reconinter.c 43.6 KB
Newer Older
John Koleszar's avatar
John Koleszar committed
1
/*
2
 *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
John Koleszar's avatar
John Koleszar committed
3
 *
4
 *  Use of this source code is governed by a BSD-style license
5 6
 *  that can be found in the LICENSE file in the root of the source
 *  tree. An additional intellectual property rights grant can be found
7
 *  in the file PATENTS.  All contributing project authors may
8
 *  be found in the AUTHORS file in the root of the source tree.
John Koleszar's avatar
John Koleszar committed
9 10 11 12
 */


#include "vpx_ports/config.h"
13
#include "vpx/vpx_integer.h"
John Koleszar's avatar
John Koleszar committed
14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
#include "recon.h"
#include "subpixel.h"
#include "blockd.h"
#include "reconinter.h"
#if CONFIG_RUNTIME_CPU_DETECT
#include "onyxc_int.h"
#endif

void vp8_copy_mem16x16_c(
    unsigned char *src,
    int src_stride,
    unsigned char *dst,
    int dst_stride)
{

    int r;

    for (r = 0; r < 16; r++)
    {
33
#if !(CONFIG_FAST_UNALIGNED)
John Koleszar's avatar
John Koleszar committed
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51
        dst[0] = src[0];
        dst[1] = src[1];
        dst[2] = src[2];
        dst[3] = src[3];
        dst[4] = src[4];
        dst[5] = src[5];
        dst[6] = src[6];
        dst[7] = src[7];
        dst[8] = src[8];
        dst[9] = src[9];
        dst[10] = src[10];
        dst[11] = src[11];
        dst[12] = src[12];
        dst[13] = src[13];
        dst[14] = src[14];
        dst[15] = src[15];

#else
52 53 54 55
        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
        ((uint32_t *)dst)[2] = ((uint32_t *)src)[2] ;
        ((uint32_t *)dst)[3] = ((uint32_t *)src)[3] ;
John Koleszar's avatar
John Koleszar committed
56 57 58 59 60 61 62 63 64

#endif
        src += src_stride;
        dst += dst_stride;

    }

}

65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86
void vp8_avg_mem16x16_c(
    unsigned char *src,
    int src_stride,
    unsigned char *dst,
    int dst_stride)
{
    int r;

    for (r = 0; r < 16; r++)
    {
        int n;

        for (n = 0; n < 16; n++)
        {
            dst[n] = (dst[n] + src[n] + 1) >> 1;
        }

        src += src_stride;
        dst += dst_stride;
    }
}

John Koleszar's avatar
John Koleszar committed
87 88 89 90 91 92 93 94 95 96
void vp8_copy_mem8x8_c(
    unsigned char *src,
    int src_stride,
    unsigned char *dst,
    int dst_stride)
{
    int r;

    for (r = 0; r < 8; r++)
    {
97
#if !(CONFIG_FAST_UNALIGNED)
John Koleszar's avatar
John Koleszar committed
98 99 100 101 102 103 104 105 106
        dst[0] = src[0];
        dst[1] = src[1];
        dst[2] = src[2];
        dst[3] = src[3];
        dst[4] = src[4];
        dst[5] = src[5];
        dst[6] = src[6];
        dst[7] = src[7];
#else
107 108
        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
John Koleszar's avatar
John Koleszar committed
109 110 111 112 113 114 115 116
#endif
        src += src_stride;
        dst += dst_stride;

    }

}

117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
void vp8_avg_mem8x8_c(
    unsigned char *src,
    int src_stride,
    unsigned char *dst,
    int dst_stride)
{
    int r;

    for (r = 0; r < 8; r++)
    {
        int n;

        for (n = 0; n < 8; n++)
        {
            dst[n] = (dst[n] + src[n] + 1) >> 1;
        }

        src += src_stride;
        dst += dst_stride;
    }
}

John Koleszar's avatar
John Koleszar committed
139 140 141 142 143 144 145 146 147 148
void vp8_copy_mem8x4_c(
    unsigned char *src,
    int src_stride,
    unsigned char *dst,
    int dst_stride)
{
    int r;

    for (r = 0; r < 4; r++)
    {
149
#if !(CONFIG_FAST_UNALIGNED)
John Koleszar's avatar
John Koleszar committed
150 151 152 153 154 155 156 157 158
        dst[0] = src[0];
        dst[1] = src[1];
        dst[2] = src[2];
        dst[3] = src[3];
        dst[4] = src[4];
        dst[5] = src[5];
        dst[6] = src[6];
        dst[7] = src[7];
#else
159 160
        ((uint32_t *)dst)[0] = ((uint32_t *)src)[0] ;
        ((uint32_t *)dst)[1] = ((uint32_t *)src)[1] ;
John Koleszar's avatar
John Koleszar committed
161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
#endif
        src += src_stride;
        dst += dst_stride;

    }

}



void vp8_build_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf)
{
    int r;
    unsigned char *ptr_base;
    unsigned char *ptr;
    unsigned char *pred_ptr = d->predictor;
177
    int_mv mv;
John Koleszar's avatar
John Koleszar committed
178 179

    ptr_base = *(d->base_pre);
180
    mv.as_int = d->bmi.as_mv.first.as_int;
John Koleszar's avatar
John Koleszar committed
181

182
    if (mv.as_mv.row & 7 || mv.as_mv.col & 7)
John Koleszar's avatar
John Koleszar committed
183
    {
184
        ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride + (mv.as_mv.col >> 3);
185
#if CONFIG_SIXTEENTH_SUBPEL_UV
186
        sppf(ptr, d->pre_stride, (mv.as_mv.col & 7)<<1, (mv.as_mv.row & 7)<<1, pred_ptr, pitch);
187
#else
188
        sppf(ptr, d->pre_stride, mv.as_mv.col & 7,mv.as_mv.row & 7, pred_ptr, pitch);
189
#endif
John Koleszar's avatar
John Koleszar committed
190 191 192
    }
    else
    {
193
        ptr_base += d->pre + (mv.as_mv.row >> 3) * d->pre_stride + (mv.as_mv.col >> 3);
John Koleszar's avatar
John Koleszar committed
194 195 196 197
        ptr = ptr_base;

        for (r = 0; r < 4; r++)
        {
198
#if !(CONFIG_FAST_UNALIGNED)
John Koleszar's avatar
John Koleszar committed
199 200 201 202 203
            pred_ptr[0]  = ptr[0];
            pred_ptr[1]  = ptr[1];
            pred_ptr[2]  = ptr[2];
            pred_ptr[3]  = ptr[3];
#else
204
            *(uint32_t *)pred_ptr = *(uint32_t *)ptr ;
John Koleszar's avatar
John Koleszar committed
205 206 207 208 209 210 211
#endif
            pred_ptr     += pitch;
            ptr         += d->pre_stride;
        }
    }
}

212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
/*
 * Similar to vp8_build_inter_predictors_b(), but instead of storing the
 * results in d->predictor, we average the contents of d->predictor (which
 * come from an earlier call to vp8_build_inter_predictors_b()) with the
 * predictor of the second reference frame / motion vector.
 */
void vp8_build_2nd_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf)
{
    int r;
    unsigned char *ptr_base;
    unsigned char *ptr;
    unsigned char *pred_ptr = d->predictor;
    int_mv mv;

    ptr_base = *(d->base_second_pre);
    mv.as_int = d->bmi.as_mv.second.as_int;

    if (mv.as_mv.row & 7 || mv.as_mv.col & 7)
    {
        ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride + (mv.as_mv.col >> 3);
#if CONFIG_SIXTEENTH_SUBPEL_UV
        sppf(ptr, d->pre_stride, (mv.as_mv.col & 7)<<1, (mv.as_mv.row & 7)<<1, pred_ptr, pitch);
#else
        sppf(ptr, d->pre_stride, mv.as_mv.col & 7,mv.as_mv.row & 7, pred_ptr, pitch);
#endif
    }
    else
    {
        ptr_base += d->pre + (mv.as_mv.row >> 3) * d->pre_stride + (mv.as_mv.col >> 3);
        ptr = ptr_base;

        for (r = 0; r < 4; r++)
        {
            pred_ptr[0]  = (pred_ptr[0] + ptr[0] + 1) >> 1;
            pred_ptr[1]  = (pred_ptr[1] + ptr[1] + 1) >> 1;
            pred_ptr[2]  = (pred_ptr[2] + ptr[2] + 1) >> 1;
            pred_ptr[3]  = (pred_ptr[3] + ptr[3] + 1) >> 1;
            pred_ptr    += pitch;
            ptr         += d->pre_stride;
        }
    }
}

255
static void build_inter_predictors4b(MACROBLOCKD *x, BLOCKD *d, int pitch)
John Koleszar's avatar
John Koleszar committed
256 257 258 259
{
    unsigned char *ptr_base;
    unsigned char *ptr;
    unsigned char *pred_ptr = d->predictor;
260
    int_mv mv;
John Koleszar's avatar
John Koleszar committed
261 262

    ptr_base = *(d->base_pre);
263 264
    mv.as_int = d->bmi.as_mv.first.as_int;
    ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride + (mv.as_mv.col >> 3);
John Koleszar's avatar
John Koleszar committed
265

266
    if (mv.as_mv.row & 7 || mv.as_mv.col & 7)
John Koleszar's avatar
John Koleszar committed
267
    {
268
#if CONFIG_SIXTEENTH_SUBPEL_UV
269
        x->subpixel_predict8x8(ptr, d->pre_stride, (mv.as_mv.col & 7)<<1, (mv.as_mv.row & 7)<<1, pred_ptr, pitch);
270
#else
271
        x->subpixel_predict8x8(ptr, d->pre_stride, mv.as_mv.col & 7, mv.as_mv.row & 7, pred_ptr, pitch);
272
#endif
John Koleszar's avatar
John Koleszar committed
273 274 275 276 277 278 279
    }
    else
    {
        RECON_INVOKE(&x->rtcd->recon, copy8x8)(ptr, d->pre_stride, pred_ptr, pitch);
    }
}

280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
/*
 * Similar to build_inter_predictors_4b(), but instead of storing the
 * results in d->predictor, we average the contents of d->predictor (which
 * come from an earlier call to build_inter_predictors_4b()) with the
 * predictor of the second reference frame / motion vector.
 */
static void build_2nd_inter_predictors4b(MACROBLOCKD *x, BLOCKD *d, int pitch)
{
    unsigned char *ptr_base;
    unsigned char *ptr;
    unsigned char *pred_ptr = d->predictor;
    int_mv mv;

    ptr_base = *(d->base_second_pre);
    mv.as_int = d->bmi.as_mv.second.as_int;
    ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride + (mv.as_mv.col >> 3);

    if (mv.as_mv.row & 7 || mv.as_mv.col & 7)
    {
#if CONFIG_SIXTEENTH_SUBPEL_UV
        x->subpixel_predict_avg8x8(ptr, d->pre_stride, (mv.as_mv.col & 7)<<1, (mv.as_mv.row & 7)<<1, pred_ptr, pitch);
#else
        x->subpixel_predict_avg8x8(ptr, d->pre_stride, mv.as_mv.col & 7, mv.as_mv.row & 7, pred_ptr, pitch);
#endif
    }
    else
    {
        RECON_INVOKE(&x->rtcd->recon, avg8x8)(ptr, d->pre_stride, pred_ptr, pitch);
    }
}

311
static void build_inter_predictors2b(MACROBLOCKD *x, BLOCKD *d, int pitch)
John Koleszar's avatar
John Koleszar committed
312 313 314 315
{
    unsigned char *ptr_base;
    unsigned char *ptr;
    unsigned char *pred_ptr = d->predictor;
316
    int_mv mv;
John Koleszar's avatar
John Koleszar committed
317 318

    ptr_base = *(d->base_pre);
319 320
    mv.as_int = d->bmi.as_mv.first.as_int;
    ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride + (mv.as_mv.col >> 3);
John Koleszar's avatar
John Koleszar committed
321

322
    if (mv.as_mv.row & 7 || mv.as_mv.col & 7)
John Koleszar's avatar
John Koleszar committed
323
    {
324
#if CONFIG_SIXTEENTH_SUBPEL_UV
325
        x->subpixel_predict8x4(ptr, d->pre_stride, (mv.as_mv.col & 7)<<1, (mv.as_mv.row & 7)<<1, pred_ptr, pitch);
326
#else
327
        x->subpixel_predict8x4(ptr, d->pre_stride, mv.as_mv.col & 7, mv.as_mv.row & 7, pred_ptr, pitch);
328
#endif
John Koleszar's avatar
John Koleszar committed
329 330 331 332 333 334 335 336
    }
    else
    {
        RECON_INVOKE(&x->rtcd->recon, copy8x4)(ptr, d->pre_stride, pred_ptr, pitch);
    }
}


337
/*encoder only*/
338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473
#if CONFIG_PRED_FILTER

// Select the thresholded or non-thresholded filter
#define USE_THRESH_FILTER 0

#define PRED_FILT_LEN 5

static const int filt_shift = 4;
static const int pred_filter[PRED_FILT_LEN] = {1, 2, 10, 2, 1};
// Alternative filter {1, 1, 4, 1, 1}

#if !USE_THRESH_FILTER
void filter_mb(unsigned char *src, int src_stride,
               unsigned char *dst, int dst_stride,
               int width, int height)
{
    int i, j, k;
    unsigned int Temp[32*32];
    unsigned int  *pTmp = Temp;
    unsigned char *pSrc = src - (1 + src_stride) * (PRED_FILT_LEN/2);

    // Horizontal
    for (i=0; i<height+PRED_FILT_LEN-1; i++)
    {
       for (j=0; j<width; j++)
       {
           int sum=0;
           for (k=0; k<PRED_FILT_LEN; k++)
               sum += pSrc[j+k] * pred_filter[k];
           pTmp[j] = sum;
       }

       pSrc += src_stride;
       pTmp += width;
    }

    // Vertical
    pTmp = Temp;
    for (i=0; i<width; i++)
    {
        unsigned char *pDst = dst + i;
        for (j=0; j<height; j++)
        {
            int sum=0;
            for (k=0; k<PRED_FILT_LEN; k++)
                sum += pTmp[(j+k)*width] * pred_filter[k];
            // Round
            sum = (sum + ((1 << (filt_shift<<1))>>1)) >> (filt_shift << 1);
            pDst[j*dst_stride] = (sum < 0 ? 0 : sum > 255 ? 255 : sum);
        }
        ++pTmp;
    }
}
#else
// Based on vp8_post_proc_down_and_across_c (postproc.c)
void filter_mb(unsigned char *src, int src_stride,
               unsigned char *dst, int dst_stride,
               int width, int height)
{
    unsigned char *pSrc, *pDst;
    int row;
    int col;
    int i;
    int v;
    unsigned char d[8];

    /* TODO flimit should be linked to the quantizer value */
    int flimit = 7;

    for (row = 0; row < height; row++)
    {
        /* post_proc_down for one row */
        pSrc = src;
        pDst = dst;

        for (col = 0; col < width; col++)
        {
            int kernel = (1 << (filt_shift-1));
            int v = pSrc[col];

            for (i = -2; i <= 2; i++)
            {
                if (abs(v - pSrc[col+i*src_stride]) > flimit)
                    goto down_skip_convolve;

                kernel += pred_filter[2+i] * pSrc[col+i*src_stride];
            }

            v = (kernel >> filt_shift);
        down_skip_convolve:
            pDst[col] = v;
        }

        /* now post_proc_across */
        pSrc = dst;
        pDst = dst;

        for (i = 0; i < 8; i++)
            d[i] = pSrc[i];

        for (col = 0; col < width; col++)
        {
            int kernel = (1 << (filt_shift-1));
            v = pSrc[col];

            d[col&7] = v;

            for (i = -2; i <= 2; i++)
            {
                if (abs(v - pSrc[col+i]) > flimit)
                    goto across_skip_convolve;

                kernel += pred_filter[2+i] * pSrc[col+i];
            }

            d[col&7] = (kernel >> filt_shift);
        across_skip_convolve:

            if (col >= 2)
                pDst[col-2] = d[(col-2)&7];
        }

        /* handle the last two pixels */
        pDst[col-2] = d[(col-2)&7];
        pDst[col-1] = d[(col-1)&7];

        /* next row */
        src += src_stride;
        dst += dst_stride;
    }
}
#endif  // !USE_THRESH_FILTER

#endif  // CONFIG_PRED_FILTER

void vp8_build_inter16x16_predictors_mbuv(MACROBLOCKD *xd)
John Koleszar's avatar
John Koleszar committed
474
{
475
    unsigned char *uptr, *vptr;
476 477
    unsigned char *upred_ptr = &xd->predictor[256];
    unsigned char *vpred_ptr = &xd->predictor[320];
John Koleszar's avatar
John Koleszar committed
478

479 480
    int omv_row = xd->mode_info_context->mbmi.mv.as_mv.row;
    int omv_col = xd->mode_info_context->mbmi.mv.as_mv.col;
481 482
    int mv_row  = omv_row;
    int mv_col  = omv_col;
483
    int offset;
484
    int pre_stride = xd->block[16].pre_stride;
John Koleszar's avatar
John Koleszar committed
485

486 487 488 489 490
    /* calc uv motion vectors */
    if (mv_row < 0)
        mv_row -= 1;
    else
        mv_row += 1;
John Koleszar's avatar
John Koleszar committed
491

492 493 494 495
    if (mv_col < 0)
        mv_col -= 1;
    else
        mv_col += 1;
John Koleszar's avatar
John Koleszar committed
496

497 498 499
    mv_row /= 2;
    mv_col /= 2;

500 501
    mv_row &= xd->fullpixel_mask;
    mv_col &= xd->fullpixel_mask;
502 503

    offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552
    uptr = xd->pre.u_buffer + offset;
    vptr = xd->pre.v_buffer + offset;

#if CONFIG_PRED_FILTER
    if (xd->mode_info_context->mbmi.pred_filter_enabled)
    {
        int i;
#if CONFIG_ENHANCED_INTERP
        int Interp_Extend = 4;  // 8-tap filter needs 3+4 pels extension
#else
        int Interp_Extend = 3;  // 6-tap filter needs 2+3 pels extension
#endif
        int len = 7 + (Interp_Extend << 1);
        unsigned char Temp[32*32];  // Input data required by sub-pel filter
        unsigned char *pTemp = Temp + (Interp_Extend-1)*(len+1);
        unsigned char *pSrc = uptr;
        unsigned char *pDst = upred_ptr;

        // U & V
        for (i=0; i<2 ; i++)
        {
#if CONFIG_SIXTEENTH_SUBPEL_UV
            if ((omv_row | omv_col) & 15)
            {
                // Copy extended MB into Temp array, applying the spatial filter
                filter_mb(pSrc-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                          Temp, len, len, len);

                // Sub-pel interpolation
                xd->subpixel_predict8x8(pTemp, len, omv_col & 15,
                                        omv_row & 15, pDst, 8);
            }
#else   /* CONFIG_SIXTEENTH_SUBPEL_UV */
            if ((mv_row | mv_col) & 7)
            {
                // Copy extended MB into Temp array, applying the spatial filter
                filter_mb(pSrc-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                          Temp, len, len, len);

                // Sub-pel interpolation
                xd->subpixel_predict8x8(pTemp, len, mv_col & 7,
                                        mv_row & 7, pDst, 8);
            }
#endif  /* CONFIG_SIXTEENTH_SUBPEL_UV */
            else
            {
                // Apply prediction filter as we copy from source to destination
                filter_mb(pSrc, pre_stride, pDst, 8, 8, 8);
            }
553

554 555 556 557 558 559 560
            // V
            pSrc = vptr;
            pDst = vpred_ptr;
        }
    }
    else
#endif
561 562 563
#if CONFIG_SIXTEENTH_SUBPEL_UV
    if ((omv_row | omv_col) & 15)
    {
564 565
        xd->subpixel_predict8x8(uptr, pre_stride, omv_col & 15, omv_row & 15, upred_ptr, 8);
        xd->subpixel_predict8x8(vptr, pre_stride, omv_col & 15, omv_row & 15, vpred_ptr, 8);
566 567
    }
#else   /* CONFIG_SIXTEENTH_SUBPEL_UV */
568 569
    if ((mv_row | mv_col) & 7)
    {
570 571
        xd->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, upred_ptr, 8);
        xd->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, vpred_ptr, 8);
John Koleszar's avatar
John Koleszar committed
572
    }
573
#endif  /* CONFIG_SIXTEENTH_SUBPEL_UV */
John Koleszar's avatar
John Koleszar committed
574 575
    else
    {
576 577
        RECON_INVOKE(&xd->rtcd->recon, copy8x8)(uptr, pre_stride, upred_ptr, 8);
        RECON_INVOKE(&xd->rtcd->recon, copy8x8)(vptr, pre_stride, vpred_ptr, 8);
578 579 580 581 582 583 584 585 586 587 588 589
    }
}

/*encoder only*/
void vp8_build_inter4x4_predictors_mbuv(MACROBLOCKD *x)
{
    int i, j;

    /* build uv mvs */
    for (i = 0; i < 2; i++)
    {
        for (j = 0; j < 2; j++)
John Koleszar's avatar
John Koleszar committed
590
        {
591 592 593 594 595
            int yoffset = i * 8 + j * 2;
            int uoffset = 16 + i * 2 + j;
            int voffset = 20 + i * 2 + j;
            int temp;

596 597 598 599
            temp = x->block[yoffset  ].bmi.as_mv.first.as_mv.row
                   + x->block[yoffset+1].bmi.as_mv.first.as_mv.row
                   + x->block[yoffset+4].bmi.as_mv.first.as_mv.row
                   + x->block[yoffset+5].bmi.as_mv.first.as_mv.row;
600 601 602 603

            if (temp < 0) temp -= 4;
            else temp += 4;

604
            x->block[uoffset].bmi.as_mv.first.as_mv.row = (temp / 8) & x->fullpixel_mask;
605

606 607 608 609
            temp = x->block[yoffset  ].bmi.as_mv.first.as_mv.col
                   + x->block[yoffset+1].bmi.as_mv.first.as_mv.col
                   + x->block[yoffset+4].bmi.as_mv.first.as_mv.col
                   + x->block[yoffset+5].bmi.as_mv.first.as_mv.col;
610 611 612 613

            if (temp < 0) temp -= 4;
            else temp += 4;

614 615 616 617 618 619
            x->block[uoffset].bmi.as_mv.first.as_mv.col = (temp / 8) & x->fullpixel_mask;

            x->block[voffset].bmi.as_mv.first.as_mv.row =
                x->block[uoffset].bmi.as_mv.first.as_mv.row ;
            x->block[voffset].bmi.as_mv.first.as_mv.col =
                x->block[uoffset].bmi.as_mv.first.as_mv.col ;
620

621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659
            if (x->mode_info_context->mbmi.second_ref_frame)
            {
                temp = x->block[yoffset  ].bmi.as_mv.second.as_mv.row
                     + x->block[yoffset+1].bmi.as_mv.second.as_mv.row
                     + x->block[yoffset+4].bmi.as_mv.second.as_mv.row
                     + x->block[yoffset+5].bmi.as_mv.second.as_mv.row;

                if (temp < 0)
                {
                    temp -= 4;
                }
                else
                {
                    temp += 4;
                }

                x->block[uoffset].bmi.as_mv.second.as_mv.row = (temp / 8) & x->fullpixel_mask;

                temp = x->block[yoffset  ].bmi.as_mv.second.as_mv.col
                     + x->block[yoffset+1].bmi.as_mv.second.as_mv.col
                     + x->block[yoffset+4].bmi.as_mv.second.as_mv.col
                     + x->block[yoffset+5].bmi.as_mv.second.as_mv.col;

                if (temp < 0)
                {
                    temp -= 4;
                }
                else
                {
                    temp += 4;
                }

                x->block[uoffset].bmi.as_mv.second.as_mv.col = (temp / 8) & x->fullpixel_mask;

                x->block[voffset].bmi.as_mv.second.as_mv.row =
                    x->block[uoffset].bmi.as_mv.second.as_mv.row ;
                x->block[voffset].bmi.as_mv.second.as_mv.col =
                    x->block[uoffset].bmi.as_mv.second.as_mv.col ;
            }
660 661 662 663 664 665 666 667
        }
    }

    for (i = 16; i < 24; i += 2)
    {
        BLOCKD *d0 = &x->block[i];
        BLOCKD *d1 = &x->block[i+1];

668
        if (d0->bmi.as_mv.first.as_int == d1->bmi.as_mv.first.as_int)
669 670 671 672 673
            build_inter_predictors2b(x, d0, 8);
        else
        {
            vp8_build_inter_predictors_b(d0, 8, x->subpixel_predict);
            vp8_build_inter_predictors_b(d1, 8, x->subpixel_predict);
John Koleszar's avatar
John Koleszar committed
674
        }
675 676 677 678 679 680

        if (x->mode_info_context->mbmi.second_ref_frame)
        {
            vp8_build_2nd_inter_predictors_b(d0, 8, x->subpixel_predict_avg);
            vp8_build_2nd_inter_predictors_b(d1, 8, x->subpixel_predict_avg);
        }
John Koleszar's avatar
John Koleszar committed
681 682 683
    }
}

684

685
/*encoder only*/
686
void vp8_build_inter16x16_predictors_mby(MACROBLOCKD *xd)
John Koleszar's avatar
John Koleszar committed
687
{
688 689
    unsigned char *ptr_base;
    unsigned char *ptr;
690 691 692 693
    unsigned char *pred_ptr = xd->predictor;
    int mv_row = xd->mode_info_context->mbmi.mv.as_mv.row;
    int mv_col = xd->mode_info_context->mbmi.mv.as_mv.col;
    int pre_stride = xd->block[0].pre_stride;
694

695
    ptr_base = xd->pre.y_buffer;
696
    ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
John Koleszar's avatar
John Koleszar committed
697

698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
#if CONFIG_PRED_FILTER
    if (xd->mode_info_context->mbmi.pred_filter_enabled)
    {
        // Produce predictor from the filtered source
        if ((mv_row | mv_col) & 7)
        {
            // Sub-pel filter needs extended input
#if CONFIG_ENHANCED_INTERP
            int Interp_Extend = 4;  // 8-tap filter needs 3+4 pels extension
#else
            int Interp_Extend = 3;  // 6-tap filter needs 2+3 pels extension
#endif
            int len = 15 + (Interp_Extend << 1);
            unsigned char Temp[32*32];  // Data required by sub-pel filter
            unsigned char *pTemp = Temp + (Interp_Extend-1)*(len+1);

            // Copy extended MB into Temp array, applying the spatial filter
            filter_mb(ptr-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                      Temp, len, len, len);

            // Sub-pel interpolation
#if CONFIG_SIXTEENTH_SUBPEL_UV
            xd->subpixel_predict16x16(pTemp, len, (mv_col & 7)<<1,
                                      (mv_row & 7)<<1, pred_ptr, 16);
#else
            xd->subpixel_predict16x16(pTemp, len, mv_col & 7,
                                      mv_row & 7, pred_ptr, 16);
#endif
        }
        else
        {
            // Apply spatial filter to create the prediction directly
            filter_mb(ptr, pre_stride, pred_ptr, 16, 16, 16);
        }
    }
    else
#endif
735 736
    if ((mv_row | mv_col) & 7)
    {
737
#if CONFIG_SIXTEENTH_SUBPEL_UV
738 739
        xd->subpixel_predict16x16(ptr, pre_stride, (mv_col & 7)<<1,
                                  (mv_row & 7)<<1, pred_ptr, 16);
740
#else
741 742
        xd->subpixel_predict16x16(ptr, pre_stride, mv_col & 7,
                                  mv_row & 7, pred_ptr, 16);
743
#endif
John Koleszar's avatar
John Koleszar committed
744 745 746
    }
    else
    {
747
        RECON_INVOKE(&xd->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
John Koleszar's avatar
John Koleszar committed
748 749 750
    }
}

751 752 753 754 755 756 757 758 759 760 761
static void clamp_mv_to_umv_border(MV *mv, const MACROBLOCKD *xd)
{
    /* If the MV points so far into the UMV border that no visible pixels
     * are used for reconstruction, the subpel part of the MV can be
     * discarded and the MV limited to 16 pixels with equivalent results.
     *
     * This limit kicks in at 19 pixels for the top and left edges, for
     * the 16 pixels plus 3 taps right of the central pixel when subpel
     * filtering. The bottom and right edges use 16 pixels plus 2 pixels
     * left of the central pixel when filtering.
     */
762
    if (mv->col < (xd->mb_to_left_edge - ((16+INTERP_EXTEND) << 3)))
763
        mv->col = xd->mb_to_left_edge - (16 << 3);
764
    else if (mv->col > xd->mb_to_right_edge + ((15+INTERP_EXTEND) << 3))
765 766
        mv->col = xd->mb_to_right_edge + (16 << 3);

767
    if (mv->row < (xd->mb_to_top_edge - ((16+INTERP_EXTEND) << 3)))
768
        mv->row = xd->mb_to_top_edge - (16 << 3);
769
    else if (mv->row > xd->mb_to_bottom_edge + ((15+INTERP_EXTEND) << 3))
770 771 772 773 774 775
        mv->row = xd->mb_to_bottom_edge + (16 << 3);
}

/* A version of the above function for chroma block MVs.*/
static void clamp_uvmv_to_umv_border(MV *mv, const MACROBLOCKD *xd)
{
776
    mv->col = (2*mv->col < (xd->mb_to_left_edge - ((16+INTERP_EXTEND) << 3))) ?
777
        (xd->mb_to_left_edge - (16 << 3)) >> 1 : mv->col;
778
    mv->col = (2*mv->col > xd->mb_to_right_edge + ((15+INTERP_EXTEND) << 3)) ?
779 780
        (xd->mb_to_right_edge + (16 << 3)) >> 1 : mv->col;

781
    mv->row = (2*mv->row < (xd->mb_to_top_edge - ((16+INTERP_EXTEND) << 3))) ?
782
        (xd->mb_to_top_edge - (16 << 3)) >> 1 : mv->row;
783
    mv->row = (2*mv->row > xd->mb_to_bottom_edge + ((15+INTERP_EXTEND) << 3)) ?
784 785 786
        (xd->mb_to_bottom_edge + (16 << 3)) >> 1 : mv->row;
}

787 788


789 790 791 792 793 794
void vp8_build_inter16x16_predictors_mb(MACROBLOCKD *x,
                                        unsigned char *dst_y,
                                        unsigned char *dst_u,
                                        unsigned char *dst_v,
                                        int dst_ystride,
                                        int dst_uvstride)
John Koleszar's avatar
John Koleszar committed
795
{
796 797 798
    int offset;
    unsigned char *ptr;
    unsigned char *uptr, *vptr;
799

800
    int_mv _o16x16mv;
801
    int_mv _16x16mv;
John Koleszar's avatar
John Koleszar committed
802

803 804
    unsigned char *ptr_base = x->pre.y_buffer;
    int pre_stride = x->block[0].pre_stride;
John Koleszar's avatar
John Koleszar committed
805

806
    _16x16mv.as_int = x->mode_info_context->mbmi.mv.as_int;
John Koleszar's avatar
John Koleszar committed
807

808 809 810 811 812
    if (x->mode_info_context->mbmi.need_to_clamp_mvs)
    {
        clamp_mv_to_umv_border(&_16x16mv.as_mv, x);
    }

813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833
    ptr = ptr_base + (_16x16mv.as_mv.row >> 3) * pre_stride +
                     (_16x16mv.as_mv.col >> 3);

#if CONFIG_PRED_FILTER
    if (x->mode_info_context->mbmi.pred_filter_enabled)
    {
        if ( _16x16mv.as_int & 0x00070007)
        {
            // Sub-pel filter needs extended input
#if CONFIG_ENHANCED_INTERP
            int Interp_Extend = 4;  // 8-tap filter needs 3+4 pels extension
#else
            int Interp_Extend = 3;  // 6-tap filter needs 2+3 pels extension
#endif
            int len = 15 + (Interp_Extend << 1);
            unsigned char Temp[32*32];  // Data required by the sub-pel filter
            unsigned char *pTemp = Temp + (Interp_Extend-1)*(len+1);

            // Copy extended MB into Temp array, applying the spatial filter
            filter_mb(ptr-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                      Temp, len, len, len);
834

835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
            // Sub-pel filter
#if CONFIG_SIXTEENTH_SUBPEL_UV
            x->subpixel_predict16x16(pTemp, len,
                                     (_16x16mv.as_mv.col & 7)<<1,
                                     (_16x16mv.as_mv.row & 7)<<1,
                                     dst_y, dst_ystride);
#else
            x->subpixel_predict16x16(pTemp, len,
                                     _16x16mv.as_mv.col & 7,
                                     _16x16mv.as_mv.row & 7,
                                     dst_y, dst_ystride);
#endif
        }
        else
        {
            // Apply spatial filter to create the prediction directly
            filter_mb(ptr, pre_stride, dst_y, dst_ystride, 16, 16);
        }
    }
    else
#endif
856
    if ( _16x16mv.as_int & 0x00070007)
857
    {
858
#if CONFIG_SIXTEENTH_SUBPEL_UV
859 860 861
        x->subpixel_predict16x16(ptr, pre_stride, (_16x16mv.as_mv.col & 7)<<1,
                                 (_16x16mv.as_mv.row & 7)<<1,
                                 dst_y, dst_ystride);
862
#else
863 864
        x->subpixel_predict16x16(ptr, pre_stride, _16x16mv.as_mv.col & 7,
                                 _16x16mv.as_mv.row & 7, dst_y, dst_ystride);
865
#endif
866 867 868
    }
    else
    {
869 870
        RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, dst_y,
                     dst_ystride);
871
    }
John Koleszar's avatar
John Koleszar committed
872

873
    _o16x16mv = _16x16mv;
874
    /* calc uv motion vectors */
875 876
    if ( _16x16mv.as_mv.row < 0)
      _16x16mv.as_mv.row -= 1;
877
    else
878
      _16x16mv.as_mv.row += 1;
879

880 881
    if (_16x16mv.as_mv.col < 0)
        _16x16mv.as_mv.col -= 1;
882
    else
883
        _16x16mv.as_mv.col += 1;
884

885 886
    _16x16mv.as_mv.row /= 2;
    _16x16mv.as_mv.col /= 2;
887

888 889
    _16x16mv.as_mv.row &= x->fullpixel_mask;
    _16x16mv.as_mv.col &= x->fullpixel_mask;
890

891
    pre_stride >>= 1;
892
    offset = ( _16x16mv.as_mv.row >> 3) * pre_stride + (_16x16mv.as_mv.col >> 3);
893 894
    uptr = x->pre.u_buffer + offset;
    vptr = x->pre.v_buffer + offset;
John Koleszar's avatar
John Koleszar committed
895

896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
#if CONFIG_PRED_FILTER
    if (x->mode_info_context->mbmi.pred_filter_enabled)
    {
        int i;
        unsigned char *pSrc = uptr;
        unsigned char *pDst = dst_u;
#if CONFIG_ENHANCED_INTERP
        int Interp_Extend = 4;  // 8-tap filter needs 3+4 pels extension
#else
        int Interp_Extend = 3;  // 6-tap filter needs 2+3 pels extension
#endif
        int len = 7 + (Interp_Extend << 1);
        unsigned char Temp[32*32];  // Data required by the sub-pel filter
        unsigned char *pTemp = Temp + (Interp_Extend-1)*(len+1);

        // U & V
        for (i=0; i<2; i++)
        {
#if CONFIG_SIXTEENTH_SUBPEL_UV
            if ( _o16x16mv.as_int & 0x000f000f)
            {
                // Copy extended MB into Temp array, applying the spatial filter
                filter_mb(pSrc-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                          Temp, len, len, len);

                // Sub-pel filter
                x->subpixel_predict8x8(pTemp, len,
                                       _o16x16mv.as_mv.col & 15,
                                       _o16x16mv.as_mv.row & 15,
                                       pDst, dst_uvstride);
            }
#else  /* CONFIG_SIXTEENTH_SUBPEL_UV */
            if ( _16x16mv.as_int & 0x00070007)
            {
                // Copy extended MB into Temp array, applying the spatial filter
                filter_mb(pSrc-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                          Temp, len, len, len);

                // Sub-pel filter
                x->subpixel_predict8x8(pTemp, len,
                                       _16x16mv.as_mv.col & 7,
                                       _16x16mv.as_mv.row & 7,
                                       pDst, dst_uvstride);
            }
#endif  /* CONFIG_SIXTEENTH_SUBPEL_UV */
            else
            {
                filter_mb(pSrc, pre_stride, pDst, dst_uvstride, 8, 8);
            }

            // V
            pSrc = vptr;
            pDst = dst_v;
        }
    }
    else
#endif
953 954 955 956 957 958 959
#if CONFIG_SIXTEENTH_SUBPEL_UV
    if ( _o16x16mv.as_int & 0x000f000f)
    {
        x->subpixel_predict8x8(uptr, pre_stride, _o16x16mv.as_mv.col & 15,  _o16x16mv.as_mv.row & 15, dst_u, dst_uvstride);
        x->subpixel_predict8x8(vptr, pre_stride, _o16x16mv.as_mv.col & 15,  _o16x16mv.as_mv.row & 15, dst_v, dst_uvstride);
    }
#else  /* CONFIG_SIXTEENTH_SUBPEL_UV */
960
    if ( _16x16mv.as_int & 0x00070007)
961
    {
962 963
        x->subpixel_predict8x8(uptr, pre_stride, _16x16mv.as_mv.col & 7,  _16x16mv.as_mv.row & 7, dst_u, dst_uvstride);
        x->subpixel_predict8x8(vptr, pre_stride, _16x16mv.as_mv.col & 7,  _16x16mv.as_mv.row & 7, dst_v, dst_uvstride);
John Koleszar's avatar
John Koleszar committed
964
    }
965
#endif  /* CONFIG_SIXTEENTH_SUBPEL_UV */
John Koleszar's avatar
John Koleszar committed
966 967
    else
    {
968 969 970
        RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, dst_u, dst_uvstride);
        RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, dst_v, dst_uvstride);
    }
John Koleszar's avatar
John Koleszar committed
971

972
}
John Koleszar's avatar
John Koleszar committed
973

974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998
/*
 * This function should be called after an initial call to
 * vp8_build_inter16x16_predictors_mb() or _mby()/_mbuv().
 * It will run a second sixtap filter on a (different) ref
 * frame and average the result with the output of the
 * first sixtap filter. The second reference frame is stored
 * in x->second_pre (the reference frame index is in
 * x->mode_info_context->mbmi.second_ref_frame). The second
 * motion vector is x->mode_info_context->mbmi.second_mv.
 *
 * This allows blending prediction from two reference frames
 * which sometimes leads to better prediction than from a
 * single reference framer.
 */
void vp8_build_2nd_inter16x16_predictors_mb(MACROBLOCKD *x,
                                            unsigned char *dst_y,
                                            unsigned char *dst_u,
                                            unsigned char *dst_v,
                                            int dst_ystride,
                                            int dst_uvstride)
{
    int offset;
    unsigned char *ptr;
    unsigned char *uptr, *vptr;

Yaowu Xu's avatar
Yaowu Xu committed
999 1000 1001 1002
    int_mv _16x16mv;
    int mv_row;
    int mv_col;

1003
    int omv_row, omv_col;
1004 1005 1006 1007

    unsigned char *ptr_base = x->second_pre.y_buffer;
    int pre_stride = x->block[0].pre_stride;

Yaowu Xu's avatar
Yaowu Xu committed
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
    _16x16mv.as_int = x->mode_info_context->mbmi.second_mv.as_int;

    if (x->mode_info_context->mbmi.need_to_clamp_secondmv)
    {
        clamp_mv_to_umv_border(&_16x16mv.as_mv, x);
    }

    mv_row = _16x16mv.as_mv.row;
    mv_col = _16x16mv.as_mv.col;

1018 1019
    ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);

1020 1021
#if CONFIG_PRED_FILTER
    if (x->mode_info_context->mbmi.pred_filter_enabled)
1022
    {
1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
        if ((mv_row | mv_col) & 7)
        {
            // Sub-pel filter needs extended input
#if CONFIG_ENHANCED_INTERP
            int Interp_Extend = 4;  // 8-tap filter needs 3+4 pels extension
#else
            int Interp_Extend = 3;  // 6-tap filter needs 2+3 pels extension
#endif
            int len = 15 + (Interp_Extend << 1);
            unsigned char Temp[32*32];  // Data required by sub-pel filter
            unsigned char *pTemp = Temp + (Interp_Extend-1)*(len+1);

            // Copy extended MB into Temp array, applying the spatial filter
            filter_mb(ptr-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                      Temp, len, len, len);

            // Sub-pel filter
1040
#if CONFIG_SIXTEENTH_SUBPEL_UV
1041 1042
            x->subpixel_predict_avg16x16(pTemp, len, (mv_col & 7)<<1,
                                         (mv_row & 7)<<1, dst_y, dst_ystride);
1043
#else
1044 1045
            x->subpixel_predict_avg16x16(pTemp, len, mv_col & 7,
                                         mv_row & 7, dst_y, dst_ystride);
1046
#endif
1047 1048 1049 1050 1051 1052 1053 1054
        }
        else
        {
            // TODO Needs to AVERAGE with the dst_y
            // For now, do not apply the prediction filter in these cases!
            RECON_INVOKE(&x->rtcd->recon, avg16x16)(ptr, pre_stride, dst_y,
                         dst_ystride);
        }
1055 1056
    }
    else
1057
#endif  // CONFIG_PRED_FILTER
1058
    {
1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
        if ((mv_row | mv_col) & 7)
        {
#if CONFIG_SIXTEENTH_SUBPEL_UV
            x->subpixel_predict_avg16x16(ptr, pre_stride, (mv_col & 7)<<1,
                                         (mv_row & 7)<<1, dst_y, dst_ystride);
#else
            x->subpixel_predict_avg16x16(ptr, pre_stride, mv_col & 7,
                                         mv_row & 7, dst_y, dst_ystride);
#endif
        }
        else
        {
            RECON_INVOKE(&x->rtcd->recon, avg16x16)(ptr, pre_stride, dst_y,
                         dst_ystride);
        }
1074 1075 1076
    }

    /* calc uv motion vectors */
1077 1078
    omv_row = mv_row;
    omv_col = mv_col;
1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089
    mv_row = (mv_row + (mv_row > 0)) >> 1;
    mv_col = (mv_col + (mv_col > 0)) >> 1;

    mv_row &= x->fullpixel_mask;
    mv_col &= x->fullpixel_mask;

    pre_stride >>= 1;
    offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
    uptr = x->second_pre.u_buffer + offset;
    vptr = x->second_pre.v_buffer + offset;

1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
#if CONFIG_PRED_FILTER
    if (x->mode_info_context->mbmi.pred_filter_enabled)
    {
        int i;
#if CONFIG_ENHANCED_INTERP
        int Interp_Extend = 4;  // 8-tap filter needs 3+4 pels extension
#else
        int Interp_Extend = 3;  // 6-tap filter needs 2+3 pels extension
#endif
        int len = 7 + (Interp_Extend << 1);
        unsigned char Temp[32*32];  // Data required by sub-pel filter
        unsigned char *pTemp = Temp + (Interp_Extend-1)*(len+1);
        unsigned char *pSrc = uptr;
        unsigned char *pDst = dst_u;

        // U & V
        for (i=0; i<2; i++)
        {
#if CONFIG_SIXTEENTH_SUBPEL_UV
            if ((omv_row | omv_col) & 15)
            {
                // Copy extended MB into Temp array, applying the spatial filter
                filter_mb(pSrc-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                          Temp, len, len, len);

                // Sub-pel filter
                x->subpixel_predict_avg8x8(pTemp, len, omv_col & 15,
                                           omv_row & 15, pDst, dst_uvstride);
            }
#else  /* CONFIG_SIXTEENTH_SUBPEL_UV */
            if ((mv_row | mv_col) & 7)
            {
                // Copy extended MB into Temp array, applying the spatial filter
                filter_mb(pSrc-(Interp_Extend-1)*(pre_stride+1), pre_stride,
                          Temp, len, len, len);

                // Sub-pel filter
                x->subpixel_predict_avg8x8(pTemp, len, mv_col & 7, mv_row & 7,
                                           pDst, dst_uvstride);
            }
#endif  /* CONFIG_SIXTEENTH_SUBPEL_UV */
            else
            {
                // TODO Needs to AVERAGE with the dst_[u|v]
                // For now, do not apply the prediction filter here!
                RECON_INVOKE(&x->rtcd->recon, avg8x8)(pSrc, pre_stride, pDst,
                                                      dst_uvstride);
            }

            // V
            pSrc = vptr;
            pDst = dst_v;
        }
    }
    else
#endif  // CONFIG_PRED_FILTER
1146 1147 1148 1149 1150 1151 1152
#if CONFIG_SIXTEENTH_SUBPEL_UV
    if ((omv_row | omv_col) & 15)
    {
        x->subpixel_predict_avg8x8(uptr, pre_stride, omv_col & 15, omv_row & 15, dst_u, dst_uvstride);
        x->subpixel_predict_avg8x8(vptr, pre_stride, omv_col & 15, omv_row & 15, dst_v, dst_uvstride);
    }
#else  /* CONFIG_SIXTEENTH_SUBPEL_UV */
1153 1154 1155 1156 1157
    if ((mv_row | mv_col) & 7)
    {
        x->subpixel_predict_avg8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, dst_u, dst_uvstride);
        x->subpixel_predict_avg8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, dst_v, dst_uvstride);
    }
1158
#endif  /* CONFIG_SIXTEENTH_SUBPEL_UV */
1159 1160 1161 1162 1163 1164 1165
    else
    {
        RECON_INVOKE(&x->rtcd->recon, avg8x8)(uptr, pre_stride, dst_u, dst_uvstride);
        RECON_INVOKE(&x->rtcd->recon, avg8x8)(vptr, pre_stride, dst_v, dst_uvstride);
    }
}

1166
static void build_inter4x4_predictors_mb(MACROBLOCKD *x)
1167 1168
{
    int i;
John Koleszar's avatar
John Koleszar committed
1169

1170 1171
    if (x->mode_info_context->mbmi.partitioning < 3)
    {
1172 1173 1174 1175 1176
        x->block[ 0].bmi = x->mode_info_context->bmi[ 0];
        x->block[ 2].bmi = x->mode_info_context->bmi[ 2];
        x->block[ 8].bmi = x->mode_info_context->bmi[ 8];
        x->block[10].bmi = x->mode_info_context->bmi[10];

1177 1178
        if (x->mode_info_context->mbmi.need_to_clamp_mvs)
        {
1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189
            clamp_mv_to_umv_border(&x->block[ 0].bmi.as_mv.first.as_mv, x);
            clamp_mv_to_umv_border(&x->block[ 2].bmi.as_mv.first.as_mv, x);
            clamp_mv_to_umv_border(&x->block[ 8].bmi.as_mv.first.as_mv, x);
            clamp_mv_to_umv_border(&x->block[10].bmi.as_mv.first.as_mv, x);
            if (x->mode_info_context->mbmi.second_ref_frame)
            {
                clamp_mv_to_umv_border(&x->block[ 0].bmi.as_mv.second.as_mv, x);
                clamp_mv_to_umv_border(&x->block[ 2].bmi.as_mv.second.as_mv, x);
                clamp_mv_to_umv_border(&x->block[ 8].bmi.as_mv.second.as_mv, x);
                clamp_mv_to_umv_border(&x->block[10].bmi.as_mv.second.as_mv, x);
            }
1190 1191
        }

1192

1193 1194 1195 1196
        build_inter_predictors4b(x, &x->block[ 0], 16);
        build_inter_predictors4b(x, &x->block[ 2], 16);
        build_inter_predictors4b(x, &x->block[ 8], 16);
        build_inter_predictors4b(x, &x->block[10], 16);
1197 1198 1199 1200 1201 1202 1203 1204

        if (x->mode_info_context->mbmi.second_ref_frame)
        {
            build_2nd_inter_predictors4b(x, &x->block[ 0], 16);
            build_2nd_inter_predictors4b(x, &x->block[ 2], 16);
            build_2nd_inter_predictors4b(x, &x->block[ 8], 16);
            build_2nd_inter_predictors4b(x, &x->block[10], 16);
        }
1205 1206 1207 1208
    }
    else
    {
        for (i = 0; i < 16; i += 2)
John Koleszar's avatar
John Koleszar committed
1209 1210 1211 1212
        {
            BLOCKD *d0 = &x->block[i];
            BLOCKD *d1 = &x->block[i+1];

1213 1214 1215
            x->block[i+0].bmi = x->mode_info_context->bmi[i+0];
            x->block[i+1].bmi = x->mode_info_context->bmi[i+1];

1216 1217
            if (x->mode_info_context->mbmi.need_to_clamp_mvs)
            {
1218 1219 1220 1221 1222 1223 1224
                clamp_mv_to_umv_border(&x->block[i+0].bmi.as_mv.first.as_mv, x);
                clamp_mv_to_umv_border(&x->block[i+1].bmi.as_mv.first.as_mv, x);
                if (x->mode_info_context->mbmi.second_ref_frame)
                {
                    clamp_mv_to_umv_border(&x->block[i+0].bmi.as_mv.second.as_mv, x);
                    clamp_mv_to_umv_border(&x->block[i+1].bmi.as_mv.second.as_mv, x);
                }
1225 1226
            }

1227
            if (d0->bmi.<