--- mbtransquant.c 2002/11/17 00:51:10 1.19 +++ mbtransquant.c 2003/04/10 13:05:54 1.21.2.7 @@ -1,63 +1,32 @@ /***************************************************************************** * * XVID MPEG-4 VIDEO CODEC - * - MacroBlock transfer and quantization - + * - MB Transfert/Quantization functions - * - * Copyright(C) 2002-2001 Christoph Lampert - * 2002-2001 Michael Militzer - * 2002-2001 Peter Ross - * 2002 Daniel Smith - * - * This file is part of XviD, a free MPEG-4 video encoder/decoder - * - * XviD 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 + * Copyright(C) 2001-2003 Peter Ross + * 2001-2003 Michael Militzer + * 2003 Edouard Gomez + * + * 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 + * 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 + * along with this program ; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * - * Under section 8 of the GNU General Public License, the copyright - * holders of XVID explicitly forbid distribution in the following - * countries: - * - * - Japan - * - United States of America - * - * Linking XviD statically or dynamically with other modules is making a - * combined work based on XviD. Thus, the terms and conditions of the - * GNU General Public License cover the whole combination. - * - * As a special exception, the copyright holders of XviD give you - * permission to link XviD with independent modules that communicate with - * XviD solely through the VFW1.1 and DShow interfaces, regardless of the - * license terms of these independent modules, and to copy and distribute - * the resulting combined work under terms of your choice, provided that - * every copy of the combined work is accompanied by a complete copy of - * the source code of XviD (the version of XviD used to produce the - * combined work), being distributed under the terms of the GNU General - * Public License plus this exception. An independent module is a module - * which is not derived from or based on XviD. - * - * Note that people who make modified versions of XviD are not obligated - * to grant this special exception for their modified versions; it is - * their choice whether to do so. The GNU General Public License gives - * permission to release a modified version without this exception; this - * exception also makes it possible to release a modified version which - * carries forward this exception. - * - * $Id: mbtransquant.c,v 1.19 2002/11/17 00:51:10 edgomez Exp $ + * $Id: mbtransquant.c,v 1.21.2.7 2003/04/10 13:05:54 edgomez Exp $ * ****************************************************************************/ #include +#include #include "../portab.h" #include "mbfunctions.h" @@ -71,524 +40,443 @@ #include "../quant/quant_h263.h" #include "../encoder.h" -#define MIN(X, Y) ((X)<(Y)?(X):(Y)) -#define MAX(X, Y) ((X)>(Y)?(X):(Y)) +#include "../image/reduced.h" -#define TOOSMALL_LIMIT 3 /* skip blocks having a coefficient sum below this value */ +MBFIELDTEST_PTR MBFieldTest; -/* this isnt pretty, but its better than 20 ifdefs */ +/* + * Skip blocks having a coefficient sum below this value. This value will be + * corrected according to the MB quantizer to avoid artifacts for quant==1 + */ +#define PVOP_TOOSMALL_LIMIT 1 +#define BVOP_TOOSMALL_LIMIT 3 -void -MBTransQuantIntra(const MBParam * pParam, - FRAMEINFO * frame, - MACROBLOCK * pMB, - const uint32_t x_pos, - const uint32_t y_pos, - int16_t data[6 * 64], - int16_t qcoeff[6 * 64]) -{ +/***************************************************************************** + * Local functions + ****************************************************************************/ - uint32_t stride = pParam->edged_width; - uint32_t stride2 = stride / 2; - uint32_t next_block = stride * 8; - uint32_t i; - uint32_t iQuant = frame->quant; - uint8_t *pY_Cur, *pU_Cur, *pV_Cur; - IMAGE *pCurrent = &frame->image; +/* permute block and return field dct choice */ +static __inline uint32_t +MBDecideFieldDCT(int16_t data[6 * 64]) +{ + uint32_t field = MBFieldTest(data); - pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); - pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); - pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); + if (field) + MBFrameToField(data); - start_timer(); - transfer_8to16copy(&data[0 * 64], pY_Cur, stride); - transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); - transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); - transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); - transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); - transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); - stop_transfer_timer(); + return field; +} +/* Performs Forward DCT on all blocks */ +static __inline void +MBfDCT(const MBParam * pParam, + FRAMEINFO * frame, + MACROBLOCK * pMB, + uint32_t x_pos, + uint32_t y_pos, + int16_t data[6 * 64]) +{ + /* Handles interlacing */ start_timer(); pMB->field_dct = 0; - if ((frame->global_flags & XVID_INTERLACING) && + if ((frame->vol_flags & XVID_VOL_INTERLACING) && (x_pos>0) && (x_posmb_width-1) && (y_pos>0) && (y_posmb_height-1)) { pMB->field_dct = MBDecideFieldDCT(data); } stop_interlacing_timer(); - for (i = 0; i < 6; i++) { - uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); - - start_timer(); - fdct(&data[i * 64]); - stop_dct_timer(); + /* Perform DCT */ + start_timer(); + fdct(&data[0 * 64]); + fdct(&data[1 * 64]); + fdct(&data[2 * 64]); + fdct(&data[3 * 64]); + fdct(&data[4 * 64]); + fdct(&data[5 * 64]); + stop_dct_timer(); +} - if (pParam->m_quant_type == H263_QUANT) { - start_timer(); - quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); - stop_quant_timer(); +/* Performs Inverse DCT on all blocks */ +static __inline void +MBiDCT(int16_t data[6 * 64], + const uint8_t cbp) +{ + start_timer(); + if(cbp & (1 << (5 - 0))) idct(&data[0 * 64]); + if(cbp & (1 << (5 - 1))) idct(&data[1 * 64]); + if(cbp & (1 << (5 - 2))) idct(&data[2 * 64]); + if(cbp & (1 << (5 - 3))) idct(&data[3 * 64]); + if(cbp & (1 << (5 - 4))) idct(&data[4 * 64]); + if(cbp & (1 << (5 - 5))) idct(&data[5 * 64]); + stop_idct_timer(); +} - start_timer(); - dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); - stop_iquant_timer(); - } else { - start_timer(); - quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); - stop_quant_timer(); +/* Quantize all blocks -- Intra mode */ +static __inline void +MBQuantIntra(const MBParam * pParam, + const MACROBLOCK * pMB, + int16_t qcoeff[6 * 64], + int16_t data[6*64]) +{ + int i; - start_timer(); - dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); - stop_iquant_timer(); - } + for (i = 0; i < 6; i++) { + uint32_t iDcScaler = get_dc_scaler(pMB->quant, i < 4); + /* Quantize the block */ start_timer(); - idct(&data[i * 64]); - stop_idct_timer(); + if (!(pParam->vol_flags & XVID_VOL_MPEGQUANT)) + quant_intra(&data[i * 64], &qcoeff[i * 64], pMB->quant, iDcScaler); + else + quant4_intra(&data[i * 64], &qcoeff[i * 64], pMB->quant, iDcScaler); + stop_quant_timer(); } +} - if (pMB->field_dct) { - next_block = stride; - stride *= 2; - } +/* DeQuantize all blocks -- Intra mode */ +static __inline void +MBDeQuantIntra(const MBParam * pParam, + const int iQuant, + int16_t qcoeff[6 * 64], + int16_t data[6*64]) +{ + int i; - start_timer(); - transfer_16to8copy(pY_Cur, &data[0 * 64], stride); - transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); - transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); - transfer_16to8copy(pY_Cur + next_block + 8, &data[3 * 64], stride); - transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); - transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); - stop_transfer_timer(); + for (i = 0; i < 6; i++) { + uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); + start_timer(); + if (!(pParam->vol_flags & XVID_VOL_MPEGQUANT)) + dequant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); + else + dequant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); + stop_iquant_timer(); + } } - -uint8_t -MBTransQuantInter(const MBParam * pParam, - FRAMEINFO * frame, - MACROBLOCK * pMB, - const uint32_t x_pos, - const uint32_t y_pos, - int16_t data[6 * 64], - int16_t qcoeff[6 * 64]) +/* Quantize all blocks -- Inter mode */ +static __inline uint8_t +MBQuantInter(const MBParam * pParam, + const MACROBLOCK * pMB, + int16_t data[6 * 64], + int16_t qcoeff[6 * 64], + int bvop, + int limit) { - uint32_t stride = pParam->edged_width; - uint32_t stride2 = stride / 2; - uint32_t next_block = stride * 8; - uint32_t i; - uint32_t iQuant = frame->quant; - uint8_t *pY_Cur, *pU_Cur, *pV_Cur; + int i; uint8_t cbp = 0; - uint32_t sum; - IMAGE *pCurrent = &frame->image; - - pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); - pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); - pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); - - start_timer(); - pMB->field_dct = 0; - if ((frame->global_flags & XVID_INTERLACING) && - (x_pos>0) && (x_posmb_width-1) && - (y_pos>0) && (y_posmb_height-1)) { - pMB->field_dct = MBDecideFieldDCT(data); - } - stop_interlacing_timer(); + int sum; + int code_block; for (i = 0; i < 6; i++) { - /* - * no need to transfer 8->16-bit - * (this is performed already in motion compensation) - */ + + /* Quantize the block */ start_timer(); - fdct(&data[i * 64]); - stop_dct_timer(); - - if (pParam->m_quant_type == 0) { - start_timer(); - sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant); - stop_quant_timer(); + if (!(pParam->vol_flags & XVID_VOL_MPEGQUANT)) + sum = quant_inter(&qcoeff[i * 64], &data[i * 64], pMB->quant); + else + sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], pMB->quant); + stop_quant_timer(); + + /* + * We code the block if the sum is higher than the limit and if the first + * two AC coefficients in zig zag order are not zero. + */ + code_block = 0; + if ((sum >= limit) || (qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) { + code_block = 1; } else { - start_timer(); - sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant); - stop_quant_timer(); - } - - if ((sum >= TOOSMALL_LIMIT) || (qcoeff[i*64] != 0) || - (qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) { - if (pParam->m_quant_type == H263_QUANT) { - start_timer(); - dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); - stop_iquant_timer(); + if (bvop && (pMB->mode == MODE_DIRECT || pMB->mode == MODE_DIRECT_NO4V)) { + /* dark blocks prevention for direct mode */ + if ((qcoeff[i*64] < -1) || (qcoeff[i*64] > 0)) + code_block = 1; } else { - start_timer(); - dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); - stop_iquant_timer(); + /* not direct mode */ + if (qcoeff[i*64] != 0) + code_block = 1; } - - cbp |= 1 << (5 - i); - - start_timer(); - idct(&data[i * 64]); - stop_idct_timer(); } - } - if (pMB->field_dct) { - next_block = stride; - stride *= 2; - } - - start_timer(); - if (cbp & 32) - transfer_16to8add(pY_Cur, &data[0 * 64], stride); - if (cbp & 16) - transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); - if (cbp & 8) - transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); - if (cbp & 4) - transfer_16to8add(pY_Cur + next_block + 8, &data[3 * 64], stride); - if (cbp & 2) - transfer_16to8add(pU_Cur, &data[4 * 64], stride2); - if (cbp & 1) - transfer_16to8add(pV_Cur, &data[5 * 64], stride2); - stop_transfer_timer(); + /* Set the corresponding cbp bit */ + cbp |= code_block << (5 - i); - return cbp; + } + return(cbp); } -void -MBTransQuantIntra2(const MBParam * pParam, - FRAMEINFO * frame, - MACROBLOCK * pMB, - const uint32_t x_pos, - const uint32_t y_pos, - int16_t data[6 * 64], - int16_t qcoeff[6 * 64]) +/* DeQuantize all blocks -- Inter mode */ +static __inline void +MBDeQuantInter(const MBParam * pParam, + const int iQuant, + int16_t data[6 * 64], + int16_t qcoeff[6 * 64], + const uint8_t cbp) { - MBTrans(pParam,frame,pMB,x_pos,y_pos,data); - MBfDCT(pParam,frame,pMB,data); - MBQuantIntra(pParam,frame,pMB,data,qcoeff); - MBDeQuantIntra(pParam,frame->quant,data,qcoeff); - MBiDCT(data,0x3F); - MBTransAdd(pParam,frame,pMB,x_pos,y_pos,data,0x3F); -} + int i; + for (i = 0; i < 6; i++) { + if (cbp & (1 << (5 - i))) { + start_timer(); + if (!(pParam->vol_flags & XVID_VOL_MPEGQUANT)) + dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); + else + dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); + stop_iquant_timer(); + } + } +} -uint8_t -MBTransQuantInter2(const MBParam * pParam, - FRAMEINFO * frame, - MACROBLOCK * pMB, - const uint32_t x_pos, - const uint32_t y_pos, - int16_t data[6 * 64], - int16_t qcoeff[6 * 64]) -{ - uint8_t cbp; - -/* there is no MBTrans for Inter block, that's done in motion compensation already */ +typedef void (transfer_operation_8to16_t) (int16_t *Dst, const uint8_t *Src, int BpS); +typedef void (transfer_operation_16to8_t) (uint8_t *Dst, const int16_t *Src, int BpS); - MBfDCT(pParam,frame,pMB,data); - cbp = MBQuantInter(pParam,frame->quant,data,qcoeff); - MBDeQuantInter(pParam,frame->quant,data,qcoeff,cbp); - MBiDCT(data,cbp); - MBTransAdd(pParam,frame,pMB,x_pos,y_pos,data,cbp); - - return cbp; -} -uint8_t -MBTransQuantInterBVOP(const MBParam * pParam, - FRAMEINFO * frame, - MACROBLOCK * pMB, - int16_t data[6 * 64], - int16_t qcoeff[6 * 64]) +static __inline void +MBTrans8to16(const MBParam * pParam, + FRAMEINFO * frame, + MACROBLOCK * pMB, + const uint32_t x_pos, + const uint32_t y_pos, + int16_t data[6 * 64]) { - uint8_t cbp; - -/* there is no MBTrans for Inter block, that's done in motion compensation already */ + uint32_t stride = pParam->edged_width; + uint32_t stride2 = stride / 2; + uint32_t next_block = stride * 8; + int32_t cst; + uint8_t *pY_Cur, *pU_Cur, *pV_Cur; + IMAGE *pCurrent = &frame->image; + transfer_operation_8to16_t *transfer_op = NULL; - MBfDCT(pParam,frame,pMB,data); - cbp = MBQuantInter(pParam,frame->quant,data,qcoeff); + if ((frame->vop_flags & XVID_VOP_REDUCED)) { -/* we don't have to DeQuant, iDCT and Transfer back data for B-frames */ + /* Image pointers */ + pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); + pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); + pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); - return cbp; -} + /* Block size */ + cst = 16; + /* Operation function */ + transfer_op = (transfer_operation_8to16_t*)filter_18x18_to_8x8; + } else { -void -MBfDCT(const MBParam * pParam, - FRAMEINFO * frame, - MACROBLOCK * pMB, - int16_t data[6 * 64]) -{ - int i; + /* Image pointers */ + pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); + pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); + pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); - start_timer(); - pMB->field_dct = 0; - if ((frame->global_flags & XVID_INTERLACING)) { - pMB->field_dct = MBDecideFieldDCT(data); - } - stop_interlacing_timer(); + /* Block size */ + cst = 8; - for (i = 0; i < 6; i++) { - start_timer(); - fdct(&data[i * 64]); - stop_dct_timer(); + /* Operation function */ + transfer_op = (transfer_operation_8to16_t*)transfer_8to16copy; } -} - -void -MBQuantDeQuantIntra(const MBParam * pParam, - FRAMEINFO * frame, - MACROBLOCK * pMB, - int16_t qcoeff[6 * 64], - int16_t data[6*64]) -{ - int i; - int iQuant = frame->quant; + /* Do the transfer */ start_timer(); - pMB->field_dct = 0; - if ((frame->global_flags & XVID_INTERLACING)) { - pMB->field_dct = MBDecideFieldDCT(data); - } - stop_interlacing_timer(); - - for (i = 0; i < 6; i++) { - uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); - - if (pParam->m_quant_type == H263_QUANT) { - start_timer(); - quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); - stop_quant_timer(); - - start_timer(); - dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); - stop_iquant_timer(); - } else { - start_timer(); - quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); - stop_quant_timer(); - - start_timer(); - dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); - stop_iquant_timer(); - } - } -} + transfer_op(&data[0 * 64], pY_Cur, stride); + transfer_op(&data[1 * 64], pY_Cur + cst, stride); + transfer_op(&data[2 * 64], pY_Cur + next_block, stride); + transfer_op(&data[3 * 64], pY_Cur + next_block + cst, stride); + transfer_op(&data[4 * 64], pU_Cur, stride2); + transfer_op(&data[5 * 64], pV_Cur, stride2); + stop_transfer_timer(); +} -void -MBQuantIntra(const MBParam * pParam, - FRAMEINFO * frame, - MACROBLOCK *pMB, +static __inline void +MBTrans16to8(const MBParam * pParam, + FRAMEINFO * frame, + MACROBLOCK * pMB, + const uint32_t x_pos, + const uint32_t y_pos, int16_t data[6 * 64], - int16_t qcoeff[6 * 64]) + const uint32_t add, + const uint8_t cbp) { - int i; - int iQuant = frame->quant; + uint8_t *pY_Cur, *pU_Cur, *pV_Cur; + uint32_t stride = pParam->edged_width; + uint32_t stride2 = stride / 2; + uint32_t next_block = stride * 8; + uint32_t cst; + IMAGE *pCurrent = &frame->image; + transfer_operation_16to8_t *transfer_op = NULL; - start_timer(); - pMB->field_dct = 0; - if ((frame->global_flags & XVID_INTERLACING)) { - pMB->field_dct = MBDecideFieldDCT(data); + if (pMB->field_dct) { + next_block = stride; + stride *= 2; } - stop_interlacing_timer(); - for (i = 0; i < 6; i++) { - uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); + if ((frame->vop_flags & XVID_VOP_REDUCED)) { - if (pParam->m_quant_type == H263_QUANT) { - start_timer(); - quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); - stop_quant_timer(); - } else { - start_timer(); - quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); - stop_quant_timer(); - } + /* Image pointers */ + pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); + pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); + pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); + + /* Block size */ + cst = 16; + + /* Operation function */ + if(add) + transfer_op = (transfer_operation_16to8_t*)add_upsampled_8x8_16to8; + else + transfer_op = (transfer_operation_16to8_t*)copy_upsampled_8x8_16to8; + } else { + + /* Image pointers */ + pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); + pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); + pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); + + /* Block size */ + cst = 8; + + /* Operation function */ + if(add) + transfer_op = (transfer_operation_16to8_t*)transfer_16to8add; + else + transfer_op = (transfer_operation_16to8_t*)transfer_16to8copy; } -} - -void -MBDeQuantIntra(const MBParam * pParam, - const int iQuant, - int16_t qcoeff[6 * 64], - int16_t data[6*64]) -{ - int i; - for (i = 0; i < 6; i++) { - uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); - - if (pParam->m_quant_type == H263_QUANT) { - start_timer(); - dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); - stop_iquant_timer(); - } else { - start_timer(); - dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); - stop_iquant_timer(); - } - } + /* Do the operation */ + start_timer(); + if (cbp&32) transfer_op(pY_Cur, &data[0 * 64], stride); + if (cbp&16) transfer_op(pY_Cur + cst, &data[1 * 64], stride); + if (cbp& 8) transfer_op(pY_Cur + next_block, &data[2 * 64], stride); + if (cbp& 4) transfer_op(pY_Cur + next_block + cst, &data[3 * 64], stride); + if (cbp& 2) transfer_op(pU_Cur, &data[4 * 64], stride2); + if (cbp& 1) transfer_op(pV_Cur, &data[5 * 64], stride2); + stop_transfer_timer(); } -uint8_t -MBQuantInter(const MBParam * pParam, - const int iQuant, +/***************************************************************************** + * Module functions + ****************************************************************************/ + +void +MBTransQuantIntra(const MBParam * pParam, + FRAMEINFO * frame, + MACROBLOCK * pMB, + const uint32_t x_pos, + const uint32_t y_pos, int16_t data[6 * 64], int16_t qcoeff[6 * 64]) { - int i; - uint8_t cbp = 0; - int sum; + /* Transfer data */ + MBTrans8to16(pParam, frame, pMB, x_pos, y_pos, data); - for (i = 0; i < 6; i++) { - - if (pParam->m_quant_type == 0) { - start_timer(); - sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant); - stop_quant_timer(); - } else { - start_timer(); - sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant); - stop_quant_timer(); - } + /* Perform DCT (and field decision) */ + MBfDCT(pParam, frame, pMB, x_pos, y_pos, data); - if (sum >= TOOSMALL_LIMIT) { // skip block ? - cbp |= 1 << (5 - i); - } - } - return cbp; -} + /* Quantize the block */ + MBQuantIntra(pParam, pMB, data, qcoeff); -void -MBDeQuantInter( const MBParam * pParam, - const int iQuant, - int16_t data[6 * 64], - int16_t qcoeff[6 * 64], - const uint8_t cbp) -{ - int i; + /* DeQuantize the block */ + MBDeQuantIntra(pParam, pMB->quant, data, qcoeff); - for (i = 0; i < 6; i++) { - if (cbp & (1 << (5 - i))) - { - if (pParam->m_quant_type == H263_QUANT) { - start_timer(); - dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); - stop_iquant_timer(); - } else { - start_timer(); - dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); - stop_iquant_timer(); - } - } - } -} - -void -MBiDCT( int16_t data[6 * 64], - const uint8_t cbp) -{ - int i; + /* Perform inverse DCT*/ + MBiDCT(data, 0x3F); - for (i = 0; i < 6; i++) { - if (cbp & (1 << (5 - i))) - { - start_timer(); - idct(&data[i * 64]); - stop_idct_timer(); - - } - } + /* Transfer back the data -- Don't add data */ + MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 0, 0x3F); } -void -MBTrans(const MBParam * pParam, +uint8_t +MBTransQuantInter(const MBParam * pParam, FRAMEINFO * frame, MACROBLOCK * pMB, const uint32_t x_pos, const uint32_t y_pos, - int16_t data[6 * 64]) + int16_t data[6 * 64], + int16_t qcoeff[6 * 64]) { - uint32_t stride = pParam->edged_width; - uint32_t stride2 = stride / 2; - uint32_t next_block = stride * 8; - uint8_t *pY_Cur, *pU_Cur, *pV_Cur; - IMAGE *pCurrent = &frame->image; + uint8_t cbp; + uint32_t limit; - pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); - pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); - pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); + /* + * There is no MBTrans8to16 for Inter block, that's done in motion compensation + * already + */ - start_timer(); - transfer_8to16copy(&data[0 * 64], pY_Cur, stride); - transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); - transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); - transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); - transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); - transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); - stop_transfer_timer(); -} + /* Perform DCT (and field decision) */ + MBfDCT(pParam, frame, pMB, x_pos, y_pos, data); + + /* Set the limit threshold */ + limit = PVOP_TOOSMALL_LIMIT + ((pMB->quant == 1)? 1 : 0); + + /* Quantize the block */ + cbp = MBQuantInter(pParam, pMB, data, qcoeff, 0, limit); + + /* DeQuantize the block */ + MBDeQuantInter(pParam, pMB->quant, data, qcoeff, cbp); + + /* Perform inverse DCT*/ + MBiDCT(data, cbp); + + /* Transfer back the data -- Add the data */ + MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 1, cbp); -void -MBTransAdd(const MBParam * pParam, + return(cbp); +} + +uint8_t +MBTransQuantInterBVOP(const MBParam * pParam, FRAMEINFO * frame, MACROBLOCK * pMB, const uint32_t x_pos, const uint32_t y_pos, int16_t data[6 * 64], - const uint8_t cbp) + int16_t qcoeff[6 * 64]) { - uint8_t *pY_Cur, *pU_Cur, *pV_Cur; - uint32_t stride = pParam->edged_width; - uint32_t stride2 = stride / 2; - uint32_t next_block = stride * 8; - IMAGE *pCurrent = &frame->image; + uint8_t cbp; + uint32_t limit; + + /* + * There is no MBTrans8to16 for Inter block, that's done in motion compensation + * already + */ - pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); - pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); - pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); + /* Perform DCT (and field decision) */ + MBfDCT(pParam, frame, pMB, x_pos, y_pos, data); - if (pMB->field_dct) { - next_block = stride; - stride *= 2; - } + /* Set the limit threshold */ + limit = BVOP_TOOSMALL_LIMIT; - start_timer(); - if (cbp & 32) - transfer_16to8add(pY_Cur, &data[0 * 64], stride); - if (cbp & 16) - transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); - if (cbp & 8) - transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); - if (cbp & 4) - transfer_16to8add(pY_Cur + next_block + 8, &data[3 * 64], stride); - if (cbp & 2) - transfer_16to8add(pU_Cur, &data[4 * 64], stride2); - if (cbp & 1) - transfer_16to8add(pV_Cur, &data[5 * 64], stride2); - stop_transfer_timer(); -} + /* Quantize the block */ + cbp = MBQuantInter(pParam, pMB, data, qcoeff, 1, limit); + /* + * History comment: + * We don't have to DeQuant, iDCT and Transfer back data for B-frames. + * + * BUT some plugins require the original frame to be passed so we have + * to take care of that here + */ + if((pParam->plugin_flags & XVID_REQORIGINAL)) { + /* DeQuantize the block */ + MBDeQuantInter(pParam, pMB->quant, data, qcoeff, cbp); -/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ + /* Perform inverse DCT*/ + MBiDCT(data, cbp); + /* Transfer back the data -- Add the data */ + MBTrans16to8(pParam, frame, pMB, x_pos, y_pos, data, 1, cbp); + } + return(cbp); +} + +/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ uint32_t -MBDecideFieldDCT(int16_t data[6 * 64]) +MBFieldTest_c(int16_t data[6 * 64]) { - const uint8_t blocks[] = { 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 }; const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 }; @@ -599,34 +487,30 @@ for (i = 0; i < 7; ++i) { for (j = 0; j < 8; ++j) { frame += - ABS(data[0 * 64 + (i + 1) * 8 + j] - data[0 * 64 + i * 8 + j]); + abs(data[0 * 64 + (i + 1) * 8 + j] - data[0 * 64 + i * 8 + j]); frame += - ABS(data[1 * 64 + (i + 1) * 8 + j] - data[1 * 64 + i * 8 + j]); + abs(data[1 * 64 + (i + 1) * 8 + j] - data[1 * 64 + i * 8 + j]); frame += - ABS(data[2 * 64 + (i + 1) * 8 + j] - data[2 * 64 + i * 8 + j]); + abs(data[2 * 64 + (i + 1) * 8 + j] - data[2 * 64 + i * 8 + j]); frame += - ABS(data[3 * 64 + (i + 1) * 8 + j] - data[3 * 64 + i * 8 + j]); + abs(data[3 * 64 + (i + 1) * 8 + j] - data[3 * 64 + i * 8 + j]); field += - ABS(data[blocks[i + 1] + lines[i + 1] + j] - + abs(data[blocks[i + 1] + lines[i + 1] + j] - data[blocks[i] + lines[i] + j]); field += - ABS(data[blocks[i + 1] + lines[i + 1] + 8 + j] - + abs(data[blocks[i + 1] + lines[i + 1] + 8 + j] - data[blocks[i] + lines[i] + 8 + j]); field += - ABS(data[blocks[i + 1] + 64 + lines[i + 1] + j] - + abs(data[blocks[i + 1] + 64 + lines[i + 1] + j] - data[blocks[i] + 64 + lines[i] + j]); field += - ABS(data[blocks[i + 1] + 64 + lines[i + 1] + 8 + j] - + abs(data[blocks[i + 1] + 64 + lines[i + 1] + 8 + j] - data[blocks[i] + 64 + lines[i] + 8 + j]); } } - if (frame > (field + 350)) { - MBFrameToField(data); - } - - return (frame > (field + 350)); + return (frame >= (field + 350)); }