/***************************************************************************** * Copyright (C) 2013-2020 MulticoreWare, Inc * * Authors: Steve Borho * * 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. *****************************************************************************/ #ifndef X265_CUDATA_H #define X265_CUDATA_H #include "common.h" #include "slice.h" #include "mv.h" #define NUM_TU_DEPTH 21 namespace X265_NS { // private namespace class FrameData; class Slice; struct TUEntropyCodingParameters; struct CUDataMemPool; #if ENABLE_SCC_EXT struct IBC; #endif enum PartSize { SIZE_2Nx2N, // symmetric motion partition, 2Nx2N SIZE_2NxN, // symmetric motion partition, 2Nx N SIZE_Nx2N, // symmetric motion partition, Nx2N SIZE_NxN, // symmetric motion partition, Nx N SIZE_2NxnU, // asymmetric motion partition, 2Nx( N/2) + 2Nx(3N/2) SIZE_2NxnD, // asymmetric motion partition, 2Nx(3N/2) + 2Nx( N/2) SIZE_nLx2N, // asymmetric motion partition, ( N/2)x2N + (3N/2)x2N SIZE_nRx2N, // asymmetric motion partition, (3N/2)x2N + ( N/2)x2N NUM_SIZES }; enum PredMode { MODE_NONE = 0, MODE_INTER = (1 << 0), MODE_INTRA = (1 << 1), MODE_SKIP = (1 << 2) | MODE_INTER }; // motion vector predictor direction used in AMVP enum MVP_DIR { MD_LEFT = 0, // MVP of left block MD_ABOVE, // MVP of above block MD_ABOVE_RIGHT, // MVP of above right block MD_BELOW_LEFT, // MVP of below left block MD_ABOVE_LEFT, // MVP of above left block MD_COLLOCATED // MVP of temporal neighbour }; struct CUGeom { enum { INTRA = 1<<0, // CU is intra predicted PRESENT = 1<<1, // CU is not completely outside the frame SPLIT_MANDATORY = 1<<2, // CU split is mandatory if CU is inside frame and can be split LEAF = 1<<3, // CU is a leaf node of the CTU SPLIT = 1<<4, // CU is currently split in four child CUs. }; // (1 + 4 + 16 + 64) = 85. enum { MAX_GEOMS = 85 }; uint32_t log2CUSize; // Log of the CU size. uint32_t childOffset; // offset of the first child CU from current CU uint32_t absPartIdx; // Part index of this CU in terms of 4x4 blocks. uint32_t numPartitions; // Number of 4x4 blocks in the CU uint32_t flags; // CU flags. uint32_t depth; // depth of this CU relative from CTU uint32_t geomRecurId; // Unique geom id from 0 to MAX_GEOMS - 1 for every depth }; struct MVField { MV mv; int refIdx; }; // Structure that keeps the neighbour's MV information. struct InterNeighbourMV { // Neighbour MV. The index represents the list. MV mv[2]; // Collocated right bottom CU addr. uint32_t cuAddr[2]; bool isAvailable; // For spatial prediction, this field contains the reference index // in each list (-1 if not available). // // For temporal prediction, the first value is used for the // prediction with list 0. The second value is used for the prediction // with list 1. For each value, the first four bits are the reference index // associated to the PMV, and the fifth bit is the list associated to the PMV. // if both reference indices are -1, then unifiedRef is also -1 union { int16_t refIdx[2]; int32_t unifiedRef; }; }; struct IBC { int m_numBVs; int m_numBV16s; MV m_BVs[64]; MV m_lastIntraBCMv[2]; }; typedef void(*cucopy_t)(uint8_t* dst, uint8_t* src); // dst and src are aligned to MIN(size, 32) typedef void(*cubcast_t)(uint8_t* dst, uint8_t val); // dst is aligned to MIN(size, 32) // Partition count table, index represents partitioning mode. const uint32_t nbPartsTable[8] = { 1, 2, 2, 4, 2, 2, 2, 2 }; // Partition table. // First index is partitioning mode. Second index is partition index. // Third index is 0 for partition sizes, 1 for partition offsets. The // sizes and offsets are encoded as two packed 4-bit values (X,Y). // X and Y represent 1/4 fractions of the block size. const uint32_t partTable[8][4][2] = { // XY { { 0x44, 0x00 }, { 0x00, 0x00 }, { 0x00, 0x00 }, { 0x00, 0x00 } }, // SIZE_2Nx2N. { { 0x42, 0x00 }, { 0x42, 0x02 }, { 0x00, 0x00 }, { 0x00, 0x00 } }, // SIZE_2NxN. { { 0x24, 0x00 }, { 0x24, 0x20 }, { 0x00, 0x00 }, { 0x00, 0x00 } }, // SIZE_Nx2N. { { 0x22, 0x00 }, { 0x22, 0x20 }, { 0x22, 0x02 }, { 0x22, 0x22 } }, // SIZE_NxN. { { 0x41, 0x00 }, { 0x43, 0x01 }, { 0x00, 0x00 }, { 0x00, 0x00 } }, // SIZE_2NxnU. { { 0x43, 0x00 }, { 0x41, 0x03 }, { 0x00, 0x00 }, { 0x00, 0x00 } }, // SIZE_2NxnD. { { 0x14, 0x00 }, { 0x34, 0x10 }, { 0x00, 0x00 }, { 0x00, 0x00 } }, // SIZE_nLx2N. { { 0x34, 0x00 }, { 0x14, 0x30 }, { 0x00, 0x00 }, { 0x00, 0x00 } } // SIZE_nRx2N. }; // Partition Address table. // First index is partitioning mode. Second index is partition address. const uint32_t partAddrTable[8][4] = { { 0x00, 0x00, 0x00, 0x00 }, // SIZE_2Nx2N. { 0x00, 0x08, 0x08, 0x08 }, // SIZE_2NxN. { 0x00, 0x04, 0x04, 0x04 }, // SIZE_Nx2N. { 0x00, 0x04, 0x08, 0x0C }, // SIZE_NxN. { 0x00, 0x02, 0x02, 0x02 }, // SIZE_2NxnU. { 0x00, 0x0A, 0x0A, 0x0A }, // SIZE_2NxnD. { 0x00, 0x01, 0x01, 0x01 }, // SIZE_nLx2N. { 0x00, 0x05, 0x05, 0x05 } // SIZE_nRx2N. }; // Holds part data for a CU of a given size, from an 8x8 CU to a CTU class CUData { public: cubcast_t s_partSet[NUM_FULL_DEPTH]; // pointer to broadcast set functions per absolute depth uint32_t s_numPartInCUSize; bool m_vbvAffected; FrameData* m_encData; const Slice* m_slice; cucopy_t m_partCopy; // pointer to function that copies m_numPartitions elements cubcast_t m_partSet; // pointer to function that sets m_numPartitions elements cucopy_t m_subPartCopy; // pointer to function that copies m_numPartitions/4 elements, may be NULL cubcast_t m_subPartSet; // pointer to function that sets m_numPartitions/4 elements, may be NULL uint32_t m_cuAddr; // address of CTU within the picture in raster order uint32_t m_absIdxInCTU; // address of CU within its CTU in Z scan order uint32_t m_cuPelX; // CU position within the picture, in pixels (X) uint32_t m_cuPelY; // CU position within the picture, in pixels (Y) uint32_t m_numPartitions; // maximum number of 4x4 partitions within this CU uint32_t m_chromaFormat; uint32_t m_hChromaShift; uint32_t m_vChromaShift; /* multiple slices informations */ uint8_t m_bFirstRowInSlice; uint8_t m_bLastRowInSlice; uint8_t m_bLastCuInSlice; /* Per-part data, stored contiguously */ int8_t* m_qp; // array of QP values int8_t* m_qpAnalysis; // array of QP values for analysis reuse uint8_t* m_log2CUSize; // array of cu log2Size TODO: seems redundant to depth uint8_t* m_lumaIntraDir; // array of intra directions (luma) uint8_t* m_tqBypass; // array of CU lossless flags int8_t* m_refIdx[2]; // array of motion reference indices per list uint8_t* m_cuDepth; // array of depths uint8_t* m_predMode; // array of prediction modes uint8_t* m_partSize; // array of partition sizes uint8_t* m_mergeFlag; // array of merge flags uint8_t* m_skipFlag[2]; uint8_t* m_interDir; // array of inter directions uint8_t* m_mvpIdx[2]; // array of motion vector predictor candidates or merge candidate indices [0] uint8_t* m_tuDepth; // array of transform indices uint8_t* m_transformSkip[3]; // array of transform skipping flags per plane uint8_t* m_cbf[3]; // array of coded block flags (CBF) per plane uint8_t* m_chromaIntraDir; // array of intra directions (chroma) enum { BytesPerPartition = 24 }; // combined sizeof() of all per-part data sse_t* m_distortion; coeff_t* m_trCoeff[3]; // transformed coefficient buffer per plane int8_t m_refTuDepth[NUM_TU_DEPTH]; // TU depth of CU at depths 0, 1 and 2 MV* m_mv[2]; // array of motion vectors per list MV* m_mvd[2]; // array of coded motion vector deltas per list enum { TMVP_UNIT_MASK = 0xF0 }; // mask for mapping index to into a compressed (reference) MV field const CUData* m_cuAboveLeft; // pointer to above-left neighbor CTU const CUData* m_cuAboveRight; // pointer to above-right neighbor CTU const CUData* m_cuAbove; // pointer to above neighbor CTU const CUData* m_cuLeft; // pointer to left neighbor CTU double m_meanQP; uint64_t m_fAc_den[3]; uint64_t m_fDc_den[3]; /* Feature values per CTU for dynamic refinement */ uint64_t* m_collectCURd; uint32_t* m_collectCUVariance; uint32_t* m_collectCUCount; #if ENABLE_SCC_EXT MV m_lastIntraBCMv[2]; #endif CUData(); void initialize(const CUDataMemPool& dataPool, uint32_t depth, const x265_param& param, int instance); static void calcCTUGeoms(uint32_t ctuWidth, uint32_t ctuHeight, uint32_t maxCUSize, uint32_t minCUSize, CUGeom cuDataArray[CUGeom::MAX_GEOMS]); void initCTU(const Frame& frame, uint32_t cuAddr, int qp, uint32_t firstRowInSlice, uint32_t lastRowInSlice, uint32_t lastCUInSlice); void initSubCU(const CUData& ctu, const CUGeom& cuGeom, int qp, MV lastIntraBCMv[2] = 0); void initLosslessCU(const CUData& cu, const CUGeom& cuGeom); void copyPartFrom(const CUData& cu, const CUGeom& childGeom, uint32_t subPartIdx); void setEmptyPart(const CUGeom& childGeom, uint32_t subPartIdx); void copyToPic(uint32_t depth) const; /* RD-0 methods called only from encodeResidue */ void copyFromPic(const CUData& ctu, const CUGeom& cuGeom, int csp, bool copyQp = true); void updatePic(uint32_t depth, int picCsp) const; void setPartSizeSubParts(PartSize size) { m_partSet(m_partSize, (uint8_t)size); } void setPredModeSubParts(PredMode mode) { m_partSet(m_predMode, (uint8_t)mode); } void clearCbf() { m_partSet(m_cbf[0], 0); if (m_chromaFormat != X265_CSP_I400) { m_partSet(m_cbf[1], 0); m_partSet(m_cbf[2], 0);} } /* these functions all take depth as an absolute depth from CTU, it is used to calculate the number of parts to copy */ void setQPSubParts(int8_t qp, uint32_t absPartIdx, uint32_t depth) { s_partSet[depth]((uint8_t*)m_qp + absPartIdx, (uint8_t)qp); } void setTUDepthSubParts(uint8_t tuDepth, uint32_t absPartIdx, uint32_t depth) { s_partSet[depth](m_tuDepth + absPartIdx, tuDepth); } void setLumaIntraDirSubParts(uint8_t dir, uint32_t absPartIdx, uint32_t depth) { s_partSet[depth](m_lumaIntraDir + absPartIdx, dir); } void setChromIntraDirSubParts(uint8_t dir, uint32_t absPartIdx, uint32_t depth) { s_partSet[depth](m_chromaIntraDir + absPartIdx, dir); } void setCbfSubParts(uint8_t cbf, TextType ttype, uint32_t absPartIdx, uint32_t depth) { s_partSet[depth](m_cbf[ttype] + absPartIdx, cbf); } void setCbfPartRange(uint8_t cbf, TextType ttype, uint32_t absPartIdx, uint32_t coveredPartIdxes) { memset(m_cbf[ttype] + absPartIdx, cbf, coveredPartIdxes); } void setTransformSkipSubParts(uint8_t tskip, TextType ttype, uint32_t absPartIdx, uint32_t depth) { s_partSet[depth](m_transformSkip[ttype] + absPartIdx, tskip); } void setTransformSkipPartRange(uint8_t tskip, TextType ttype, uint32_t absPartIdx, uint32_t coveredPartIdxes) { memset(m_transformSkip[ttype] + absPartIdx, tskip, coveredPartIdxes); } bool setQPSubCUs(int8_t qp, uint32_t absPartIdx, uint32_t depth); void setPUInterDir(uint8_t dir, uint32_t absPartIdx, uint32_t puIdx); void setPUMv(int list, const MV& mv, int absPartIdx, int puIdx); void setPURefIdx(int list, int8_t refIdx, int absPartIdx, int puIdx); uint8_t getCbf(uint32_t absPartIdx, TextType ttype, uint32_t tuDepth) const { return (m_cbf[ttype][absPartIdx] >> tuDepth) & 0x1; } bool getQtRootCbf(uint32_t absPartIdx) const { return (m_cbf[0][absPartIdx] || ((m_chromaFormat != X265_CSP_I400) && (m_cbf[1][absPartIdx] || m_cbf[2][absPartIdx]))); } int8_t getRefQP(uint32_t currAbsIdxInCTU) const; uint32_t getInterMergeCandidates(uint32_t absPartIdx, uint32_t puIdx, MVField (*candMvField)[2], uint8_t* candDir) const; void clipMv(MV& outMV) const; int getPMV(InterNeighbourMV* neighbours, uint32_t reference_list, uint32_t refIdx, MV* amvpCand, MV* pmv, uint32_t puIdx = 0, uint32_t absPartIdx = 0) const; void getNeighbourMV(uint32_t puIdx, uint32_t absPartIdx, InterNeighbourMV* neighbours) const; void getIntraTUQtDepthRange(uint32_t tuDepthRange[2], uint32_t absPartIdx) const; void getInterTUQtDepthRange(uint32_t tuDepthRange[2], uint32_t absPartIdx) const; uint32_t getBestRefIdx(uint32_t subPartIdx) const { return ((m_interDir[subPartIdx] & 1) << m_refIdx[0][subPartIdx]) | (((m_interDir[subPartIdx] >> 1) & 1) << (m_refIdx[1][subPartIdx] + 16)); } uint32_t getPUOffset(uint32_t puIdx, uint32_t absPartIdx) const { return (partAddrTable[(int)m_partSize[absPartIdx]][puIdx] << (m_slice->m_param->unitSizeDepth - m_cuDepth[absPartIdx]) * 2) >> 4; } uint32_t getNumPartInter(uint32_t absPartIdx) const { return nbPartsTable[(int)m_partSize[absPartIdx]]; } bool isIntra(uint32_t absPartIdx) const { return m_predMode[absPartIdx] == MODE_INTRA; } bool isInter(uint32_t absPartIdx) const { return !!(m_predMode[absPartIdx] & MODE_INTER); } bool isSkipped(uint32_t absPartIdx) const { return m_predMode[absPartIdx] == MODE_SKIP; } bool isBipredRestriction() const { return m_log2CUSize[0] == 3 && m_partSize[0] != SIZE_2Nx2N; } void getPartIndexAndSize(uint32_t puIdx, uint32_t& absPartIdx, int& puWidth, int& puHeight) const; void getMvField(const CUData* cu, uint32_t absPartIdx, int picList, MVField& mvField) const; void getAllowedChromaDir(uint32_t absPartIdx, uint32_t* modeList) const; int getIntraDirLumaPredictor(uint32_t absPartIdx, uint32_t* intraDirPred) const; uint32_t getSCUAddr() const { return (m_cuAddr << m_slice->m_param->unitSizeDepth * 2) + m_absIdxInCTU; } uint32_t getCtxSplitFlag(uint32_t absPartIdx, uint32_t depth) const; uint32_t getCtxSkipFlag(uint32_t absPartIdx) const; void getTUEntropyCodingParameters(TUEntropyCodingParameters &result, uint32_t absPartIdx, uint32_t log2TrSize, bool bIsLuma) const; const CUData* getPULeft(uint32_t& lPartUnitIdx, uint32_t curPartUnitIdx) const; const CUData* getPUAbove(uint32_t& aPartUnitIdx, uint32_t curPartUnitIdx) const; const CUData* getPUAboveLeft(uint32_t& alPartUnitIdx, uint32_t curPartUnitIdx) const; const CUData* getPUAboveRight(uint32_t& arPartUnitIdx, uint32_t curPartUnitIdx) const; const CUData* getPUBelowLeft(uint32_t& blPartUnitIdx, uint32_t curPartUnitIdx) const; const CUData* getQpMinCuLeft(uint32_t& lPartUnitIdx, uint32_t currAbsIdxInCTU) const; const CUData* getQpMinCuAbove(uint32_t& aPartUnitIdx, uint32_t currAbsIdxInCTU) const; const CUData* getPUAboveRightAdi(uint32_t& arPartUnitIdx, uint32_t curPartUnitIdx, uint32_t partUnitOffset) const; const CUData* getPUBelowLeftAdi(uint32_t& blPartUnitIdx, uint32_t curPartUnitIdx, uint32_t partUnitOffset) const; #if ENABLE_SCC_EXT void getIntraBCMVPsEncOnly(uint32_t absPartIdx, MV* MvPred, int& nbPred, int puIdx); bool getDerivedBV(uint32_t absPartIdx, const MV& currentMv, MV& derivedMv, uint32_t width, uint32_t height); bool isIntraBC(const CUData* cu, uint32_t absPartIdx) const; bool getColMVPIBC(int ctuRsAddr, int partUnitIdx, MV& rcMv); void roundMergeCandidates(MVField(*candMvField)[2], int iCount) const; bool is8x8BipredRestriction(MV mvL0, MV mvL1, int iRefIdxL0, int iRefIdxL1) const; #endif protected: template void setAllPU(T *p, const T& val, int absPartIdx, int puIdx); int8_t getLastCodedQP(uint32_t absPartIdx) const; int getLastValidPartIdx(int absPartIdx) const; bool hasEqualMotion(uint32_t absPartIdx, const CUData& candCU, uint32_t candAbsPartIdx) const; /* Check whether the current PU and a spatial neighboring PU are in same merge region */ bool isDiffMER(int xN, int yN, int xP, int yP) const { return ((xN >> 2) != (xP >> 2)) || ((yN >> 2) != (yP >> 2)); } // add possible motion vector predictor candidates bool getDirectPMV(MV& pmv, InterNeighbourMV *neighbours, uint32_t picList, uint32_t refIdx) const; bool getIndirectPMV(MV& outMV, InterNeighbourMV *neighbours, uint32_t reference_list, uint32_t refIdx) const; void getInterNeighbourMV(InterNeighbourMV *neighbour, uint32_t partUnitIdx, MVP_DIR dir) const; bool getColMVP(MV& outMV, int& outRefIdx, int picList, int cuAddr, int absPartIdx) const; bool getCollocatedMV(int cuAddr, int partUnitIdx, InterNeighbourMV *neighbour) const; MV scaleMvByPOCDist(const MV& inMV, int curPOC, int curRefPOC, int colPOC, int colRefPOC) const; void deriveLeftRightTopIdx(uint32_t puIdx, uint32_t& partIdxLT, uint32_t& partIdxRT) const; uint32_t deriveCenterIdx(uint32_t puIdx) const; uint32_t deriveRightBottomIdx(uint32_t puIdx) const; uint32_t deriveLeftBottomIdx(uint32_t puIdx) const; }; // TU settings for entropy encoding struct TUEntropyCodingParameters { const uint16_t *scan; const uint16_t *scanCG; ScanType scanType; uint32_t firstSignificanceMapContext; }; struct CUDataMemPool { uint8_t* charMemBlock; coeff_t* trCoeffMemBlock; MV* mvMemBlock; sse_t* distortionMemBlock; uint64_t* dynRefineRdBlock; uint32_t* dynRefCntBlock; uint32_t* dynRefVarBlock; CUDataMemPool() { charMemBlock = NULL; trCoeffMemBlock = NULL; mvMemBlock = NULL; distortionMemBlock = NULL; dynRefineRdBlock = NULL; dynRefCntBlock = NULL; dynRefVarBlock = NULL;} bool create(uint32_t depth, uint32_t csp, uint32_t numInstances, const x265_param& param) { uint32_t numPartition = param.num4x4Partitions >> (depth * 2); uint32_t cuSize = param.maxCUSize >> depth; uint32_t sizeL = cuSize * cuSize; if (csp == X265_CSP_I400) { CHECKED_MALLOC(trCoeffMemBlock, coeff_t, (sizeL) * numInstances); } else { uint32_t sizeC = sizeL >> (CHROMA_H_SHIFT(csp) + CHROMA_V_SHIFT(csp)); CHECKED_MALLOC(trCoeffMemBlock, coeff_t, (sizeL + sizeC * 2) * numInstances); } CHECKED_MALLOC(charMemBlock, uint8_t, numPartition * numInstances * CUData::BytesPerPartition); CHECKED_MALLOC_ZERO(mvMemBlock, MV, numPartition * 4 * numInstances); CHECKED_MALLOC(distortionMemBlock, sse_t, numPartition * numInstances); return true; fail: return false; } void destroy() { X265_FREE(trCoeffMemBlock); X265_FREE(mvMemBlock); X265_FREE(charMemBlock); X265_FREE(distortionMemBlock); } }; } #endif // ifndef X265_CUDATA_H