pixel.c 33.8 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;
}

143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167
static uint64_t pixel_ssd_nv12_core( pixel *pixuv1, int stride1, pixel *pixuv2, int stride2, int width, int height )
{
    uint32_t ssd_u=0, ssd_v=0;
    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];
            ssd_u += du*du;
            ssd_v += dv*dv;
        }
    return ssd_u + ((uint64_t)ssd_v<<32);
}

// SSD in uint32 (i.e. packing two into uint64) can potentially overflow on
// image widths >= 11008 (or 6604 if interlaced), since this is called on blocks
// of height up to 12 (resp 20). Though it will probably take significantly more
// than that at sane distortion levels.
uint64_t 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 = pf->ssd_nv12_core( pix1, i_pix1, pix2, i_pix2, i_width&~7, i_height );
    if( i_width&7 )
        ssd += pixel_ssd_nv12_core( pix1+(i_width&~7), i_pix1, pix2+(i_width&~7), i_pix2, i_width&7, i_height );
    return ssd;
}
168

169 170 171
/****************************************************************************
 * pixel_var_wxh
 ****************************************************************************/
Fiona Glaser's avatar
Fiona Glaser committed
172
#define PIXEL_VAR_C( name, w ) \
173
static uint64_t name( pixel *pix, int i_stride ) \
174
{                                             \
Fiona Glaser's avatar
Fiona Glaser committed
175
    uint32_t sum = 0, sqr = 0;                \
176
    for( int y = 0; y < w; y++ )              \
177
    {                                         \
178
        for( int x = 0; x < w; x++ )          \
179 180 181 182 183 184
        {                                     \
            sum += pix[x];                    \
            sqr += pix[x] * pix[x];           \
        }                                     \
        pix += i_stride;                      \
    }                                         \
Fiona Glaser's avatar
Fiona Glaser committed
185
    return sum + ((uint64_t)sqr << 32);       \
186 187
}

Fiona Glaser's avatar
Fiona Glaser committed
188 189
PIXEL_VAR_C( x264_pixel_var_16x16, 16 )
PIXEL_VAR_C( x264_pixel_var_8x8,    8 )
190

191 192 193
/****************************************************************************
 * pixel_var2_wxh
 ****************************************************************************/
194
static int pixel_var2_8x8( pixel *pix1, int i_stride1, pixel *pix2, int i_stride2, int *ssd )
195 196
{
    uint32_t var = 0, sum = 0, sqr = 0;
197
    for( int y = 0; y < 8; y++ )
198
    {
199
        for( int x = 0; x < 8; x++ )
200 201 202 203 204 205 206 207 208
        {
            int diff = pix1[x] - pix2[x];
            sum += diff;
            sqr += diff * diff;
        }
        pix1 += i_stride1;
        pix2 += i_stride2;
    }
    sum = abs(sum);
209
    var = sqr - ((uint64_t)sum * sum >> 6);
210 211 212 213
    *ssd = sqr;
    return var;
}

214

215
#define HADAMARD4(d0, d1, d2, d3, s0, s1, s2, s3) {\
Loren Merritt's avatar
Loren Merritt committed
216 217 218 219 220 221 222 223 224
    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;\
}
225

226 227 228 229 230 231 232 233
// 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;
}

234 235 236
/****************************************************************************
 * pixel_satd_WxH: sum of 4x4 Hadamard transformed differences
 ****************************************************************************/
Laurent Aimar's avatar
Laurent Aimar committed
237

238
static NOINLINE int x264_pixel_satd_4x4( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
239 240
{
    uint32_t tmp[4][2];
241 242 243
    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
244
    {
245 246 247 248 249 250 251 252
        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
253
    }
254
    for( int i = 0; i < 2; i++ )
255
    {
256
        HADAMARD4( a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i] );
257 258 259 260
        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
261 262
}

263
static NOINLINE int x264_pixel_satd_8x4( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
264 265
{
    uint32_t tmp[4][4];
266 267 268
    uint32_t a0, a1, a2, a3;
    int sum = 0;
    for( int i = 0; i < 4; i++, pix1 += i_pix1, pix2 += i_pix2 )
269 270 271 272 273 274 275
    {
        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 );
    }
276
    for( int i = 0; i < 4; i++ )
277
    {
278
        HADAMARD4( a0, a1, a2, a3, tmp[0][i], tmp[1][i], tmp[2][i], tmp[3][i] );
279 280 281 282
        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
283

284
#define PIXEL_SATD_C( w, h, sub )\
285
static int x264_pixel_satd_##w##x##h( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )\
286 287 288 289 290 291 292 293 294 295 296 297 298
{\
    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;\
299
}
300 301 302 303 304
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 )
305 306


307
static NOINLINE int sa8d_8x8( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
308 309
{
    uint32_t tmp[8][4];
310 311 312
    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 )
313
    {
314 315 316 317 318 319 320 321 322 323 324 325 326
        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 );
327
    }
328
    for( int i = 0; i < 4; i++ )
329
    {
330 331
        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] );
332 333 334 335 336 337 338 339
        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;
}
340

341
static int x264_pixel_sa8d_8x8( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
342 343 344
{
    int sum = sa8d_8x8( pix1, i_pix1, pix2, i_pix2 );
    return (sum+2)>>2;
345 346
}

347
static int x264_pixel_sa8d_16x16( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
348 349 350 351 352 353
{
    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;
354 355
}

Loren Merritt's avatar
Loren Merritt committed
356

357
static NOINLINE uint64_t pixel_hadamard_ac( pixel *pix, int stride )
Loren Merritt's avatar
Loren Merritt committed
358
{
359
    uint32_t tmp[32];
360 361 362
    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
363
    {
364 365 366 367 368 369 370 371 372
        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
373
    }
374
    for( int i = 0; i < 8; i++ )
Loren Merritt's avatar
Loren Merritt committed
375
    {
376
        HADAMARD4( a0, a1, a2, a3, tmp[i*4+0], tmp[i*4+1], tmp[i*4+2], tmp[i*4+3] );
377 378 379 380 381
        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
382
    }
383
    for( int i = 0; i < 8; i++ )
Loren Merritt's avatar
Loren Merritt committed
384
    {
385 386
        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
387
    }
388 389 390
    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
391 392 393 394
    return ((uint64_t)sum8<<32) + sum4;
}

#define HADAMARD_AC(w,h) \
395
static uint64_t x264_pixel_hadamard_ac_##w##x##h( pixel *pix, int stride )\
Loren Merritt's avatar
Loren Merritt committed
396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411
{\
    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
412 413 414
/****************************************************************************
 * pixel_sad_x4
 ****************************************************************************/
415
#define SAD_X( size ) \
416
static void x264_pixel_sad_x3_##size( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, int i_stride, int scores[3] )\
417 418 419 420 421
{\
    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 );\
}\
422
static void x264_pixel_sad_x4_##size( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, pixel *pix3, int i_stride, int scores[4] )\
423 424 425 426 427 428 429 430 431 432 433 434 435 436 437
{\
    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 )

438
#if !X264_HIGH_BIT_DEPTH
Steven Walters's avatar
Steven Walters committed
439
#if ARCH_UltraSparc
440 441 442 443 444
SAD_X( 16x16_vis )
SAD_X( 16x8_vis )
SAD_X( 8x16_vis )
SAD_X( 8x8_vis )
#endif
445
#endif // !X264_HIGH_BIT_DEPTH
446

447 448 449 450 451 452
/****************************************************************************
 * 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 ) \
453
static void x264_pixel_satd_x3_##size##cpu( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, int i_stride, int scores[3] )\
454 455 456 457 458
{\
    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 );\
}\
459
static void x264_pixel_satd_x4_##size##cpu( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, pixel *pix3, int i_stride, int scores[4] )\
460 461 462 463 464 465
{\
    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 );\
}
466
#define SATD_X_DECL6( cpu )\
467 468 469 470
SATD_X( 16x16, cpu )\
SATD_X( 16x8, cpu )\
SATD_X( 8x16, cpu )\
SATD_X( 8x8, cpu )\
471 472
SATD_X( 8x4, cpu )\
SATD_X( 4x8, cpu )
473
#define SATD_X_DECL7( cpu )\
474
SATD_X_DECL6( cpu )\
475 476 477
SATD_X( 4x4, cpu )

SATD_X_DECL7()
478
#if !X264_HIGH_BIT_DEPTH
Steven Walters's avatar
Steven Walters committed
479
#if HAVE_MMX
480
SATD_X_DECL7( _mmxext )
481
SATD_X_DECL6( _sse2 )
Fiona Glaser's avatar
Fiona Glaser committed
482
SATD_X_DECL7( _ssse3 )
483
SATD_X_DECL7( _sse4 )
484 485
#endif

Steven Walters's avatar
Steven Walters committed
486
#if HAVE_ARMV6
487 488
SATD_X_DECL7( _neon )
#endif
489
#endif // !X264_HIGH_BIT_DEPTH
490

491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523
#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
524 525 526
/****************************************************************************
 * structural similarity metric
 ****************************************************************************/
527 528
static void ssim_4x4x2_core( const pixel *pix1, int stride1,
                             const pixel *pix2, int stride2,
529 530
                             int sums[2][4])
{
531
    for( int z = 0; z < 2; z++ )
532
    {
533 534 535
        uint32_t s1 = 0, s2 = 0, ss = 0, s12 = 0;
        for( int y = 0; y < 4; y++ )
            for( int x = 0; x < 4; x++ )
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555
            {
                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 )
{
556 557
    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);
558 559
    int vars = ss*64 - s1*s1 - s2*s2;
    int covar = s12*64 - s1*s2;
560 561
    return (float)(2*s1*s2 + ssim_c1) * (float)(2*covar + ssim_c2)
         / ((float)(s1*s1 + s2*s2 + ssim_c1) * (float)(vars + ssim_c2));
562 563 564 565 566
}

static float ssim_end4( int sum0[5][4], int sum1[5][4], int width )
{
    float ssim = 0.0;
567
    for( int i = 0; i < width; i++ )
568 569 570 571 572 573 574 575
        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,
576 577
                           pixel *pix1, int stride1,
                           pixel *pix2, int stride2,
578
                           int width, int height, void *buf )
579
{
580
    int z = 0;
581
    float ssim = 0.0;
582
    int (*sum0)[4] = buf;
583
    int (*sum1)[4] = sum0 + (width >> 2) + 3;
584 585
    width >>= 2;
    height >>= 2;
586
    for( int y = 1; y < height; y++ )
587 588 589 590
    {
        for( ; z <= y; z++ )
        {
            XCHG( void*, sum0, sum1 );
591
            for( int x = 0; x < width; x+=2 )
592 593
                pf->ssim_4x4x2_core( &pix1[4*(x+z*stride1)], stride1, &pix2[4*(x+z*stride2)], stride2, &sum0[x] );
        }
594
        for( int x = 0; x < width-1; x += 4 )
595 596
            ssim += pf->ssim_end4( sum0+x, sum1+x, X264_MIN(4,width-x-1) );
    }
597
    return ssim;
598 599 600
}


Loren Merritt's avatar
Loren Merritt committed
601 602 603
/****************************************************************************
 * successive elimination
 ****************************************************************************/
604 605
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
606
{
607 608
    int nmv = 0;
    for( int i = 0; i < width; i++, sums++ )
609 610 611 612 613 614 615 616 617 618
    {
        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
619 620
}

621 622
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
623
{
624 625
    int nmv = 0;
    for( int i = 0; i < width; i++, sums++ )
626 627 628 629 630 631 632 633
    {
        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
634 635
}

636 637
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
638
{
639 640
    int nmv = 0;
    for( int i = 0; i<width; i++, sums++ )
641 642 643 644 645 646 647
    {
        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
648 649 650
}


Laurent Aimar's avatar
Laurent Aimar committed
651 652 653 654 655
/****************************************************************************
 * x264_pixel_init:
 ****************************************************************************/
void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
{
656 657
    memset( pixf, 0, sizeof(*pixf) );

658 659 660 661 662 663 664 665 666 667
#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;
668
#define INIT6_NAME( name1, name2, cpu ) \
669
    INIT5_NAME( name1, name2, cpu ) \
670 671 672
    pixf->name1[PIXEL_4x8]   = x264_pixel_##name2##_4x8##cpu;
#define INIT7_NAME( name1, name2, cpu ) \
    INIT6_NAME( name1, name2, cpu ) \
673 674 675 676
    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 )
677
#define INIT6( name, cpu ) INIT6_NAME( name, name, cpu )
678
#define INIT7( name, cpu ) INIT7_NAME( name, name, cpu )
679

680 681 682 683 684
#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;

685
    INIT7( sad, );
686
    INIT7_NAME( sad_aligned, sad, );
687 688 689 690
    INIT7( sad_x3, );
    INIT7( sad_x4, );
    INIT7( ssd, );
    INIT7( satd, );
691 692
    INIT7( satd_x3, );
    INIT7( satd_x4, );
Loren Merritt's avatar
Loren Merritt committed
693
    INIT4( hadamard_ac, );
694
    INIT_ADS( );
695

696 697
    pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16;
    pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8;
698 699 700
    pixf->var[PIXEL_16x16] = x264_pixel_var_16x16;
    pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8;

701
    pixf->ssd_nv12_core = pixel_ssd_nv12_core;
702 703
    pixf->ssim_4x4x2_core = ssim_4x4x2_core;
    pixf->ssim_end4 = ssim_end4;
704
    pixf->var2_8x8 = pixel_var2_8x8;
705

706 707 708 709 710 711 712 713 714
    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;

715
#if !X264_HIGH_BIT_DEPTH
Steven Walters's avatar
Steven Walters committed
716
#if HAVE_MMX
717 718
    if( cpu&X264_CPU_MMX )
    {
719
        INIT7( ssd, _mmx );
720 721
    }

Laurent Aimar's avatar
Laurent Aimar committed
722 723
    if( cpu&X264_CPU_MMXEXT )
    {
724
        INIT7( sad, _mmxext );
725
        INIT7_NAME( sad_aligned, sad, _mmxext );
726 727 728
        INIT7( sad_x3, _mmxext );
        INIT7( sad_x4, _mmxext );
        INIT7( satd, _mmxext );
729 730
        INIT7( satd_x3, _mmxext );
        INIT7( satd_x4, _mmxext );
Loren Merritt's avatar
Loren Merritt committed
731
        INIT4( hadamard_ac, _mmxext );
732
        INIT_ADS( _mmxext );
733 734
        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmxext;
        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_mmxext;
735
        pixf->ssd_nv12_core    = x264_pixel_ssd_nv12_core_mmxext;
Steven Walters's avatar
Steven Walters committed
736
#if ARCH_X86
737 738
        pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_mmxext;
        pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_mmxext;
739
        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_mmxext;
740
        pixf->ssim_4x4x2_core  = x264_pixel_ssim_4x4x2_core_mmxext;
741
        pixf->var2_8x8 = x264_pixel_var2_8x8_mmxext;
742

743
        if( cpu&X264_CPU_CACHELINE_32 )
744
        {
745 746 747 748 749 750 751 752 753
            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 );
754 755
        }
#else
756
        if( cpu&X264_CPU_CACHELINE_64 )
757 758 759 760 761 762 763 764 765
        {
            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;
        }
766
#endif
767
        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmxext;
Fiona Glaser's avatar
Fiona Glaser committed
768
        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_mmxext;
769
        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_mmxext;
Fiona Glaser's avatar
Fiona Glaser committed
770
        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_mmxext;
Fiona Glaser's avatar
Fiona Glaser committed
771
        pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8_mmxext;
772
        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_mmxext;
Fiona Glaser's avatar
Fiona Glaser committed
773
        pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_mmxext;
Laurent Aimar's avatar
Laurent Aimar committed
774
    }
775

776 777 778 779 780
    if( cpu&X264_CPU_SSE2 )
    {
        INIT5( ssd, _sse2slow );
        INIT2_NAME( sad_aligned, sad, _sse2_aligned );
        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_sse2;
781
        pixf->ssd_nv12_core    = x264_pixel_ssd_nv12_core_sse2;
782 783 784 785
        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
786
#if ARCH_X86_64
787 788
        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_sse2;
#endif
789
        pixf->var2_8x8 = x264_pixel_var2_8x8_sse2;
790 791
    }

792
    if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_SSE2_IS_SLOW) )
793
    {
794 795 796
        INIT2( sad, _sse2 );
        INIT2( sad_x3, _sse2 );
        INIT2( sad_x4, _sse2 );
797 798 799
        INIT6( satd, _sse2 );
        INIT6( satd_x3, _sse2 );
        INIT6( satd_x4, _sse2 );
800 801 802 803
        if( !(cpu&X264_CPU_STACK_MOD4) )
        {
            INIT4( hadamard_ac, _sse2 );
        }
804
        INIT_ADS( _sse2 );
805
        pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_sse2;
806
        pixf->intra_sad_x3_16x16 = x264_intra_sad_x3_16x16_sse2;
807
        if( cpu&X264_CPU_CACHELINE_64 )
808
        {
809
            INIT2( ssd, _sse2); /* faster for width 16 on p4 */
Steven Walters's avatar
Steven Walters committed
810
#if ARCH_X86
811 812 813 814
            INIT2( sad, _cache64_sse2 );
            INIT2( sad_x3, _cache64_sse2 );
            INIT2( sad_x4, _cache64_sse2 );
#endif
815 816 817 818 819 820 821
           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
822 823 824 825 826
        if( cpu&X264_CPU_SSE_MISALIGN )
        {
            INIT2( sad_x3, _sse2_misalign );
            INIT2( sad_x4, _sse2_misalign );
        }
827
    }
828

829 830 831 832 833 834 835 836 837 838 839 840
    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;
    }

841
    if( (cpu&X264_CPU_SSE3) && (cpu&X264_CPU_CACHELINE_64) )
842 843 844 845 846 847
    {
        INIT2( sad, _sse3 );
        INIT2( sad_x3, _sse3 );
        INIT2( sad_x4, _sse3 );
    }

848 849
    if( cpu&X264_CPU_SSSE3 )
    {
850 851 852 853
        if( !(cpu&X264_CPU_STACK_MOD4) )
        {
            INIT4( hadamard_ac, _ssse3 );
        }
854
        INIT_ADS( _ssse3 );
855 856 857 858 859 860 861 862 863
        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 );
        }
864
        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_ssse3;
865
        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_ssse3;
866
        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_ssse3;
Fiona Glaser's avatar
Fiona Glaser committed
867
        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_ssse3;
868
        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_ssse3;
Steven Walters's avatar
Steven Walters committed
869
#if ARCH_X86_64
870
        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_ssse3;
Loren Merritt's avatar
Loren Merritt committed
871
#endif
872
        pixf->var2_8x8 = x264_pixel_var2_8x8_ssse3;
873
        if( cpu&X264_CPU_CACHELINE_64 )
874 875 876 877 878
        {
            INIT2( sad, _cache64_ssse3 );
            INIT2( sad_x3, _cache64_ssse3 );
            INIT2( sad_x4, _cache64_ssse3 );
        }
879
        if( cpu&X264_CPU_SLOW_ATOM || !(cpu&X264_CPU_SHUFFLE_IS_FAST) )
880
        {
881
            INIT5( ssd, _sse2 ); /* on conroe, sse2 is faster for width8/16 */
882
        }
883
    }
884 885 886

    if( cpu&X264_CPU_SSE4 )
    {
887 888 889 890 891 892 893 894 895
        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;
896 897 898
        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;
899
    }
900
#endif //HAVE_MMX
901

Steven Walters's avatar
Steven Walters committed
902
#if HAVE_ARMV6
903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942
    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 );
        INIT7( sad_x3, _neon );
        INIT7( sad_x4, _neon );
        INIT7( ssd, _neon );
        INIT7( satd, _neon );
        INIT7( satd_x3, _neon );
        INIT7( satd_x4, _neon );
        INIT4( hadamard_ac, _neon );
        pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_neon;
        pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_neon;
        pixf->var[PIXEL_8x8]    = x264_pixel_var_8x8_neon;
        pixf->var[PIXEL_16x16]  = x264_pixel_var_16x16_neon;
        pixf->var2_8x8          = x264_pixel_var2_8x8_neon;

        pixf->ssim_4x4x2_core   = x264_pixel_ssim_4x4x2_core_neon;
        pixf->ssim_end4         = x264_pixel_ssim_end4_neon;

        if( cpu&X264_CPU_FAST_NEON_MRC )
        {
            pixf->sad[PIXEL_4x8] = x264_pixel_sad_4x8_neon;
            pixf->sad[PIXEL_4x4] = x264_pixel_sad_4x4_neon;
            pixf->sad_aligned[PIXEL_4x8] = x264_pixel_sad_aligned_4x8_neon;
            pixf->sad_aligned[PIXEL_4x4] = x264_pixel_sad_aligned_4x4_neon;
        }
        else    // really just scheduled for dual issue / A8
        {
            INIT5( sad_aligned, _neon_dual );
        }
    }
#endif
943
#endif // !X264_HIGH_BIT_DEPTH
Steven Walters's avatar
Steven Walters committed
944
#if HAVE_ALTIVEC
Laurent Aimar's avatar
Laurent Aimar committed
945 946 947 948 949
    if( cpu&X264_CPU_ALTIVEC )
    {
        x264_pixel_altivec_init( pixf );
    }
#endif
950
#if !X264_HIGH_BIT_DEPTH
Steven Walters's avatar
Steven Walters committed
951
#if ARCH_UltraSparc