1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* 8bit<->16bit transfer |
* - 8bit<->16bit transfer - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
|
* 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. |
|
7 |
* |
* |
8 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
9 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
17 |
* |
* |
18 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
|
* |
|
|
*************************************************************************/ |
|
|
|
|
|
/************************************************************************** |
|
21 |
* |
* |
22 |
* History: |
* $Id$ |
23 |
* |
* |
24 |
* 14.04.2002 added transfer_8to16sub2 |
****************************************************************************/ |
25 |
* 07.01.2002 merge functions from compensate; rename functions |
|
26 |
* 22.12.2001 transfer_8to8add16 limit fix |
#include "../global.h" |
27 |
* 07.11.2001 initial version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
#include "mem_transfer.h" |
28 |
|
|
29 |
*************************************************************************/ |
/* Function pointers - Initialized in the xvid.c module */ |
30 |
|
|
31 |
|
TRANSFER_8TO16COPY_PTR transfer_8to16copy; |
32 |
|
TRANSFER_16TO8COPY_PTR transfer_16to8copy; |
33 |
|
|
34 |
TRANSFER_8TO16SUB_PTR transfer_8to16sub; |
TRANSFER_8TO16SUB_PTR transfer_8to16sub; |
35 |
|
TRANSFER_8TO16SUBRO_PTR transfer_8to16subro; |
36 |
TRANSFER_8TO16SUB2_PTR transfer_8to16sub2; |
TRANSFER_8TO16SUB2_PTR transfer_8to16sub2; |
37 |
|
TRANSFER_8TO16SUB2RO_PTR transfer_8to16sub2ro; |
38 |
TRANSFER_16TO8ADD_PTR transfer_16to8add; |
TRANSFER_16TO8ADD_PTR transfer_16to8add; |
39 |
|
|
40 |
// function pointers |
TRANSFER8X8_COPY_PTR transfer8x8_copy; |
41 |
|
|
42 |
|
#define USE_REFERENCE_C |
43 |
|
|
44 |
/***************************************************************************** |
/***************************************************************************** |
45 |
* |
* |
46 |
TRANSFER_8TO16SUB_PTR transfer_8to16sub; |
* All these functions are used to transfer data from a 8 bit data array |
47 |
* to a 16 bit data array. |
* to a 16 bit data array. |
48 |
TRANSFER_16TO8ADD_PTR transfer_16to8add; |
* |
49 |
* This is typically used during motion compensation, that's why some |
* This is typically used during motion compensation, that's why some |
50 |
* functions also do the addition/substraction of another buffer during the |
* functions also do the addition/substraction of another buffer during the |
51 |
* so called transfer. |
* so called transfer. |
52 |
* |
* |
53 |
|
****************************************************************************/ |
54 |
|
|
55 |
|
/* |
56 |
|
* SRC - the source buffer |
57 |
|
* DST - the destination buffer |
58 |
|
* |
59 |
|
* Then the function does the 8->16 bit transfer and this serie of operations : |
60 |
|
* |
61 |
|
* SRC (8bit) = SRC |
62 |
|
* DST (16bit) = SRC |
63 |
*/ |
*/ |
64 |
|
void |
65 |
|
transfer_8to16copy_c(int16_t * const dst, |
66 |
|
const uint8_t * const src, |
67 |
|
uint32_t stride) |
68 |
|
{ |
69 |
|
int i, j; |
70 |
|
for (j = 0; j < 8; j++) { |
71 |
|
for (i = 0; i < 8; i++) { |
72 |
dst[j * 8 + i] = (int16_t) src[j * stride + i]; |
dst[j * 8 + i] = (int16_t) src[j * stride + i]; |
73 |
} |
} |
74 |
} |
} |
83 |
* SRC (16bit) = SRC |
* SRC (16bit) = SRC |
84 |
* DST (8bit) = max(min(SRC, 255), 0) |
* DST (8bit) = max(min(SRC, 255), 0) |
85 |
*/ |
*/ |
86 |
|
void |
87 |
|
transfer_16to8copy_c(uint8_t * const dst, |
88 |
|
const int16_t * const src, |
89 |
|
uint32_t stride) |
90 |
|
{ |
91 |
|
int i, j; |
92 |
|
|
93 |
|
for (j = 0; j < 8; j++) { |
94 |
|
for (i = 0; i < 8; i++) { |
95 |
|
#ifdef USE_REFERENCE_C |
96 |
int16_t pixel = src[j * 8 + i]; |
int16_t pixel = src[j * 8 + i]; |
97 |
|
|
98 |
if (pixel < 0) { |
if (pixel < 0) { |
101 |
pixel = 255; |
pixel = 255; |
102 |
} |
} |
103 |
dst[j * stride + i] = (uint8_t) pixel; |
dst[j * stride + i] = (uint8_t) pixel; |
104 |
|
#else |
105 |
|
const int16_t pixel = src[j * 8 + i]; |
106 |
|
const uint8_t value = (uint8_t)( (pixel&~255) ? (-pixel)>>(8*sizeof(pixel)-1) : pixel ); |
107 |
|
dst[j*stride + i] = value; |
108 |
|
#endif |
109 |
} |
} |
110 |
} |
} |
111 |
} |
} |
123 |
* R (8bit) = R |
* R (8bit) = R |
124 |
* C (8bit) = R |
* C (8bit) = R |
125 |
* DCT (16bit) = C - R |
* DCT (16bit) = C - R |
|
perform motion compensation (and 8bit->16bit dct transfer) |
|
|
... with write back |
|
126 |
*/ |
*/ |
127 |
|
void |
128 |
|
transfer_8to16sub_c(int16_t * const dct, |
129 |
|
uint8_t * const cur, |
130 |
|
const uint8_t * ref, |
131 |
|
const uint32_t stride) |
132 |
|
{ |
133 |
|
int i, j; |
134 |
|
|
135 |
|
for (j = 0; j < 8; j++) { |
136 |
for (i = 0; i < 8; i++) { |
for (i = 0; i < 8; i++) { |
137 |
uint8_t c = cur[j * stride + i]; |
const uint8_t c = cur[j * stride + i]; |
138 |
uint8_t r = ref[j * stride + i]; |
const uint8_t r = ref[j * stride + i]; |
139 |
|
|
140 |
cur[j * stride + i] = r; |
cur[j * stride + i] = r; |
141 |
dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
144 |
} |
} |
145 |
|
|
146 |
|
|
147 |
|
void |
148 |
|
transfer_8to16subro_c(int16_t * const dct, |
149 |
|
const uint8_t * const cur, |
150 |
|
const uint8_t * ref, |
151 |
|
const uint32_t stride) |
152 |
|
{ |
153 |
|
int i, j; |
154 |
|
|
155 |
|
for (j = 0; j < 8; j++) { |
156 |
|
for (i = 0; i < 8; i++) { |
157 |
|
const uint8_t c = cur[j * stride + i]; |
158 |
|
const uint8_t r = ref[j * stride + i]; |
159 |
|
dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
160 |
|
} |
161 |
|
} |
162 |
|
} |
163 |
|
|
164 |
|
|
165 |
|
|
166 |
/* |
/* |
167 |
* C - the current buffer |
* C - the current buffer |
168 |
* R1 - the 1st reference buffer |
* R1 - the 1st reference buffer |
173 |
* |
* |
174 |
* R1 (8bit) = R1 |
* R1 (8bit) = R1 |
175 |
* R2 (8bit) = R2 |
* R2 (8bit) = R2 |
176 |
performs bi motion compensation (and 8bit->16bit dct transfer) |
* R (temp) = min((R1 + R2)/2, 255) |
177 |
.. but does not write back |
* DCT (16bit)= C - R |
178 |
|
* C (8bit) = R |
179 |
*/ |
*/ |
180 |
|
void |
181 |
|
transfer_8to16sub2_c(int16_t * const dct, |
182 |
|
uint8_t * const cur, |
183 |
|
const uint8_t * ref1, |
184 |
|
const uint8_t * ref2, |
185 |
|
const uint32_t stride) |
186 |
|
{ |
187 |
|
uint32_t i, j; |
188 |
|
|
189 |
for (j = 0; j < 8; j++) { |
for (j = 0; j < 8; j++) { |
190 |
for (i = 0; i < 8; i++) { |
for (i = 0; i < 8; i++) { |
191 |
uint8_t c = cur[j * stride + i]; |
const uint8_t c = cur[j * stride + i]; |
192 |
int r = (ref1[j * stride + i] + ref2[j * stride + i] + 1) / 2; |
const uint8_t r = (ref1[j * stride + i] + ref2[j * stride + i] + 1) >> 1; |
193 |
|
cur[j * stride + i] = r; |
194 |
if (r > 255) { |
dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
|
r = 255; |
|
195 |
} |
} |
196 |
//cur[j * stride + i] = r; |
} |
197 |
|
} |
198 |
|
|
199 |
|
void |
200 |
|
transfer_8to16sub2ro_c(int16_t * const dct, |
201 |
|
const uint8_t * const cur, |
202 |
|
const uint8_t * ref1, |
203 |
|
const uint8_t * ref2, |
204 |
|
const uint32_t stride) |
205 |
|
{ |
206 |
|
uint32_t i, j; |
207 |
|
|
208 |
|
for (j = 0; j < 8; j++) { |
209 |
|
for (i = 0; i < 8; i++) { |
210 |
|
const uint8_t c = cur[j * stride + i]; |
211 |
|
const uint8_t r = (ref1[j * stride + i] + ref2[j * stride + i] + 1) >> 1; |
212 |
dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
213 |
} |
} |
214 |
} |
} |
224 |
* SRC (16bit) = SRC |
* SRC (16bit) = SRC |
225 |
* DST (8bit) = max(min(DST+SRC, 255), 0) |
* DST (8bit) = max(min(DST+SRC, 255), 0) |
226 |
*/ |
*/ |
227 |
|
void |
228 |
|
transfer_16to8add_c(uint8_t * const dst, |
229 |
|
const int16_t * const src, |
230 |
|
uint32_t stride) |
231 |
|
{ |
232 |
|
int i, j; |
233 |
|
|
234 |
|
for (j = 0; j < 8; j++) { |
235 |
|
for (i = 0; i < 8; i++) { |
236 |
|
#ifdef USE_REFERENCE_C |
237 |
int16_t pixel = (int16_t) dst[j * stride + i] + src[j * 8 + i]; |
int16_t pixel = (int16_t) dst[j * stride + i] + src[j * 8 + i]; |
238 |
|
|
239 |
if (pixel < 0) { |
if (pixel < 0) { |
242 |
pixel = 255; |
pixel = 255; |
243 |
} |
} |
244 |
dst[j * stride + i] = (uint8_t) pixel; |
dst[j * stride + i] = (uint8_t) pixel; |
245 |
|
#else |
246 |
|
const int16_t pixel = (int16_t) dst[j * stride + i] + src[j * 8 + i]; |
247 |
|
const uint8_t value = (uint8_t)( (pixel&~255) ? (-pixel)>>(8*sizeof(pixel)-1) : pixel ); |
248 |
|
dst[j*stride + i] = value; |
249 |
|
#endif |
250 |
|
|
251 |
} |
} |
252 |
} |
} |
253 |
} |
} |
261 |
* SRC (8bit) = SRC |
* SRC (8bit) = SRC |
262 |
* DST (8bit) = SRC |
* DST (8bit) = SRC |
263 |
*/ |
*/ |
264 |
|
void |
265 |
|
transfer8x8_copy_c(uint8_t * const dst, |
266 |
|
const uint8_t * const src, |
267 |
dst[j * stride + i] = src[j * stride + i]; |
const uint32_t stride) |
268 |
|
{ |
269 |
|
int j, i; |
270 |
|
|
271 |
|
for (j = 0; j < 8; ++j) { |
272 |
|
uint8_t *d = dst + j*stride; |
273 |
|
const uint8_t *s = src + j*stride; |
274 |
|
|
275 |
|
for (i = 0; i < 8; ++i) |
276 |
|
{ |
277 |
|
*d++ = *s++; |
278 |
} |
} |
279 |
} |
} |
280 |
} |
} |