--- mbtransquant.c 2002/12/09 10:47:05 1.11.2.6 +++ mbtransquant.c 2002/09/21 03:07:56 1.13 @@ -1,55 +1,37 @@ - /****************************************************************************** - * * - * This file is part of XviD, a free MPEG-4 video encoder/decoder * - * * - * XviD is an implementation of a part of one or more MPEG-4 Video tools * - * as specified in ISO/IEC 14496-2 standard. Those intending to use this * - * software module in hardware or software products are advised that its * - * use may infringe existing patents or copyrights, and any such use * - * would be at such party's own risk. The original developer of this * - * software module and his/her company, and subsequent editors and their * - * companies, will have no liability for use of this software or * - * modifications or derivatives thereof. * - * * - * 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 * - * (at your option) any later version. * - * * - * XviD 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * - * * - ******************************************************************************/ - - /****************************************************************************** - * * - * mbtransquant.c * - * * - * Copyright (C) 2001 - Peter Ross * - * Copyright (C) 2001 - Michael Militzer * - * * - * For more information visit the XviD homepage: http://www.xvid.org * - * * - ******************************************************************************/ - - /****************************************************************************** - * * - * Revision history: * - * * - * 29.03.2002 interlacing speedup - used transfer strides instead of * - * manual field-to-frame conversion * - * 26.03.2002 interlacing support - moved transfers outside loops * - * 22.12.2001 get_dc_scaler() moved to common.h * - * 19.11.2001 introduced coefficient thresholding (Isibaar) * - * 17.11.2001 initial version * - * * - ******************************************************************************/ +/***************************************************************************** + * + * XVID MPEG-4 VIDEO CODEC + * - MacroBlock transfer and quantization - + * + * Copyright(C) 2002-2001 Michael Militzer + * 2002-2001 Peter Ross + * + * This program is an implementation of a part of one or more MPEG-4 + * Video tools as specified in ISO/IEC 14496-2 standard. Those intending + * to use this software module in hardware or software products are + * advised that its use may infringe existing patents or copyrights, and + * any such use would be at such party's own risk. The original + * developer of this software module and his/her company, and subsequent + * editors and their companies, will have no liability for use of this + * software or modifications or derivatives thereof. + * + * 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + * $Id: mbtransquant.c,v 1.13 2002/09/21 03:07:56 suxen_drol Exp $ + * + ****************************************************************************/ #include @@ -65,14 +47,12 @@ #include "../quant/quant_h263.h" #include "../encoder.h" -#include "../image/reduced.h" - -MBFIELDTEST_PTR MBFieldTest; - #define MIN(X, Y) ((X)<(Y)?(X):(Y)) #define MAX(X, Y) ((X)>(Y)?(X):(Y)) -#define TOOSMALL_LIMIT 1 /* skip blocks having a coefficient sum below this value */ +#define TOOSMALL_LIMIT 3 /* skip blocks having a coefficient sum below this value */ + +/* this isnt pretty, but its better than 20 ifdefs */ void MBTransQuantIntra(const MBParam * pParam, @@ -86,40 +66,25 @@ uint32_t stride = pParam->edged_width; uint32_t stride2 = stride / 2; - uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); + 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; + 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(); - if ((frame->global_flags & XVID_REDUCED)) - { - 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); - - filter_18x18_to_8x8(&data[0 * 64], pY_Cur, stride); - filter_18x18_to_8x8(&data[1 * 64], pY_Cur + 16, stride); - filter_18x18_to_8x8(&data[2 * 64], pY_Cur + next_block, stride); - filter_18x18_to_8x8(&data[3 * 64], pY_Cur + next_block + 16, stride); - filter_18x18_to_8x8(&data[4 * 64], pU_Cur, stride2); - filter_18x18_to_8x8(&data[5 * 64], pV_Cur, stride2); - }else{ - 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); - - 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); - } + 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(); - /* XXX: rrv+interlacing is buggy */ start_timer(); pMB->field_dct = 0; if ((frame->global_flags & XVID_INTERLACING) && @@ -140,61 +105,39 @@ 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(); - } - - /* speedup: dont decode when encoding only ivops */ - if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) - { - 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(); - } start_timer(); - idct(&data[i * 64]); - stop_idct_timer(); - } - } - - /* speedup: dont decode when encoding only ivops */ - if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) - { - - if (pMB->field_dct) { - next_block = stride; - stride *= 2; + dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); + stop_iquant_timer(); } start_timer(); - if ((frame->global_flags & XVID_REDUCED)) - { - copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); - copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); - copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); - copy_upsampled_8x8_16to8(pY_Cur + next_block + 16, &data[3 * 64], stride); - copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); - copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); - - }else{ - 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(); + idct(&data[i * 64]); + stop_idct_timer(); } + if (pMB->field_dct) { + next_block = stride; + stride *= 2; + } + + 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(); + } @@ -210,7 +153,7 @@ uint32_t stride = pParam->edged_width; uint32_t stride2 = stride / 2; - uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); + uint32_t next_block = stride * 8; uint32_t i; uint32_t iQuant = frame->quant; uint8_t *pY_Cur, *pU_Cur, *pV_Cur; @@ -218,16 +161,9 @@ uint32_t sum; IMAGE *pCurrent = &frame->image; - if ((frame->global_flags & XVID_REDUCED)) - { - 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); - }else{ - 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); - } + 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; @@ -239,8 +175,6 @@ stop_interlacing_timer(); for (i = 0; i < 6; i++) { - uint32_t increase_limit = (iQuant == 1) ? 1 : 0; - /* * no need to transfer 8->16-bit * (this is performed already in motion compensation) @@ -259,7 +193,7 @@ stop_quant_timer(); } - if ((sum >= TOOSMALL_LIMIT + increase_limit) || (qcoeff[i*64] != 0) || + 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) { @@ -286,34 +220,18 @@ } start_timer(); - if ((frame->global_flags & XVID_REDUCED)) - { - if (cbp & 32) - add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); - if (cbp & 16) - add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); - if (cbp & 8) - add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); - if (cbp & 4) - add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); - if (cbp & 2) - add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); - if (cbp & 1) - add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); - }else{ - 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); - } + 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(); return cbp; @@ -640,27 +558,13 @@ -/* permute block and return field dct choice */ +/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ uint32_t MBDecideFieldDCT(int16_t data[6 * 64]) { - uint32_t field = MBFieldTest(data); - if (field) { - MBFrameToField(data); - } - - return field; -} - - -/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ - -uint32_t -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 }; @@ -694,7 +598,11 @@ } } - return (frame >= (field + 350)); + if (frame > field) { + MBFrameToField(data); + } + + return (frame > field); }