FFmpeg
Loading...
Searching...
No Matches
xvididct.c
Go to the documentation of this file.
1/*
2 * Xvid MPEG-4 IDCT
3 *
4 * Copyright (C) 2006-2011 Xvid Solutions GmbH
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23/**
24 * @file
25 * Walken IDCT
26 * Alternative IDCT implementation for decoding compatibility.
27 *
28 * @author Skal
29 * @note This C version is not the original IDCT, but a modified one that
30 * yields the same error profile as the MMX/MMXEXT/SSE2 versions.
31 */
32
33#include "config.h"
34#include "libavutil/attributes.h"
35#include "idctdsp.h"
36#include "xvididct.h"
37
38#define ROW_SHIFT 11
39#define COL_SHIFT 6
40
41// #define FIX(x) (int)((x) * (1 << ROW_SHIFT))
42#define RND0 65536 // 1 << (COL_SHIFT + ROW_SHIFT - 1);
43#define RND1 3597 // FIX (1.75683487303);
44#define RND2 2260 // FIX (1.10355339059);
45#define RND3 1203 // FIX (0.587788325588);
46#define RND4 0
47#define RND5 120 // FIX (0.058658283817);
48#define RND6 512 // FIX (0.25);
49#define RND7 512 // FIX (0.25);
50
51static const int TAB04[] = { 22725, 21407, 19266, 16384, 12873, 8867, 4520 };
52static const int TAB17[] = { 31521, 29692, 26722, 22725, 17855, 12299, 6270 };
53static const int TAB26[] = { 29692, 27969, 25172, 21407, 16819, 11585, 5906 };
54static const int TAB35[] = { 26722, 25172, 22654, 19266, 15137, 10426, 5315 };
55
56static int idct_row(short *in, const int *const tab, int rnd)
57{
58 const unsigned c1 = tab[0];
59 const unsigned c2 = tab[1];
60 const unsigned c3 = tab[2];
61 const unsigned c4 = tab[3];
62 const unsigned c5 = tab[4];
63 const unsigned c6 = tab[5];
64 const unsigned c7 = tab[6];
65
66 const int right = in[5] | in[6] | in[7];
67 const int left = in[1] | in[2] | in[3];
68 if (!(right | in[4])) {
69 const int k = c4 * in[0] + rnd;
70 if (left) {
71 const unsigned a0 = k + c2 * in[2];
72 const unsigned a1 = k + c6 * in[2];
73 const unsigned a2 = k - c6 * in[2];
74 const unsigned a3 = k - c2 * in[2];
75
76 const int b0 = c1 * in[1] + c3 * in[3];
77 const int b1 = c3 * in[1] - c7 * in[3];
78 const int b2 = c5 * in[1] - c1 * in[3];
79 const int b3 = c7 * in[1] - c5 * in[3];
80
81 in[0] = (int)(a0 + b0) >> ROW_SHIFT;
82 in[1] = (int)(a1 + b1) >> ROW_SHIFT;
83 in[2] = (int)(a2 + b2) >> ROW_SHIFT;
84 in[3] = (int)(a3 + b3) >> ROW_SHIFT;
85 in[4] = (int)(a3 - b3) >> ROW_SHIFT;
86 in[5] = (int)(a2 - b2) >> ROW_SHIFT;
87 in[6] = (int)(a1 - b1) >> ROW_SHIFT;
88 in[7] = (int)(a0 - b0) >> ROW_SHIFT;
89 } else {
90 const int a0 = k >> ROW_SHIFT;
91 if (a0) {
92 in[0] =
93 in[1] =
94 in[2] =
95 in[3] =
96 in[4] =
97 in[5] =
98 in[6] =
99 in[7] = a0;
100 } else
101 return 0;
102 }
103 } else if (!(left | right)) {
104 const int a0 = (int)(rnd + c4 * (in[0] + in[4])) >> ROW_SHIFT;
105 const int a1 = (int)(rnd + c4 * (in[0] - in[4])) >> ROW_SHIFT;
106
107 in[0] = a0;
108 in[3] = a0;
109 in[4] = a0;
110 in[7] = a0;
111 in[1] = a1;
112 in[2] = a1;
113 in[5] = a1;
114 in[6] = a1;
115 } else {
116 const unsigned int k = c4 * in[0] + rnd;
117 const unsigned int a0 = k + c2 * in[2] + c4 * in[4] + c6 * in[6];
118 const unsigned int a1 = k + c6 * in[2] - c4 * in[4] - c2 * in[6];
119 const unsigned int a2 = k - c6 * in[2] - c4 * in[4] + c2 * in[6];
120 const unsigned int a3 = k - c2 * in[2] + c4 * in[4] - c6 * in[6];
121
122 const unsigned int b0 = c1 * in[1] + c3 * in[3] + c5 * in[5] + c7 * in[7];
123 const unsigned int b1 = c3 * in[1] - c7 * in[3] - c1 * in[5] - c5 * in[7];
124 const unsigned int b2 = c5 * in[1] - c1 * in[3] + c7 * in[5] + c3 * in[7];
125 const unsigned int b3 = c7 * in[1] - c5 * in[3] + c3 * in[5] - c1 * in[7];
126
127 in[0] = (int)(a0 + b0) >> ROW_SHIFT;
128 in[1] = (int)(a1 + b1) >> ROW_SHIFT;
129 in[2] = (int)(a2 + b2) >> ROW_SHIFT;
130 in[3] = (int)(a3 + b3) >> ROW_SHIFT;
131 in[4] = (int)(a3 - b3) >> ROW_SHIFT;
132 in[5] = (int)(a2 - b2) >> ROW_SHIFT;
133 in[6] = (int)(a1 - b1) >> ROW_SHIFT;
134 in[7] = (int)(a0 - b0) >> ROW_SHIFT;
135 }
136 return 1;
137}
138
139#define TAN1 0x32EC
140#define TAN2 0x6A0A
141#define TAN3 0xAB0E
142#define SQRT2 0x5A82
143
144#define MULT(c, x, n) ((unsigned)((int)((c) * (unsigned)(x)) >> (n)))
145// 12b version => #define MULT(c,x, n) ((((c) >> 3) * (x)) >> ((n) - 3))
146// 12b zero-testing version:
147
148#define BUTTERFLY(a, b, tmp) \
149 (tmp) = (a) + (b); \
150 (b) = (a) - (b); \
151 (a) = (tmp)
152
153#define LOAD_BUTTERFLY(m1, m2, a, b, tmp, s) \
154 (m1) = (s)[(a)] + (s)[(b)]; \
155 (m2) = (s)[(a)] - (s)[(b)]
156
157static void idct_col_8(short *const in)
158{
159 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, spill;
160
161 // odd
162
163 mm4 = (int) in[7 * 8];
164 mm5 = (int) in[5 * 8];
165 mm6 = (int) in[3 * 8];
166 mm7 = (int) in[1 * 8];
167
168 mm0 = MULT(TAN1, mm4, 16) + mm7;
169 mm1 = MULT(TAN1, mm7, 16) - mm4;
170 mm2 = MULT(TAN3, mm5, 16) + mm6;
171 mm3 = MULT(TAN3, mm6, 16) - mm5;
172
173 mm7 = mm0 + mm2;
174 mm4 = mm1 - mm3;
175 mm0 = mm0 - mm2;
176 mm1 = mm1 + mm3;
177 mm6 = mm0 + mm1;
178 mm5 = mm0 - mm1;
179 mm5 = 2 * MULT(SQRT2, mm5, 16); // 2*sqrt2
180 mm6 = 2 * MULT(SQRT2, mm6, 16); // Watch out: precision loss but done to match
181 // the pmulhw used in MMX/MMXEXT/SSE2 versions
182
183 // even
184
185 mm1 = (int) in[2 * 8];
186 mm2 = (int) in[6 * 8];
187 mm3 = MULT(TAN2, mm2, 16) + mm1;
188 mm2 = MULT(TAN2, mm1, 16) - mm2;
189
190 LOAD_BUTTERFLY(mm0, mm1, 0 * 8, 4 * 8, spill, in);
191
192 BUTTERFLY(mm0, mm3, spill);
193 BUTTERFLY(mm0, mm7, spill);
194 in[8 * 0] = (int16_t) (mm0 >> COL_SHIFT);
195 in[8 * 7] = (int16_t) (mm7 >> COL_SHIFT);
196 BUTTERFLY(mm3, mm4, mm0);
197 in[8 * 3] = (int16_t) (mm3 >> COL_SHIFT);
198 in[8 * 4] = (int16_t) (mm4 >> COL_SHIFT);
199
200 BUTTERFLY(mm1, mm2, mm0);
201 BUTTERFLY(mm1, mm6, mm0);
202 in[8 * 1] = (int16_t) (mm1 >> COL_SHIFT);
203 in[8 * 6] = (int16_t) (mm6 >> COL_SHIFT);
204 BUTTERFLY(mm2, mm5, mm0);
205 in[8 * 2] = (int16_t) (mm2 >> COL_SHIFT);
206 in[8 * 5] = (int16_t) (mm5 >> COL_SHIFT);
207}
208
209static void idct_col_4(short *const in)
210{
211 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, spill;
212
213 // odd
214
215 mm0 = (int) in[1 * 8];
216 mm2 = (int) in[3 * 8];
217
218 mm1 = MULT(TAN1, mm0, 16);
219 mm3 = MULT(TAN3, mm2, 16);
220
221 mm7 = mm0 + mm2;
222 mm4 = mm1 - mm3;
223 mm0 = mm0 - mm2;
224 mm1 = mm1 + mm3;
225 mm6 = mm0 + mm1;
226 mm5 = mm0 - mm1;
227 mm6 = 2 * MULT(SQRT2, mm6, 16); // 2*sqrt2
228 mm5 = 2 * MULT(SQRT2, mm5, 16);
229
230 // even
231
232 mm0 = mm1 = (int) in[0 * 8];
233 mm3 = (int) in[2 * 8];
234 mm2 = MULT(TAN2, mm3, 16);
235
236 BUTTERFLY(mm0, mm3, spill);
237 BUTTERFLY(mm0, mm7, spill);
238 in[8 * 0] = (int16_t) (mm0 >> COL_SHIFT);
239 in[8 * 7] = (int16_t) (mm7 >> COL_SHIFT);
240 BUTTERFLY(mm3, mm4, mm0);
241 in[8 * 3] = (int16_t) (mm3 >> COL_SHIFT);
242 in[8 * 4] = (int16_t) (mm4 >> COL_SHIFT);
243
244 BUTTERFLY(mm1, mm2, mm0);
245 BUTTERFLY(mm1, mm6, mm0);
246 in[8 * 1] = (int16_t) (mm1 >> COL_SHIFT);
247 in[8 * 6] = (int16_t) (mm6 >> COL_SHIFT);
248 BUTTERFLY(mm2, mm5, mm0);
249 in[8 * 2] = (int16_t) (mm2 >> COL_SHIFT);
250 in[8 * 5] = (int16_t) (mm5 >> COL_SHIFT);
251}
252
253static void idct_col_3(short *const in)
254{
255 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, spill;
256
257 // odd
258
259 mm7 = (int) in[1 * 8];
260 mm4 = MULT(TAN1, mm7, 16);
261
262 mm6 = mm7 + mm4;
263 mm5 = mm7 - mm4;
264 mm6 = 2 * MULT(SQRT2, mm6, 16); // 2*sqrt2
265 mm5 = 2 * MULT(SQRT2, mm5, 16);
266
267 // even
268
269 mm0 = mm1 = (int) in[0 * 8];
270 mm3 = (int) in[2 * 8];
271 mm2 = MULT(TAN2, mm3, 16);
272
273 BUTTERFLY(mm0, mm3, spill);
274 BUTTERFLY(mm0, mm7, spill);
275 in[8 * 0] = (int16_t) (mm0 >> COL_SHIFT);
276 in[8 * 7] = (int16_t) (mm7 >> COL_SHIFT);
277 BUTTERFLY(mm3, mm4, mm0);
278 in[8 * 3] = (int16_t) (mm3 >> COL_SHIFT);
279 in[8 * 4] = (int16_t) (mm4 >> COL_SHIFT);
280
281 BUTTERFLY(mm1, mm2, mm0);
282 BUTTERFLY(mm1, mm6, mm0);
283 in[8 * 1] = (int16_t) (mm1 >> COL_SHIFT);
284 in[8 * 6] = (int16_t) (mm6 >> COL_SHIFT);
285 BUTTERFLY(mm2, mm5, mm0);
286 in[8 * 2] = (int16_t) (mm2 >> COL_SHIFT);
287 in[8 * 5] = (int16_t) (mm5 >> COL_SHIFT);
288}
289
290void ff_xvid_idct(int16_t *const in)
291{
292 int i, rows = 0x07;
293
294 idct_row(in + 0 * 8, TAB04, RND0);
295 idct_row(in + 1 * 8, TAB17, RND1);
296 idct_row(in + 2 * 8, TAB26, RND2);
297 if (idct_row(in + 3 * 8, TAB35, RND3))
298 rows |= 0x08;
299 if (idct_row(in + 4 * 8, TAB04, RND4))
300 rows |= 0x10;
301 if (idct_row(in + 5 * 8, TAB35, RND5))
302 rows |= 0x20;
303 if (idct_row(in + 6 * 8, TAB26, RND6))
304 rows |= 0x40;
305 if (idct_row(in + 7 * 8, TAB17, RND7))
306 rows |= 0x80;
307
308 if (rows & 0xF0) {
309 for (i = 0; i < 8; i++)
310 idct_col_8(in + i);
311 } else if (rows & 0x08) {
312 for (i = 0; i < 8; i++)
313 idct_col_4(in + i);
314 } else {
315 for (i = 0; i < 8; i++)
316 idct_col_3(in + i);
317 }
318}
319
320static void xvid_idct_put(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
321{
322 ff_xvid_idct(block);
323 ff_put_pixels_clamped_c(block, dest, line_size);
324}
325
326static void xvid_idct_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
327{
328 ff_xvid_idct(block);
329 ff_add_pixels_clamped_c(block, dest, line_size);
330}
331
332av_cold void ff_xvid_idct_init(IDCTDSPContext *c)
333{
334 c->idct_put = xvid_idct_put;
335 c->idct_add = xvid_idct_add;
336 c->idct = ff_xvid_idct;
337 c->perm_type = FF_IDCT_PERM_NONE;
338
339#if ARCH_X86 && HAVE_X86ASM
340 ff_xvid_idct_init_x86(c);
341#elif ARCH_MIPS
342 ff_xvid_idct_init_mips(c);
343#endif
344}
left
static int BS_FUNC left(const BSCTX *bc)
Return the number of the bits left in a buffer.
Definition bitstream_template.h:170
i
#define i(width, name, range_min, range_max)
Definition cbs_h264.c:63
rnd
#define rnd
Definition checkasm.h:136
block
static int16_t block[64]
Definition dct.c:125
FF_IDCT_PERM_NONE
@ FF_IDCT_PERM_NONE
Definition idctdsp.h:28
ff_add_pixels_clamped_c
void ff_add_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels, ptrdiff_t line_size)
Definition idctdsp.c:147
ff_put_pixels_clamped_c
void ff_put_pixels_clamped_c(const int16_t *block, uint8_t *restrict pixels, ptrdiff_t line_size)
Definition idctdsp.c:73
attributes.h
Macro definitions for various function/variable attributes.
av_cold
#define av_cold
Definition attributes.h:117
c2
static const uint64_t c2
Definition murmur3.c:53
c1
static const uint64_t c1
Definition murmur3.c:52
COL_SHIFT
#define COL_SHIFT
Definition simple_idct_mmi.c:40
ROW_SHIFT
#define ROW_SHIFT
Definition simple_idct_mmi.c:39
tab
static const struct twinvq_data tab
Definition twinvq_data.h:10345
b1
static double b1(void *priv, double x, double y)
Definition vf_xfade.c:2034
a0
static double a0(void *priv, double x, double y)
Definition vf_xfade.c:2028
b2
static double b2(void *priv, double x, double y)
Definition vf_xfade.c:2035
b3
static double b3(void *priv, double x, double y)
Definition vf_xfade.c:2036
a3
static double a3(void *priv, double x, double y)
Definition vf_xfade.c:2031
b0
static double b0(void *priv, double x, double y)
Definition vf_xfade.c:2033
a2
static double a2(void *priv, double x, double y)
Definition vf_xfade.c:2030
a1
static double a1(void *priv, double x, double y)
Definition vf_xfade.c:2029
c
static double c[64]
Definition vsrc_mptestsrc.c:89
TAN1
#define TAN1
Definition xvididct.c:139
RND3
#define RND3
Definition xvididct.c:45
MULT
#define MULT(c, x, n)
Definition xvididct.c:144
BUTTERFLY
#define BUTTERFLY(a, b, tmp)
Definition xvididct.c:148
RND5
#define RND5
Definition xvididct.c:47
idct_col_3
static void idct_col_3(short *const in)
Definition xvididct.c:253
TAB35
static const int TAB35[]
Definition xvididct.c:54
RND4
#define RND4
Definition xvididct.c:46
TAN2
#define TAN2
Definition xvididct.c:140
SQRT2
#define SQRT2
Definition xvididct.c:142
TAB04
static const int TAB04[]
Definition xvididct.c:51
idct_col_4
static void idct_col_4(short *const in)
Definition xvididct.c:209
ff_xvid_idct
void ff_xvid_idct(int16_t *const in)
Definition xvididct.c:290
TAB26
static const int TAB26[]
Definition xvididct.c:53
RND0
#define RND0
Definition xvididct.c:42
RND7
#define RND7
Definition xvididct.c:49
idct_row
static int idct_row(short *in, const int *const tab, int rnd)
Definition xvididct.c:56
xvid_idct_add
static void xvid_idct_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
Definition xvididct.c:326
RND2
#define RND2
Definition xvididct.c:44
idct_col_8
static void idct_col_8(short *const in)
Definition xvididct.c:157
LOAD_BUTTERFLY
#define LOAD_BUTTERFLY(m1, m2, a, b, tmp, s)
Definition xvididct.c:153
TAN3
#define TAN3
Definition xvididct.c:141
RND6
#define RND6
Definition xvididct.c:48
xvid_idct_put
static void xvid_idct_put(uint8_t *dest, ptrdiff_t line_size, int16_t *block)
Definition xvididct.c:320
ff_xvid_idct_init
av_cold void ff_xvid_idct_init(IDCTDSPContext *c)
Definition xvididct.c:332
TAB17
static const int TAB17[]
Definition xvididct.c:52
RND1
#define RND1
Definition xvididct.c:43
ff_xvid_idct_init_x86
av_cold void ff_xvid_idct_init_x86(IDCTDSPContext *c)
Definition xvididct_init.c:27
ff_xvid_idct_init_mips
av_cold void ff_xvid_idct_init_mips(IDCTDSPContext *c)
Definition xvididct_init_mips.c:25
Generated on Sun Aug 16 2026 20:20:54 for FFmpeg by doxygen 1.13.2