1 /*****************************************************************************
2 * dct.c: h264 encoder library
3 *****************************************************************************
4 * Copyright (C) 2003 Laurent Aimar
5 * $Id: dct.c,v 1.1 2004/06/03 19:27:06 fenrir Exp $
7 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program; if not, write to the Free Software
21 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
22 *****************************************************************************/
36 # include "i386/dct.h"
43 static inline int clip_uint8( int a )
52 * XXX For all dct dc : input could be equal to output so ...
55 static void dct2x2dc( int16_t d[2][2] )
59 tmp[0][0] = d[0][0] + d[0][1];
60 tmp[1][0] = d[0][0] - d[0][1];
61 tmp[0][1] = d[1][0] + d[1][1];
62 tmp[1][1] = d[1][0] - d[1][1];
64 d[0][0] = tmp[0][0] + tmp[0][1];
65 d[0][1] = tmp[1][0] + tmp[1][1];
66 d[1][0] = tmp[0][0] - tmp[0][1];
67 d[1][1] = tmp[1][0] - tmp[1][1];
70 static void dct4x4dc( int16_t d[4][4] )
77 for( i = 0; i < 4; i++ )
79 s01 = d[i][0] + d[i][1];
80 d01 = d[i][0] - d[i][1];
81 s23 = d[i][2] + d[i][3];
82 d23 = d[i][2] - d[i][3];
84 tmp[0][i] = s01 + s23;
85 tmp[1][i] = s01 - s23;
86 tmp[2][i] = d01 - d23;
87 tmp[3][i] = d01 + d23;
90 for( i = 0; i < 4; i++ )
92 s01 = tmp[i][0] + tmp[i][1];
93 d01 = tmp[i][0] - tmp[i][1];
94 s23 = tmp[i][2] + tmp[i][3];
95 d23 = tmp[i][2] - tmp[i][3];
97 d[0][i] = ( s01 + s23 + 1 ) >> 1;
98 d[1][i] = ( s01 - s23 + 1 ) >> 1;
99 d[2][i] = ( d01 - d23 + 1 ) >> 1;
100 d[3][i] = ( d01 + d23 + 1 ) >> 1;
104 static void idct4x4dc( int16_t d[4][4] )
111 for( i = 0; i < 4; i++ )
113 s01 = d[0][i] + d[1][i];
114 d01 = d[0][i] - d[1][i];
115 s23 = d[2][i] + d[3][i];
116 d23 = d[2][i] - d[3][i];
118 tmp[0][i] = s01 + s23;
119 tmp[1][i] = s01 - s23;
120 tmp[2][i] = d01 - d23;
121 tmp[3][i] = d01 + d23;
124 for( i = 0; i < 4; i++ )
126 s01 = tmp[i][0] + tmp[i][1];
127 d01 = tmp[i][0] - tmp[i][1];
128 s23 = tmp[i][2] + tmp[i][3];
129 d23 = tmp[i][2] - tmp[i][3];
138 static void sub4x4_dct( int16_t dct[4][4], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
145 for( y = 0; y < 4; y++ )
147 for( x = 0; x < 4; x++ )
149 d[y][x] = pix1[x] - pix2[x];
155 for( i = 0; i < 4; i++ )
157 const int s03 = d[i][0] + d[i][3];
158 const int s12 = d[i][1] + d[i][2];
159 const int d03 = d[i][0] - d[i][3];
160 const int d12 = d[i][1] - d[i][2];
162 tmp[0][i] = s03 + s12;
163 tmp[1][i] = 2*d03 + d12;
164 tmp[2][i] = s03 - s12;
165 tmp[3][i] = d03 - 2*d12;
168 for( i = 0; i < 4; i++ )
170 const int s03 = tmp[i][0] + tmp[i][3];
171 const int s12 = tmp[i][1] + tmp[i][2];
172 const int d03 = tmp[i][0] - tmp[i][3];
173 const int d12 = tmp[i][1] - tmp[i][2];
175 dct[0][i] = s03 + s12;
176 dct[1][i] = 2*d03 + d12;
177 dct[2][i] = s03 - s12;
178 dct[3][i] = d03 - 2*d12;
182 static void sub8x8_dct( int16_t dct[4][4][4], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
184 sub4x4_dct( dct[0], &pix1[0], i_pix1, &pix2[0], i_pix2 );
185 sub4x4_dct( dct[1], &pix1[4], i_pix1, &pix2[4], i_pix2 );
186 sub4x4_dct( dct[2], &pix1[4*i_pix1+0], i_pix1, &pix2[4*i_pix2+0], i_pix2 );
187 sub4x4_dct( dct[3], &pix1[4*i_pix1+4], i_pix1, &pix2[4*i_pix2+4], i_pix2 );
190 static void sub16x16_dct( int16_t dct[16][4][4], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
192 sub8x8_dct( &dct[ 0], pix1, i_pix1, pix2, i_pix2 );
193 sub8x8_dct( &dct[ 4], &pix1[8], i_pix1, &pix2[8], i_pix2 );
194 sub8x8_dct( &dct[ 8], &pix1[8*i_pix1], i_pix1, &pix2[8*i_pix2], i_pix2 );
195 sub8x8_dct( &dct[12], &pix1[8*i_pix1+8], i_pix1, &pix2[8*i_pix2+8], i_pix2 );
199 static void add4x4_idct( uint8_t *p_dst, int i_dst, int16_t dct[4][4] )
206 for( i = 0; i < 4; i++ )
208 const int s02 = dct[i][0] + dct[i][2];
209 const int d02 = dct[i][0] - dct[i][2];
210 const int s13 = dct[i][1] + (dct[i][3]>>1);
211 const int d13 = (dct[i][1]>>1) - dct[i][3];
213 tmp[i][0] = s02 + s13;
214 tmp[i][1] = d02 + d13;
215 tmp[i][2] = d02 - d13;
216 tmp[i][3] = s02 - s13;
219 for( i = 0; i < 4; i++ )
221 const int s02 = tmp[0][i] + tmp[2][i];
222 const int d02 = tmp[0][i] - tmp[2][i];
223 const int s13 = tmp[1][i] + (tmp[3][i]>>1);
224 const int d13 = (tmp[1][i]>>1) - tmp[3][i];
226 d[0][i] = ( s02 + s13 + 32 ) >> 6;
227 d[1][i] = ( d02 + d13 + 32 ) >> 6;
228 d[2][i] = ( d02 - d13 + 32 ) >> 6;
229 d[3][i] = ( s02 - s13 + 32 ) >> 6;
233 for( y = 0; y < 4; y++ )
235 for( x = 0; x < 4; x++ )
237 p_dst[x] = clip_uint8( p_dst[x] + d[y][x] );
243 static void add8x8_idct( uint8_t *p_dst, int i_dst, int16_t dct[4][4][4] )
245 add4x4_idct( p_dst, i_dst, dct[0] );
246 add4x4_idct( &p_dst[4], i_dst, dct[1] );
247 add4x4_idct( &p_dst[4*i_dst+0], i_dst, dct[2] );
248 add4x4_idct( &p_dst[4*i_dst+4], i_dst, dct[3] );
251 static void add16x16_idct( uint8_t *p_dst, int i_dst, int16_t dct[16][4][4] )
253 add8x8_idct( &p_dst[0], i_dst, &dct[0] );
254 add8x8_idct( &p_dst[8], i_dst, &dct[4] );
255 add8x8_idct( &p_dst[8*i_dst], i_dst, &dct[8] );
256 add8x8_idct( &p_dst[8*i_dst+8], i_dst, &dct[12] );
259 /****************************************************************************
261 ****************************************************************************/
263 static inline void dct8_1d( int16_t src[8][8], int16_t dst[8][8] )
266 for( i = 0; i < 8; i++ )
268 const int s07 = src[i][0] + src[i][7];
269 const int s16 = src[i][1] + src[i][6];
270 const int s25 = src[i][2] + src[i][5];
271 const int s34 = src[i][3] + src[i][4];
273 const int a0 = s07 + s34;
274 const int a1 = s16 + s25;
275 const int a2 = s07 - s34;
276 const int a3 = s16 - s25;
278 const int d07 = src[i][0] - src[i][7];
279 const int d16 = src[i][1] - src[i][6];
280 const int d25 = src[i][2] - src[i][5];
281 const int d34 = src[i][3] - src[i][4];
283 const int a4 = d16 + d25 + (d07 + (d07>>1));
284 const int a5 = d07 - d34 - (d25 + (d25>>1));
285 const int a6 = d07 + d34 - (d16 + (d16>>1));
286 const int a7 = d16 - d25 + (d34 + (d34>>1));
289 dst[1][i] = a4 + (a7>>2);
290 dst[2][i] = a2 + (a3>>1);
291 dst[3][i] = a5 + (a6>>2);
293 dst[5][i] = a6 - (a5>>2);
294 dst[6][i] = (a2>>1) - a3;
295 dst[7][i] = (a4>>2) - a7;
299 static void sub8x8_dct8( int16_t dct[8][8], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
305 for( y = 0; y < 8; y++ )
307 for( x = 0; x < 8; x++ )
309 d[y][x] = pix1[x] - pix2[x];
319 static void sub16x16_dct8( int16_t dct[4][8][8], uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
321 sub8x8_dct8( dct[0], pix1, i_pix1, pix2, i_pix2 );
322 sub8x8_dct8( dct[1], &pix1[8], i_pix1, &pix2[8], i_pix2 );
323 sub8x8_dct8( dct[2], &pix1[8*i_pix1], i_pix1, &pix2[8*i_pix2], i_pix2 );
324 sub8x8_dct8( dct[3], &pix1[8*i_pix1+8], i_pix1, &pix2[8*i_pix2+8], i_pix2 );
327 static inline void idct8_1d( int16_t src[8][8], int16_t dst[8][8] )
330 for( i = 0; i < 8; i++ )
332 const int a0 = src[i][0] + src[i][4];
333 const int a2 = src[i][0] - src[i][4];
334 const int a4 = (src[i][2]>>1) - src[i][6];
335 const int a6 = (src[i][6]>>1) + src[i][2];
337 const int b0 = a0 + a6;
338 const int b2 = a2 + a4;
339 const int b4 = a2 - a4;
340 const int b6 = a0 - a6;
342 const int a1 = -src[i][3] + src[i][5] - src[i][7] - (src[i][7]>>1);
343 const int a3 = src[i][1] + src[i][7] - src[i][3] - (src[i][3]>>1);
344 const int a5 = -src[i][1] + src[i][7] + src[i][5] + (src[i][5]>>1);
345 const int a7 = src[i][3] + src[i][5] + src[i][1] + (src[i][1]>>1);
347 const int b1 = (a7>>2) + a1;
348 const int b3 = a3 + (a5>>2);
349 const int b5 = (a3>>2) - a5;
350 const int b7 = a7 - (a1>>2);
363 static void add8x8_idct8( uint8_t *p_dst, int i_dst, int16_t dct[8][8] )
369 idct8_1d( dct, tmp );
372 for( y = 0; y < 8; y++ )
374 for( x = 0; x < 8; x++ )
376 p_dst[x] = clip_uint8( p_dst[x] + ((d[y][x] + 32) >> 6) );
382 static void add16x16_idct8( uint8_t *p_dst, int i_dst, int16_t dct[4][8][8] )
384 add8x8_idct8( &p_dst[0], i_dst, dct[0] );
385 add8x8_idct8( &p_dst[8], i_dst, dct[1] );
386 add8x8_idct8( &p_dst[8*i_dst], i_dst, dct[2] );
387 add8x8_idct8( &p_dst[8*i_dst+8], i_dst, dct[3] );
391 /****************************************************************************
393 ****************************************************************************/
394 void x264_dct_init( int cpu, x264_dct_function_t *dctf )
396 dctf->sub4x4_dct = sub4x4_dct;
397 dctf->add4x4_idct = add4x4_idct;
399 dctf->sub8x8_dct = sub8x8_dct;
400 dctf->add8x8_idct = add8x8_idct;
402 dctf->sub16x16_dct = sub16x16_dct;
403 dctf->add16x16_idct = add16x16_idct;
405 dctf->sub8x8_dct8 = sub8x8_dct8;
406 dctf->add8x8_idct8 = add8x8_idct8;
408 dctf->sub16x16_dct8 = sub16x16_dct8;
409 dctf->add16x16_idct8 = add16x16_idct8;
411 dctf->dct4x4dc = dct4x4dc;
412 dctf->idct4x4dc = idct4x4dc;
414 dctf->dct2x2dc = dct2x2dc;
415 dctf->idct2x2dc = dct2x2dc;
418 if( cpu&X264_CPU_MMXEXT )
420 dctf->sub4x4_dct = x264_sub4x4_dct_mmxext;
421 dctf->sub8x8_dct = x264_sub8x8_dct_mmxext;
422 dctf->sub16x16_dct = x264_sub16x16_dct_mmxext;
424 dctf->add4x4_idct = x264_add4x4_idct_mmxext;
425 dctf->add8x8_idct = x264_add8x8_idct_mmxext;
426 dctf->add16x16_idct = x264_add16x16_idct_mmxext;
428 dctf->dct4x4dc = x264_dct4x4dc_mmxext;
429 dctf->idct4x4dc = x264_idct4x4dc_mmxext;
433 if( cpu&X264_CPU_ALTIVEC )
435 dctf->sub4x4_dct = x264_sub4x4_dct_altivec;
436 dctf->sub8x8_dct = x264_sub8x8_dct_altivec;
437 dctf->sub16x16_dct = x264_sub16x16_dct_altivec;