1 /*****************************************************************************
\r
3 * XVID MPEG-4 VIDEO CODEC
\r
6 * Copyright (C) 2006-2011 Xvid Solutions GmbH
\r
8 * This program is free software ; you can redistribute it and/or modify
\r
9 * it under the terms of the GNU General Public License as published by
\r
10 * the Free Software Foundation ; either version 2 of the License, or
\r
11 * (at your option) any later version.
\r
13 * This program is distributed in the hope that it will be useful,
\r
14 * but WITHOUT ANY WARRANTY ; without even the implied warranty of
\r
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
\r
16 * GNU General Public License for more details.
\r
18 * You should have received a copy of the GNU General Public License
\r
19 * along with this program ; if not, write to the Free Software
\r
20 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
\r
22 * $Id: idct.c 1986 2011-05-18 09:07:40Z Isibaar $
\r
24 ****************************************************************************/
\r
30 * Alternative idct implementations for decoding compatibility
\r
32 * NOTE: this "C" version is not the original one,
\r
33 * but is modified to yield the same error profile
\r
34 * than the MMX version.
\r
36 ************************************************************************/
\r
42 /* function pointer */
\r
45 #define XVID_DSP_CLIP_255(x) ( ((x)&~255) ? ((-(x)) >> (8*sizeof((x))-1))&0xff : (x) )
\r
47 #define ROW_SHIFT 11
\r
50 // #define FIX(x) (int)((x) * (1<<ROW_SHIFT))
\r
51 #define Rnd0 65536 // 1<<(COL_SHIFT+ROW_SHIFT-1);
\r
52 #define Rnd1 3597 // FIX (1.75683487303);
\r
53 #define Rnd2 2260 // FIX (1.10355339059);
\r
54 #define Rnd3 1203 // FIX (0.587788325588);
\r
56 #define Rnd5 120 // FIX (0.058658283817);
\r
57 #define Rnd6 512 // FIX (0.25);
\r
58 #define Rnd7 512 // FIX (0.25);
\r
61 static const int Tab04[] = { 22725, 21407, 19266, 16384, 12873, 8867, 4520 };
\r
62 static const int Tab17[] = { 31521, 29692, 26722, 22725, 17855, 12299, 6270 };
\r
63 static const int Tab26[] = { 29692, 27969, 25172, 21407, 16819, 11585, 5906 };
\r
64 static const int Tab35[] = { 26722, 25172, 22654, 19266, 15137, 10426, 5315 };
\r
66 static int Idct_Row(short * In, const int * const Tab, int Rnd)
\r
68 const int C1 = Tab[0];
\r
69 const int C2 = Tab[1];
\r
70 const int C3 = Tab[2];
\r
71 const int C4 = Tab[3];
\r
72 const int C5 = Tab[4];
\r
73 const int C6 = Tab[5];
\r
74 const int C7 = Tab[6];
\r
76 const int Right = In[5]|In[6]|In[7];
\r
77 const int Left = In[1]|In[2]|In[3];
\r
78 if (!(Right | In[4]))
\r
80 const int K = C4*In[0] + Rnd;
\r
83 const int a0 = K + C2*In[2];
\r
84 const int a1 = K + C6*In[2];
\r
85 const int a2 = K - C6*In[2];
\r
86 const int a3 = K - C2*In[2];
\r
88 const int b0 = C1*In[1] + C3*In[3];
\r
89 const int b1 = C3*In[1] - C7*In[3];
\r
90 const int b2 = C5*In[1] - C1*In[3];
\r
91 const int b3 = C7*In[1] - C5*In[3];
\r
93 In[0] = (a0 + b0) >> ROW_SHIFT;
\r
94 In[1] = (a1 + b1) >> ROW_SHIFT;
\r
95 In[2] = (a2 + b2) >> ROW_SHIFT;
\r
96 In[3] = (a3 + b3) >> ROW_SHIFT;
\r
97 In[4] = (a3 - b3) >> ROW_SHIFT;
\r
98 In[5] = (a2 - b2) >> ROW_SHIFT;
\r
99 In[6] = (a1 - b1) >> ROW_SHIFT;
\r
100 In[7] = (a0 - b0) >> ROW_SHIFT;
\r
104 const int a0 = K >> ROW_SHIFT;
\r
106 In[0] = In[1] = In[2] = In[3] =
\r
107 In[4] = In[5] = In[6] = In[7] = a0;
\r
112 else if (!(Left|Right))
\r
114 const int a0 = (Rnd + C4*(In[0]+In[4])) >> ROW_SHIFT;
\r
115 const int a1 = (Rnd + C4*(In[0]-In[4])) >> ROW_SHIFT;
\r
128 const int K = C4*In[0] + Rnd;
\r
129 const int a0 = K + C2*In[2] + C4*In[4] + C6*In[6];
\r
130 const int a1 = K + C6*In[2] - C4*In[4] - C2*In[6];
\r
131 const int a2 = K - C6*In[2] - C4*In[4] + C2*In[6];
\r
132 const int a3 = K - C2*In[2] + C4*In[4] - C6*In[6];
\r
134 const int b0 = C1*In[1] + C3*In[3] + C5*In[5] + C7*In[7];
\r
135 const int b1 = C3*In[1] - C7*In[3] - C1*In[5] - C5*In[7];
\r
136 const int b2 = C5*In[1] - C1*In[3] + C7*In[5] + C3*In[7];
\r
137 const int b3 = C7*In[1] - C5*In[3] + C3*In[5] - C1*In[7];
\r
139 In[0] = (a0 + b0) >> ROW_SHIFT;
\r
140 In[1] = (a1 + b1) >> ROW_SHIFT;
\r
141 In[2] = (a2 + b2) >> ROW_SHIFT;
\r
142 In[3] = (a3 + b3) >> ROW_SHIFT;
\r
143 In[4] = (a3 - b3) >> ROW_SHIFT;
\r
144 In[5] = (a2 - b2) >> ROW_SHIFT;
\r
145 In[6] = (a1 - b1) >> ROW_SHIFT;
\r
146 In[7] = (a0 - b0) >> ROW_SHIFT;
\r
151 #define Tan1 0x32ec
\r
152 #define Tan2 0x6a0a
\r
153 #define Tan3 0xab0e
\r
154 #define Sqrt2 0x5a82
\r
156 #define MULT(c,x, n) ( ((c) * (x)) >> (n) )
\r
157 // 12b version => #define MULT(c,x, n) ( (((c)>>3) * (x)) >> ((n)-3) )
\r
158 // 12b zero-testing version:
\r
160 #define BUTF(a, b, tmp) \
\r
165 #define LOAD_BUTF(m1, m2, a, b, tmp, S) \
\r
166 (m1) = (S)[(a)] + (S)[(b)]; \
\r
167 (m2) = (S)[(a)] - (S)[(b)]
\r
169 static void Idct_Col_8(short * const In)
\r
171 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, Spill;
\r
175 mm4 = (int)In[7*8];
\r
176 mm5 = (int)In[5*8];
\r
177 mm6 = (int)In[3*8];
\r
178 mm7 = (int)In[1*8];
\r
180 mm0 = MULT(Tan1, mm4, 16) + mm7;
\r
181 mm1 = MULT(Tan1, mm7, 16) - mm4;
\r
182 mm2 = MULT(Tan3, mm5, 16) + mm6;
\r
183 mm3 = MULT(Tan3, mm6, 16) - mm5;
\r
191 mm5 = 2*MULT(Sqrt2, mm5, 16); // 2*sqrt2
\r
192 mm6 = 2*MULT(Sqrt2, mm6, 16); // Watch out: precision loss but done to match
\r
193 // the pmulhw used in mmx/sse versions
\r
197 mm1 = (int)In[2*8];
\r
198 mm2 = (int)In[6*8];
\r
199 mm3 = MULT(Tan2,mm2, 16) + mm1;
\r
200 mm2 = MULT(Tan2,mm1, 16) - mm2;
\r
202 LOAD_BUTF(mm0, mm1, 0*8, 4*8, Spill, In);
\r
204 BUTF(mm0, mm3, Spill);
\r
205 BUTF(mm0, mm7, Spill);
\r
206 In[8*0] = (int16_t) (mm0 >> COL_SHIFT);
\r
207 In[8*7] = (int16_t) (mm7 >> COL_SHIFT);
\r
208 BUTF(mm3, mm4, mm0);
\r
209 In[8*3] = (int16_t) (mm3 >> COL_SHIFT);
\r
210 In[8*4] = (int16_t) (mm4 >> COL_SHIFT);
\r
212 BUTF(mm1, mm2, mm0);
\r
213 BUTF(mm1, mm6, mm0);
\r
214 In[8*1] = (int16_t) (mm1 >> COL_SHIFT);
\r
215 In[8*6] = (int16_t) (mm6 >> COL_SHIFT);
\r
216 BUTF(mm2, mm5, mm0);
\r
217 In[8*2] = (int16_t) (mm2 >> COL_SHIFT);
\r
218 In[8*5] = (int16_t) (mm5 >> COL_SHIFT);
\r
221 static void Idct_Col_4(short * const In)
\r
223 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, Spill;
\r
227 mm0 = (int)In[1*8];
\r
228 mm2 = (int)In[3*8];
\r
230 mm1 = MULT(Tan1, mm0, 16);
\r
231 mm3 = MULT(Tan3, mm2, 16);
\r
239 mm6 = 2*MULT(Sqrt2, mm6, 16); // 2*sqrt2
\r
240 mm5 = 2*MULT(Sqrt2, mm5, 16);
\r
244 mm0 = mm1 = (int)In[0*8];
\r
245 mm3 = (int)In[2*8];
\r
246 mm2 = MULT(Tan2,mm3, 16);
\r
248 BUTF(mm0, mm3, Spill);
\r
249 BUTF(mm0, mm7, Spill);
\r
250 In[8*0] = (int16_t) (mm0 >> COL_SHIFT);
\r
251 In[8*7] = (int16_t) (mm7 >> COL_SHIFT);
\r
252 BUTF(mm3, mm4, mm0);
\r
253 In[8*3] = (int16_t) (mm3 >> COL_SHIFT);
\r
254 In[8*4] = (int16_t) (mm4 >> COL_SHIFT);
\r
256 BUTF(mm1, mm2, mm0);
\r
257 BUTF(mm1, mm6, mm0);
\r
258 In[8*1] = (int16_t) (mm1 >> COL_SHIFT);
\r
259 In[8*6] = (int16_t) (mm6 >> COL_SHIFT);
\r
260 BUTF(mm2, mm5, mm0);
\r
261 In[8*2] = (int16_t) (mm2 >> COL_SHIFT);
\r
262 In[8*5] = (int16_t) (mm5 >> COL_SHIFT);
\r
265 static void Idct_Col_3(short * const In)
\r
267 int mm0, mm1, mm2, mm3, mm4, mm5, mm6, mm7, Spill;
\r
271 mm7 = (int)In[1*8];
\r
272 mm4 = MULT(Tan1, mm7, 16);
\r
276 mm6 = 2*MULT(Sqrt2, mm6, 16); // 2*sqrt2
\r
277 mm5 = 2*MULT(Sqrt2, mm5, 16);
\r
281 mm0 = mm1 = (int)In[0*8];
\r
282 mm3 = (int)In[2*8];
\r
283 mm2 = MULT(Tan2,mm3, 16);
\r
285 BUTF(mm0, mm3, Spill);
\r
286 BUTF(mm0, mm7, Spill);
\r
287 In[8*0] = (int16_t) (mm0 >> COL_SHIFT);
\r
288 In[8*7] = (int16_t) (mm7 >> COL_SHIFT);
\r
289 BUTF(mm3, mm4, mm0);
\r
290 In[8*3] = (int16_t) (mm3 >> COL_SHIFT);
\r
291 In[8*4] = (int16_t) (mm4 >> COL_SHIFT);
\r
293 BUTF(mm1, mm2, mm0);
\r
294 BUTF(mm1, mm6, mm0);
\r
295 In[8*1] = (int16_t) (mm1 >> COL_SHIFT);
\r
296 In[8*6] = (int16_t) (mm6 >> COL_SHIFT);
\r
297 BUTF(mm2, mm5, mm0);
\r
298 In[8*2] = (int16_t) (mm2 >> COL_SHIFT);
\r
299 In[8*5] = (int16_t) (mm5 >> COL_SHIFT);
\r
310 //////////////////////////////////////////////////////////
\r
312 void idct_int32(short *const In)
\r
314 int i, Rows = 0x07;
\r
316 Idct_Row(In + 0*8, Tab04, Rnd0);
\r
317 Idct_Row(In + 1*8, Tab17, Rnd1);
\r
318 Idct_Row(In + 2*8, Tab26, Rnd2);
\r
319 if (Idct_Row(In + 3*8, Tab35, Rnd3)) Rows |= 0x08;
\r
320 if (Idct_Row(In + 4*8, Tab04, Rnd4)) Rows |= 0x10;
\r
321 if (Idct_Row(In + 5*8, Tab35, Rnd5)) Rows |= 0x20;
\r
322 if (Idct_Row(In + 6*8, Tab26, Rnd6)) Rows |= 0x40;
\r
323 if (Idct_Row(In + 7*8, Tab17, Rnd7)) Rows |= 0x80;
\r
327 Idct_Col_8(In + i);
\r
329 else if (Rows&0x08) {
\r
331 Idct_Col_4(In + i);
\r
335 Idct_Col_3(In + i);
\r