1 |
/***************************************************************************** |
/****************************************************************************** |
2 |
* |
* * |
3 |
* XVID MPEG-4 VIDEO CODEC |
* This file is part of XviD, a free MPEG-4 video encoder/decoder * |
4 |
* - MacroBlock transfer and quantization - |
* * |
5 |
* |
* XviD is an implementation of a part of one or more MPEG-4 Video tools * |
6 |
* Copyright(C) 2002-2001 Christoph Lampert <gruel@web.de> |
* as specified in ISO/IEC 14496-2 standard. Those intending to use this * |
7 |
* 2002-2001 Michael Militzer <isibaar@xvid.org> |
* software module in hardware or software products are advised that its * |
8 |
* 2002-2001 Peter Ross <pross@xvid.org> |
* use may infringe existing patents or copyrights, and any such use * |
9 |
* 2002 Daniel Smith <danielsmith@astroboymail.com> |
* would be at such party's own risk. The original developer of this * |
10 |
* |
* software module and his/her company, and subsequent editors and their * |
11 |
* This program is an implementation of a part of one or more MPEG-4 |
* companies, will have no liability for use of this software or * |
12 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* modifications or derivatives thereof. * |
13 |
* to use this software module in hardware or software products are |
* * |
14 |
* advised that its use may infringe existing patents or copyrights, and |
* XviD is free software; you can redistribute it and/or modify it * |
15 |
* any such use would be at such party's own risk. The original |
* under the terms of the GNU General Public License as published by * |
16 |
* developer of this software module and his/her company, and subsequent |
* the Free Software Foundation; either version 2 of the License, or * |
17 |
* editors and their companies, will have no liability for use of this |
* (at your option) any later version. * |
18 |
* software or modifications or derivatives thereof. |
* * |
19 |
* |
* XviD is distributed in the hope that it will be useful, but * |
20 |
* This program is free software; you can redistribute it and/or modify |
* WITHOUT ANY WARRANTY; without even the implied warranty of * |
21 |
* it under the terms of the GNU General Public License as published by |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * |
22 |
* the Free Software Foundation; either version 2 of the License, or |
* GNU General Public License for more details. * |
23 |
* (at your option) any later version. |
* * |
24 |
* |
* You should have received a copy of the GNU General Public License * |
25 |
* This program is distributed in the hope that it will be useful, |
* along with this program; if not, write to the Free Software * |
26 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * |
27 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
* * |
28 |
* GNU General Public License for more details. |
******************************************************************************/ |
29 |
* |
|
30 |
* You should have received a copy of the GNU General Public License |
/****************************************************************************** |
31 |
* along with this program; if not, write to the Free Software |
* * |
32 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* mbtransquant.c * |
33 |
* |
* * |
34 |
* $Id: mbtransquant.c,v 1.18 2002-10-19 12:20:33 edgomez Exp $ |
* Copyright (C) 2001 - Peter Ross <pross@cs.rmit.edu.au> * |
35 |
* |
* Copyright (C) 2001 - Michael Militzer <isibaar@xvid.org> * |
36 |
****************************************************************************/ |
* * |
37 |
|
* For more information visit the XviD homepage: http://www.xvid.org * |
38 |
|
* * |
39 |
|
******************************************************************************/ |
40 |
|
|
41 |
|
/****************************************************************************** |
42 |
|
* * |
43 |
|
* Revision history: * |
44 |
|
* * |
45 |
|
* 29.03.2002 interlacing speedup - used transfer strides instead of * |
46 |
|
* manual field-to-frame conversion * |
47 |
|
* 26.03.2002 interlacing support - moved transfers outside loops * |
48 |
|
* 22.12.2001 get_dc_scaler() moved to common.h * |
49 |
|
* 19.11.2001 introduced coefficient thresholding (Isibaar) * |
50 |
|
* 17.11.2001 initial version * |
51 |
|
* * |
52 |
|
******************************************************************************/ |
53 |
|
|
54 |
#include <string.h> |
#include <string.h> |
55 |
|
|
65 |
#include "../quant/quant_h263.h" |
#include "../quant/quant_h263.h" |
66 |
#include "../encoder.h" |
#include "../encoder.h" |
67 |
|
|
68 |
#define MIN(X, Y) ((X)<(Y)?(X):(Y)) |
#include "../image/reduced.h" |
|
#define MAX(X, Y) ((X)>(Y)?(X):(Y)) |
|
69 |
|
|
70 |
#define TOOSMALL_LIMIT 3 /* skip blocks having a coefficient sum below this value */ |
MBFIELDTEST_PTR MBFieldTest; |
71 |
|
|
72 |
/* this isnt pretty, but its better than 20 ifdefs */ |
#define TOOSMALL_LIMIT 1 /* skip blocks having a coefficient sum below this value */ |
73 |
|
|
74 |
|
static __inline void |
75 |
|
MBfDCT(int16_t data[6 * 64]) |
76 |
|
{ |
77 |
|
start_timer(); |
78 |
|
fdct(&data[0 * 64]); |
79 |
|
fdct(&data[1 * 64]); |
80 |
|
fdct(&data[2 * 64]); |
81 |
|
fdct(&data[3 * 64]); |
82 |
|
fdct(&data[4 * 64]); |
83 |
|
fdct(&data[5 * 64]); |
84 |
|
stop_dct_timer(); |
85 |
|
} |
86 |
|
|
87 |
|
|
88 |
|
static __inline uint32_t |
89 |
|
QuantizeInterBlock( int16_t qcoeff[64], |
90 |
|
const int16_t data[64], |
91 |
|
const uint32_t iQuant, |
92 |
|
const uint32_t quant_type) |
93 |
|
{ |
94 |
|
uint32_t sum; |
95 |
|
|
96 |
|
start_timer(); |
97 |
|
if (quant_type == H263_QUANT) |
98 |
|
sum = quant_inter(qcoeff, data, iQuant); |
99 |
|
else |
100 |
|
sum = quant4_inter(qcoeff, data, iQuant); |
101 |
|
|
102 |
|
stop_quant_timer(); |
103 |
|
return sum; |
104 |
|
} |
105 |
|
|
106 |
void |
void |
107 |
MBTransQuantIntra(const MBParam * pParam, |
MBTransQuantIntra(const MBParam * const pParam, |
108 |
FRAMEINFO * frame, |
FRAMEINFO * const frame, |
109 |
MACROBLOCK * pMB, |
MACROBLOCK * const pMB, |
110 |
const uint32_t x_pos, |
const uint32_t x_pos, |
111 |
const uint32_t y_pos, |
const uint32_t y_pos, |
112 |
int16_t data[6 * 64], |
int16_t data[6 * 64], |
114 |
{ |
{ |
115 |
|
|
116 |
uint32_t stride = pParam->edged_width; |
uint32_t stride = pParam->edged_width; |
117 |
uint32_t stride2 = stride / 2; |
const uint32_t stride2 = stride / 2; |
118 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
119 |
uint32_t i; |
int i; |
120 |
uint32_t iQuant = frame->quant; |
const uint32_t iQuant = pMB->quant; |
121 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
122 |
IMAGE *pCurrent = &frame->image; |
const IMAGE * const pCurrent = &frame->image; |
123 |
|
|
124 |
|
start_timer(); |
125 |
|
if ((frame->global_flags & XVID_REDUCED)) |
126 |
|
{ |
127 |
|
pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); |
128 |
|
pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); |
129 |
|
pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); |
130 |
|
|
131 |
|
filter_18x18_to_8x8(&data[0 * 64], pY_Cur, stride); |
132 |
|
filter_18x18_to_8x8(&data[1 * 64], pY_Cur + 16, stride); |
133 |
|
filter_18x18_to_8x8(&data[2 * 64], pY_Cur + next_block, stride); |
134 |
|
filter_18x18_to_8x8(&data[3 * 64], pY_Cur + next_block + 16, stride); |
135 |
|
filter_18x18_to_8x8(&data[4 * 64], pU_Cur, stride2); |
136 |
|
filter_18x18_to_8x8(&data[5 * 64], pV_Cur, stride2); |
137 |
|
} else { |
138 |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
139 |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
140 |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
141 |
|
|
|
start_timer(); |
|
142 |
transfer_8to16copy(&data[0 * 64], pY_Cur, stride); |
transfer_8to16copy(&data[0 * 64], pY_Cur, stride); |
143 |
transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); |
transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); |
144 |
transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); |
transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); |
145 |
transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); |
transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); |
146 |
transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); |
transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); |
147 |
transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); |
transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); |
148 |
|
} |
149 |
stop_transfer_timer(); |
stop_transfer_timer(); |
150 |
|
|
151 |
|
/* XXX: rrv+interlacing is buggy */ |
152 |
start_timer(); |
start_timer(); |
153 |
pMB->field_dct = 0; |
pMB->field_dct = 0; |
154 |
if ((frame->global_flags & XVID_INTERLACING) && |
if ((frame->global_flags & XVID_INTERLACING) && |
158 |
} |
} |
159 |
stop_interlacing_timer(); |
stop_interlacing_timer(); |
160 |
|
|
161 |
for (i = 0; i < 6; i++) { |
MBfDCT(data); |
|
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
162 |
|
|
163 |
start_timer(); |
for (i = 0; i < 6; i++) { |
164 |
fdct(&data[i * 64]); |
const uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
stop_dct_timer(); |
|
165 |
|
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
166 |
start_timer(); |
start_timer(); |
167 |
|
if (pParam->m_quant_type == H263_QUANT) |
168 |
quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
169 |
stop_quant_timer(); |
else |
|
|
|
|
start_timer(); |
|
|
dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
|
|
stop_iquant_timer(); |
|
|
} else { |
|
|
start_timer(); |
|
170 |
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
171 |
stop_quant_timer(); |
stop_quant_timer(); |
172 |
|
|
173 |
|
/* speedup: dont decode when encoding only ivops */ |
174 |
|
if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) |
175 |
|
{ |
176 |
start_timer(); |
start_timer(); |
177 |
|
if (pParam->m_quant_type == H263_QUANT) |
178 |
|
dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
179 |
|
else |
180 |
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
181 |
stop_iquant_timer(); |
stop_iquant_timer(); |
|
} |
|
182 |
|
|
183 |
start_timer(); |
start_timer(); |
184 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
185 |
stop_idct_timer(); |
stop_idct_timer(); |
186 |
} |
} |
187 |
|
} |
188 |
|
|
189 |
|
/* speedup: dont decode when encoding only ivops */ |
190 |
|
if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) |
191 |
|
{ |
192 |
|
|
193 |
if (pMB->field_dct) { |
if (pMB->field_dct) { |
194 |
next_block = stride; |
next_block = stride; |
196 |
} |
} |
197 |
|
|
198 |
start_timer(); |
start_timer(); |
199 |
|
if ((frame->global_flags & XVID_REDUCED)) { |
200 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
201 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
202 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
203 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block + 16, &data[3 * 64], stride); |
204 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
205 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
206 |
|
} else { |
207 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
208 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
209 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
210 |
transfer_16to8copy(pY_Cur + next_block + 8, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block + 8, &data[3 * 64], stride); |
211 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
212 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
213 |
|
} |
214 |
stop_transfer_timer(); |
stop_transfer_timer(); |
|
|
|
215 |
} |
} |
216 |
|
|
217 |
|
} |
218 |
|
|
219 |
uint8_t |
uint8_t |
220 |
MBTransQuantInter(const MBParam * pParam, |
MBTransQuantInter(const MBParam * const pParam, |
221 |
FRAMEINFO * frame, |
FRAMEINFO * const frame, |
222 |
MACROBLOCK * pMB, |
MACROBLOCK * const pMB, |
223 |
const uint32_t x_pos, |
const uint32_t x_pos, |
224 |
const uint32_t y_pos, |
const uint32_t y_pos, |
225 |
int16_t data[6 * 64], |
int16_t data[6 * 64], |
226 |
int16_t qcoeff[6 * 64]) |
int16_t qcoeff[6 * 64]) |
227 |
{ |
{ |
|
|
|
228 |
uint32_t stride = pParam->edged_width; |
uint32_t stride = pParam->edged_width; |
229 |
uint32_t stride2 = stride / 2; |
const uint32_t stride2 = stride / 2; |
230 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
231 |
uint32_t i; |
int i; |
232 |
uint32_t iQuant = frame->quant; |
const uint32_t iQuant = pMB->quant; |
233 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
234 |
uint8_t cbp = 0; |
int cbp = 0; |
235 |
uint32_t sum; |
uint32_t sum; |
236 |
IMAGE *pCurrent = &frame->image; |
const IMAGE * const pCurrent = &frame->image; |
237 |
|
|
238 |
|
if ((frame->global_flags & XVID_REDUCED)) { |
239 |
|
pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); |
240 |
|
pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); |
241 |
|
pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); |
242 |
|
} else { |
243 |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
244 |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
245 |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
246 |
|
} |
247 |
|
|
248 |
start_timer(); |
start_timer(); |
249 |
pMB->field_dct = 0; |
pMB->field_dct = 0; |
254 |
} |
} |
255 |
stop_interlacing_timer(); |
stop_interlacing_timer(); |
256 |
|
|
257 |
|
MBfDCT(data); |
258 |
|
|
259 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
260 |
|
const uint32_t limit = TOOSMALL_LIMIT + ((iQuant == 1) ? 1 : 0); |
261 |
/* |
/* |
262 |
* no need to transfer 8->16-bit |
* no need to transfer 8->16-bit |
263 |
* (this is performed already in motion compensation) |
* (this is performed already in motion compensation) |
264 |
*/ |
*/ |
|
start_timer(); |
|
|
fdct(&data[i * 64]); |
|
|
stop_dct_timer(); |
|
265 |
|
|
266 |
if (pParam->m_quant_type == 0) { |
sum = QuantizeInterBlock(&qcoeff[i * 64], &data[i * 64], iQuant, pParam->m_quant_type); |
267 |
start_timer(); |
|
268 |
sum = quant_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
if(frame->global_flags & XVID_CARTOON_MODE) { |
269 |
stop_quant_timer(); |
limit *= 3; |
|
} else { |
|
|
start_timer(); |
|
|
sum = quant4_inter(&qcoeff[i * 64], &data[i * 64], iQuant); |
|
|
stop_quant_timer(); |
|
270 |
} |
} |
271 |
|
|
272 |
if ((sum >= TOOSMALL_LIMIT) || (qcoeff[i*64] != 0) || |
if (sum >= limit) { |
|
(qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) { |
|
273 |
|
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
274 |
start_timer(); |
start_timer(); |
275 |
|
if (pParam->m_quant_type == H263_QUANT) |
276 |
dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
dequant_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
277 |
stop_iquant_timer(); |
else |
|
} else { |
|
|
start_timer(); |
|
278 |
dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
dequant4_inter(&data[i * 64], &qcoeff[i * 64], iQuant); |
279 |
stop_iquant_timer(); |
stop_iquant_timer(); |
|
} |
|
280 |
|
|
281 |
cbp |= 1 << (5 - i); |
cbp |= 1 << (5 - i); |
282 |
|
|
292 |
} |
} |
293 |
|
|
294 |
start_timer(); |
start_timer(); |
295 |
|
if ((frame->global_flags & XVID_REDUCED)) { |
296 |
|
if (cbp & 32) |
297 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
298 |
|
if (cbp & 16) |
299 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
300 |
|
if (cbp & 8) |
301 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
302 |
|
if (cbp & 4) |
303 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
304 |
|
if (cbp & 2) |
305 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
306 |
|
if (cbp & 1) |
307 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
308 |
|
} else { |
309 |
if (cbp & 32) |
if (cbp & 32) |
310 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
311 |
if (cbp & 16) |
if (cbp & 16) |
318 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
319 |
if (cbp & 1) |
if (cbp & 1) |
320 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
stop_transfer_timer(); |
|
|
|
|
|
return cbp; |
|
|
|
|
321 |
} |
} |
322 |
|
stop_transfer_timer(); |
323 |
|
|
324 |
void |
return (uint8_t) cbp; |
|
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]) |
|
|
{ |
|
|
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); |
|
|
} |
|
|
|
|
|
|
|
|
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 */ |
|
|
|
|
|
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; |
|
325 |
} |
} |
326 |
|
|
327 |
uint8_t |
uint8_t |
331 |
int16_t data[6 * 64], |
int16_t data[6 * 64], |
332 |
int16_t qcoeff[6 * 64]) |
int16_t qcoeff[6 * 64]) |
333 |
{ |
{ |
334 |
uint8_t cbp; |
int cbp = 0; |
|
|
|
|
/* there is no MBTrans for Inter block, that's done in motion compensation already */ |
|
|
|
|
|
MBfDCT(pParam,frame,pMB,data); |
|
|
cbp = MBQuantInter(pParam,frame->quant,data,qcoeff); |
|
|
|
|
|
/* we don't have to DeQuant, iDCT and Transfer back data for B-frames */ |
|
|
|
|
|
return cbp; |
|
|
} |
|
|
|
|
|
|
|
|
void |
|
|
MBfDCT(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
int16_t data[6 * 64]) |
|
|
{ |
|
|
int i; |
|
|
|
|
|
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++) { |
|
|
start_timer(); |
|
|
fdct(&data[i * 64]); |
|
|
stop_dct_timer(); |
|
|
} |
|
|
} |
|
|
|
|
|
void |
|
|
MBQuantDeQuantIntra(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
int16_t qcoeff[6 * 64], |
|
|
int16_t data[6*64]) |
|
|
{ |
|
335 |
int i; |
int i; |
|
int iQuant = frame->quant; |
|
|
|
|
|
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(); |
|
336 |
|
|
337 |
start_timer(); |
/* there is no MBTrans for Inter block, that's done in motion compensation already */ |
|
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
|
|
stop_iquant_timer(); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
void |
|
|
MBQuantIntra(const MBParam * pParam, |
|
|
FRAMEINFO * frame, |
|
|
MACROBLOCK *pMB, |
|
|
int16_t data[6 * 64], |
|
|
int16_t qcoeff[6 * 64]) |
|
|
{ |
|
|
int i; |
|
|
int iQuant = frame->quant; |
|
338 |
|
|
339 |
start_timer(); |
start_timer(); |
340 |
pMB->field_dct = 0; |
pMB->field_dct = 0; |
343 |
} |
} |
344 |
stop_interlacing_timer(); |
stop_interlacing_timer(); |
345 |
|
|
346 |
for (i = 0; i < 6; i++) { |
MBfDCT(data); |
|
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(); |
|
|
} else { |
|
|
start_timer(); |
|
|
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
|
|
stop_quant_timer(); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
void |
|
|
MBDeQuantIntra(const MBParam * pParam, |
|
|
const int iQuant, |
|
|
int16_t qcoeff[6 * 64], |
|
|
int16_t data[6*64]) |
|
|
{ |
|
|
int i; |
|
347 |
|
|
348 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
349 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
int codedecision = 0; |
350 |
|
|
351 |
if (pParam->m_quant_type == H263_QUANT) { |
int sum = QuantizeInterBlock(&qcoeff[i * 64], &data[i * 64], pMB->quant, pParam->m_quant_type); |
|
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(); |
|
|
} |
|
|
} |
|
|
} |
|
|
|
|
|
uint8_t |
|
|
MBQuantInter(const MBParam * pParam, |
|
|
const int iQuant, |
|
|
int16_t data[6 * 64], |
|
|
int16_t qcoeff[6 * 64]) |
|
|
{ |
|
|
|
|
|
int i; |
|
|
uint8_t cbp = 0; |
|
|
int sum; |
|
352 |
|
|
353 |
for (i = 0; i < 6; i++) { |
if(frame->global_flags & XVID_CARTOON_MODE) { |
354 |
|
limit *= 2; |
|
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(); |
|
355 |
} |
} |
356 |
|
|
357 |
if (sum >= TOOSMALL_LIMIT) { // skip block ? |
if ((sum > 2) || (qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0) ) codedecision = 1; |
358 |
cbp |= 1 << (5 - i); |
else { |
359 |
} |
if (pMB->mode == MODE_DIRECT || pMB->mode == MODE_DIRECT_NO4V) { |
360 |
} |
// dark blocks prevention for direct mode |
361 |
return cbp; |
if ( (qcoeff[i*64] < -1) || (qcoeff[i*64] > 0) ) codedecision = 1; |
362 |
|
} else |
363 |
|
if (qcoeff[i*64] != 0) codedecision = 1; // not direct mode |
364 |
} |
} |
365 |
|
|
366 |
void |
if (codedecision) cbp |= 1 << (5 - i); |
|
MBDeQuantInter( const MBParam * pParam, |
|
|
const int iQuant, |
|
|
int16_t data[6 * 64], |
|
|
int16_t qcoeff[6 * 64], |
|
|
const uint8_t cbp) |
|
|
{ |
|
|
int i; |
|
|
|
|
|
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(); |
|
|
} |
|
|
} |
|
|
} |
|
367 |
} |
} |
368 |
|
|
369 |
void |
/* we don't have to DeQuant, iDCT and Transfer back data for B-frames if we don't reconstruct this frame */ |
370 |
MBiDCT( int16_t data[6 * 64], |
/* warning: reconstruction not supported yet */ |
371 |
const uint8_t cbp) |
return (uint8_t) cbp; |
|
{ |
|
|
int i; |
|
|
|
|
|
for (i = 0; i < 6; i++) { |
|
|
if (cbp & (1 << (5 - i))) |
|
|
{ |
|
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
|
|
|
} |
|
|
} |
|
372 |
} |
} |
373 |
|
|
374 |
|
/* permute block and return field dct choice */ |
375 |
|
|
376 |
void |
static uint32_t |
377 |
MBTrans(const MBParam * pParam, |
MBDecideFieldDCT(int16_t data[6 * 64]) |
|
FRAMEINFO * frame, |
|
|
MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
int16_t data[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; |
|
|
|
|
|
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(); |
|
|
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(); |
|
|
} |
|
|
|
|
|
void |
|
|
MBTransAdd(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) |
|
378 |
{ |
{ |
379 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
const uint32_t field = MBFieldTest(data); |
380 |
uint32_t stride = pParam->edged_width; |
if (field) MBFrameToField(data); |
|
uint32_t stride2 = stride / 2; |
|
|
uint32_t next_block = stride * 8; |
|
|
IMAGE *pCurrent = &frame->image; |
|
381 |
|
|
382 |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
return field; |
|
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
|
|
|
if (pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
383 |
} |
} |
384 |
|
|
|
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(); |
|
|
} |
|
|
|
|
|
|
|
|
|
|
385 |
/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ |
/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ |
386 |
|
|
|
|
|
387 |
uint32_t |
uint32_t |
388 |
MBDecideFieldDCT(int16_t data[6 * 64]) |
MBFieldTest_c(int16_t data[6 * 64]) |
389 |
{ |
{ |
|
|
|
390 |
const uint8_t blocks[] = |
const uint8_t blocks[] = |
391 |
{ 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 }; |
{ 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 }; |
392 |
const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 }; |
const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 }; |
420 |
} |
} |
421 |
} |
} |
422 |
|
|
423 |
if (frame > (field + 350)) { |
return (frame >= (field + 350)); |
|
MBFrameToField(data); |
|
|
} |
|
|
|
|
|
return (frame > (field + 350)); |
|
424 |
} |
} |
425 |
|
|
426 |
|
|