mc.c 18.3 KB
Newer Older
Laurent Aimar's avatar
Laurent Aimar committed
1 2 3
/*****************************************************************************
 * mc.c: h264 encoder library (Motion Compensation)
 *****************************************************************************
4
 * Copyright (C) 2003-2008 x264 project
Laurent Aimar's avatar
Laurent Aimar committed
5 6
 *
 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7
 *          Loren Merritt <lorenm@u.washington.edu>
Laurent Aimar's avatar
Laurent Aimar committed
8 9 10 11 12 13 14 15 16 17 18 19 20
 *
 * 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
21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
Laurent Aimar's avatar
Laurent Aimar committed
22 23
 *****************************************************************************/

Loren Merritt's avatar
Loren Merritt committed
24
#include "common.h"
Laurent Aimar's avatar
Laurent Aimar committed
25

26
#ifdef HAVE_MMX
27
#include "x86/mc.h"
Laurent Aimar's avatar
Laurent Aimar committed
28
#endif
29
#ifdef ARCH_PPC
30
#include "ppc/mc.h"
Laurent Aimar's avatar
Laurent Aimar committed
31
#endif
32 33 34
#ifdef ARCH_ARM
#include "arm/mc.h"
#endif
Laurent Aimar's avatar
Laurent Aimar committed
35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54


static inline void pixel_avg( uint8_t *dst,  int i_dst_stride,
                              uint8_t *src1, int i_src1_stride,
                              uint8_t *src2, int i_src2_stride,
                              int i_width, int i_height )
{
    int x, y;
    for( y = 0; y < i_height; y++ )
    {
        for( x = 0; x < i_width; x++ )
        {
            dst[x] = ( src1[x] + src2[x] + 1 ) >> 1;
        }
        dst  += i_dst_stride;
        src1 += i_src1_stride;
        src2 += i_src2_stride;
    }
}

55
static inline void pixel_avg_wxh( uint8_t *dst, int i_dst, uint8_t *src1, int i_src1, uint8_t *src2, int i_src2, int width, int height )
56 57 58 59 60 61
{
    int x, y;
    for( y = 0; y < height; y++ )
    {
        for( x = 0; x < width; x++ )
        {
62
            dst[x] = ( src1[x] + src2[x] + 1 ) >> 1;
63
        }
64 65
        src1 += i_src1;
        src2 += i_src2;
66 67 68 69 70 71
        dst += i_dst;
    }
}

/* Implicit weighted bipred only:
 * assumes log2_denom = 5, offset = 0, weight1 + weight2 = 64 */
72 73 74
#define op_scale2(x) dst[x] = x264_clip_uint8( (src1[x]*i_weight1 + src2[x]*i_weight2 + (1<<5)) >> 6 )
static inline void pixel_avg_weight_wxh( uint8_t *dst, int i_dst, uint8_t *src1, int i_src1, uint8_t *src2, int i_src2, int width, int height, int i_weight1 )
{
75 76
    int y;
    const int i_weight2 = 64 - i_weight1;
77 78
    for( y = 0; y<height; y++, dst += i_dst, src1 += i_src1, src2 += i_src2 )
    {
79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
        op_scale2(0);
        op_scale2(1);
        if(width==2) continue;
        op_scale2(2);
        op_scale2(3);
        if(width==4) continue;
        op_scale2(4);
        op_scale2(5);
        op_scale2(6);
        op_scale2(7);
        if(width==8) continue;
        op_scale2(8);
        op_scale2(9);
        op_scale2(10);
        op_scale2(11);
        op_scale2(12);
        op_scale2(13);
        op_scale2(14);
        op_scale2(15);
    }
}
Fiona Glaser's avatar
Fiona Glaser committed
100
#undef op_scale2
101

Fiona Glaser's avatar
Fiona Glaser committed
102 103 104 105
#define PIXEL_AVG_C( name, width, height ) \
static void name( uint8_t *pix1, int i_stride_pix1, \
                  uint8_t *pix2, int i_stride_pix2, \
                  uint8_t *pix3, int i_stride_pix3, int weight ) \
106
{ \
Fiona Glaser's avatar
Fiona Glaser committed
107 108 109 110
    if( weight == 32 )\
        pixel_avg_wxh( pix1, i_stride_pix1, pix2, i_stride_pix2, pix3, i_stride_pix3, width, height ); \
    else\
        pixel_avg_weight_wxh( pix1, i_stride_pix1, pix2, i_stride_pix2, pix3, i_stride_pix3, width, height, weight ); \
111
}
Fiona Glaser's avatar
Fiona Glaser committed
112 113 114 115 116 117 118 119 120 121
PIXEL_AVG_C( pixel_avg_16x16, 16, 16 )
PIXEL_AVG_C( pixel_avg_16x8,  16, 8 )
PIXEL_AVG_C( pixel_avg_8x16,  8, 16 )
PIXEL_AVG_C( pixel_avg_8x8,   8, 8 )
PIXEL_AVG_C( pixel_avg_8x4,   8, 4 )
PIXEL_AVG_C( pixel_avg_4x8,   4, 8 )
PIXEL_AVG_C( pixel_avg_4x4,   4, 4 )
PIXEL_AVG_C( pixel_avg_4x2,   4, 2 )
PIXEL_AVG_C( pixel_avg_2x4,   2, 4 )
PIXEL_AVG_C( pixel_avg_2x2,   2, 2 )
122

Dylan Yudaken's avatar
Dylan Yudaken committed
123 124 125 126
static void x264_weight_cache( x264_t *h, x264_weight_t *w )
{
    w->weightfn = h->mc.weight;
}
Loren Merritt's avatar
Loren Merritt committed
127 128
#define opscale(x) dst[x] = x264_clip_uint8( ((src[x] * weight->i_scale + (1<<(weight->i_denom - 1))) >> weight->i_denom) + weight->i_offset )
#define opscale_noden(x) dst[x] = x264_clip_uint8( src[x] * weight->i_scale + weight->i_offset )
Dylan Yudaken's avatar
Dylan Yudaken committed
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
static inline void mc_weight( uint8_t *dst, int i_dst_stride, uint8_t *src, int i_src_stride, const x264_weight_t *weight, int i_width, int i_height )
{

    int x, y;
    if( weight->i_denom >= 1 )
    {
        for( y = 0; y < i_height; y++, dst += i_dst_stride, src += i_src_stride )
        {
            for( x = 0; x < i_width; x++ )
                opscale( x );
        }
    }
    else
    {
        for( y = 0; y < i_height; y++, dst += i_dst_stride, src += i_src_stride )
            for( x = 0; x < i_width; x++ )
                opscale_noden( x );
    }
}

#define MC_WEIGHT_C( name, lx ) \
    static void name( uint8_t *dst, int i_dst_stride, uint8_t *src, int i_src_stride, const x264_weight_t *weight, int height ) \
{ \
    int x, y; \
    if( weight->i_denom >= 1 ) \
    { \
        for( y = 0; y < height; y++, dst += i_dst_stride, src += i_src_stride ) \
            for( x = 0; x < lx; x++ ) \
                opscale( x ); \
    } \
    else \
    { \
        for( y = 0; y < height; y++, dst += i_dst_stride, src += i_src_stride ) \
            for( x = 0; x < lx; x++ ) \
                opscale_noden( x ); \
    } \
}

MC_WEIGHT_C( mc_weight_w20, 20 )
MC_WEIGHT_C( mc_weight_w16, 16 )
MC_WEIGHT_C( mc_weight_w12, 12 )
MC_WEIGHT_C( mc_weight_w8,   8 )
MC_WEIGHT_C( mc_weight_w4,   4 )
MC_WEIGHT_C( mc_weight_w2,   2 )

static weight_fn_t x264_mc_weight_wtab[6] =
{
    mc_weight_w2,
    mc_weight_w4,
    mc_weight_w8,
    mc_weight_w12,
    mc_weight_w16,
    mc_weight_w20,
};
const x264_weight_t weight_none[3] = { {{0}} };
Laurent Aimar's avatar
Laurent Aimar committed
184 185 186 187 188 189 190 191 192 193 194 195 196
static void mc_copy( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_width, int i_height )
{
    int y;

    for( y = 0; y < i_height; y++ )
    {
        memcpy( dst, src, i_width );

        src += i_src_stride;
        dst += i_dst_stride;
    }
}

Loren Merritt's avatar
Loren Merritt committed
197 198
#define TAPFILTER(pix, d) ((pix)[x-2*d] + (pix)[x+3*d] - 5*((pix)[x-d] + (pix)[x+2*d]) + 20*((pix)[x] + (pix)[x+d]))
static void hpel_filter( uint8_t *dsth, uint8_t *dstv, uint8_t *dstc, uint8_t *src,
199
                         int stride, int width, int height, int16_t *buf )
Laurent Aimar's avatar
Laurent Aimar committed
200 201
{
    int x, y;
Loren Merritt's avatar
Loren Merritt committed
202
    for( y=0; y<height; y++ )
Laurent Aimar's avatar
Laurent Aimar committed
203
    {
Loren Merritt's avatar
Loren Merritt committed
204
        for( x=-2; x<width+3; x++ )
Laurent Aimar's avatar
Laurent Aimar committed
205
        {
Loren Merritt's avatar
Loren Merritt committed
206 207 208
            int v = TAPFILTER(src,stride);
            dstv[x] = x264_clip_uint8((v + 16) >> 5);
            buf[x+2] = v;
Laurent Aimar's avatar
Laurent Aimar committed
209
        }
Loren Merritt's avatar
Loren Merritt committed
210 211 212 213 214 215 216 217
        for( x=0; x<width; x++ )
            dstc[x] = x264_clip_uint8((TAPFILTER(buf+2,1) + 512) >> 10);
        for( x=0; x<width; x++ )
            dsth[x] = x264_clip_uint8((TAPFILTER(src,1) + 16) >> 5);
        dsth += stride;
        dstv += stride;
        dstc += stride;
        src += stride;
Laurent Aimar's avatar
Laurent Aimar committed
218
    }
219 220
}

Loren Merritt's avatar
Loren Merritt committed
221 222 223
static const int hpel_ref0[16] = {0,1,1,1,0,1,1,1,2,3,3,3,0,1,1,1};
static const int hpel_ref1[16] = {0,0,0,0,2,2,3,2,2,2,3,2,2,2,3,2};

224 225
static void mc_luma( uint8_t *dst,    int i_dst_stride,
                     uint8_t *src[4], int i_src_stride,
Loren Merritt's avatar
Loren Merritt committed
226
                     int mvx, int mvy,
Dylan Yudaken's avatar
Dylan Yudaken committed
227
                     int i_width, int i_height, const x264_weight_t *weight )
228
{
Loren Merritt's avatar
Loren Merritt committed
229 230 231
    int qpel_idx = ((mvy&3)<<2) + (mvx&3);
    int offset = (mvy>>2)*i_src_stride + (mvx>>2);
    uint8_t *src1 = src[hpel_ref0[qpel_idx]] + offset + ((mvy&3) == 3) * i_src_stride;
232

Loren Merritt's avatar
Loren Merritt committed
233
    if( qpel_idx & 5 ) /* qpel interpolation needed */
234
    {
Loren Merritt's avatar
Loren Merritt committed
235
        uint8_t *src2 = src[hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
236 237
        pixel_avg( dst, i_dst_stride, src1, i_src_stride,
                   src2, i_src_stride, i_width, i_height );
Dylan Yudaken's avatar
Dylan Yudaken committed
238 239
        if( weight->weightfn )
            mc_weight( dst, i_dst_stride, dst, i_dst_stride, weight, i_width, i_height );
240
    }
Dylan Yudaken's avatar
Dylan Yudaken committed
241 242
    else if( weight->weightfn )
        mc_weight( dst, i_dst_stride, src1, i_src_stride, weight, i_width, i_height );
243 244 245 246
    else
        mc_copy( src1, i_src_stride, dst, i_dst_stride, i_width, i_height );
}

247 248
static uint8_t *get_ref( uint8_t *dst,   int *i_dst_stride,
                         uint8_t *src[4], int i_src_stride,
Loren Merritt's avatar
Loren Merritt committed
249
                         int mvx, int mvy,
Dylan Yudaken's avatar
Dylan Yudaken committed
250
                         int i_width, int i_height, const x264_weight_t *weight )
251
{
Loren Merritt's avatar
Loren Merritt committed
252 253 254
    int qpel_idx = ((mvy&3)<<2) + (mvx&3);
    int offset = (mvy>>2)*i_src_stride + (mvx>>2);
    uint8_t *src1 = src[hpel_ref0[qpel_idx]] + offset + ((mvy&3) == 3) * i_src_stride;
255

Loren Merritt's avatar
Loren Merritt committed
256
    if( qpel_idx & 5 ) /* qpel interpolation needed */
257
    {
Loren Merritt's avatar
Loren Merritt committed
258
        uint8_t *src2 = src[hpel_ref1[qpel_idx]] + offset + ((mvx&3) == 3);
259 260
        pixel_avg( dst, *i_dst_stride, src1, i_src_stride,
                   src2, i_src_stride, i_width, i_height );
Dylan Yudaken's avatar
Dylan Yudaken committed
261 262 263 264 265 266 267
        if( weight->weightfn )
            mc_weight( dst, *i_dst_stride, dst, *i_dst_stride, weight, i_width, i_height );
        return dst;
    }
    else if( weight->weightfn )
    {
        mc_weight( dst, *i_dst_stride, src1, i_src_stride, weight, i_width, i_height );
268 269 270 271 272 273 274 275 276
        return dst;
    }
    else
    {
        *i_dst_stride = i_src_stride;
        return src1;
    }
}

Laurent Aimar's avatar
Laurent Aimar committed
277
/* full chroma mc (ie until 1/8 pixel)*/
278 279 280 281
static void mc_chroma( uint8_t *dst, int i_dst_stride,
                       uint8_t *src, int i_src_stride,
                       int mvx, int mvy,
                       int i_width, int i_height )
Laurent Aimar's avatar
Laurent Aimar committed
282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
{
    uint8_t *srcp;
    int x, y;

    const int d8x = mvx&0x07;
    const int d8y = mvy&0x07;

    const int cA = (8-d8x)*(8-d8y);
    const int cB = d8x    *(8-d8y);
    const int cC = (8-d8x)*d8y;
    const int cD = d8x    *d8y;

    src  += (mvy >> 3) * i_src_stride + (mvx >> 3);
    srcp = &src[i_src_stride];

    for( y = 0; y < i_height; y++ )
    {
        for( x = 0; x < i_width; x++ )
        {
            dst[x] = ( cA*src[x]  + cB*src[x+1] +
                       cC*srcp[x] + cD*srcp[x+1] + 32 ) >> 6;
        }
        dst  += i_dst_stride;

        src   = srcp;
        srcp += i_src_stride;
    }
}

Loren Merritt's avatar
Loren Merritt committed
311 312 313 314 315 316 317 318 319
#define MC_COPY(W) \
static void mc_copy_w##W( uint8_t *dst, int i_dst, uint8_t *src, int i_src, int i_height ) \
{ \
    mc_copy( src, i_src, dst, i_dst, W, i_height ); \
}
MC_COPY( 16 )
MC_COPY( 8 )
MC_COPY( 4 )

320 321 322 323 324 325 326 327 328 329 330
static void plane_copy( uint8_t *dst, int i_dst,
                        uint8_t *src, int i_src, int w, int h)
{
    while( h-- )
    {
        memcpy( dst, src, w );
        dst += i_dst;
        src += i_src;
    }
}

Fiona Glaser's avatar
Fiona Glaser committed
331 332
static void prefetch_fenc_null( uint8_t *pix_y, int stride_y,
                                uint8_t *pix_uv, int stride_uv, int mb_x )
333 334
{}

Fiona Glaser's avatar
Fiona Glaser committed
335
static void prefetch_ref_null( uint8_t *pix, int stride, int parity )
336 337
{}

Fiona Glaser's avatar
Fiona Glaser committed
338 339 340 341 342
static void memzero_aligned( void * dst, int n )
{
    memset( dst, 0, n );
}

Loren Merritt's avatar
Loren Merritt committed
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378
static void integral_init4h( uint16_t *sum, uint8_t *pix, int stride )
{
    int x, v = pix[0]+pix[1]+pix[2]+pix[3];
    for( x=0; x<stride-4; x++ )
    {
        sum[x] = v + sum[x-stride];
        v += pix[x+4] - pix[x];
    }
}

static void integral_init8h( uint16_t *sum, uint8_t *pix, int stride )
{
    int x, v = pix[0]+pix[1]+pix[2]+pix[3]+pix[4]+pix[5]+pix[6]+pix[7];
    for( x=0; x<stride-8; x++ )
    {
        sum[x] = v + sum[x-stride];
        v += pix[x+8] - pix[x];
    }
}

static void integral_init4v( uint16_t *sum8, uint16_t *sum4, int stride )
{
    int x;
    for( x=0; x<stride-8; x++ )
        sum4[x] = sum8[x+4*stride] - sum8[x];
    for( x=0; x<stride-8; x++ )
        sum8[x] = sum8[x+8*stride] + sum8[x+8*stride+4] - sum8[x] - sum8[x+4];
}

static void integral_init8v( uint16_t *sum8, int stride )
{
    int x;
    for( x=0; x<stride-8; x++ )
        sum8[x] = sum8[x+8*stride] - sum8[x];
}

Loren Merritt's avatar
Loren Merritt committed
379 380 381 382 383 384 385 386 387 388 389
void x264_frame_init_lowres( x264_t *h, x264_frame_t *frame )
{
    uint8_t *src = frame->plane[0];
    int i_stride = frame->i_stride[0];
    int i_height = frame->i_lines[0];
    int i_width  = frame->i_width[0];
    int x, y;

    // duplicate last row and column so that their interpolation doesn't have to be special-cased
    for( y=0; y<i_height; y++ )
        src[i_width+y*i_stride] = src[i_width-1+y*i_stride];
390
    memcpy( src+i_stride*i_height, src+i_stride*(i_height-1), i_width+1 );
Loren Merritt's avatar
Loren Merritt committed
391 392 393 394
    h->mc.frame_init_lowres_core( src, frame->lowres[0], frame->lowres[1], frame->lowres[2], frame->lowres[3],
                                  i_stride, frame->i_stride_lowres, frame->i_width_lowres, frame->i_lines_lowres );
    x264_frame_expand_border_lowres( frame );

395 396 397 398 399
    memset( frame->i_cost_est, -1, sizeof(frame->i_cost_est) );

    for( x = 0; x < h->param.i_bframe + 2; x++ )
        for( y = 0; y < h->param.i_bframe + 2; y++ )
            frame->i_row_satds[y][x][0] = -1;
400 401 402 403

    for( y = 0; y <= !!h->param.i_bframe; y++ )
        for( x = 0; x <= h->param.i_bframe; x++ )
            frame->lowres_mvs[y][x][0][0] = 0x7FFF;
Loren Merritt's avatar
Loren Merritt committed
404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
}

static void frame_init_lowres_core( uint8_t *src0, uint8_t *dst0, uint8_t *dsth, uint8_t *dstv, uint8_t *dstc,
                                    int src_stride, int dst_stride, int width, int height )
{
    int x,y;
    for( y=0; y<height; y++ )
    {
        uint8_t *src1 = src0+src_stride;
        uint8_t *src2 = src1+src_stride;
        for( x=0; x<width; x++ )
        {
            // slower than naive bilinear, but matches asm
#define FILTER(a,b,c,d) ((((a+b+1)>>1)+((c+d+1)>>1)+1)>>1)
            dst0[x] = FILTER(src0[2*x  ], src1[2*x  ], src0[2*x+1], src1[2*x+1]);
            dsth[x] = FILTER(src0[2*x+1], src1[2*x+1], src0[2*x+2], src1[2*x+2]);
            dstv[x] = FILTER(src1[2*x  ], src2[2*x  ], src1[2*x+1], src2[2*x+1]);
            dstc[x] = FILTER(src1[2*x+1], src2[2*x+1], src1[2*x+2], src2[2*x+2]);
#undef FILTER
        }
        src0 += src_stride*2;
        dst0 += dst_stride;
        dsth += dst_stride;
        dstv += dst_stride;
        dstc += dst_stride;
    }
}

432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458
#if defined(__GNUC__) && (defined(ARCH_X86) || defined(ARCH_X86_64))
// gcc isn't smart enough to use the "idiv" instruction
static ALWAYS_INLINE int32_t div_64_32(int64_t x, int32_t y) {
    int32_t quotient, remainder;
    asm("idiv %4"
        :"=a"(quotient), "=d"(remainder)
        :"a"((uint32_t)x), "d"((int32_t)(x>>32)), "r"(y)
    );
    return quotient;
}
#else
#define div_64_32(x,y) ((x)/(y))
#endif

/* Estimate the total amount of influence on future quality that could be had if we
 * were to improve the reference samples used to inter predict any given macroblock. */
static void mbtree_propagate_cost( int *dst, uint16_t *propagate_in, uint16_t *intra_costs,
                                   uint16_t *inter_costs, uint16_t *inv_qscales, int len )
{
    int i;
    for( i=0; i<len; i++ )
    {
        int propagate_amount = propagate_in[i] + ((intra_costs[i] * inv_qscales[i] + 128)>>8);
        dst[i] = div_64_32((int64_t)propagate_amount * (intra_costs[i] - inter_costs[i]), intra_costs[i]);
    }
}

459
void x264_mc_init( int cpu, x264_mc_functions_t *pf )
Laurent Aimar's avatar
Laurent Aimar committed
460
{
461 462
    pf->mc_luma   = mc_luma;
    pf->get_ref   = get_ref;
463
    pf->mc_chroma = mc_chroma;
Laurent Aimar's avatar
Laurent Aimar committed
464

465 466 467 468 469 470 471 472 473 474
    pf->avg[PIXEL_16x16]= pixel_avg_16x16;
    pf->avg[PIXEL_16x8] = pixel_avg_16x8;
    pf->avg[PIXEL_8x16] = pixel_avg_8x16;
    pf->avg[PIXEL_8x8]  = pixel_avg_8x8;
    pf->avg[PIXEL_8x4]  = pixel_avg_8x4;
    pf->avg[PIXEL_4x8]  = pixel_avg_4x8;
    pf->avg[PIXEL_4x4]  = pixel_avg_4x4;
    pf->avg[PIXEL_4x2]  = pixel_avg_4x2;
    pf->avg[PIXEL_2x4]  = pixel_avg_2x4;
    pf->avg[PIXEL_2x2]  = pixel_avg_2x2;
Loren Merritt's avatar
Loren Merritt committed
475

Dylan Yudaken's avatar
Dylan Yudaken committed
476 477 478 479 480
    pf->weight    = x264_mc_weight_wtab;
    pf->offsetadd = x264_mc_weight_wtab;
    pf->offsetsub = x264_mc_weight_wtab;
    pf->weight_cache = x264_weight_cache;

481
    pf->copy_16x16_unaligned = mc_copy_w16;
Loren Merritt's avatar
Loren Merritt committed
482 483 484 485
    pf->copy[PIXEL_16x16] = mc_copy_w16;
    pf->copy[PIXEL_8x8]   = mc_copy_w8;
    pf->copy[PIXEL_4x4]   = mc_copy_w4;

486
    pf->plane_copy = plane_copy;
487
    pf->hpel_filter = hpel_filter;
488

489 490
    pf->prefetch_fenc = prefetch_fenc_null;
    pf->prefetch_ref  = prefetch_ref_null;
Fiona Glaser's avatar
Fiona Glaser committed
491
    pf->memcpy_aligned = memcpy;
Fiona Glaser's avatar
Fiona Glaser committed
492
    pf->memzero_aligned = memzero_aligned;
Loren Merritt's avatar
Loren Merritt committed
493
    pf->frame_init_lowres_core = frame_init_lowres_core;
494

Loren Merritt's avatar
Loren Merritt committed
495 496 497 498 499
    pf->integral_init4h = integral_init4h;
    pf->integral_init8h = integral_init8h;
    pf->integral_init4v = integral_init4v;
    pf->integral_init8v = integral_init8v;

500 501
    pf->mbtree_propagate_cost = mbtree_propagate_cost;

502
#ifdef HAVE_MMX
Loren Merritt's avatar
Loren Merritt committed
503
    x264_mc_init_mmx( cpu, pf );
504
#endif
505
#ifdef ARCH_PPC
Laurent Aimar's avatar
Laurent Aimar committed
506 507 508
    if( cpu&X264_CPU_ALTIVEC )
        x264_mc_altivec_init( pf );
#endif
509 510 511
#ifdef HAVE_ARMV6
    x264_mc_init_arm( cpu, pf );
#endif
Laurent Aimar's avatar
Laurent Aimar committed
512 513
}

514
void x264_frame_filter( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
515
{
516
    const int b_interlaced = h->sh.b_mbaff;
517
    const int stride = frame->i_stride[0] << b_interlaced;
518
    const int width = frame->i_width[0];
519
    int start = (mb_y*16 >> b_interlaced) - 8; // buffer = 4 for deblock + 3 for 6tap, rounded to 8
Loren Merritt's avatar
Loren Merritt committed
520
    int height = ((b_end ? frame->i_lines[0] : mb_y*16) >> b_interlaced) + 8;
521
    int offs = start*stride - 8; // buffer = 3 for 6tap, aligned to 8 for simd
Loren Merritt's avatar
Loren Merritt committed
522
    int y;
523

Loren Merritt's avatar
Loren Merritt committed
524 525
    if( mb_y & b_interlaced )
        return;
526

527
    for( y=0; y<=b_interlaced; y++, offs+=frame->i_stride[0] )
528
    {
529
        h->mc.hpel_filter(
530 531 532 533
            frame->filtered[1] + offs,
            frame->filtered[2] + offs,
            frame->filtered[3] + offs,
            frame->plane[0] + offs,
534 535
            stride, width + 16, height - start,
            h->scratch_buffer );
536
    }
537 538

    /* generate integral image:
Loren Merritt's avatar
Loren Merritt committed
539 540
     * frame->integral contains 2 planes. in the upper plane, each element is
     * the sum of an 8x8 pixel region with top-left corner on that point.
Loren Merritt's avatar
Loren Merritt committed
541
     * in the lower plane, 4x4 sums (needed only with --partitions p4x4). */
542

Loren Merritt's avatar
Loren Merritt committed
543
    if( frame->integral )
544
    {
Loren Merritt's avatar
Loren Merritt committed
545 546
        if( start < 0 )
        {
547 548
            memset( frame->integral - PADV * stride - PADH, 0, stride * sizeof(uint16_t) );
            start = -PADV;
Loren Merritt's avatar
Loren Merritt committed
549 550
        }
        if( b_end )
551
            height += PADV-9;
Loren Merritt's avatar
Loren Merritt committed
552
        for( y = start; y < height; y++ )
553
        {
Loren Merritt's avatar
Loren Merritt committed
554 555 556 557 558 559 560 561 562 563 564 565
            uint8_t  *pix  = frame->plane[0] + y * stride - PADH;
            uint16_t *sum8 = frame->integral + (y+1) * stride - PADH;
            uint16_t *sum4;
            if( h->frames.b_have_sub8x8_esa )
            {
                h->mc.integral_init4h( sum8, pix, stride );
                sum8 -= 8*stride;
                sum4 = sum8 + stride * (frame->i_lines[0] + PADV*2);
                if( y >= 8-PADV )
                    h->mc.integral_init4v( sum8, sum4, stride );
            }
            else
Loren Merritt's avatar
Loren Merritt committed
566
            {
Loren Merritt's avatar
Loren Merritt committed
567 568 569
                h->mc.integral_init8h( sum8, pix, stride );
                if( y >= 8-PADV )
                    h->mc.integral_init8v( sum8-8*stride, stride );
Loren Merritt's avatar
Loren Merritt committed
570 571
            }
        }
572
    }
573
}