1 /*****************************************************************************
2 * pixel.c: h264 encoder
3 *****************************************************************************
4 * Copyright (C) 2003-2008 x264 project
6 * Authors: Loren Merritt <lorenm@u.washington.edu>
7 * 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., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
22 *****************************************************************************/
27 # include "x86/pixel.h"
30 # include "ppc/pixel.h"
33 # include "arm/pixel.h"
35 #ifdef ARCH_UltraSparc
36 # include "sparc/pixel.h"
40 /****************************************************************************
42 ****************************************************************************/
43 #define PIXEL_SAD_C( name, lx, ly ) \
44 static int name( uint8_t *pix1, int i_stride_pix1, \
45 uint8_t *pix2, int i_stride_pix2 ) \
48 for( int y = 0; y < ly; y++ ) \
50 for( int x = 0; x < lx; x++ ) \
52 i_sum += abs( pix1[x] - pix2[x] ); \
54 pix1 += i_stride_pix1; \
55 pix2 += i_stride_pix2; \
61 PIXEL_SAD_C( x264_pixel_sad_16x16, 16, 16 )
62 PIXEL_SAD_C( x264_pixel_sad_16x8, 16, 8 )
63 PIXEL_SAD_C( x264_pixel_sad_8x16, 8, 16 )
64 PIXEL_SAD_C( x264_pixel_sad_8x8, 8, 8 )
65 PIXEL_SAD_C( x264_pixel_sad_8x4, 8, 4 )
66 PIXEL_SAD_C( x264_pixel_sad_4x8, 4, 8 )
67 PIXEL_SAD_C( x264_pixel_sad_4x4, 4, 4 )
70 /****************************************************************************
72 ****************************************************************************/
73 #define PIXEL_SSD_C( name, lx, ly ) \
74 static int name( uint8_t *pix1, int i_stride_pix1, \
75 uint8_t *pix2, int i_stride_pix2 ) \
78 for( int y = 0; y < ly; y++ ) \
80 for( int x = 0; x < lx; x++ ) \
82 int d = pix1[x] - pix2[x]; \
85 pix1 += i_stride_pix1; \
86 pix2 += i_stride_pix2; \
91 PIXEL_SSD_C( x264_pixel_ssd_16x16, 16, 16 )
92 PIXEL_SSD_C( x264_pixel_ssd_16x8, 16, 8 )
93 PIXEL_SSD_C( x264_pixel_ssd_8x16, 8, 16 )
94 PIXEL_SSD_C( x264_pixel_ssd_8x8, 8, 8 )
95 PIXEL_SSD_C( x264_pixel_ssd_8x4, 8, 4 )
96 PIXEL_SSD_C( x264_pixel_ssd_4x8, 4, 8 )
97 PIXEL_SSD_C( x264_pixel_ssd_4x4, 4, 4 )
99 int64_t x264_pixel_ssd_wxh( x264_pixel_function_t *pf, uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2, int i_width, int i_height )
103 int align = !(((intptr_t)pix1 | (intptr_t)pix2 | i_pix1 | i_pix2) & 15);
105 #define SSD(size) i_ssd += pf->ssd[size]( pix1 + y*i_pix1 + x, i_pix1, \
106 pix2 + y*i_pix2 + x, i_pix2 );
107 for( y = 0; y < i_height-15; y += 16 )
111 for( ; x < i_width-15; x += 16 )
113 for( ; x < i_width-7; x += 8 )
117 for( int x = 0; x < i_width-7; x += 8 )
121 #define SSD1 { int d = pix1[y*i_pix1+x] - pix2[y*i_pix2+x]; i_ssd += d*d; }
124 for( y = 0; y < (i_height & ~7); y++ )
125 for( int x = i_width & ~7; x < i_width; x++ )
130 for( y = i_height & ~7; y < i_height; y++ )
131 for( int x = 0; x < i_width; x++ )
140 /****************************************************************************
142 ****************************************************************************/
143 #define PIXEL_VAR_C( name, w ) \
144 static uint64_t name( uint8_t *pix, int i_stride ) \
146 uint32_t sum = 0, sqr = 0; \
147 for( int y = 0; y < w; y++ ) \
149 for( int x = 0; x < w; x++ ) \
152 sqr += pix[x] * pix[x]; \
156 return sum + ((uint64_t)sqr << 32); \
159 PIXEL_VAR_C( x264_pixel_var_16x16, 16 )
160 PIXEL_VAR_C( x264_pixel_var_8x8, 8 )
162 /****************************************************************************
164 ****************************************************************************/
165 static int pixel_var2_8x8( uint8_t *pix1, int i_stride1, uint8_t *pix2, int i_stride2, int *ssd )
167 uint32_t var = 0, sum = 0, sqr = 0;
168 for( int y = 0; y < 8; y++ )
170 for( int x = 0; x < 8; x++ )
172 int diff = pix1[x] - pix2[x];
180 var = sqr - (sum * sum >> 6);
186 #define HADAMARD4(d0, d1, d2, d3, s0, s1, s2, s3) {\
197 // in: a pseudo-simd number of the form x+(y<<16)
198 // return: abs(x)+(abs(y)<<16)
199 static ALWAYS_INLINE uint32_t abs2( uint32_t a )
201 uint32_t s = ((a>>15)&0x10001)*0xffff;
205 /****************************************************************************
206 * pixel_satd_WxH: sum of 4x4 Hadamard transformed differences
207 ****************************************************************************/
209 static NOINLINE int x264_pixel_satd_4x4( uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
212 uint32_t a0, a1, a2, a3, b0, b1;
214 for( int i = 0; i < 4; i++, pix1 += i_pix1, pix2 += i_pix2 )
216 a0 = pix1[0] - pix2[0];
217 a1 = pix1[1] - pix2[1];
218 b0 = (a0+a1) + ((a0-a1)<<16);
219 a2 = pix1[2] - pix2[2];
220 a3 = pix1[3] - pix2[3];
221 b1 = (a2+a3) + ((a2-a3)<<16);
225 for( int i = 0; i < 2; i++ )
227 HADAMARD4( a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i] );
228 a0 = abs2(a0) + abs2(a1) + abs2(a2) + abs2(a3);
229 sum += ((uint16_t)a0) + (a0>>16);
234 static NOINLINE int x264_pixel_satd_8x4( uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
237 uint32_t a0, a1, a2, a3;
239 for( int i = 0; i < 4; i++, pix1 += i_pix1, pix2 += i_pix2 )
241 a0 = (pix1[0] - pix2[0]) + ((pix1[4] - pix2[4]) << 16);
242 a1 = (pix1[1] - pix2[1]) + ((pix1[5] - pix2[5]) << 16);
243 a2 = (pix1[2] - pix2[2]) + ((pix1[6] - pix2[6]) << 16);
244 a3 = (pix1[3] - pix2[3]) + ((pix1[7] - pix2[7]) << 16);
245 HADAMARD4( tmp[i][0], tmp[i][1], tmp[i][2], tmp[i][3], a0,a1,a2,a3 );
247 for( int i = 0; i < 4; i++ )
249 HADAMARD4( a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i] );
250 sum += abs2(a0) + abs2(a1) + abs2(a2) + abs2(a3);
252 return (((uint16_t)sum) + ((uint32_t)sum>>16)) >> 1;
255 #define PIXEL_SATD_C( w, h, sub )\
256 static int x264_pixel_satd_##w##x##h( uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )\
258 int sum = sub( pix1, i_pix1, pix2, i_pix2 )\
259 + sub( pix1+4*i_pix1, i_pix1, pix2+4*i_pix2, i_pix2 );\
261 sum+= sub( pix1+8, i_pix1, pix2+8, i_pix2 )\
262 + sub( pix1+8+4*i_pix1, i_pix1, pix2+8+4*i_pix2, i_pix2 );\
264 sum+= sub( pix1+8*i_pix1, i_pix1, pix2+8*i_pix2, i_pix2 )\
265 + sub( pix1+12*i_pix1, i_pix1, pix2+12*i_pix2, i_pix2 );\
266 if( w==16 && h==16 )\
267 sum+= sub( pix1+8+8*i_pix1, i_pix1, pix2+8+8*i_pix2, i_pix2 )\
268 + sub( pix1+8+12*i_pix1, i_pix1, pix2+8+12*i_pix2, i_pix2 );\
271 PIXEL_SATD_C( 16, 16, x264_pixel_satd_8x4 )
272 PIXEL_SATD_C( 16, 8, x264_pixel_satd_8x4 )
273 PIXEL_SATD_C( 8, 16, x264_pixel_satd_8x4 )
274 PIXEL_SATD_C( 8, 8, x264_pixel_satd_8x4 )
275 PIXEL_SATD_C( 4, 8, x264_pixel_satd_4x4 )
278 static NOINLINE int sa8d_8x8( uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
281 uint32_t a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3;
283 for( int i = 0; i < 8; i++, pix1 += i_pix1, pix2 += i_pix2 )
285 a0 = pix1[0] - pix2[0];
286 a1 = pix1[1] - pix2[1];
287 b0 = (a0+a1) + ((a0-a1)<<16);
288 a2 = pix1[2] - pix2[2];
289 a3 = pix1[3] - pix2[3];
290 b1 = (a2+a3) + ((a2-a3)<<16);
291 a4 = pix1[4] - pix2[4];
292 a5 = pix1[5] - pix2[5];
293 b2 = (a4+a5) + ((a4-a5)<<16);
294 a6 = pix1[6] - pix2[6];
295 a7 = pix1[7] - pix2[7];
296 b3 = (a6+a7) + ((a6-a7)<<16);
297 HADAMARD4( tmp[i][0], tmp[i][1], tmp[i][2], tmp[i][3], b0,b1,b2,b3 );
299 for( int i = 0; i < 4; i++ )
301 HADAMARD4( a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i] );
302 HADAMARD4( a4, a5, a6, a7, tmp[4][i], tmp[5][i], tmp[6][i], tmp[7][i] );
303 b0 = abs2(a0+a4) + abs2(a0-a4);
304 b0 += abs2(a1+a5) + abs2(a1-a5);
305 b0 += abs2(a2+a6) + abs2(a2-a6);
306 b0 += abs2(a3+a7) + abs2(a3-a7);
307 sum += (uint16_t)b0 + (b0>>16);
312 static int x264_pixel_sa8d_8x8( uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
314 int sum = sa8d_8x8( pix1, i_pix1, pix2, i_pix2 );
318 static int x264_pixel_sa8d_16x16( uint8_t *pix1, int i_pix1, uint8_t *pix2, int i_pix2 )
320 int sum = sa8d_8x8( pix1, i_pix1, pix2, i_pix2 )
321 + sa8d_8x8( pix1+8, i_pix1, pix2+8, i_pix2 )
322 + sa8d_8x8( pix1+8*i_pix1, i_pix1, pix2+8*i_pix2, i_pix2 )
323 + sa8d_8x8( pix1+8+8*i_pix1, i_pix1, pix2+8+8*i_pix2, i_pix2 );
328 static NOINLINE uint64_t pixel_hadamard_ac( uint8_t *pix, int stride )
331 uint32_t a0, a1, a2, a3, dc;
332 int sum4 = 0, sum8 = 0;
333 for( int i = 0; i < 8; i++, pix+=stride )
335 uint32_t *t = tmp + (i&3) + (i&4)*4;
336 a0 = (pix[0]+pix[1]) + ((pix[0]-pix[1])<<16);
337 a1 = (pix[2]+pix[3]) + ((pix[2]-pix[3])<<16);
340 a2 = (pix[4]+pix[5]) + ((pix[4]-pix[5])<<16);
341 a3 = (pix[6]+pix[7]) + ((pix[6]-pix[7])<<16);
345 for( int i = 0; i < 8; i++ )
347 HADAMARD4( a0, a1, a2, a3, tmp[i*4+0], tmp[i*4+1], tmp[i*4+2], tmp[i*4+3] );
352 sum4 += abs2(a0) + abs2(a1) + abs2(a2) + abs2(a3);
354 for( int i = 0; i < 8; i++ )
356 HADAMARD4( a0,a1,a2,a3, tmp[i], tmp[8+i], tmp[16+i], tmp[24+i] );
357 sum8 += abs2(a0) + abs2(a1) + abs2(a2) + abs2(a3);
359 dc = (uint16_t)(tmp[0] + tmp[8] + tmp[16] + tmp[24]);
360 sum4 = (uint16_t)sum4 + ((uint32_t)sum4>>16) - dc;
361 sum8 = (uint16_t)sum8 + ((uint32_t)sum8>>16) - dc;
362 return ((uint64_t)sum8<<32) + sum4;
365 #define HADAMARD_AC(w,h) \
366 static uint64_t x264_pixel_hadamard_ac_##w##x##h( uint8_t *pix, int stride )\
368 uint64_t sum = pixel_hadamard_ac( pix, stride );\
370 sum += pixel_hadamard_ac( pix+8, stride );\
372 sum += pixel_hadamard_ac( pix+8*stride, stride );\
373 if( w==16 && h==16 )\
374 sum += pixel_hadamard_ac( pix+8*stride+8, stride );\
375 return ((sum>>34)<<32) + ((uint32_t)sum>>1);\
377 HADAMARD_AC( 16, 16 )
383 /****************************************************************************
385 ****************************************************************************/
386 #define SAD_X( size ) \
387 static void x264_pixel_sad_x3_##size( uint8_t *fenc, uint8_t *pix0, uint8_t *pix1, uint8_t *pix2, int i_stride, int scores[3] )\
389 scores[0] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix0, i_stride );\
390 scores[1] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix1, i_stride );\
391 scores[2] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix2, i_stride );\
393 static void x264_pixel_sad_x4_##size( uint8_t *fenc, uint8_t *pix0, uint8_t *pix1, uint8_t *pix2, uint8_t *pix3, int i_stride, int scores[4] )\
395 scores[0] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix0, i_stride );\
396 scores[1] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix1, i_stride );\
397 scores[2] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix2, i_stride );\
398 scores[3] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix3, i_stride );\
409 #ifdef ARCH_UltraSparc
416 /****************************************************************************
418 * no faster than single satd, but needed for satd to be a drop-in replacement for sad
419 ****************************************************************************/
421 #define SATD_X( size, cpu ) \
422 static void x264_pixel_satd_x3_##size##cpu( uint8_t *fenc, uint8_t *pix0, uint8_t *pix1, uint8_t *pix2, int i_stride, int scores[3] )\
424 scores[0] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix0, i_stride );\
425 scores[1] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix1, i_stride );\
426 scores[2] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix2, i_stride );\
428 static void x264_pixel_satd_x4_##size##cpu( uint8_t *fenc, uint8_t *pix0, uint8_t *pix1, uint8_t *pix2, uint8_t *pix3, int i_stride, int scores[4] )\
430 scores[0] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix0, i_stride );\
431 scores[1] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix1, i_stride );\
432 scores[2] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix2, i_stride );\
433 scores[3] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix3, i_stride );\
435 #define SATD_X_DECL6( cpu )\
436 SATD_X( 16x16, cpu )\
442 #define SATD_X_DECL7( cpu )\
448 SATD_X_DECL7( _mmxext )
449 SATD_X_DECL6( _sse2 )
450 SATD_X_DECL7( _ssse3 )
451 SATD_X_DECL7( _sse4 )
455 SATD_X_DECL7( _neon )
458 /****************************************************************************
459 * structural similarity metric
460 ****************************************************************************/
461 static void ssim_4x4x2_core( const uint8_t *pix1, int stride1,
462 const uint8_t *pix2, int stride2,
465 for( int z = 0; z < 2; z++ )
467 uint32_t s1 = 0, s2 = 0, ss = 0, s12 = 0;
468 for( int y = 0; y < 4; y++ )
469 for( int x = 0; x < 4; x++ )
471 int a = pix1[x+y*stride1];
472 int b = pix2[x+y*stride2];
488 static float ssim_end1( int s1, int s2, int ss, int s12 )
490 static const int ssim_c1 = (int)(.01*.01*255*255*64 + .5);
491 static const int ssim_c2 = (int)(.03*.03*255*255*64*63 + .5);
492 int vars = ss*64 - s1*s1 - s2*s2;
493 int covar = s12*64 - s1*s2;
494 return (float)(2*s1*s2 + ssim_c1) * (float)(2*covar + ssim_c2)
495 / ((float)(s1*s1 + s2*s2 + ssim_c1) * (float)(vars + ssim_c2));
498 static float ssim_end4( int sum0[5][4], int sum1[5][4], int width )
501 for( int i = 0; i < width; i++ )
502 ssim += ssim_end1( sum0[i][0] + sum0[i+1][0] + sum1[i][0] + sum1[i+1][0],
503 sum0[i][1] + sum0[i+1][1] + sum1[i][1] + sum1[i+1][1],
504 sum0[i][2] + sum0[i+1][2] + sum1[i][2] + sum1[i+1][2],
505 sum0[i][3] + sum0[i+1][3] + sum1[i][3] + sum1[i+1][3] );
509 float x264_pixel_ssim_wxh( x264_pixel_function_t *pf,
510 uint8_t *pix1, int stride1,
511 uint8_t *pix2, int stride2,
512 int width, int height, void *buf )
516 int (*sum0)[4] = buf;
517 int (*sum1)[4] = sum0 + width/4+3;
520 for( int y = 1; y < height; y++ )
524 XCHG( void*, sum0, sum1 );
525 for( int x = 0; x < width; x+=2 )
526 pf->ssim_4x4x2_core( &pix1[4*(x+z*stride1)], stride1, &pix2[4*(x+z*stride2)], stride2, &sum0[x] );
528 for( int x = 0; x < width-1; x += 4 )
529 ssim += pf->ssim_end4( sum0+x, sum1+x, X264_MIN(4,width-x-1) );
535 /****************************************************************************
536 * successive elimination
537 ****************************************************************************/
538 static int x264_pixel_ads4( int enc_dc[4], uint16_t *sums, int delta,
539 uint16_t *cost_mvx, int16_t *mvs, int width, int thresh )
542 for( int i = 0; i < width; i++, sums++ )
544 int ads = abs( enc_dc[0] - sums[0] )
545 + abs( enc_dc[1] - sums[8] )
546 + abs( enc_dc[2] - sums[delta] )
547 + abs( enc_dc[3] - sums[delta+8] )
555 static int x264_pixel_ads2( int enc_dc[2], uint16_t *sums, int delta,
556 uint16_t *cost_mvx, int16_t *mvs, int width, int thresh )
559 for( int i = 0; i < width; i++, sums++ )
561 int ads = abs( enc_dc[0] - sums[0] )
562 + abs( enc_dc[1] - sums[delta] )
570 static int x264_pixel_ads1( int enc_dc[1], uint16_t *sums, int delta,
571 uint16_t *cost_mvx, int16_t *mvs, int width, int thresh )
574 for( int i = 0; i<width; i++, sums++ )
576 int ads = abs( enc_dc[0] - sums[0] )
585 /****************************************************************************
587 ****************************************************************************/
588 void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
590 memset( pixf, 0, sizeof(*pixf) );
592 #define INIT2_NAME( name1, name2, cpu ) \
593 pixf->name1[PIXEL_16x16] = x264_pixel_##name2##_16x16##cpu;\
594 pixf->name1[PIXEL_16x8] = x264_pixel_##name2##_16x8##cpu;
595 #define INIT4_NAME( name1, name2, cpu ) \
596 INIT2_NAME( name1, name2, cpu ) \
597 pixf->name1[PIXEL_8x16] = x264_pixel_##name2##_8x16##cpu;\
598 pixf->name1[PIXEL_8x8] = x264_pixel_##name2##_8x8##cpu;
599 #define INIT5_NAME( name1, name2, cpu ) \
600 INIT4_NAME( name1, name2, cpu ) \
601 pixf->name1[PIXEL_8x4] = x264_pixel_##name2##_8x4##cpu;
602 #define INIT6_NAME( name1, name2, cpu ) \
603 INIT5_NAME( name1, name2, cpu ) \
604 pixf->name1[PIXEL_4x8] = x264_pixel_##name2##_4x8##cpu;
605 #define INIT7_NAME( name1, name2, cpu ) \
606 INIT6_NAME( name1, name2, cpu ) \
607 pixf->name1[PIXEL_4x4] = x264_pixel_##name2##_4x4##cpu;
608 #define INIT2( name, cpu ) INIT2_NAME( name, name, cpu )
609 #define INIT4( name, cpu ) INIT4_NAME( name, name, cpu )
610 #define INIT5( name, cpu ) INIT5_NAME( name, name, cpu )
611 #define INIT6( name, cpu ) INIT6_NAME( name, name, cpu )
612 #define INIT7( name, cpu ) INIT7_NAME( name, name, cpu )
614 #define INIT_ADS( cpu ) \
615 pixf->ads[PIXEL_16x16] = x264_pixel_ads4##cpu;\
616 pixf->ads[PIXEL_16x8] = x264_pixel_ads2##cpu;\
617 pixf->ads[PIXEL_8x8] = x264_pixel_ads1##cpu;
620 INIT7_NAME( sad_aligned, sad, );
627 INIT4( hadamard_ac, );
630 pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16;
631 pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8;
632 pixf->var[PIXEL_16x16] = x264_pixel_var_16x16;
633 pixf->var[PIXEL_8x8] = x264_pixel_var_8x8;
635 pixf->ssim_4x4x2_core = ssim_4x4x2_core;
636 pixf->ssim_end4 = ssim_end4;
637 pixf->var2_8x8 = pixel_var2_8x8;
640 if( cpu&X264_CPU_MMX )
645 if( cpu&X264_CPU_MMXEXT )
647 INIT7( sad, _mmxext );
648 INIT7_NAME( sad_aligned, sad, _mmxext );
649 INIT7( sad_x3, _mmxext );
650 INIT7( sad_x4, _mmxext );
651 INIT7( satd, _mmxext );
652 INIT7( satd_x3, _mmxext );
653 INIT7( satd_x4, _mmxext );
654 INIT4( hadamard_ac, _mmxext );
656 pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmxext;
657 pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_mmxext;
659 pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_mmxext;
660 pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_mmxext;
661 pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_mmxext;
662 pixf->ssim_4x4x2_core = x264_pixel_ssim_4x4x2_core_mmxext;
663 pixf->var2_8x8 = x264_pixel_var2_8x8_mmxext;
665 if( cpu&X264_CPU_CACHELINE_32 )
667 INIT5( sad, _cache32_mmxext );
668 INIT4( sad_x3, _cache32_mmxext );
669 INIT4( sad_x4, _cache32_mmxext );
671 else if( cpu&X264_CPU_CACHELINE_64 )
673 INIT5( sad, _cache64_mmxext );
674 INIT4( sad_x3, _cache64_mmxext );
675 INIT4( sad_x4, _cache64_mmxext );
678 if( cpu&X264_CPU_CACHELINE_64 )
680 pixf->sad[PIXEL_8x16] = x264_pixel_sad_8x16_cache64_mmxext;
681 pixf->sad[PIXEL_8x8] = x264_pixel_sad_8x8_cache64_mmxext;
682 pixf->sad[PIXEL_8x4] = x264_pixel_sad_8x4_cache64_mmxext;
683 pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_cache64_mmxext;
684 pixf->sad_x3[PIXEL_8x8] = x264_pixel_sad_x3_8x8_cache64_mmxext;
685 pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_mmxext;
686 pixf->sad_x4[PIXEL_8x8] = x264_pixel_sad_x4_8x8_cache64_mmxext;
689 pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmxext;
690 pixf->intra_sad_x3_16x16 = x264_intra_sad_x3_16x16_mmxext;
691 pixf->intra_satd_x3_8x8c = x264_intra_satd_x3_8x8c_mmxext;
692 pixf->intra_sad_x3_8x8c = x264_intra_sad_x3_8x8c_mmxext;
693 pixf->intra_sad_x3_8x8 = x264_intra_sad_x3_8x8_mmxext;
694 pixf->intra_satd_x3_4x4 = x264_intra_satd_x3_4x4_mmxext;
695 pixf->intra_sad_x3_4x4 = x264_intra_sad_x3_4x4_mmxext;
698 if( cpu&X264_CPU_SSE2 )
700 INIT5( ssd, _sse2slow );
701 INIT2_NAME( sad_aligned, sad, _sse2_aligned );
702 pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_sse2;
703 pixf->ssim_4x4x2_core = x264_pixel_ssim_4x4x2_core_sse2;
704 pixf->ssim_end4 = x264_pixel_ssim_end4_sse2;
705 pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_sse2;
706 pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_sse2;
708 pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_sse2;
710 pixf->var2_8x8 = x264_pixel_var2_8x8_sse2;
713 if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_SSE2_IS_SLOW) )
716 INIT2( sad_x3, _sse2 );
717 INIT2( sad_x4, _sse2 );
718 INIT6( satd, _sse2 );
719 INIT6( satd_x3, _sse2 );
720 INIT6( satd_x4, _sse2 );
721 if( !(cpu&X264_CPU_STACK_MOD4) )
723 INIT4( hadamard_ac, _sse2 );
726 pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_sse2;
727 pixf->intra_sad_x3_16x16 = x264_intra_sad_x3_16x16_sse2;
728 if( cpu&X264_CPU_CACHELINE_64 )
730 INIT2( ssd, _sse2); /* faster for width 16 on p4 */
732 INIT2( sad, _cache64_sse2 );
733 INIT2( sad_x3, _cache64_sse2 );
734 INIT2( sad_x4, _cache64_sse2 );
736 if( cpu&X264_CPU_SSE2_IS_FAST )
738 pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_cache64_sse2;
739 pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_sse2;
743 if( cpu&X264_CPU_SSE_MISALIGN )
745 INIT2( sad_x3, _sse2_misalign );
746 INIT2( sad_x4, _sse2_misalign );
750 if( cpu&X264_CPU_SSE2_IS_FAST && !(cpu&X264_CPU_CACHELINE_64) )
752 pixf->sad_aligned[PIXEL_8x16] = x264_pixel_sad_8x16_sse2;
753 pixf->sad[PIXEL_8x16] = x264_pixel_sad_8x16_sse2;
754 pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_sse2;
755 pixf->sad_x3[PIXEL_8x8] = x264_pixel_sad_x3_8x8_sse2;
756 pixf->sad_x3[PIXEL_8x4] = x264_pixel_sad_x3_8x4_sse2;
757 pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_sse2;
758 pixf->sad_x4[PIXEL_8x8] = x264_pixel_sad_x4_8x8_sse2;
759 pixf->sad_x4[PIXEL_8x4] = x264_pixel_sad_x4_8x4_sse2;
762 if( (cpu&X264_CPU_SSE3) && (cpu&X264_CPU_CACHELINE_64) )
765 INIT2( sad_x3, _sse3 );
766 INIT2( sad_x4, _sse3 );
769 if( cpu&X264_CPU_SSSE3 )
771 INIT7( ssd, _ssse3 );
772 INIT7( satd, _ssse3 );
773 INIT7( satd_x3, _ssse3 );
774 INIT7( satd_x4, _ssse3 );
775 if( !(cpu&X264_CPU_STACK_MOD4) )
777 INIT4( hadamard_ac, _ssse3 );
780 pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_ssse3;
781 pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_ssse3;
782 pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_ssse3;
783 pixf->intra_sad_x3_16x16 = x264_intra_sad_x3_16x16_ssse3;
784 pixf->intra_satd_x3_8x8c = x264_intra_satd_x3_8x8c_ssse3;
785 pixf->intra_sad_x3_8x8c = x264_intra_sad_x3_8x8c_ssse3;
786 pixf->intra_satd_x3_4x4 = x264_intra_satd_x3_4x4_ssse3;
788 pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_ssse3;
790 pixf->var2_8x8 = x264_pixel_var2_8x8_ssse3;
791 if( cpu&X264_CPU_CACHELINE_64 )
793 INIT2( sad, _cache64_ssse3 );
794 INIT2( sad_x3, _cache64_ssse3 );
795 INIT2( sad_x4, _cache64_ssse3 );
797 if( !(cpu&X264_CPU_SHUFFLE_IS_FAST) )
799 INIT5( ssd, _sse2 ); /* on conroe, sse2 is faster for width8/16 */
803 if( cpu&X264_CPU_SSE4 )
805 INIT7( satd, _sse4 );
806 INIT7( satd_x3, _sse4 );
807 INIT7( satd_x4, _sse4 );
808 if( !(cpu&X264_CPU_STACK_MOD4) )
810 INIT4( hadamard_ac, _sse4 );
812 pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_sse4;
813 pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_sse4;
818 if( cpu&X264_CPU_ARMV6 )
820 pixf->sad[PIXEL_4x8] = x264_pixel_sad_4x8_armv6;
821 pixf->sad[PIXEL_4x4] = x264_pixel_sad_4x4_armv6;
822 pixf->sad_aligned[PIXEL_4x8] = x264_pixel_sad_4x8_armv6;
823 pixf->sad_aligned[PIXEL_4x4] = x264_pixel_sad_4x4_armv6;
825 if( cpu&X264_CPU_NEON )
828 INIT5( sad_aligned, _neon );
829 INIT7( sad_x3, _neon );
830 INIT7( sad_x4, _neon );
832 INIT7( satd, _neon );
833 INIT7( satd_x3, _neon );
834 INIT7( satd_x4, _neon );
835 INIT4( hadamard_ac, _neon );
836 pixf->sa8d[PIXEL_8x8] = x264_pixel_sa8d_8x8_neon;
837 pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_neon;
838 pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_neon;
839 pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_neon;
840 pixf->var2_8x8 = x264_pixel_var2_8x8_neon;
842 pixf->ssim_4x4x2_core = x264_pixel_ssim_4x4x2_core_neon;
843 pixf->ssim_end4 = x264_pixel_ssim_end4_neon;
845 if( cpu&X264_CPU_FAST_NEON_MRC )
847 pixf->sad[PIXEL_4x8] = x264_pixel_sad_4x8_neon;
848 pixf->sad[PIXEL_4x4] = x264_pixel_sad_4x4_neon;
849 pixf->sad_aligned[PIXEL_4x8] = x264_pixel_sad_aligned_4x8_neon;
850 pixf->sad_aligned[PIXEL_4x4] = x264_pixel_sad_aligned_4x4_neon;
852 else // really just scheduled for dual issue / A8
854 INIT5( sad_aligned, _neon_dual );
859 if( cpu&X264_CPU_ALTIVEC )
861 x264_pixel_altivec_init( pixf );
864 #ifdef ARCH_UltraSparc
866 INIT4( sad_x3, _vis );
867 INIT4( sad_x4, _vis );
870 pixf->ads[PIXEL_8x16] =
871 pixf->ads[PIXEL_8x4] =
872 pixf->ads[PIXEL_4x8] = pixf->ads[PIXEL_16x8];
873 pixf->ads[PIXEL_4x4] = pixf->ads[PIXEL_8x8];