
    ruiDist += m_pcRdCost->getDistPart( piOrg, piReco, bitDepth, compID, DF_SSE );
  }
}

void IntraSearch::xRecurIntraCodingLumaQT( CodingStructure &cs, Partitioner &partitioner )
{
  const UnitArea &currArea = partitioner.currArea();
  const CodingUnit &cu     = *cs.getCU(currArea.lumaPos(), partitioner.chType);
  uint32_t     currDepth       = partitioner.currTrDepth;
  const PPS &pps           = *cs.pps;
  const bool keepResi      = pps.getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() || KEEP_PRED_AND_RESI_SIGNALS;
  bool bCheckFull          = true;
  bool bCheckSplit         = false;
  bCheckFull               = cs.pcv->noRQT && !partitioner.canSplit( TU_MAX_TR_SPLIT, cs );
  bCheckSplit              = cs.pcv->noRQT &&  partitioner.canSplit( TU_MAX_TR_SPLIT, cs );

  uint32_t    numSig           = 0;

  if( !cs.pcv->noRQT )
  {
  }

  bool    checkInitTrDepth = false, checkInitTrDepthTransformSkipWinner = false;

  double     dSingleCost                        = MAX_DOUBLE;
  Distortion uiSingleDistLuma                   = 0;
  uint64_t     singleFracBits                     = 0;
  bool       checkTransformSkip                 = pps.getUseTransformSkip();
  int        bestModeId[MAX_NUM_COMPONENT]      = {0, 0, 0};
  uint8_t      nNumTransformCands                 = cu.emtFlag ? 4 : 1; //4 is the number of transforms of emt
  bool       isAllIntra                         = m_pcEncCfg->getIntraPeriod() == 1;

  uint8_t numTransformIndexCands                  = nNumTransformCands;

  const TempCtx ctxStart  ( m_CtxCache, m_CABACEstimator->getCtx() );
  TempCtx       ctxBest   ( m_CtxCache );

  CodingStructure *csSplit = nullptr;
  CodingStructure *csFull  = nullptr;

  if( bCheckSplit )
  {
    csSplit = &cs;
  }
  else if( bCheckFull )
  {
    csFull = &cs;
  }

  if( bCheckFull )
  {
    csFull->cost = 0.0;

    TransformUnit &tu = csFull->addTU( CS::getArea( *csFull, currArea, partitioner.chType ), partitioner.chType );
    tu.depth = currDepth;

    checkTransformSkip &= TU::hasTransformSkipFlag( *tu.cs, tu.Y() );
    checkTransformSkip &= !cu.transQuantBypass;
    checkTransformSkip &= !cu.emtFlag;

    CHECK( !tu.Y().valid(), "Invalid TU" );

    //this prevents transformSkip from being checked because we already know it's not the best mode
    checkTransformSkip = ( checkInitTrDepth && !checkInitTrDepthTransformSkipWinner ) ? false : checkTransformSkip;


    CHECK( checkInitTrDepthTransformSkipWinner && !checkTransformSkip, "Transform Skip must be enabled if it was the winner in the previous call of xRecurIntraCodingLumaQT!" );

    CodingStructure &saveCS = *m_pSaveCS[0];

    TransformUnit *tmpTU = nullptr;

    Distortion singleDistTmpLuma = 0;
    uint64_t     singleTmpFracBits = 0;
    double     singleCostTmp     = 0;
    int        firstCheckId      = 0;

    //we add the EMT candidates to the loop. TransformSkip will still be the last one to be checked (when modeId == lastCheckId) as long as checkTransformSkip is true
    int        lastCheckId       = numTransformIndexCands - ( firstCheckId + 1 ) + ( int ) checkTransformSkip;
    bool isNotOnlyOneMode        = lastCheckId != firstCheckId && !checkInitTrDepthTransformSkipWinner;

    if( isNotOnlyOneMode )
    {
      saveCS.pcv     = cs.pcv;
      saveCS.picture = cs.picture;
      saveCS.area.repositionTo(cs.area);
      saveCS.clearTUs();
      tmpTU = &saveCS.addTU(currArea, partitioner.chType);
    }

    bool cbfBestMode = false;


    for( int modeId = firstCheckId; modeId <= lastCheckId; modeId++ )
    {
      if( checkInitTrDepthTransformSkipWinner )
      {
        //If this is a full RQT call and the winner of the first call (checkFirst=true) was transformSkip, then we skip the first iteration of the loop, since transform skip always comes at the end
        if( modeId == firstCheckId )
        {
          continue;
        }
      }

      uint8_t transformIndex = modeId;


      if( ( transformIndex < lastCheckId ) || ( ( transformIndex == lastCheckId ) && !checkTransformSkip ) ) //we avoid this if the mode is transformSkip
      {
        // Skip checking other transform candidates if zero CBF is encountered and it is the best transform so far
        if( m_pcEncCfg->getFastIntraEMT() && isAllIntra && transformIndex && !cbfBestMode )
        {
          continue;
        }
      }

      if ((modeId != firstCheckId) && isNotOnlyOneMode)
      {
        m_CABACEstimator->getCtx() = ctxStart;
      }

      int default0Save1Load2 = 0;
      singleDistTmpLuma = 0;

      if (modeId == firstCheckId && modeId != lastCheckId && !checkInitTrDepthTransformSkipWinner )
      {
        default0Save1Load2 = 1;
      }
      else if (modeId != firstCheckId)
      {
        default0Save1Load2 = 2;
      }

      if (cu.emtFlag)
      {
        tu.emtIdx = transformIndex;
      }
      if( !checkTransformSkip )
      {
        tu.transformSkip[COMPONENT_Y] = false;
      }
      else
      {
        tu.transformSkip[COMPONENT_Y] = modeId == lastCheckId;
      }

      xIntraCodingTUBlock( tu, COMPONENT_Y, false, singleDistTmpLuma, default0Save1Load2, &numSig );

      //----- determine rate and r-d cost -----
      //the condition (transformIndex != DCT2_EMT) seems to be irrelevant, since DCT2_EMT=7 and the highest value of transformIndex is 4
1578
1579
1580
#if JVET_L0059_MTS_SIMP
      if( ( modeId == lastCheckId && checkTransformSkip && !TU::getCbfAtDepth( tu, COMPONENT_Y, currDepth ) ) )
#else
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
      if( ( modeId == lastCheckId && checkTransformSkip && !TU::getCbfAtDepth( tu, COMPONENT_Y, currDepth ) )
        || ( tu.emtIdx > 0 && ( checkTransformSkip ? transformIndex != lastCheckId : true ) && tu.emtIdx != DCT2_EMT && numSig <= g_EmtSigNumThr ) )
#endif
      {
        //In order not to code TS flag when cbf is zero, the case for TS with cbf being zero is forbidden.
        singleCostTmp = MAX_DOUBLE;
      }
      else
      {
        singleTmpFracBits = xGetIntraFracBitsQT( *csFull, partitioner, true, false );
        singleCostTmp     = m_pcRdCost->calcRdCost( singleTmpFracBits, singleDistTmpLuma );
      }

      if (singleCostTmp < dSingleCost)
      {
        dSingleCost       = singleCostTmp;
        uiSingleDistLuma  = singleDistTmpLuma;
        singleFracBits    = singleTmpFracBits;

        bestModeId[COMPONENT_Y] = modeId;
        cbfBestMode       = TU::getCbfAtDepth( tu, COMPONENT_Y, currDepth );


        if( bestModeId[COMPONENT_Y] != lastCheckId )
        {
#if KEEP_PRED_AND_RESI_SIGNALS
          saveCS.getPredBuf( tu.Y() ).copyFrom( csFull->getPredBuf( tu.Y() ) );
#endif
          saveCS.getRecoBuf( tu.Y() ).copyFrom( csFull->getRecoBuf( tu.Y() ) );

          if( keepResi )
          {
            saveCS.getResiBuf   ( tu.Y() ).copyFrom( csFull->getResiBuf   ( tu.Y() ) );
            saveCS.getOrgResiBuf( tu.Y() ).copyFrom( csFull->getOrgResiBuf( tu.Y() ) );
          }

          tmpTU->copyComponentFrom( tu, COMPONENT_Y );

          ctxBest = m_CABACEstimator->getCtx();
        }
      }
    }

    if( bestModeId[COMPONENT_Y] != lastCheckId )
    {
#if KEEP_PRED_AND_RESI_SIGNALS
      csFull->getPredBuf( tu.Y() ).copyFrom( saveCS.getPredBuf( tu.Y() ) );
#endif
      csFull->getRecoBuf( tu.Y() ).copyFrom( saveCS.getRecoBuf( tu.Y() ) );

      if( keepResi )
      {
        csFull->getResiBuf   ( tu.Y() ).copyFrom( saveCS.getResiBuf   ( tu.Y() ) );
        csFull->getOrgResiBuf( tu.Y() ).copyFrom( saveCS.getOrgResiBuf( tu.Y() ) );
      }

      tu.copyComponentFrom( *tmpTU, COMPONENT_Y );

      if( !bCheckSplit )
      {
        m_CABACEstimator->getCtx() = ctxBest;
      }
    }
    else if( bCheckSplit )
    {
      ctxBest = m_CABACEstimator->getCtx();
    }

    csFull->cost     += dSingleCost;
    csFull->dist     += uiSingleDistLuma;
    csFull->fracBits += singleFracBits;
  }

  if( bCheckSplit )
  {
    //----- store full entropy coding status, load original entropy coding status -----
    if( bCheckFull )
    {
      m_CABACEstimator->getCtx() = ctxStart;
    }
    //----- code splitted block -----
    csSplit->cost = 0;

    bool uiSplitCbfLuma  = false;
    bool splitIsSelected = true;
    if( cs.pcv->noRQT && partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
    {
      partitioner.splitCurrArea( TU_MAX_TR_SPLIT, cs );
    }

    do
    {
      xRecurIntraCodingLumaQT( *csSplit, partitioner );

      csSplit->setDecomp( partitioner.currArea().Y() );

      uiSplitCbfLuma |= TU::getCbfAtDepth( *csSplit->getTU( partitioner.currArea().lumaPos(), partitioner.chType ), COMPONENT_Y, partitioner.currTrDepth );



    } while( partitioner.nextPart( *csSplit ) );

    partitioner.exitCurrSplit();

    if( splitIsSelected )
    {
      for( auto &ptu : csSplit->tus )
      {
        if( currArea.Y().contains( ptu->Y() ) )
        {
          TU::setCbfAtDepth( *ptu, COMPONENT_Y, currDepth, uiSplitCbfLuma ? 1 : 0 );
        }
      }

      //----- restore context states -----
      m_CABACEstimator->getCtx() = ctxStart;

      //----- determine rate and r-d cost -----
      csSplit->fracBits = xGetIntraFracBitsQT(*csSplit, partitioner, true, false);

      //--- update cost ---
      csSplit->cost     = m_pcRdCost->calcRdCost(csSplit->fracBits, csSplit->dist);
    }
  }

  if( csFull || csSplit )
  {
    {
      // otherwise this would've happened in useSubStructure
      cs.picture->getRecoBuf( currArea.Y() ).copyFrom( cs.getRecoBuf( currArea.Y() ) );
    }

    cs.cost = m_pcRdCost->calcRdCost( cs.fracBits, cs.dist );
  }
}

ChromaCbfs IntraSearch::xRecurIntraChromaCodingQT(CodingStructure &cs, Partitioner& partitioner)
{
  UnitArea currArea                   = partitioner.currArea();
  const bool keepResi                 = cs.sps->getSpsNext().getUseLMChroma() || KEEP_PRED_AND_RESI_SIGNALS;
  if( !currArea.Cb().valid() ) return ChromaCbfs( false );


  TransformUnit &currTU               = *cs.getTU( currArea.chromaPos(), CHANNEL_TYPE_CHROMA );
  const PredictionUnit &pu            = *cs.getPU( currArea.chromaPos(), CHANNEL_TYPE_CHROMA );
  const TransformUnit &currTULuma     = CS::isDualITree( cs ) ? *cs.picture->cs->getTU( currArea.lumaPos(), CHANNEL_TYPE_LUMA ) : currTU;

  uint32_t     currDepth                  = partitioner.currTrDepth;
  const PPS &pps                      = *cs.pps;
  ChromaCbfs cbfs                     ( false );

  if (currDepth == currTU.depth)
  {
    if (!currArea.Cb().valid() || !currArea.Cr().valid())
    {
      return cbfs;
    }

    bool checkTransformSkip = pps.getUseTransformSkip();
    checkTransformSkip &= TU::hasTransformSkipFlag( *currTU.cs, partitioner.currArea().Cb() );

    if( m_pcEncCfg->getUseTransformSkipFast() )
    {
      checkTransformSkip &= TU::hasTransformSkipFlag( *currTU.cs, partitioner.currArea().Y() );

      if( checkTransformSkip && cs.pcv->noChroma2x2 )
      {
        int nbLumaSkip = currTULuma.transformSkip[0] ? 1 : 0;

        {
          // the chroma blocks are co-located with the last luma block, so backwards references are needed
          nbLumaSkip += cs.getTU( currTULuma.Y().topLeft().offset( -1,  0 ), partitioner.chType )->transformSkip[0] ? 1 : 0;
          nbLumaSkip += cs.getTU( currTULuma.Y().topLeft().offset( -1, -1 ), partitioner.chType )->transformSkip[0] ? 1 : 0;
          nbLumaSkip += cs.getTU( currTULuma.Y().topLeft().offset(  0, -1 ), partitioner.chType )->transformSkip[0] ? 1 : 0;
        }

        checkTransformSkip &= ( nbLumaSkip > 0 );
      }
    }

    CodingStructure &saveCS = *m_pSaveCS[1];
    saveCS.pcv      = cs.pcv;
    saveCS.picture  = cs.picture;
    saveCS.area.repositionTo( cs.area );
    saveCS.initStructData( -1, false, true );

    TransformUnit &tmpTU = saveCS.addTU(currArea, partitioner.chType);


    cs.setDecomp(currArea.Cb(), true); // set in advance (required for Cb2/Cr2 in 4:2:2 video)

    const unsigned      numTBlocks  = ::getNumberValidTBlocks( *cs.pcv );

    for( uint32_t c = COMPONENT_Cb; c < numTBlocks; c++)
    {
      const ComponentID compID  = ComponentID(c);
      const CompArea&   area    = currTU.blocks[compID];

      double     dSingleCost    = MAX_DOUBLE;
      int        bestModeId     = 0;
      Distortion singleDistC    = 0;
      Distortion singleDistCTmp = 0;
      double     singleCostTmp  = 0;

      const bool checkCrossComponentPrediction = PU::isChromaIntraModeCrossCheckMode( pu ) && pps.getPpsRangeExtension().getCrossComponentPredictionEnabledFlag() && TU::getCbf( currTU, COMPONENT_Y );

      const int  crossCPredictionModesToTest = checkCrossComponentPrediction ? 2 : 1;
      const int  transformSkipModesToTest    = checkTransformSkip ? 2 : 1;
      const int  totalModesToTest            = crossCPredictionModesToTest * transformSkipModesToTest;
      const bool isOneMode                   = (totalModesToTest == 1);

      int currModeId = 0;
      int default0Save1Load2 = 0;

      TempCtx ctxStart  ( m_CtxCache );
      TempCtx ctxBest   ( m_CtxCache );

      if (!isOneMode)
      {
        ctxStart = m_CABACEstimator->getCtx();
      }

      for (int transformSkipModeId = 0; transformSkipModeId < transformSkipModesToTest; transformSkipModeId++)
      {
        for (int crossCPredictionModeId = 0; crossCPredictionModeId < crossCPredictionModesToTest; crossCPredictionModeId++)
        {
          currTU.compAlpha    [compID] = 0;
          currTU.transformSkip[compID] = transformSkipModeId;

          currModeId++;

          const bool isFirstMode = (currModeId == 1);
          const bool isLastMode  = (currModeId == totalModesToTest); // currModeId is indexed from 1

          if (isOneMode)
          {
            default0Save1Load2 = 0;
          }
          else if (!isOneMode && (transformSkipModeId == 0) && (crossCPredictionModeId == 0))
          {
            default0Save1Load2 = 1; //save prediction on first mode
          }
          else
          {
            default0Save1Load2 = 2; //load it on subsequent modes
          }

          if (!isFirstMode) // if not first mode to be tested
          {
            m_CABACEstimator->getCtx() = ctxStart;
          }

          singleDistCTmp = 0;

          xIntraCodingTUBlock( currTU, compID, crossCPredictionModeId != 0, singleDistCTmp, default0Save1Load2 );

          if( ( ( crossCPredictionModeId == 1 ) && ( currTU.compAlpha[compID] == 0 ) ) || ( ( transformSkipModeId == 1 ) && !TU::getCbf( currTU, compID ) ) ) //In order not to code TS flag when cbf is zero, the case for TS with cbf being zero is forbidden.
          {
            singleCostTmp = MAX_DOUBLE;
          }
          else if( !isOneMode )
          {
            uint64_t fracBitsTmp = xGetIntraFracBitsQTChroma( currTU, compID );
            singleCostTmp = m_pcRdCost->calcRdCost( fracBitsTmp, singleDistCTmp );
          }

          if( singleCostTmp < dSingleCost )
          {
            dSingleCost = singleCostTmp;
            singleDistC = singleDistCTmp;
            bestModeId  = currModeId;

            if( !isLastMode )
            {
#if KEEP_PRED_AND_RESI_SIGNALS
              saveCS.getPredBuf   (area).copyFrom(cs.getPredBuf   (area));
              saveCS.getOrgResiBuf(area).copyFrom(cs.getOrgResiBuf(area));
#endif
              if( keepResi )
              {
                saveCS.getResiBuf (area).copyFrom(cs.getResiBuf   (area));
              }
              saveCS.getRecoBuf   (area).copyFrom(cs.getRecoBuf   (area));

              tmpTU.copyComponentFrom(currTU, compID);

              ctxBest = m_CABACEstimator->getCtx();
            }
          }
        }
      }

      if (bestModeId < totalModesToTest)
      {
#if KEEP_PRED_AND_RESI_SIGNALS
        cs.getPredBuf   (area).copyFrom(saveCS.getPredBuf   (area));
        cs.getOrgResiBuf(area).copyFrom(saveCS.getOrgResiBuf(area));
#endif
        if( keepResi )
        {
          cs.getResiBuf (area).copyFrom(saveCS.getResiBuf   (area));
        }
        cs.getRecoBuf   (area).copyFrom(saveCS.getRecoBuf   (area));

        currTU.copyComponentFrom(tmpTU, compID);

        m_CABACEstimator->getCtx() = ctxBest;
      }

      cs.picture->getRecoBuf(area).copyFrom(cs.getRecoBuf(area));

      cbfs.cbf(compID) = TU::getCbf(currTU, compID);

      cs.dist += singleDistC;
    }
  }
  else
  {
    unsigned    numValidTBlocks   = ::getNumberValidTBlocks( *cs.pcv );
    ChromaCbfs  SplitCbfs         ( false );

    if( partitioner.canSplit( TU_MAX_TR_SPLIT, cs ) )
    {
      partitioner.splitCurrArea( TU_MAX_TR_SPLIT, cs );
    }
    else
      THROW( "Implicit TU split not available" );

    do
    {
      ChromaCbfs subCbfs = xRecurIntraChromaCodingQT( cs, partitioner );

      for( uint32_t ch = COMPONENT_Cb; ch < numValidTBlocks; ch++ )
      {
        const ComponentID compID = ComponentID( ch );
        SplitCbfs.cbf( compID ) |= subCbfs.cbf( compID );
      }
    } while( partitioner.nextPart( cs ) );

    partitioner.exitCurrSplit();

    {

      cbfs.Cb |= SplitCbfs.Cb;
      cbfs.Cr |= SplitCbfs.Cr;

      for( auto &ptu : cs.tus )
      {
        if( currArea.Cb().contains( ptu->Cb() ) || ( !ptu->Cb().valid() && currArea.Y().contains( ptu->Y() ) ) )
        {
          TU::setCbfAtDepth( *ptu, COMPONENT_Cb, currDepth, SplitCbfs.Cb );
          TU::setCbfAtDepth( *ptu, COMPONENT_Cr, currDepth, SplitCbfs.Cr );
        }
      }
    }
  }

  return cbfs;
}

uint64_t IntraSearch::xFracModeBitsIntra(PredictionUnit &pu, const uint32_t &uiMode, const ChannelType &chType)
{
  uint32_t orgMode = uiMode;

1945
#if JVET_L0100_MULTI_HYPOTHESIS_INTRA
1946
  if (!pu.mhIntraFlag)
1947
#endif
1948
1949
1950
1951
1952
1953
  std::swap(orgMode, pu.intraDir[chType]);

  m_CABACEstimator->resetBits();

  if( isLuma( chType ) )
  {
1954
#if JVET_L0100_MULTI_HYPOTHESIS_INTRA
1955
    if ( pu.mhIntraFlag )
1956
1957
1958
      m_CABACEstimator->MHIntra_luma_pred_modes(*pu.cu);
    else
    {
1959
1960
1961
#if JVET_L0283_MULTI_REF_LINE
      m_CABACEstimator->extend_ref_line(pu);
#endif
1962
1963
1964
      m_CABACEstimator->intra_luma_pred_mode(pu);
    }
#else
1965
1966
1967
#if JVET_L0283_MULTI_REF_LINE
    m_CABACEstimator->extend_ref_line(pu);
#endif
1968
    m_CABACEstimator->intra_luma_pred_mode( pu );
1969
#endif
1970
1971
1972
1973
1974
1975
  }
  else
  {
    m_CABACEstimator->intra_chroma_pred_mode( pu );
  }

1976
#if JVET_L0100_MULTI_HYPOTHESIS_INTRA
1977
  if ( !pu.mhIntraFlag )
1978
#endif
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
  std::swap(orgMode, pu.intraDir[chType]);

  return m_CABACEstimator->getEstFracBits();
}



void IntraSearch::encPredIntraDPCM( const ComponentID &compID, PelBuf &pOrg, PelBuf &pDst, const uint32_t &uiDirMode )
{
  CHECK( pOrg.buf == 0, "Encoder DPCM called without original buffer" );

  const int srcStride = m_topRefLength + 1;
  CPelBuf   pSrc = CPelBuf(getPredictorPtr(compID), srcStride, m_leftRefLength + 1);

  // Sample Adaptive intra-Prediction (SAP)
  if( uiDirMode == HOR_IDX )
  {
    // left column filled with reference samples, remaining columns filled with pOrg data
    for( int y = 0; y < pDst.height; y++ )
    {
      pDst.at( 0, y ) = pSrc.at( 0, 1 + y );
    }
    CPelBuf orgRest  = pOrg.subBuf( 0, 0, pOrg.width - 1, pOrg.height );
    PelBuf  predRest = pDst.subBuf( 1, 0, pDst.width - 1, pDst.height );

    predRest.copyFrom( orgRest );
  }
  else // VER_IDX
  {
    // top row filled with reference samples, remaining rows filled with pOrg data
    for( int x = 0; x < pDst.width; x++ )
    {
      pDst.at( x, 0 ) = pSrc.at( 1 + x, 0 );
    }
    CPelBuf orgRest  = pOrg.subBuf( 0, 0, pOrg.width, pOrg.height - 1 );
    PelBuf  predRest = pDst.subBuf( 0, 1, pDst.width, pDst.height - 1 );

    predRest.copyFrom( orgRest );
  }
}

bool IntraSearch::useDPCMForFirstPassIntraEstimation( const PredictionUnit &pu, const uint32_t &uiDirMode )
{
  return CU::isRDPCMEnabled( *pu.cu ) && pu.cu->transQuantBypass && (uiDirMode == HOR_IDX || uiDirMode == VER_IDX);
}