pixel.c 33.6 KB
Newer Older
Laurent Aimar's avatar
Laurent Aimar committed
1
/*****************************************************************************
Fiona Glaser's avatar
Fiona Glaser committed
2
 * pixel.c: pixel metrics
Laurent Aimar's avatar
Laurent Aimar committed
3
 *****************************************************************************
Fiona Glaser's avatar
Fiona Glaser committed
4
 * Copyright (C) 2003-2010 x264 project
Laurent Aimar's avatar
Laurent Aimar committed
5
 *
6 7
 * Authors: Loren Merritt <lorenm@u.washington.edu>
 *          Laurent Aimar <fenrir@via.ecp.fr>
Fiona Glaser's avatar
Fiona Glaser committed
8
 *          Fiona Glaser <fiona@x264.com>
Laurent Aimar's avatar
Laurent Aimar committed
9 10 11 12 13 14 15 16 17 18 19 20 21
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
22
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
Fiona Glaser's avatar
Fiona Glaser committed
23 24 25
 *
 * This program is also available under a commercial proprietary license.
 * For more information, contact us at licensing@x264.com.
Laurent Aimar's avatar
Laurent Aimar committed
26 27
 *****************************************************************************/

Loren Merritt's avatar
Loren Merritt committed
28
#include "common.h"
Laurent Aimar's avatar
Laurent Aimar committed
29

Steven Walters's avatar
Steven Walters committed
30
#if HAVE_MMX
31
#   include "x86/pixel.h"
Laurent Aimar's avatar
Laurent Aimar committed
32
#endif
Steven Walters's avatar
Steven Walters committed
33
#if ARCH_PPC
Laurent Aimar's avatar
Laurent Aimar committed
34 35
#   include "ppc/pixel.h"
#endif
Steven Walters's avatar
Steven Walters committed
36
#if ARCH_ARM
37 38
#   include "arm/pixel.h"
#endif
Steven Walters's avatar
Steven Walters committed
39
#if ARCH_UltraSparc
40 41
#   include "sparc/pixel.h"
#endif
Laurent Aimar's avatar
Laurent Aimar committed
42 43 44 45 46 47


/****************************************************************************
 * pixel_sad_WxH
 ****************************************************************************/
#define PIXEL_SAD_C( name, lx, ly ) \
48 49
static int name( pixel *pix1, int i_stride_pix1,  \
                 pixel *pix2, int i_stride_pix2 ) \
Laurent Aimar's avatar
Laurent Aimar committed
50 51
{                                                   \
    int i_sum = 0;                                  \
52
    for( int y = 0; y < ly; y++ )                   \
Laurent Aimar's avatar
Laurent Aimar committed
53
    {                                               \
54
        for( int x = 0; x < lx; x++ )               \
Laurent Aimar's avatar
Laurent Aimar committed
55 56 57 58 59 60 61 62 63 64
        {                                           \
            i_sum += abs( pix1[x] - pix2[x] );      \
        }                                           \
        pix1 += i_stride_pix1;                      \
        pix2 += i_stride_pix2;                      \
    }                                               \
    return i_sum;                                   \
}


65 66 67 68 69 70 71
PIXEL_SAD_C( x264_pixel_sad_16x16, 16, 16 )
PIXEL_SAD_C( x264_pixel_sad_16x8,  16,  8 )
PIXEL_SAD_C( x264_pixel_sad_8x16,   8, 16 )
PIXEL_SAD_C( x264_pixel_sad_8x8,    8,  8 )
PIXEL_SAD_C( x264_pixel_sad_8x4,    8,  4 )
PIXEL_SAD_C( x264_pixel_sad_4x8,    4,  8 )
PIXEL_SAD_C( x264_pixel_sad_4x4,    4,  4 )
Laurent Aimar's avatar
Laurent Aimar committed
72

73 74 75 76 77

/****************************************************************************
 * pixel_ssd_WxH
 ****************************************************************************/
#define PIXEL_SSD_C( name, lx, ly ) \
78 79
static int name( pixel *pix1, int i_stride_pix1,  \
                 pixel *pix2, int i_stride_pix2 ) \
80 81
{                                                   \
    int i_sum = 0;                                  \
82
    for( int y = 0; y < ly; y++ )                   \
83
    {                                               \
84
        for( int x = 0; x < lx; x++ )               \
85 86 87 88 89 90 91 92 93 94
        {                                           \
            int d = pix1[x] - pix2[x];              \
            i_sum += d*d;                           \
        }                                           \
        pix1 += i_stride_pix1;                      \
        pix2 += i_stride_pix2;                      \
    }                                               \
    return i_sum;                                   \
}

95 96 97 98 99 100 101
PIXEL_SSD_C( x264_pixel_ssd_16x16, 16, 16 )
PIXEL_SSD_C( x264_pixel_ssd_16x8,  16,  8 )
PIXEL_SSD_C( x264_pixel_ssd_8x16,   8, 16 )
PIXEL_SSD_C( x264_pixel_ssd_8x8,    8,  8 )
PIXEL_SSD_C( x264_pixel_ssd_8x4,    8,  4 )
PIXEL_SSD_C( x264_pixel_ssd_4x8,    4,  8 )
PIXEL_SSD_C( x264_pixel_ssd_4x4,    4,  4 )
102

103
uint64_t x264_pixel_ssd_wxh( x264_pixel_function_t *pf, pixel *pix1, int i_pix1, pixel *pix2, int i_pix2, int i_width, int i_height )
Loren Merritt's avatar
Loren Merritt committed
104
{
105
    uint64_t i_ssd = 0;
106
    int y;
Anton Mitrofanov's avatar
Anton Mitrofanov committed
107
    int align = !(((intptr_t)pix1 | (intptr_t)pix2 | i_pix1 | i_pix2) & 15);
Loren Merritt's avatar
Loren Merritt committed
108 109 110 111 112

#define SSD(size) i_ssd += pf->ssd[size]( pix1 + y*i_pix1 + x, i_pix1, \
                                          pix2 + y*i_pix2 + x, i_pix2 );
    for( y = 0; y < i_height-15; y += 16 )
    {
113
        int x = 0;
Loren Merritt's avatar
Loren Merritt committed
114 115 116 117
        if( align )
            for( ; x < i_width-15; x += 16 )
                SSD(PIXEL_16x16);
        for( ; x < i_width-7; x += 8 )
Loren Merritt's avatar
Loren Merritt committed
118 119 120
            SSD(PIXEL_8x16);
    }
    if( y < i_height-7 )
121
        for( int x = 0; x < i_width-7; x += 8 )
Loren Merritt's avatar
Loren Merritt committed
122 123 124 125
            SSD(PIXEL_8x8);
#undef SSD

#define SSD1 { int d = pix1[y*i_pix1+x] - pix2[y*i_pix2+x]; i_ssd += d*d; }
126
    if( i_width & 7 )
Loren Merritt's avatar
Loren Merritt committed
127 128
    {
        for( y = 0; y < (i_height & ~7); y++ )
129
            for( int x = i_width & ~7; x < i_width; x++ )
Loren Merritt's avatar
Loren Merritt committed
130 131
                SSD1;
    }
132
    if( i_height & 7 )
Loren Merritt's avatar
Loren Merritt committed
133 134
    {
        for( y = i_height & ~7; y < i_height; y++ )
135
            for( int x = 0; x < i_width; x++ )
Loren Merritt's avatar
Loren Merritt committed
136 137 138 139 140 141 142
                SSD1;
    }
#undef SSD1

    return i_ssd;
}

143
static void pixel_ssd_nv12_core( pixel *pixuv1, int stride1, pixel *pixuv2, int stride2, int width, int height, uint64_t *ssd_u, uint64_t *ssd_v )
144
{
145
    *ssd_u = 0, *ssd_v = 0;
146 147 148 149 150
    for( int y = 0; y < height; y++, pixuv1+=stride1, pixuv2+=stride2 )
        for( int x = 0; x < width; x++ )
        {
            int du = pixuv1[2*x]   - pixuv2[2*x];
            int dv = pixuv1[2*x+1] - pixuv2[2*x+1];
151 152
            *ssd_u += du*du;
            *ssd_v += dv*dv;
153 154 155
        }
}

156
void x264_pixel_ssd_nv12( x264_pixel_function_t *pf, pixel *pix1, int i_pix1, pixel *pix2, int i_pix2, int i_width, int i_height, uint64_t *ssd_u, uint64_t *ssd_v )
157
{
158
    pf->ssd_nv12_core( pix1, i_pix1, pix2, i_pix2, i_width&~7, i_height, ssd_u, ssd_v );
159
    if( i_width&7 )
160 161 162 163 164 165
    {
        uint64_t tmp[2];
        pixel_ssd_nv12_core( pix1+(i_width&~7), i_pix1, pix2+(i_width&~7), i_pix2, i_width&7, i_height, &tmp[0], &tmp[1] );
        *ssd_u += tmp[0];
        *ssd_v += tmp[1];
    }
166
}
167

168 169 170
/****************************************************************************
 * pixel_var_wxh
 ****************************************************************************/
Fiona Glaser's avatar
Fiona Glaser committed
171
#define PIXEL_VAR_C( name, w ) \
172
static uint64_t name( pixel *pix, int i_stride ) \
173
{                                             \
Fiona Glaser's avatar
Fiona Glaser committed
174
    uint32_t sum = 0, sqr = 0;                \
175
    for( int y = 0; y < w; y++ )              \
176
    {                                         \
177
        for( int x = 0; x < w; x++ )          \
178 179 180 181 182 183
        {                                     \
            sum += pix[x];                    \
            sqr += pix[x] * pix[x];           \
        }                                     \
        pix += i_stride;                      \
    }                                         \
Fiona Glaser's avatar
Fiona Glaser committed
184
    return sum + ((uint64_t)sqr << 32);       \
185 186
}

Fiona Glaser's avatar
Fiona Glaser committed
187 188
PIXEL_VAR_C( x264_pixel_var_16x16, 16 )
PIXEL_VAR_C( x264_pixel_var_8x8,    8 )
189

190 191 192
/****************************************************************************
 * pixel_var2_wxh
 ****************************************************************************/
193
static int pixel_var2_8x8( pixel *pix1, int i_stride1, pixel *pix2, int i_stride2, int *ssd )
194 195
{
    uint32_t var = 0, sum = 0, sqr = 0;
196
    for( int y = 0; y < 8; y++ )
197
    {
198
        for( int x = 0; x < 8; x++ )
199 200 201 202 203 204 205 206 207
        {
            int diff = pix1[x] - pix2[x];
            sum += diff;
            sqr += diff * diff;
        }
        pix1 += i_stride1;
        pix2 += i_stride2;
    }
    sum = abs(sum);
208
    var = sqr - ((uint64_t)sum * sum >> 6);
209 210 211 212
    *ssd = sqr;
    return var;
}

213

214
#define HADAMARD4(d0, d1, d2, d3, s0, s1, s2, s3) {\
Loren Merritt's avatar
Loren Merritt committed
215 216 217 218 219 220 221 222 223
    int t0 = s0 + s1;\
    int t1 = s0 - s1;\
    int t2 = s2 + s3;\
    int t3 = s2 - s3;\
    d0 = t0 + t2;\
    d2 = t0 - t2;\
    d1 = t1 + t3;\
    d3 = t1 - t3;\
}
224

225 226 227 228 229 230 231 232
// in: a pseudo-simd number of the form x+(y<<16)
// return: abs(x)+(abs(y)<<16)
static ALWAYS_INLINE uint32_t abs2( uint32_t a )
{
    uint32_t s = ((a>>15)&0x10001)*0xffff;
    return (a+s)^s;
}

233 234 235
/****************************************************************************
 * pixel_satd_WxH: sum of 4x4 Hadamard transformed differences
 ****************************************************************************/
Laurent Aimar's avatar
Laurent Aimar committed
236

237
static NOINLINE int x264_pixel_satd_4x4( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
238 239
{
    uint32_t tmp[4][2];
240 241 242
    uint32_t a0, a1, a2, a3, b0, b1;
    int sum = 0;
    for( int i = 0; i < 4; i++, pix1 += i_pix1, pix2 += i_pix2 )
Laurent Aimar's avatar
Laurent Aimar committed
243
    {
244 245 246 247 248 249 250 251
        a0 = pix1[0] - pix2[0];
        a1 = pix1[1] - pix2[1];
        b0 = (a0+a1) + ((a0-a1)<<16);
        a2 = pix1[2] - pix2[2];
        a3 = pix1[3] - pix2[3];
        b1 = (a2+a3) + ((a2-a3)<<16);
        tmp[i][0] = b0 + b1;
        tmp[i][1] = b0 - b1;
Laurent Aimar's avatar
Laurent Aimar committed
252
    }
253
    for( int i = 0; i < 2; i++ )
254
    {
255
        HADAMARD4( a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i] );
256 257 258 259
        a0 = abs2(a0) + abs2(a1) + abs2(a2) + abs2(a3);
        sum += ((uint16_t)a0) + (a0>>16);
    }
    return sum >> 1;
Laurent Aimar's avatar
Laurent Aimar committed
260 261
}

262
static NOINLINE int x264_pixel_satd_8x4( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
263 264
{
    uint32_t tmp[4][4];
265 266 267
    uint32_t a0, a1, a2, a3;
    int sum = 0;
    for( int i = 0; i < 4; i++, pix1 += i_pix1, pix2 += i_pix2 )
268 269 270 271 272 273 274
    {
        a0 = (pix1[0] - pix2[0]) + ((pix1[4] - pix2[4]) << 16);
        a1 = (pix1[1] - pix2[1]) + ((pix1[5] - pix2[5]) << 16);
        a2 = (pix1[2] - pix2[2]) + ((pix1[6] - pix2[6]) << 16);
        a3 = (pix1[3] - pix2[3]) + ((pix1[7] - pix2[7]) << 16);
        HADAMARD4( tmp[i][0], tmp[i][1], tmp[i][2], tmp[i][3], a0,a1,a2,a3 );
    }
275
    for( int i = 0; i < 4; i++ )
276
    {
277
        HADAMARD4( a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i] );
278 279 280 281
        sum += abs2(a0) + abs2(a1) + abs2(a2) + abs2(a3);
    }
    return (((uint16_t)sum) + ((uint32_t)sum>>16)) >> 1;
}
Laurent Aimar's avatar
Laurent Aimar committed
282

283
#define PIXEL_SATD_C( w, h, sub )\
284
static int x264_pixel_satd_##w##x##h( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )\
285 286 287 288 289 290 291 292 293 294 295 296 297
{\
    int sum = sub( pix1, i_pix1, pix2, i_pix2 )\
            + sub( pix1+4*i_pix1, i_pix1, pix2+4*i_pix2, i_pix2 );\
    if( w==16 )\
        sum+= sub( pix1+8, i_pix1, pix2+8, i_pix2 )\
            + sub( pix1+8+4*i_pix1, i_pix1, pix2+8+4*i_pix2, i_pix2 );\
    if( h==16 )\
        sum+= sub( pix1+8*i_pix1, i_pix1, pix2+8*i_pix2, i_pix2 )\
            + sub( pix1+12*i_pix1, i_pix1, pix2+12*i_pix2, i_pix2 );\
    if( w==16 && h==16 )\
        sum+= sub( pix1+8+8*i_pix1, i_pix1, pix2+8+8*i_pix2, i_pix2 )\
            + sub( pix1+8+12*i_pix1, i_pix1, pix2+8+12*i_pix2, i_pix2 );\
    return sum;\
298
}
299 300 301 302 303
PIXEL_SATD_C( 16, 16, x264_pixel_satd_8x4 )
PIXEL_SATD_C( 16, 8,  x264_pixel_satd_8x4 )
PIXEL_SATD_C( 8,  16, x264_pixel_satd_8x4 )
PIXEL_SATD_C( 8,  8,  x264_pixel_satd_8x4 )
PIXEL_SATD_C( 4,  8,  x264_pixel_satd_4x4 )
304 305


306
static NOINLINE int sa8d_8x8( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
307 308
{
    uint32_t tmp[8][4];
309 310 311
    uint32_t a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3;
    int sum = 0;
    for( int i = 0; i < 8; i++, pix1 += i_pix1, pix2 += i_pix2 )
312
    {
313 314 315 316 317 318 319 320 321 322 323 324 325
        a0 = pix1[0] - pix2[0];
        a1 = pix1[1] - pix2[1];
        b0 = (a0+a1) + ((a0-a1)<<16);
        a2 = pix1[2] - pix2[2];
        a3 = pix1[3] - pix2[3];
        b1 = (a2+a3) + ((a2-a3)<<16);
        a4 = pix1[4] - pix2[4];
        a5 = pix1[5] - pix2[5];
        b2 = (a4+a5) + ((a4-a5)<<16);
        a6 = pix1[6] - pix2[6];
        a7 = pix1[7] - pix2[7];
        b3 = (a6+a7) + ((a6-a7)<<16);
        HADAMARD4( tmp[i][0], tmp[i][1], tmp[i][2], tmp[i][3], b0,b1,b2,b3 );
326
    }
327
    for( int i = 0; i < 4; i++ )
328
    {
329 330
        HADAMARD4( a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i] );
        HADAMARD4( a4, a5, a6, a7, tmp[4][i], tmp[5][i], tmp[6][i], tmp[7][i] );
331 332 333 334 335 336 337 338
        b0  = abs2(a0+a4) + abs2(a0-a4);
        b0 += abs2(a1+a5) + abs2(a1-a5);
        b0 += abs2(a2+a6) + abs2(a2-a6);
        b0 += abs2(a3+a7) + abs2(a3-a7);
        sum += (uint16_t)b0 + (b0>>16);
    }
    return sum;
}
339

340
static int x264_pixel_sa8d_8x8( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
341 342 343
{
    int sum = sa8d_8x8( pix1, i_pix1, pix2, i_pix2 );
    return (sum+2)>>2;
344 345
}

346
static int x264_pixel_sa8d_16x16( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
347 348 349 350 351 352
{
    int sum = sa8d_8x8( pix1, i_pix1, pix2, i_pix2 )
            + sa8d_8x8( pix1+8, i_pix1, pix2+8, i_pix2 )
            + sa8d_8x8( pix1+8*i_pix1, i_pix1, pix2+8*i_pix2, i_pix2 )
            + sa8d_8x8( pix1+8+8*i_pix1, i_pix1, pix2+8+8*i_pix2, i_pix2 );
    return (sum+2)>>2;
353 354
}

Loren Merritt's avatar
Loren Merritt committed
355

356
static NOINLINE uint64_t pixel_hadamard_ac( pixel *pix, int stride )
Loren Merritt's avatar
Loren Merritt committed
357
{
358
    uint32_t tmp[32];
359 360 361
    uint32_t a0, a1, a2, a3, dc;
    int sum4 = 0, sum8 = 0;
    for( int i = 0; i < 8; i++, pix+=stride )
Loren Merritt's avatar
Loren Merritt committed
362
    {
363 364 365 366 367 368 369 370 371
        uint32_t *t = tmp + (i&3) + (i&4)*4;
        a0 = (pix[0]+pix[1]) + ((pix[0]-pix[1])<<16);
        a1 = (pix[2]+pix[3]) + ((pix[2]-pix[3])<<16);
        t[0] = a0 + a1;
        t[4] = a0 - a1;
        a2 = (pix[4]+pix[5]) + ((pix[4]-pix[5])<<16);
        a3 = (pix[6]+pix[7]) + ((pix[6]-pix[7])<<16);
        t[8] = a2 + a3;
        t[12] = a2 - a3;
Loren Merritt's avatar
Loren Merritt committed
372
    }
373
    for( int i = 0; i < 8; i++ )
Loren Merritt's avatar
Loren Merritt committed
374
    {
375
        HADAMARD4( a0, a1, a2, a3, tmp[i*4+0], tmp[i*4+1], tmp[i*4+2], tmp[i*4+3] );
376 377 378 379 380
        tmp[i*4+0] = a0;
        tmp[i*4+1] = a1;
        tmp[i*4+2] = a2;
        tmp[i*4+3] = a3;
        sum4 += abs2(a0) + abs2(a1) + abs2(a2) + abs2(a3);
Loren Merritt's avatar
Loren Merritt committed
381
    }
382
    for( int i = 0; i < 8; i++ )
Loren Merritt's avatar
Loren Merritt committed
383
    {
384 385
        HADAMARD4( a0,a1,a2,a3, tmp[i], tmp[8+i], tmp[16+i], tmp[24+i] );
        sum8 += abs2(a0) + abs2(a1) + abs2(a2) + abs2(a3);
Loren Merritt's avatar
Loren Merritt committed
386
    }
387 388 389
    dc = (uint16_t)(tmp[0] + tmp[8] + tmp[16] + tmp[24]);
    sum4 = (uint16_t)sum4 + ((uint32_t)sum4>>16) - dc;
    sum8 = (uint16_t)sum8 + ((uint32_t)sum8>>16) - dc;
Loren Merritt's avatar
Loren Merritt committed
390 391 392 393
    return ((uint64_t)sum8<<32) + sum4;
}

#define HADAMARD_AC(w,h) \
394
static uint64_t x264_pixel_hadamard_ac_##w##x##h( pixel *pix, int stride )\
Loren Merritt's avatar
Loren Merritt committed
395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410
{\
    uint64_t sum = pixel_hadamard_ac( pix, stride );\
    if( w==16 )\
        sum += pixel_hadamard_ac( pix+8, stride );\
    if( h==16 )\
        sum += pixel_hadamard_ac( pix+8*stride, stride );\
    if( w==16 && h==16 )\
        sum += pixel_hadamard_ac( pix+8*stride+8, stride );\
    return ((sum>>34)<<32) + ((uint32_t)sum>>1);\
}
HADAMARD_AC( 16, 16 )
HADAMARD_AC( 16, 8 )
HADAMARD_AC( 8, 16 )
HADAMARD_AC( 8, 8 )


Loren Merritt's avatar
Loren Merritt committed
411 412 413
/****************************************************************************
 * pixel_sad_x4
 ****************************************************************************/
414
#define SAD_X( size ) \
415
static void x264_pixel_sad_x3_##size( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, int i_stride, int scores[3] )\
416 417 418 419 420
{\
    scores[0] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix0, i_stride );\
    scores[1] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix1, i_stride );\
    scores[2] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix2, i_stride );\
}\
421
static void x264_pixel_sad_x4_##size( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, pixel *pix3, int i_stride, int scores[4] )\
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
{\
    scores[0] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix0, i_stride );\
    scores[1] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix1, i_stride );\
    scores[2] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix2, i_stride );\
    scores[3] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix3, i_stride );\
}

SAD_X( 16x16 )
SAD_X( 16x8 )
SAD_X( 8x16 )
SAD_X( 8x8 )
SAD_X( 8x4 )
SAD_X( 4x8 )
SAD_X( 4x4 )

437
#if !X264_HIGH_BIT_DEPTH
Steven Walters's avatar
Steven Walters committed
438
#if ARCH_UltraSparc
439 440 441 442 443
SAD_X( 16x16_vis )
SAD_X( 16x8_vis )
SAD_X( 8x16_vis )
SAD_X( 8x8_vis )
#endif
444
#endif // !X264_HIGH_BIT_DEPTH
445

446 447 448 449 450 451
/****************************************************************************
 * pixel_satd_x4
 * no faster than single satd, but needed for satd to be a drop-in replacement for sad
 ****************************************************************************/

#define SATD_X( size, cpu ) \
452
static void x264_pixel_satd_x3_##size##cpu( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, int i_stride, int scores[3] )\
453 454 455 456 457
{\
    scores[0] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix0, i_stride );\
    scores[1] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix1, i_stride );\
    scores[2] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix2, i_stride );\
}\
458
static void x264_pixel_satd_x4_##size##cpu( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, pixel *pix3, int i_stride, int scores[4] )\
459 460 461 462 463 464
{\
    scores[0] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix0, i_stride );\
    scores[1] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix1, i_stride );\
    scores[2] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix2, i_stride );\
    scores[3] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix3, i_stride );\
}
465
#define SATD_X_DECL6( cpu )\
466 467 468 469
SATD_X( 16x16, cpu )\
SATD_X( 16x8, cpu )\
SATD_X( 8x16, cpu )\
SATD_X( 8x8, cpu )\
470 471
SATD_X( 8x4, cpu )\
SATD_X( 4x8, cpu )
472
#define SATD_X_DECL7( cpu )\
473
SATD_X_DECL6( cpu )\
474 475 476
SATD_X( 4x4, cpu )

SATD_X_DECL7()
477
#if !X264_HIGH_BIT_DEPTH
Steven Walters's avatar
Steven Walters committed
478
#if HAVE_MMX
479
SATD_X_DECL7( _mmxext )
480
SATD_X_DECL6( _sse2 )
Fiona Glaser's avatar
Fiona Glaser committed
481
SATD_X_DECL7( _ssse3 )
482
SATD_X_DECL7( _sse4 )
483 484
#endif

Steven Walters's avatar
Steven Walters committed
485
#if HAVE_ARMV6
486 487
SATD_X_DECL7( _neon )
#endif
488
#endif // !X264_HIGH_BIT_DEPTH
489

490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
#define INTRA_MBCMP_8x8( mbcmp )\
void x264_intra_##mbcmp##_x3_8x8( pixel *fenc, pixel edge[33], int res[3] )\
{\
    pixel pix[8*FDEC_STRIDE];\
    x264_predict_8x8_v_c( pix, edge );\
    res[0] = x264_pixel_##mbcmp##_8x8( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
    x264_predict_8x8_h_c( pix, edge );\
    res[1] = x264_pixel_##mbcmp##_8x8( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
    x264_predict_8x8_dc_c( pix, edge );\
    res[2] = x264_pixel_##mbcmp##_8x8( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
}

INTRA_MBCMP_8x8(sad)
INTRA_MBCMP_8x8(sa8d)

#define INTRA_MBCMP( mbcmp, size, pred1, pred2, pred3, chroma )\
void x264_intra_##mbcmp##_x3_##size##x##size##chroma( pixel *fenc, pixel *fdec, int res[3] )\
{\
    x264_predict_##size##x##size##chroma##_##pred1##_c( fdec );\
    res[0] = x264_pixel_##mbcmp##_##size##x##size( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
    x264_predict_##size##x##size##chroma##_##pred2##_c( fdec );\
    res[1] = x264_pixel_##mbcmp##_##size##x##size( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
    x264_predict_##size##x##size##chroma##_##pred3##_c( fdec );\
    res[2] = x264_pixel_##mbcmp##_##size##x##size( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
}

INTRA_MBCMP(sad, 4, v, h, dc, )
INTRA_MBCMP(satd, 4, v, h, dc, )
INTRA_MBCMP(sad, 8, dc, h, v, c )
INTRA_MBCMP(satd, 8, dc, h, v, c )
INTRA_MBCMP(sad, 16, v, h, dc, )
INTRA_MBCMP(satd, 16, v, h, dc, )

Loren Merritt's avatar
Loren Merritt committed
523 524 525
/****************************************************************************
 * structural similarity metric
 ****************************************************************************/
526 527
static void ssim_4x4x2_core( const pixel *pix1, int stride1,
                             const pixel *pix2, int stride2,
528 529
                             int sums[2][4])
{
530
    for( int z = 0; z < 2; z++ )
531
    {
532 533 534
        uint32_t s1 = 0, s2 = 0, ss = 0, s12 = 0;
        for( int y = 0; y < 4; y++ )
            for( int x = 0; x < 4; x++ )
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
            {
                int a = pix1[x+y*stride1];
                int b = pix2[x+y*stride2];
                s1  += a;
                s2  += b;
                ss  += a*a;
                ss  += b*b;
                s12 += a*b;
            }
        sums[z][0] = s1;
        sums[z][1] = s2;
        sums[z][2] = ss;
        sums[z][3] = s12;
        pix1 += 4;
        pix2 += 4;
    }
}

static float ssim_end1( int s1, int s2, int ss, int s12 )
{
555 556
    static const int ssim_c1 = (int)(.01*.01*PIXEL_MAX*PIXEL_MAX*64 + .5);
    static const int ssim_c2 = (int)(.03*.03*PIXEL_MAX*PIXEL_MAX*64*63 + .5);
557 558
    int vars = ss*64 - s1*s1 - s2*s2;
    int covar = s12*64 - s1*s2;
559 560
    return (float)(2*s1*s2 + ssim_c1) * (float)(2*covar + ssim_c2)
         / ((float)(s1*s1 + s2*s2 + ssim_c1) * (float)(vars + ssim_c2));
561 562 563 564 565
}

static float ssim_end4( int sum0[5][4], int sum1[5][4], int width )
{
    float ssim = 0.0;
566
    for( int i = 0; i < width; i++ )
567 568 569 570 571 572 573 574
        ssim += ssim_end1( sum0[i][0] + sum0[i+1][0] + sum1[i][0] + sum1[i+1][0],
                           sum0[i][1] + sum0[i+1][1] + sum1[i][1] + sum1[i+1][1],
                           sum0[i][2] + sum0[i+1][2] + sum1[i][2] + sum1[i+1][2],
                           sum0[i][3] + sum0[i+1][3] + sum1[i][3] + sum1[i+1][3] );
    return ssim;
}

float x264_pixel_ssim_wxh( x264_pixel_function_t *pf,
575 576
                           pixel *pix1, int stride1,
                           pixel *pix2, int stride2,
577
                           int width, int height, void *buf )
578
{
579
    int z = 0;
580
    float ssim = 0.0;
581
    int (*sum0)[4] = buf;
582
    int (*sum1)[4] = sum0 + (width >> 2) + 3;
583 584
    width >>= 2;
    height >>= 2;
585
    for( int y = 1; y < height; y++ )
586 587 588 589
    {
        for( ; z <= y; z++ )
        {
            XCHG( void*, sum0, sum1 );
590
            for( int x = 0; x < width; x+=2 )
591 592
                pf->ssim_4x4x2_core( &pix1[4*(x+z*stride1)], stride1, &pix2[4*(x+z*stride2)], stride2, &sum0[x] );
        }
593
        for( int x = 0; x < width-1; x += 4 )
594 595
            ssim += pf->ssim_end4( sum0+x, sum1+x, X264_MIN(4,width-x-1) );
    }
596
    return ssim;
597 598 599
}


Loren Merritt's avatar
Loren Merritt committed
600 601 602
/****************************************************************************
 * successive elimination
 ****************************************************************************/
603 604
static int x264_pixel_ads4( int enc_dc[4], uint16_t *sums, int delta,
                            uint16_t *cost_mvx, int16_t *mvs, int width, int thresh )
Loren Merritt's avatar
Loren Merritt committed
605
{
606 607
    int nmv = 0;
    for( int i = 0; i < width; i++, sums++ )
608 609 610 611 612 613 614 615 616 617
    {
        int ads = abs( enc_dc[0] - sums[0] )
                + abs( enc_dc[1] - sums[8] )
                + abs( enc_dc[2] - sums[delta] )
                + abs( enc_dc[3] - sums[delta+8] )
                + cost_mvx[i];
        if( ads < thresh )
            mvs[nmv++] = i;
    }
    return nmv;
Loren Merritt's avatar
Loren Merritt committed
618 619
}

620 621
static int x264_pixel_ads2( int enc_dc[2], uint16_t *sums, int delta,
                            uint16_t *cost_mvx, int16_t *mvs, int width, int thresh )
Loren Merritt's avatar
Loren Merritt committed
622
{
623 624
    int nmv = 0;
    for( int i = 0; i < width; i++, sums++ )
625 626 627 628 629 630 631 632
    {
        int ads = abs( enc_dc[0] - sums[0] )
                + abs( enc_dc[1] - sums[delta] )
                + cost_mvx[i];
        if( ads < thresh )
            mvs[nmv++] = i;
    }
    return nmv;
Loren Merritt's avatar
Loren Merritt committed
633 634
}

635 636
static int x264_pixel_ads1( int enc_dc[1], uint16_t *sums, int delta,
                            uint16_t *cost_mvx, int16_t *mvs, int width, int thresh )
Loren Merritt's avatar
Loren Merritt committed
637
{
638 639
    int nmv = 0;
    for( int i = 0; i<width; i++, sums++ )
640 641 642 643 644 645 646
    {
        int ads = abs( enc_dc[0] - sums[0] )
                + cost_mvx[i];
        if( ads < thresh )
            mvs[nmv++] = i;
    }
    return nmv;
Loren Merritt's avatar
Loren Merritt committed
647 648 649
}


Laurent Aimar's avatar
Laurent Aimar committed
650 651 652 653 654
/****************************************************************************
 * x264_pixel_init:
 ****************************************************************************/
void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
{
655 656
    memset( pixf, 0, sizeof(*pixf) );

657 658 659 660 661 662 663 664 665 666
#define INIT2_NAME( name1, name2, cpu ) \
    pixf->name1[PIXEL_16x16] = x264_pixel_##name2##_16x16##cpu;\
    pixf->name1[PIXEL_16x8]  = x264_pixel_##name2##_16x8##cpu;
#define INIT4_NAME( name1, name2, cpu ) \
    INIT2_NAME( name1, name2, cpu ) \
    pixf->name1[PIXEL_8x16]  = x264_pixel_##name2##_8x16##cpu;\
    pixf->name1[PIXEL_8x8]   = x264_pixel_##name2##_8x8##cpu;
#define INIT5_NAME( name1, name2, cpu ) \
    INIT4_NAME( name1, name2, cpu ) \
    pixf->name1[PIXEL_8x4]   = x264_pixel_##name2##_8x4##cpu;
667
#define INIT6_NAME( name1, name2, cpu ) \
668
    INIT5_NAME( name1, name2, cpu ) \
669 670 671
    pixf->name1[PIXEL_4x8]   = x264_pixel_##name2##_4x8##cpu;
#define INIT7_NAME( name1, name2, cpu ) \
    INIT6_NAME( name1, name2, cpu ) \
672 673 674 675
    pixf->name1[PIXEL_4x4]   = x264_pixel_##name2##_4x4##cpu;
#define INIT2( name, cpu ) INIT2_NAME( name, name, cpu )
#define INIT4( name, cpu ) INIT4_NAME( name, name, cpu )
#define INIT5( name, cpu ) INIT5_NAME( name, name, cpu )
676
#define INIT6( name, cpu ) INIT6_NAME( name, name, cpu )
677
#define INIT7( name, cpu ) INIT7_NAME( name, name, cpu )
678

679 680 681 682 683
#define INIT_ADS( cpu ) \
    pixf->ads[PIXEL_16x16] = x264_pixel_ads4##cpu;\
    pixf->ads[PIXEL_16x8] = x264_pixel_ads2##cpu;\
    pixf->ads[PIXEL_8x8] = x264_pixel_ads1##cpu;

684
    INIT7( sad, );
685
    INIT7_NAME( sad_aligned, sad, );
686 687 688 689
    INIT7( sad_x3, );
    INIT7( sad_x4, );
    INIT7( ssd, );
    INIT7( satd, );
690 691
    INIT7( satd_x3, );
    INIT7( satd_x4, );
Loren Merritt's avatar
Loren Merritt committed
692
    INIT4( hadamard_ac, );
693
    INIT_ADS( );
694

695 696
    pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16;
    pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8;
697 698 699
    pixf->var[PIXEL_16x16] = x264_pixel_var_16x16;
    pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8;

700
    pixf->ssd_nv12_core = pixel_ssd_nv12_core;
701 702
    pixf->ssim_4x4x2_core = ssim_4x4x2_core;
    pixf->ssim_end4 = ssim_end4;
703
    pixf->var2_8x8 = pixel_var2_8x8;
704

705 706 707 708 709 710 711 712 713
    pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4;
    pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4;
    pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8;
    pixf->intra_sa8d_x3_8x8   = x264_intra_sa8d_x3_8x8;
    pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c;
    pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c;
    pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16;
    pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16;

714
#if !X264_HIGH_BIT_DEPTH
Steven Walters's avatar
Steven Walters committed
715
#if HAVE_MMX
716 717
    if( cpu&X264_CPU_MMX )
    {
718
        INIT7( ssd, _mmx );
719 720
    }

Laurent Aimar's avatar
Laurent Aimar committed
721 722
    if( cpu&X264_CPU_MMXEXT )
    {
723
        INIT7( sad, _mmxext );
724
        INIT7_NAME( sad_aligned, sad, _mmxext );
725 726 727
        INIT7( sad_x3, _mmxext );
        INIT7( sad_x4, _mmxext );
        INIT7( satd, _mmxext );
728 729
        INIT7( satd_x3, _mmxext );
        INIT7( satd_x4, _mmxext );
Loren Merritt's avatar
Loren Merritt committed
730
        INIT4( hadamard_ac, _mmxext );
731
        INIT_ADS( _mmxext );
732 733
        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmxext;
        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_mmxext;
734
        pixf->ssd_nv12_core    = x264_pixel_ssd_nv12_core_mmxext;
Steven Walters's avatar
Steven Walters committed
735
#if ARCH_X86
736 737
        pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_mmxext;
        pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_mmxext;
738
        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_mmxext;
739
        pixf->ssim_4x4x2_core  = x264_pixel_ssim_4x4x2_core_mmxext;
740
        pixf->var2_8x8 = x264_pixel_var2_8x8_mmxext;
741

742
        if( cpu&X264_CPU_CACHELINE_32 )
743
        {
744 745 746 747 748 749 750 751 752
            INIT5( sad, _cache32_mmxext );
            INIT4( sad_x3, _cache32_mmxext );
            INIT4( sad_x4, _cache32_mmxext );
        }
        else if( cpu&X264_CPU_CACHELINE_64 )
        {
            INIT5( sad, _cache64_mmxext );
            INIT4( sad_x3, _cache64_mmxext );
            INIT4( sad_x4, _cache64_mmxext );
753 754
        }
#else
755
        if( cpu&X264_CPU_CACHELINE_64 )
756 757 758 759 760 761 762 763 764
        {
            pixf->sad[PIXEL_8x16] = x264_pixel_sad_8x16_cache64_mmxext;
            pixf->sad[PIXEL_8x8]  = x264_pixel_sad_8x8_cache64_mmxext;
            pixf->sad[PIXEL_8x4]  = x264_pixel_sad_8x4_cache64_mmxext;
            pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_cache64_mmxext;
            pixf->sad_x3[PIXEL_8x8]  = x264_pixel_sad_x3_8x8_cache64_mmxext;
            pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_mmxext;
            pixf->sad_x4[PIXEL_8x8]  = x264_pixel_sad_x4_8x8_cache64_mmxext;
        }
765
#endif
766
        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmxext;
Fiona Glaser's avatar
Fiona Glaser committed
767
        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_mmxext;
768
        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_mmxext;
Fiona Glaser's avatar
Fiona Glaser committed
769
        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_mmxext;
Fiona Glaser's avatar
Fiona Glaser committed
770
        pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8_mmxext;
771
        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_mmxext;
Fiona Glaser's avatar
Fiona Glaser committed
772
        pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_mmxext;
Laurent Aimar's avatar
Laurent Aimar committed
773
    }
774

775 776 777 778 779
    if( cpu&X264_CPU_SSE2 )
    {
        INIT5( ssd, _sse2slow );
        INIT2_NAME( sad_aligned, sad, _sse2_aligned );
        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_sse2;
780
        pixf->ssd_nv12_core    = x264_pixel_ssd_nv12_core_sse2;
781 782 783 784
        pixf->ssim_4x4x2_core  = x264_pixel_ssim_4x4x2_core_sse2;
        pixf->ssim_end4        = x264_pixel_ssim_end4_sse2;
        pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_sse2;
        pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_sse2;
Steven Walters's avatar
Steven Walters committed
785
#if ARCH_X86_64
786 787
        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_sse2;
#endif
788
        pixf->var2_8x8 = x264_pixel_var2_8x8_sse2;
789 790
    }

791
    if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_SSE2_IS_SLOW) )
792
    {
793 794 795
        INIT2( sad, _sse2 );
        INIT2( sad_x3, _sse2 );
        INIT2( sad_x4, _sse2 );
796 797 798
        INIT6( satd, _sse2 );
        INIT6( satd_x3, _sse2 );
        INIT6( satd_x4, _sse2 );
799 800 801 802
        if( !(cpu&X264_CPU_STACK_MOD4) )
        {
            INIT4( hadamard_ac, _sse2 );
        }
803
        INIT_ADS( _sse2 );
804
        pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_sse2;
805
        pixf->intra_sad_x3_16x16 = x264_intra_sad_x3_16x16_sse2;
806
        if( cpu&X264_CPU_CACHELINE_64 )
807
        {
808
            INIT2( ssd, _sse2); /* faster for width 16 on p4 */
Steven Walters's avatar
Steven Walters committed
809
#if ARCH_X86
810 811 812 813
            INIT2( sad, _cache64_sse2 );
            INIT2( sad_x3, _cache64_sse2 );
            INIT2( sad_x4, _cache64_sse2 );
#endif
814 815 816 817 818 819 820
           if( cpu&X264_CPU_SSE2_IS_FAST )
           {
               pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_cache64_sse2;
               pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_sse2;
           }
        }

Fiona Glaser's avatar
Fiona Glaser committed
821 822 823 824 825
        if( cpu&X264_CPU_SSE_MISALIGN )
        {
            INIT2( sad_x3, _sse2_misalign );
            INIT2( sad_x4, _sse2_misalign );
        }
826
    }
827

828 829 830 831 832 833 834 835 836 837 838 839
    if( cpu&X264_CPU_SSE2_IS_FAST && !(cpu&X264_CPU_CACHELINE_64) )
    {
        pixf->sad_aligned[PIXEL_8x16] = x264_pixel_sad_8x16_sse2;
        pixf->sad[PIXEL_8x16] = x264_pixel_sad_8x16_sse2;
        pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_sse2;
        pixf->sad_x3[PIXEL_8x8] = x264_pixel_sad_x3_8x8_sse2;
        pixf->sad_x3[PIXEL_8x4] = x264_pixel_sad_x3_8x4_sse2;
        pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_sse2;
        pixf->sad_x4[PIXEL_8x8] = x264_pixel_sad_x4_8x8_sse2;
        pixf->sad_x4[PIXEL_8x4] = x264_pixel_sad_x4_8x4_sse2;
    }

840
    if( (cpu&X264_CPU_SSE3) && (cpu&X264_CPU_CACHELINE_64) )
841 842 843 844 845 846
    {
        INIT2( sad, _sse3 );
        INIT2( sad_x3, _sse3 );
        INIT2( sad_x4, _sse3 );
    }

847 848
    if( cpu&X264_CPU_SSSE3 )
    {
849 850 851 852
        if( !(cpu&X264_CPU_STACK_MOD4) )
        {
            INIT4( hadamard_ac, _ssse3 );
        }
853
        INIT_ADS( _ssse3 );
854 855 856 857 858 859 860 861 862
        if( !(cpu&X264_CPU_SLOW_ATOM) )
        {
            INIT7( ssd, _ssse3 );
            pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_ssse3;
            pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_ssse3;
            INIT7( satd, _ssse3 );
            INIT7( satd_x3, _ssse3 );
            INIT7( satd_x4, _ssse3 );
        }
863
        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_ssse3;
864
        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_ssse3;
865
        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_ssse3;
Fiona Glaser's avatar
Fiona Glaser committed
866
        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_ssse3;
867
        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_ssse3;
Steven Walters's avatar
Steven Walters committed
868
#if ARCH_X86_64
869
        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_ssse3;
Loren Merritt's avatar
Loren Merritt committed
870
#endif
871
        pixf->var2_8x8 = x264_pixel_var2_8x8_ssse3;
872
        if( cpu&X264_CPU_CACHELINE_64 )
873 874 875 876 877
        {
            INIT2( sad, _cache64_ssse3 );
            INIT2( sad_x3, _cache64_ssse3 );
            INIT2( sad_x4, _cache64_ssse3 );
        }
878
        if( cpu&X264_CPU_SLOW_ATOM || !(cpu&X264_CPU_SHUFFLE_IS_FAST) )
879
        {
880
            INIT5( ssd, _sse2 ); /* on conroe, sse2 is faster for width8/16 */
881
        }
882
    }
883 884 885

    if( cpu&X264_CPU_SSE4 )
    {
886 887 888 889 890 891 892 893 894
        INIT7( satd, _sse4 );
        INIT7( satd_x3, _sse4 );
        INIT7( satd_x4, _sse4 );
        if( !(cpu&X264_CPU_STACK_MOD4) )
        {
            INIT4( hadamard_ac, _sse4 );
        }
        pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_sse4;
        pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_sse4;
895 896 897
        pixf->intra_sad_x3_4x4 = x264_intra_sad_x3_4x4_sse4;
        /* Slower on Conroe, so only enable under SSE4 */
        pixf->intra_sad_x3_8x8  = x264_intra_sad_x3_8x8_ssse3;
898
    }
899
#endif //HAVE_MMX
900

Steven Walters's avatar
Steven Walters committed
901
#if HAVE_ARMV6
902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941
    if( cpu&X264_CPU_ARMV6 )
    {
        pixf->sad[PIXEL_4x8] = x264_pixel_sad_4x8_armv6;
        pixf->sad[PIXEL_4x4] = x264_pixel_sad_4x4_armv6;
        pixf->sad_aligned[PIXEL_4x8] = x264_pixel_sad_4x8_armv6;
        pixf->sad_aligned[PIXEL_4x4] = x264_pixel_sad_4x4_armv6;
    }
    if( cpu&X264_CPU_NEON )
    {
        INIT5( sad, _neon );
        INIT5( sad_aligned, _neon );