1
|
1 /*
|
|
2 * idct.c
|
|
3 * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
|
|
4 *
|
|
5 * Portions of this code are from the MPEG software simulation group
|
|
6 * idct implementation. This code will be replaced with a new
|
|
7 * implementation soon.
|
|
8 *
|
|
9 * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
|
|
10 *
|
|
11 * mpeg2dec is free software; you can redistribute it and/or modify
|
|
12 * it under the terms of the GNU General Public License as published by
|
|
13 * the Free Software Foundation; either version 2 of the License, or
|
|
14 * (at your option) any later version.
|
|
15 *
|
|
16 * mpeg2dec is distributed in the hope that it will be useful,
|
|
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
19 * GNU General Public License for more details.
|
|
20 *
|
|
21 * You should have received a copy of the GNU General Public License
|
|
22 * along with this program; if not, write to the Free Software
|
|
23 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
24 */
|
|
25
|
|
26 /**********************************************************/
|
|
27 /* inverse two dimensional DCT, Chen-Wang algorithm */
|
|
28 /* (cf. IEEE ASSP-32, pp. 803-816, Aug. 1984) */
|
|
29 /* 32-bit integer arithmetic (8 bit coefficients) */
|
|
30 /* 11 mults, 29 adds per DCT */
|
|
31 /* sE, 18.8.91 */
|
|
32 /**********************************************************/
|
|
33 /* coefficients extended to 12 bit for IEEE1180-1990 */
|
|
34 /* compliance sE, 2.1.94 */
|
|
35 /**********************************************************/
|
|
36
|
|
37 /* this code assumes >> to be a two's-complement arithmetic */
|
|
38 /* right shift: (-2)>>1 == -1 , (-3)>>1 == -2 */
|
|
39
|
|
40 #include "config.h"
|
|
41
|
|
42 #include <stdio.h>
|
|
43 #include <inttypes.h>
|
|
44
|
|
45 #include "mpeg2_internal.h"
|
|
46 #include "mm_accel.h"
|
|
47
|
|
48 #define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */
|
|
49 #define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */
|
|
50 #define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */
|
|
51 #define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */
|
|
52 #define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */
|
|
53 #define W7 565 /* 2048*sqrt (2)*cos (7*pi/16) */
|
|
54
|
|
55
|
|
56 // idct main entry point
|
|
57 void (*idct_block_copy) (int16_t * block, uint8_t * dest, int stride);
|
|
58 void (*idct_block_add) (int16_t * block, uint8_t * dest, int stride);
|
|
59
|
|
60 static void idct_block_copy_c (int16_t *block, uint8_t * dest, int stride);
|
|
61 static void idct_block_add_c (int16_t *block, uint8_t * dest, int stride);
|
|
62
|
|
63 static uint8_t clip_lut[1024];
|
|
64 #define CLIP(i) ((clip_lut+384)[ (i)])
|
|
65
|
|
66 void idct_init (void)
|
|
67 {
|
|
68 #ifdef ARCH_X86
|
|
69 if (config.flags & MM_ACCEL_X86_MMXEXT) {
|
|
70 fprintf (stderr, "Using MMXEXT for IDCT transform\n");
|
|
71 idct_block_copy = idct_block_copy_mmxext;
|
|
72 idct_block_add = idct_block_add_mmxext;
|
|
73 idct_mmx_init ();
|
|
74 } else if (config.flags & MM_ACCEL_X86_MMX) {
|
|
75 fprintf (stderr, "Using MMX for IDCT transform\n");
|
|
76 idct_block_copy = idct_block_copy_mmx;
|
|
77 idct_block_add = idct_block_add_mmx;
|
|
78 idct_mmx_init ();
|
|
79 } else
|
|
80 #endif
|
|
81 #ifdef LIBMPEG2_MLIB
|
|
82 if (config.flags & MM_ACCEL_MLIB) {
|
|
83 fprintf (stderr, "Using mlib for IDCT transform\n");
|
|
84 idct_block_copy = idct_block_copy_mlib;
|
|
85 idct_block_add = idct_block_add_mlib;
|
|
86 } else
|
|
87 #endif
|
|
88 {
|
|
89 int i;
|
|
90
|
|
91 fprintf (stderr, "No accelerated IDCT transform found\n");
|
|
92 idct_block_copy = idct_block_copy_c;
|
|
93 idct_block_add = idct_block_add_c;
|
|
94 for (i = -384; i < 640; i++)
|
|
95 clip_lut[i+384] = (i < 0) ? 0 : ((i > 255) ? 255 : i);
|
|
96 }
|
|
97 }
|
|
98
|
|
99 /* row (horizontal) IDCT
|
|
100 *
|
|
101 * 7 pi 1
|
|
102 * dst[k] = sum c[l] * src[l] * cos ( -- * ( k + - ) * l )
|
|
103 * l=0 8 2
|
|
104 *
|
|
105 * where: c[0] = 128
|
|
106 * c[1..7] = 128*sqrt (2)
|
|
107 */
|
|
108
|
|
109 static void inline idct_row (int16_t * block)
|
|
110 {
|
|
111 int x0, x1, x2, x3, x4, x5, x6, x7, x8;
|
|
112
|
|
113 x1 = block[4] << 11;
|
|
114 x2 = block[6];
|
|
115 x3 = block[2];
|
|
116 x4 = block[1];
|
|
117 x5 = block[7];
|
|
118 x6 = block[5];
|
|
119 x7 = block[3];
|
|
120
|
|
121 /* shortcut */
|
|
122 if (! (x1 | x2 | x3 | x4 | x5 | x6 | x7 )) {
|
|
123 block[0] = block[1] = block[2] = block[3] = block[4] =
|
|
124 block[5] = block[6] = block[7] = block[0]<<3;
|
|
125 return;
|
|
126 }
|
|
127
|
|
128 x0 = (block[0] << 11) + 128; /* for proper rounding in the fourth stage */
|
|
129
|
|
130 /* first stage */
|
|
131 x8 = W7 * (x4 + x5);
|
|
132 x4 = x8 + (W1 - W7) * x4;
|
|
133 x5 = x8 - (W1 + W7) * x5;
|
|
134 x8 = W3 * (x6 + x7);
|
|
135 x6 = x8 - (W3 - W5) * x6;
|
|
136 x7 = x8 - (W3 + W5) * x7;
|
|
137
|
|
138 /* second stage */
|
|
139 x8 = x0 + x1;
|
|
140 x0 -= x1;
|
|
141 x1 = W6 * (x3 + x2);
|
|
142 x2 = x1 - (W2 + W6) * x2;
|
|
143 x3 = x1 + (W2 - W6) * x3;
|
|
144 x1 = x4 + x6;
|
|
145 x4 -= x6;
|
|
146 x6 = x5 + x7;
|
|
147 x5 -= x7;
|
|
148
|
|
149 /* third stage */
|
|
150 x7 = x8 + x3;
|
|
151 x8 -= x3;
|
|
152 x3 = x0 + x2;
|
|
153 x0 -= x2;
|
|
154 x2 = (181 * (x4 + x5) + 128) >> 8;
|
|
155 x4 = (181 * (x4 - x5) + 128) >> 8;
|
|
156
|
|
157 /* fourth stage */
|
|
158 block[0] = (x7 + x1) >> 8;
|
|
159 block[1] = (x3 + x2) >> 8;
|
|
160 block[2] = (x0 + x4) >> 8;
|
|
161 block[3] = (x8 + x6) >> 8;
|
|
162 block[4] = (x8 - x6) >> 8;
|
|
163 block[5] = (x0 - x4) >> 8;
|
|
164 block[6] = (x3 - x2) >> 8;
|
|
165 block[7] = (x7 - x1) >> 8;
|
|
166 }
|
|
167
|
|
168 /* column (vertical) IDCT
|
|
169 *
|
|
170 * 7 pi 1
|
|
171 * dst[8*k] = sum c[l] * src[8*l] * cos ( -- * ( k + - ) * l )
|
|
172 * l=0 8 2
|
|
173 *
|
|
174 * where: c[0] = 1/1024
|
|
175 * c[1..7] = (1/1024)*sqrt (2)
|
|
176 */
|
|
177
|
|
178 static void inline idct_col (int16_t *block)
|
|
179 {
|
|
180 int x0, x1, x2, x3, x4, x5, x6, x7, x8;
|
|
181
|
|
182 /* shortcut */
|
|
183 x1 = block [8*4] << 8;
|
|
184 x2 = block [8*6];
|
|
185 x3 = block [8*2];
|
|
186 x4 = block [8*1];
|
|
187 x5 = block [8*7];
|
|
188 x6 = block [8*5];
|
|
189 x7 = block [8*3];
|
|
190
|
|
191 #if 0
|
|
192 if (! (x1 | x2 | x3 | x4 | x5 | x6 | x7 )) {
|
|
193 block[8*0] = block[8*1] = block[8*2] = block[8*3] = block[8*4] =
|
|
194 block[8*5] = block[8*6] = block[8*7] = (block[8*0] + 32) >> 6;
|
|
195 return;
|
|
196 }
|
|
197 #endif
|
|
198
|
|
199 x0 = (block[8*0] << 8) + 8192;
|
|
200
|
|
201 /* first stage */
|
|
202 x8 = W7 * (x4 + x5) + 4;
|
|
203 x4 = (x8 + (W1 - W7) * x4) >> 3;
|
|
204 x5 = (x8 - (W1 + W7) * x5) >> 3;
|
|
205 x8 = W3 * (x6 + x7) + 4;
|
|
206 x6 = (x8 - (W3 - W5) * x6) >> 3;
|
|
207 x7 = (x8 - (W3 + W5) * x7) >> 3;
|
|
208
|
|
209 /* second stage */
|
|
210 x8 = x0 + x1;
|
|
211 x0 -= x1;
|
|
212 x1 = W6 * (x3 + x2) + 4;
|
|
213 x2 = (x1 - (W2 + W6) * x2) >> 3;
|
|
214 x3 = (x1 + (W2 - W6) * x3) >> 3;
|
|
215 x1 = x4 + x6;
|
|
216 x4 -= x6;
|
|
217 x6 = x5 + x7;
|
|
218 x5 -= x7;
|
|
219
|
|
220 /* third stage */
|
|
221 x7 = x8 + x3;
|
|
222 x8 -= x3;
|
|
223 x3 = x0 + x2;
|
|
224 x0 -= x2;
|
|
225 x2 = (181 * (x4 + x5) + 128) >> 8;
|
|
226 x4 = (181 * (x4 - x5) + 128) >> 8;
|
|
227
|
|
228 /* fourth stage */
|
|
229 block[8*0] = (x7 + x1) >> 14;
|
|
230 block[8*1] = (x3 + x2) >> 14;
|
|
231 block[8*2] = (x0 + x4) >> 14;
|
|
232 block[8*3] = (x8 + x6) >> 14;
|
|
233 block[8*4] = (x8 - x6) >> 14;
|
|
234 block[8*5] = (x0 - x4) >> 14;
|
|
235 block[8*6] = (x3 - x2) >> 14;
|
|
236 block[8*7] = (x7 - x1) >> 14;
|
|
237 }
|
|
238
|
|
239 void idct_block_copy_c (int16_t * block, uint8_t * dest, int stride)
|
|
240 {
|
|
241 int i;
|
|
242
|
|
243 for (i = 0; i < 8; i++)
|
|
244 idct_row (block + 8 * i);
|
|
245
|
|
246 for (i = 0; i < 8; i++)
|
|
247 idct_col (block + i);
|
|
248
|
|
249 i = 8;
|
|
250 do {
|
|
251 dest[0] = CLIP (block[0]);
|
|
252 dest[1] = CLIP (block[1]);
|
|
253 dest[2] = CLIP (block[2]);
|
|
254 dest[3] = CLIP (block[3]);
|
|
255 dest[4] = CLIP (block[4]);
|
|
256 dest[5] = CLIP (block[5]);
|
|
257 dest[6] = CLIP (block[6]);
|
|
258 dest[7] = CLIP (block[7]);
|
|
259
|
|
260 dest += stride;
|
|
261 block += 8;
|
|
262 } while (--i);
|
|
263 }
|
|
264
|
|
265 void idct_block_add_c (int16_t * block, uint8_t * dest, int stride)
|
|
266 {
|
|
267 int i;
|
|
268
|
|
269 for (i = 0; i < 8; i++)
|
|
270 idct_row (block + 8 * i);
|
|
271
|
|
272 for (i = 0; i < 8; i++)
|
|
273 idct_col (block + i);
|
|
274
|
|
275 i = 8;
|
|
276 do {
|
|
277 dest[0] = CLIP (block[0] + dest[0]);
|
|
278 dest[1] = CLIP (block[1] + dest[1]);
|
|
279 dest[2] = CLIP (block[2] + dest[2]);
|
|
280 dest[3] = CLIP (block[3] + dest[3]);
|
|
281 dest[4] = CLIP (block[4] + dest[4]);
|
|
282 dest[5] = CLIP (block[5] + dest[5]);
|
|
283 dest[6] = CLIP (block[6] + dest[6]);
|
|
284 dest[7] = CLIP (block[7] + dest[7]);
|
|
285
|
|
286 dest += stride;
|
|
287 block += 8;
|
|
288 } while (--i);
|
|
289 }
|