1 /*****************************************************************************
3 * XVID MPEG-4 VIDEO CODEC
6 * Copyright (C) 2006-2011 Xvid Solutions GmbH
8 * This file is part of FFmpeg.
10 * FFmpeg is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
15 * FFmpeg is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 ****************************************************************************/
30 * Alternative idct implementations for decoding compatibility
33 * @note this "C" version is not the original one, but is modified to
34 * yield the same error profile as the MMX version.
44 // #define FIX(x) (int)((x) * (1<<ROW_SHIFT))
45 #define Rnd0 65536 // 1<<(COL_SHIFT+ROW_SHIFT-1);
46 #define Rnd1 3597 // FIX (1.75683487303);
47 #define Rnd2 2260 // FIX (1.10355339059);
48 #define Rnd3 1203 // FIX (0.587788325588);
50 #define Rnd5 120 // FIX (0.058658283817);
51 #define Rnd6 512 // FIX (0.25);
52 #define Rnd7 512 // FIX (0.25);
54 static const int Tab04[] = { 22725, 21407, 19266, 16384, 12873, 8867, 4520 };
55 static const int Tab17[] = { 31521, 29692, 26722, 22725, 17855, 12299, 6270 };
56 static const int Tab26[] = { 29692, 27969, 25172, 21407, 16819, 11585, 5906 };
57 static const int Tab35[] = { 26722, 25172, 22654, 19266, 15137, 10426, 5315 };
59 static int Idct_Row(short * In, const int * const Tab, int Rnd)
61 const int C1 = Tab[0];
62 const int C2 = Tab[1];
63 const int C3 = Tab[2];
64 const int C4 = Tab[3];
65 const int C5 = Tab[4];
66 const int C6 = Tab[5];
67 const int C7 = Tab[6];
69 const int Right = In[5]|In[6]|In[7];
70 const int Left = In[1]|In[2]|In[3];
73 const int K = C4*In[0] + Rnd;
76 const int a0 = K + C2*In[2];
77 const int a1 = K + C6*In[2];
78 const int a2 = K - C6*In[2];
79 const int a3 = K - C2*In[2];
81 const int b0 = C1*In[1] + C3*In[3];
82 const int b1 = C3*In[1] - C7*In[3];
83 const int b2 = C5*In[1] - C1*In[3];
84 const int b3 = C7*In[1] - C5*In[3];
86 In[0] = (a0 + b0) >> ROW_SHIFT;
87 In[1] = (a1 + b1) >> ROW_SHIFT;
88 In[2] = (a2 + b2) >> ROW_SHIFT;
89 In[3] = (a3 + b3) >> ROW_SHIFT;
90 In[4] = (a3 - b3) >> ROW_SHIFT;
91 In[5] = (a2 - b2) >> ROW_SHIFT;
92 In[6] = (a1 - b1) >> ROW_SHIFT;
93 In[7] = (a0 - b0) >> ROW_SHIFT;
97 const int a0 = K >> ROW_SHIFT;
99 In[0] = In[1] = In[2] = In[3] =
100 In[4] = In[5] = In[6] = In[7] = a0;
105 else if (!(Left|Right))
107 const int a0 = (Rnd + C4*(In[0]+In[4])) >> ROW_SHIFT;
108 const int a1 = (Rnd + C4*(In[0]-In[4])) >> ROW_SHIFT;
121 const int K = C4*In[0] + Rnd;
122 const int a0 = K + C2*In[2] + C4*In[4] + C6*In[6];
123 const int a1 = K + C6*In[2] - C4*In[4] - C2*In[6];
124 const int a2 = K - C6*In[2] - C4*In[4] + C2*In[6];
125 const int a3 = K - C2*In[2] + C4*In[4] - C6*In[6];
127 const int b0 = C1*In[1] + C3*In[3] + C5*In[5] + C7*In[7];
128 const int b1 = C3*In[1] - C7*In[3] - C1*In[5] - C5*In[7];
129 const int b2 = C5*In[1] - C1*In[3] + C7*In[5] + C3*In[7];
130 const int b3 = C7*In[1] - C5*In[3] + C3*In[5] - C1*In[7];
132 In[0] = (a0 + b0) >> ROW_SHIFT;
133 In[1] = (a1 + b1) >> ROW_SHIFT;
134 In[2] = (a2 + b2) >> ROW_SHIFT;
135 In[3] = (a3 + b3) >> ROW_SHIFT;
136 In[4] = (a3 - b3) >> ROW_SHIFT;
137 In[5] = (a2 - b2) >> ROW_SHIFT;
138 In[6] = (a1 - b1) >> ROW_SHIFT;
139 In[7] = (a0 - b0) >> ROW_SHIFT;
149 #define MULT(c,x, n) ( ((c) * (x)) >> (n) )
150 // 12b version => #define MULT(c,x, n) ( (((c)>>3) * (x)) >> ((n)-3) )
151 // 12b zero-testing version:
153 #define BUTF(a, b, tmp) \
158 #define LOAD_BUTF(m1, m2, a, b, tmp, S) \
159 (m1) = (S)[(a)] + (S)[(b)]; \
160 (m2) = (S)[(a)] - (S)[(b)]
162 static void Idct_Col_8(short * const In)
164 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, Spill;
173 mm0 = MULT(Tan1, mm4, 16) + mm7;
174 mm1 = MULT(Tan1, mm7, 16) - mm4;
175 mm2 = MULT(Tan3, mm5, 16) + mm6;
176 mm3 = MULT(Tan3, mm6, 16) - mm5;
184 mm5 = 2*MULT(Sqrt2, mm5, 16); // 2*sqrt2
185 mm6 = 2*MULT(Sqrt2, mm6, 16); // Watch out: precision loss but done to match
186 // the pmulhw used in mmx/sse versions
192 mm3 = MULT(Tan2,mm2, 16) + mm1;
193 mm2 = MULT(Tan2,mm1, 16) - mm2;
195 LOAD_BUTF(mm0, mm1, 0*8, 4*8, Spill, In);
197 BUTF(mm0, mm3, Spill);
198 BUTF(mm0, mm7, Spill);
199 In[8*0] = (int16_t) (mm0 >> COL_SHIFT);
200 In[8*7] = (int16_t) (mm7 >> COL_SHIFT);
202 In[8*3] = (int16_t) (mm3 >> COL_SHIFT);
203 In[8*4] = (int16_t) (mm4 >> COL_SHIFT);
207 In[8*1] = (int16_t) (mm1 >> COL_SHIFT);
208 In[8*6] = (int16_t) (mm6 >> COL_SHIFT);
210 In[8*2] = (int16_t) (mm2 >> COL_SHIFT);
211 In[8*5] = (int16_t) (mm5 >> COL_SHIFT);
214 static void Idct_Col_4(short * const In)
216 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, Spill;
223 mm1 = MULT(Tan1, mm0, 16);
224 mm3 = MULT(Tan3, mm2, 16);
232 mm6 = 2*MULT(Sqrt2, mm6, 16); // 2*sqrt2
233 mm5 = 2*MULT(Sqrt2, mm5, 16);
237 mm0 = mm1 = (int)In[0*8];
239 mm2 = MULT(Tan2,mm3, 16);
241 BUTF(mm0, mm3, Spill);
242 BUTF(mm0, mm7, Spill);
243 In[8*0] = (int16_t) (mm0 >> COL_SHIFT);
244 In[8*7] = (int16_t) (mm7 >> COL_SHIFT);
246 In[8*3] = (int16_t) (mm3 >> COL_SHIFT);
247 In[8*4] = (int16_t) (mm4 >> COL_SHIFT);
251 In[8*1] = (int16_t) (mm1 >> COL_SHIFT);
252 In[8*6] = (int16_t) (mm6 >> COL_SHIFT);
254 In[8*2] = (int16_t) (mm2 >> COL_SHIFT);
255 In[8*5] = (int16_t) (mm5 >> COL_SHIFT);
258 static void Idct_Col_3(short * const In)
260 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, Spill;
265 mm4 = MULT(Tan1, mm7, 16);
269 mm6 = 2*MULT(Sqrt2, mm6, 16); // 2*sqrt2
270 mm5 = 2*MULT(Sqrt2, mm5, 16);
274 mm0 = mm1 = (int)In[0*8];
276 mm2 = MULT(Tan2,mm3, 16);
278 BUTF(mm0, mm3, Spill);
279 BUTF(mm0, mm7, Spill);
280 In[8*0] = (int16_t) (mm0 >> COL_SHIFT);
281 In[8*7] = (int16_t) (mm7 >> COL_SHIFT);
283 In[8*3] = (int16_t) (mm3 >> COL_SHIFT);
284 In[8*4] = (int16_t) (mm4 >> COL_SHIFT);
288 In[8*1] = (int16_t) (mm1 >> COL_SHIFT);
289 In[8*6] = (int16_t) (mm6 >> COL_SHIFT);
291 In[8*2] = (int16_t) (mm2 >> COL_SHIFT);
292 In[8*5] = (int16_t) (mm5 >> COL_SHIFT);
303 //////////////////////////////////////////////////////////
305 void ff_idct_xvid(int16_t *const In)
309 Idct_Row(In + 0*8, Tab04, Rnd0);
310 Idct_Row(In + 1*8, Tab17, Rnd1);
311 Idct_Row(In + 2*8, Tab26, Rnd2);
312 if (Idct_Row(In + 3*8, Tab35, Rnd3)) Rows |= 0x08;
313 if (Idct_Row(In + 4*8, Tab04, Rnd4)) Rows |= 0x10;
314 if (Idct_Row(In + 5*8, Tab35, Rnd5)) Rows |= 0x20;
315 if (Idct_Row(In + 6*8, Tab26, Rnd6)) Rows |= 0x40;
316 if (Idct_Row(In + 7*8, Tab17, Rnd7)) Rows |= 0x80;
322 else if (Rows&0x08) {