mc_tmpl.c 25.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
/*
 * Copyright © 2018, VideoLAN and dav1d authors
 * Copyright © 2018, Two Orioles, LLC
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 * 1. Redistributions of source code must retain the above copyright notice, this
 *    list of conditions and the following disclaimer.
 *
 * 2. Redistributions in binary form must reproduce the above copyright notice,
 *    this list of conditions and the following disclaimer in the documentation
 *    and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

#include "config.h"

30
#include <assert.h>
31 32 33
#include <stdlib.h>
#include <string.h>

34
#include "common/attributes.h"
35 36 37 38 39
#include "common/intops.h"

#include "src/mc.h"
#include "src/tables.h"

40
static NOINLINE void
41 42 43 44 45 46 47 48 49 50 51
put_c(pixel *dst, const ptrdiff_t dst_stride,
      const pixel *src, const ptrdiff_t src_stride, const int w, int h)
{
    do {
        pixel_copy(dst, src, w);

        dst += dst_stride;
        src += src_stride;
    } while (--h);
}

52
static NOINLINE void
53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
prep_c(coef *tmp, const pixel *src, const ptrdiff_t src_stride,
       const int w, int h)
{
    do {
        for (int x = 0; x < w; x++)
            tmp[x] = src[x] << 4;

        tmp += w;
        src += src_stride;
    } while (--h);
}

#define FILTER_8TAP(src, x, F, stride) \
    (F[0] * src[x + -3 * stride] + \
     F[1] * src[x + -2 * stride] + \
     F[2] * src[x + -1 * stride] + \
     F[3] * src[x + +0 * stride] + \
     F[4] * src[x + +1 * stride] + \
     F[5] * src[x + +2 * stride] + \
     F[6] * src[x + +3 * stride] + \
     F[7] * src[x + +4 * stride])

#define FILTER_8TAP_RND(src, x, F, stride, sh) \
    ((FILTER_8TAP(src, x, F, stride) + ((1 << sh) >> 1)) >> sh)

#define FILTER_8TAP_CLIP(src, x, F, stride, sh) \
    iclip_pixel(FILTER_8TAP_RND(src, x, F, stride, sh))

81 82 83 84 85 86 87 88 89 90
#define GET_H_FILTER(mx) \
    const int8_t *const fh = !(mx) ? NULL : w > 4 ? \
        dav1d_mc_subpel_filters[filter_type & 3][(mx) - 1] : \
        dav1d_mc_subpel_filters[3 + (filter_type & 1)][(mx) - 1]

#define GET_V_FILTER(my) \
    const int8_t *const fv = !(my) ? NULL : h > 4 ? \
        dav1d_mc_subpel_filters[filter_type >> 2][(my) - 1] : \
        dav1d_mc_subpel_filters[3 + ((filter_type >> 2) & 1)][(my) - 1]

91
#define GET_FILTERS() \
92 93
    GET_H_FILTER(mx); \
    GET_V_FILTER(my)
94

95
static NOINLINE void
96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112
put_8tap_c(pixel *dst, ptrdiff_t dst_stride,
           const pixel *src, ptrdiff_t src_stride,
           const int w, int h, const int mx, const int my,
           const int filter_type)
{
    GET_FILTERS();
    dst_stride = PXSTRIDE(dst_stride);
    src_stride = PXSTRIDE(src_stride);

    if (fh) {
        if (fv) {
            int tmp_h = h + 7;
            coef mid[128 * 135], *mid_ptr = mid;

            src -= src_stride * 3;
            do {
                for (int x = 0; x < w; x++)
113
                    mid_ptr[x] = FILTER_8TAP_RND(src, x, fh, 1, 2);
114 115 116 117 118 119 120 121

                mid_ptr += 128;
                src += src_stride;
            } while (--tmp_h);

            mid_ptr = mid + 128 * 3;
            do {
                for (int x = 0; x < w; x++)
122
                    dst[x] = FILTER_8TAP_CLIP(mid_ptr, x, fv, 128, 10);
123 124 125 126 127 128 129

                mid_ptr += 128;
                dst += dst_stride;
            } while (--h);
        } else {
            do {
                for (int x = 0; x < w; x++) {
130
                    const int px = FILTER_8TAP_RND(src, x, fh, 1, 2);
131 132 133 134 135 136 137 138 139 140
                    dst[x] = iclip_pixel((px + 8) >> 4);
                }

                dst += dst_stride;
                src += src_stride;
            } while (--h);
        }
    } else if (fv) {
        do {
            for (int x = 0; x < w; x++)
141
                dst[x] = FILTER_8TAP_CLIP(src, x, fv, src_stride, 6);
142 143 144 145 146 147 148 149

            dst += dst_stride;
            src += src_stride;
        } while (--h);
    } else
        put_c(dst, dst_stride, src, src_stride, w, h);
}

150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
static NOINLINE void
put_8tap_scaled_c(pixel *dst, const ptrdiff_t dst_stride,
                  const pixel *src, const ptrdiff_t src_stride,
                  const int w, int h, const int mx, int my,
                  const int dx, const int dy, const int filter_type)
{
    int tmp_h = (((h - 1) * dy + my) >> 10) + 8;
    coef mid[128 * (256 + 7)], *mid_ptr = mid;

    src -= src_stride * 3;
    do {
        int x;
        int imx = mx, ioff = 0;

        for (x = 0; x < w; x++) {
            GET_H_FILTER(imx >> 6);
            mid_ptr[x] = fh ? FILTER_8TAP_RND(src, ioff, fh, 1, 2) : src[ioff] << 4;
            imx += dx;
            ioff += imx >> 10;
            imx &= 0x3ff;
        }

        mid_ptr += 128;
        src += PXSTRIDE(src_stride);
    } while (--tmp_h);

    mid_ptr = mid + 128 * 3;
    for (int y = 0; y < h; y++) {
        int x;
        GET_V_FILTER(my >> 6);

        for (x = 0; x < w; x++)
            dst[x] = fv ? FILTER_8TAP_CLIP(mid_ptr, x, fv, 128, 10) :
                          (mid_ptr[x] + 8) >> 4;

        my += dy;
        mid_ptr += (my >> 10) * 128;
        my &= 0x3ff;
        dst += PXSTRIDE(dst_stride);
    }
}

192
static NOINLINE void
193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
prep_8tap_c(coef *tmp, const pixel *src, ptrdiff_t src_stride,
            const int w, int h, const int mx, const int my,
            const int filter_type)
{
    GET_FILTERS();
    src_stride = PXSTRIDE(src_stride);

    if (fh) {
        if (fv) {
            int tmp_h = h + 7;
            coef mid[128 * 135], *mid_ptr = mid;

            src -= src_stride * 3;
            do {
                for (int x = 0; x < w; x++)
208
                    mid_ptr[x] = FILTER_8TAP_RND(src, x, fh, 1, 2);
209 210 211 212 213 214 215 216

                mid_ptr += 128;
                src += src_stride;
            } while (--tmp_h);

            mid_ptr = mid + 128 * 3;
            do {
                for (int x = 0; x < w; x++)
217
                    tmp[x] = FILTER_8TAP_RND(mid_ptr, x, fv, 128, 6);
218 219 220 221 222 223 224

                mid_ptr += 128;
                tmp += w;
            } while (--h);
        } else {
            do {
                for (int x = 0; x < w; x++)
225
                    tmp[x] = FILTER_8TAP_RND(src, x, fh, 1, 2);
226 227 228 229 230 231 232 233

                tmp += w;
                src += src_stride;
            } while (--h);
        }
    } else if (fv) {
        do {
            for (int x = 0; x < w; x++)
234
                tmp[x] = FILTER_8TAP_RND(src, x, fv, src_stride, 2);
235 236 237 238 239 240 241 242

            tmp += w;
            src += src_stride;
        } while (--h);
    } else
        prep_c(tmp, src, src_stride, w, h);
}

243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
static NOINLINE void
prep_8tap_scaled_c(coef *tmp, const pixel *src, const ptrdiff_t src_stride,
                   const int w, int h, const int mx, int my,
                   const int dx, const int dy, const int filter_type)
{
    int tmp_h = (((h - 1) * dy + my) >> 10) + 8;
    coef mid[128 * (256 + 7)], *mid_ptr = mid;

    src -= src_stride * 3;
    do {
        int x;
        int imx = mx, ioff = 0;

        for (x = 0; x < w; x++) {
            GET_H_FILTER(imx >> 6);
            mid_ptr[x] = fh ? FILTER_8TAP_RND(src, ioff, fh, 1, 2) : src[ioff] << 4;
            imx += dx;
            ioff += imx >> 10;
            imx &= 0x3ff;
        }

        mid_ptr += 128;
        src += PXSTRIDE(src_stride);
    } while (--tmp_h);

    mid_ptr = mid + 128 * 3;
    for (int y = 0; y < h; y++) {
        int x;
        GET_V_FILTER(my >> 6);

        for (x = 0; x < w; x++)
            tmp[x] = fv ? FILTER_8TAP_CLIP(mid_ptr, x, fv, 128, 6) : mid_ptr[x];

        my += dy;
        mid_ptr += (my >> 10) * 128;
        my &= 0x3ff;
        tmp += w;
    }
}

283 284 285 286 287 288 289 290 291 292 293
#define filter_fns(type, type_h, type_v) \
static void put_8tap_##type##_c(pixel *const dst, \
                                const ptrdiff_t dst_stride, \
                                const pixel *const src, \
                                const ptrdiff_t src_stride, \
                                const int w, const int h, \
                                const int mx, const int my) \
{ \
    put_8tap_c(dst, dst_stride, src, src_stride, w, h, mx, my, \
               type_h | (type_v << 2)); \
} \
294 295 296 297 298 299 300 301 302 303 304
static void put_8tap_##type##_scaled_c(pixel *const dst, \
                                       const ptrdiff_t dst_stride, \
                                       const pixel *const src, \
                                       const ptrdiff_t src_stride, \
                                       const int w, const int h, \
                                       const int mx, const int my, \
                                       const int dx, const int dy) \
{ \
    put_8tap_scaled_c(dst, dst_stride, src, src_stride, w, h, mx, my, dx, dy, \
                      type_h | (type_v << 2)); \
} \
305 306 307 308 309 310 311 312
static void prep_8tap_##type##_c(coef *const tmp, \
                                 const pixel *const src, \
                                 const ptrdiff_t src_stride, \
                                 const int w, const int h, \
                                 const int mx, const int my) \
{ \
    prep_8tap_c(tmp, src, src_stride, w, h, mx, my, \
                type_h | (type_v << 2)); \
313 314 315 316 317 318 319 320 321 322
} \
static void prep_8tap_##type##_scaled_c(coef *const tmp, \
                                        const pixel *const src, \
                                        const ptrdiff_t src_stride, \
                                        const int w, const int h, \
                                        const int mx, const int my, \
                                        const int dx, const int dy) \
{ \
    prep_8tap_scaled_c(tmp, src, src_stride, w, h, mx, my, dx, dy, \
                       type_h | (type_v << 2)); \
323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
}

filter_fns(regular,        FILTER_8TAP_REGULAR, FILTER_8TAP_REGULAR)
filter_fns(regular_sharp,  FILTER_8TAP_REGULAR, FILTER_8TAP_SHARP)
filter_fns(regular_smooth, FILTER_8TAP_REGULAR, FILTER_8TAP_SMOOTH)
filter_fns(smooth,         FILTER_8TAP_SMOOTH,  FILTER_8TAP_SMOOTH)
filter_fns(smooth_regular, FILTER_8TAP_SMOOTH,  FILTER_8TAP_REGULAR)
filter_fns(smooth_sharp,   FILTER_8TAP_SMOOTH,  FILTER_8TAP_SHARP)
filter_fns(sharp,          FILTER_8TAP_SHARP,   FILTER_8TAP_SHARP)
filter_fns(sharp_regular,  FILTER_8TAP_SHARP,   FILTER_8TAP_REGULAR)
filter_fns(sharp_smooth,   FILTER_8TAP_SHARP,   FILTER_8TAP_SMOOTH)

#define FILTER_BILIN(src, x, mxy, stride) \
    (16 * src[x] + (mxy * (src[x + stride] - src[x])))

#define FILTER_BILIN_RND(src, x, mxy, stride, sh) \
    ((FILTER_BILIN(src, x, mxy, stride) + ((1 << sh) >> 1)) >> sh)

#define FILTER_BILIN_CLIP(src, x, mxy, stride, sh) \
    iclip_pixel(FILTER_BILIN_RND(src, x, mxy, stride, sh))

static void put_bilin_c(pixel *dst, ptrdiff_t dst_stride,
                        const pixel *src, ptrdiff_t src_stride,
                        const int w, int h, const int mx, const int my)
{
    dst_stride = PXSTRIDE(dst_stride);
    src_stride = PXSTRIDE(src_stride);

    if (mx) {
        if (my) {
            coef mid[128 * 129], *mid_ptr = mid;
            int tmp_h = h + 1;

            do {
                for (int x = 0; x < w; x++)
                    mid_ptr[x] = FILTER_BILIN(src, x, mx, 1);

                mid_ptr += 128;
                src += src_stride;
            } while (--tmp_h);

            mid_ptr = mid;
            do {
                for (int x = 0; x < w; x++)
                    dst[x] = FILTER_BILIN_CLIP(mid_ptr, x, my, 128, 8);

                mid_ptr += 128;
                dst += dst_stride;
            } while (--h);
        } else {
            do {
                for (int x = 0; x < w; x++)
                    dst[x] = FILTER_BILIN_CLIP(src, x, mx, 1, 4);

                dst += dst_stride;
                src += src_stride;
            } while (--h);
        }
    } else if (my) {
        do {
            for (int x = 0; x < w; x++)
                dst[x] = FILTER_BILIN_CLIP(src, x, my, src_stride, 4);

            dst += dst_stride;
            src += src_stride;
        } while (--h);
    } else
        put_c(dst, dst_stride, src, src_stride, w, h);
}

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
static void put_bilin_scaled_c(pixel *dst, ptrdiff_t dst_stride,
                               const pixel *src, ptrdiff_t src_stride,
                               const int w, int h, const int mx, int my,
                               const int dx, const int dy)
{
    int tmp_h = (((h - 1) * dy + my) >> 10) + 2;
    coef mid[128 * (256 + 1)], *mid_ptr = mid;

    do {
        int x;
        int imx = mx, ioff = 0;

        for (x = 0; x < w; x++) {
            mid_ptr[x] = FILTER_BILIN(src, ioff, imx >> 6, 1);
            imx += dx;
            ioff += imx >> 10;
            imx &= 0x3ff;
        }

        mid_ptr += 128;
        src += PXSTRIDE(src_stride);
    } while (--tmp_h);

    mid_ptr = mid;
    do {
        int x;

        for (x = 0; x < w; x++)
            dst[x] = FILTER_BILIN_CLIP(mid_ptr, x, my >> 6, 128, 8);

        my += dy;
        mid_ptr += (my >> 10) * 128;
        my &= 0x3ff;
        dst += PXSTRIDE(dst_stride);
    } while (--h);
}

430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477
static void prep_bilin_c(coef *tmp,
                         const pixel *src, ptrdiff_t src_stride,
                         const int w, int h, const int mx, const int my)
{
    src_stride = PXSTRIDE(src_stride);

    if (mx) {
        if (my) {
            coef mid[128 * 129], *mid_ptr = mid;
            int tmp_h = h + 1;

            do {
                for (int x = 0; x < w; x++)
                    mid_ptr[x] = FILTER_BILIN(src, x, mx, 1);

                mid_ptr += 128;
                src += src_stride;
            } while (--tmp_h);

            mid_ptr = mid;
            do {
                for (int x = 0; x < w; x++)
                    tmp[x] = FILTER_BILIN_RND(mid_ptr, x, my, 128, 4);

                mid_ptr += 128;
                tmp += w;
            } while (--h);
        } else {
            do {
                for (int x = 0; x < w; x++)
                    tmp[x] = FILTER_BILIN(src, x, mx, 1);

                tmp += w;
                src += src_stride;
            } while (--h);
        }
    } else if (my) {
        do {
            for (int x = 0; x < w; x++)
                tmp[x] = FILTER_BILIN(src, x, my, src_stride);

            tmp += w;
            src += src_stride;
        } while (--h);
    } else
        prep_c(tmp, src, src_stride, w, h);
}

478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514
static void prep_bilin_scaled_c(coef *tmp,
                                const pixel *src, ptrdiff_t src_stride,
                                const int w, int h, const int mx, int my,
                                const int dx, const int dy)
{
    int tmp_h = (((h - 1) * dy + my) >> 10) + 2;
    coef mid[128 * (256 + 1)], *mid_ptr = mid;

    do {
        int x;
        int imx = mx, ioff = 0;

        for (x = 0; x < w; x++) {
            mid_ptr[x] = FILTER_BILIN(src, ioff, imx >> 6, 1);
            imx += dx;
            ioff += imx >> 10;
            imx &= 0x3ff;
        }

        mid_ptr += 128;
        src += PXSTRIDE(src_stride);
    } while (--tmp_h);

    mid_ptr = mid;
    do {
        int x;

        for (x = 0; x < w; x++)
            tmp[x] = FILTER_BILIN_RND(mid_ptr, x, my >> 6, 128, 4);

        my += dy;
        mid_ptr += (my >> 10) * 128;
        my &= 0x3ff;
        tmp += w;
    } while (--h);
}

515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558
static void avg_c(pixel *dst, const ptrdiff_t dst_stride,
                  const coef *tmp1, const coef *tmp2, const int w, int h)
{
    do {
        for (int x = 0; x < w; x++)
            dst[x] = iclip_pixel((tmp1[x] + tmp2[x] + 16) >> 5);

        tmp1 += w;
        tmp2 += w;
        dst += PXSTRIDE(dst_stride);
    } while (--h);
}

static void w_avg_c(pixel *dst, const ptrdiff_t dst_stride,
                    const coef *tmp1, const coef *tmp2, const int w, int h,
                    const int weight)
{
    do {
        for (int x = 0; x < w; x++)
            dst[x] = iclip_pixel((tmp1[x] * weight +
                                  tmp2[x] * (16 - weight) + 128) >> 8);

        tmp1 += w;
        tmp2 += w;
        dst += PXSTRIDE(dst_stride);
    } while (--h);
}

static void mask_c(pixel *dst, const ptrdiff_t dst_stride,
                   const coef *tmp1, const coef *tmp2, const int w, int h,
                   const uint8_t *mask)
{
    do {
        for (int x = 0; x < w; x++)
            dst[x] = iclip_pixel((tmp1[x] * mask[x] +
                                  tmp2[x] * (64 - mask[x]) + 512) >> 10);

        tmp1 += w;
        tmp2 += w;
        mask += w;
        dst += PXSTRIDE(dst_stride);
    } while (--h);
}

Henrik Gramner's avatar
Henrik Gramner committed
559 560 561 562 563
#define blend_px(a, b, m) (((a * (64 - m) + b * m) + 32) >> 6)
static NOINLINE void
blend_internal_c(pixel *dst, const ptrdiff_t dst_stride, const pixel *tmp,
                 const int w, int h, const uint8_t *mask,
                 const ptrdiff_t mask_stride)
564
{
Henrik Gramner's avatar
Henrik Gramner committed
565
    do {
566
        for (int x = 0; x < w; x++) {
Henrik Gramner's avatar
Henrik Gramner committed
567
            dst[x] = blend_px(dst[x], tmp[x], mask[x]);
568 569
        }
        dst += PXSTRIDE(dst_stride);
570
        tmp += w;
Henrik Gramner's avatar
Henrik Gramner committed
571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598
        mask += mask_stride;
    } while (--h);
}

static void blend_c(pixel *dst, const ptrdiff_t dst_stride, const pixel *tmp,
                    const int w, const int h, const uint8_t *mask)
{
    blend_internal_c(dst, dst_stride, tmp, w, h, mask, w);
}

static void blend_v_c(pixel *dst, const ptrdiff_t dst_stride, const pixel *tmp,
                      const int w, const int h)
{
    blend_internal_c(dst, dst_stride, tmp, w, h, &dav1d_obmc_masks[w], 0);
}

static void blend_h_c(pixel *dst, const ptrdiff_t dst_stride, const pixel *tmp,
                      const int w, int h)
{
    const uint8_t *mask = &dav1d_obmc_masks[h];
    do {
        const int m = *mask++;
        for (int x = 0; x < w; x++) {
            dst[x] = blend_px(dst[x], tmp[x], m);
        }
        dst += PXSTRIDE(dst_stride);
        tmp += w;
    } while (--h);
599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
}

static void w_mask_c(pixel *dst, const ptrdiff_t dst_stride,
                     const coef *tmp1, const coef *tmp2, const int w, int h,
                     uint8_t *mask, const int sign,
                     const int ss_hor, const int ss_ver)
{
    // store mask at 2x2 resolution, i.e. store 2x1 sum for even rows,
    // and then load this intermediate to calculate final value for odd rows
    const int rnd = 8 << (BITDEPTH - 8);
    do {
        for (int x = 0; x < w; x++) {
            const int m = imin(38 + ((abs(tmp1[x] - tmp2[x]) + rnd) >> BITDEPTH), 64);
            dst[x] = iclip_pixel((tmp1[x] * m +
                                  tmp2[x] * (64 - m) + 512) >> 10);

            if (ss_hor) {
                x++;

                const int n = imin(38 + ((abs(tmp1[x] - tmp2[x]) + rnd) >> BITDEPTH), 64);
                dst[x] = iclip_pixel((tmp1[x] * n +
                                      tmp2[x] * (64 - n) + 512) >> 10);

                if (h & ss_ver) {
                    mask[x >> 1] = (m + n + mask[x >> 1] + 2 - sign) >> 2;
                } else if (ss_ver) {
                    mask[x >> 1] = m + n;
                } else {
                    mask[x >> 1] = (m + n + 1 - sign) >> 1;
                }
            } else {
                mask[x] = m;
            }
        }

        tmp1 += w;
        tmp2 += w;
        dst += PXSTRIDE(dst_stride);
        if (!ss_ver || (h & 1)) mask += w >> ss_hor;
    } while (--h);
}

#define w_mask_fns(ssn, ss_hor, ss_ver) \
static void w_mask_##ssn##_c(pixel *const dst, const ptrdiff_t dst_stride, \
                             const coef *const tmp1, const coef *const tmp2, \
                             const int w, const int h, uint8_t *mask, \
                             const int sign) \
{ \
    w_mask_c(dst, dst_stride, tmp1, tmp2, w, h, mask, sign, ss_hor, ss_ver); \
}

w_mask_fns(444, 0, 0);
w_mask_fns(422, 1, 0);
w_mask_fns(420, 1, 1);

#undef w_mask_fns

656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
#define FILTER_WARP(src, x, F, stride) \
    (F[0] * src[x + -3 * stride] + \
     F[4] * src[x + -2 * stride] + \
     F[1] * src[x + -1 * stride] + \
     F[5] * src[x + +0 * stride] + \
     F[2] * src[x + +1 * stride] + \
     F[6] * src[x + +2 * stride] + \
     F[3] * src[x + +3 * stride] + \
     F[7] * src[x + +4 * stride])

#define FILTER_WARP_RND(src, x, F, stride, sh) \
    ((FILTER_WARP(src, x, F, stride) + ((1 << sh) >> 1)) >> sh)

#define FILTER_WARP_CLIP(src, x, F, stride, sh) \
    iclip_pixel(FILTER_WARP_RND(src, x, F, stride, sh))

672 673 674 675 676 677 678 679 680 681 682 683
static void warp_affine_8x8_c(pixel *dst, const ptrdiff_t dst_stride,
                              const pixel *src, const ptrdiff_t src_stride,
                              const int16_t *const abcd, int mx, int my)
{
    coef mid[15 * 8], *mid_ptr = mid;

    src -= 3 * PXSTRIDE(src_stride);
    for (int y = 0; y < 15; y++, mx += abcd[1]) {
        for (int x = 0, tmx = mx; x < 8; x++, tmx += abcd[0]) {
            const int8_t *const filter =
                dav1d_mc_warp_filter[64 + ((tmx + 512) >> 10)];

684
            mid_ptr[x] = FILTER_WARP_RND(src, x, filter, 1, 3);
685 686 687 688 689 690 691 692 693 694 695
        }
        src += PXSTRIDE(src_stride);
        mid_ptr += 8;
    }

    mid_ptr = &mid[3 * 8];
    for (int y = 0; y < 8; y++, my += abcd[3]) {
        for (int x = 0, tmy = my; x < 8; x++, tmy += abcd[2]) {
            const int8_t *const filter =
                dav1d_mc_warp_filter[64 + ((tmy + 512) >> 10)];

696
            dst[x] = FILTER_WARP_CLIP(mid_ptr, x, filter, 8, 11);
697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714
        }
        mid_ptr += 8;
        dst += PXSTRIDE(dst_stride);
    }
}

static void warp_affine_8x8t_c(coef *tmp, const ptrdiff_t tmp_stride,
                               const pixel *src, const ptrdiff_t src_stride,
                               const int16_t *const abcd, int mx, int my)
{
    coef mid[15 * 8], *mid_ptr = mid;

    src -= 3 * PXSTRIDE(src_stride);
    for (int y = 0; y < 15; y++, mx += abcd[1]) {
        for (int x = 0, tmx = mx; x < 8; x++, tmx += abcd[0]) {
            const int8_t *const filter =
                dav1d_mc_warp_filter[64 + ((tmx + 512) >> 10)];

715
            mid_ptr[x] = FILTER_WARP_RND(src, x, filter, 1, 3);
716 717 718 719 720 721 722 723 724 725 726
        }
        src += PXSTRIDE(src_stride);
        mid_ptr += 8;
    }

    mid_ptr = &mid[3 * 8];
    for (int y = 0; y < 8; y++, my += abcd[3]) {
        for (int x = 0, tmy = my; x < 8; x++, tmy += abcd[2]) {
            const int8_t *const filter =
                dav1d_mc_warp_filter[64 + ((tmy + 512) >> 10)];

727
            tmp[x] = FILTER_WARP_RND(mid_ptr, x, filter, 8, 7);
728 729 730 731 732 733
        }
        mid_ptr += 8;
        tmp += tmp_stride;
    }
}

734 735 736 737 738
static void emu_edge_c(const intptr_t bw, const intptr_t bh,
                       const intptr_t iw, const intptr_t ih,
                       const intptr_t x, const intptr_t y,
                       pixel *dst, const ptrdiff_t dst_stride,
                       const pixel *ref, const ptrdiff_t ref_stride)
739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782
{
    // find offset in reference of visible block to copy
    ref += iclip(y, 0, ih - 1) * PXSTRIDE(ref_stride) + iclip(x, 0, iw - 1);

    // number of pixels to extend (left, right, top, bottom)
    const int left_ext = iclip(-x, 0, bw - 1);
    const int right_ext = iclip(x + bw - iw, 0, bw - 1);
    assert(left_ext + right_ext < bw);
    const int top_ext = iclip(-y, 0, bh - 1);
    const int bottom_ext = iclip(y + bh - ih, 0, bh - 1);
    assert(top_ext + bottom_ext < bh);

    // copy visible portion first
    pixel *blk = dst + top_ext * PXSTRIDE(dst_stride);
    const int center_w = bw - left_ext - right_ext;
    const int center_h = bh - top_ext - bottom_ext;
    for (int y = 0; y < center_h; y++) {
        pixel_copy(blk + left_ext, ref, center_w);
        // extend left edge for this line
        if (left_ext)
            pixel_set(blk, blk[left_ext], left_ext);
        // extend right edge for this line
        if (right_ext)
            pixel_set(blk + left_ext + center_w, blk[left_ext + center_w - 1],
                      right_ext);
        ref += PXSTRIDE(ref_stride);
        blk += PXSTRIDE(dst_stride);
    }

    // copy top
    blk = dst + top_ext * PXSTRIDE(dst_stride);
    for (int y = 0; y < top_ext; y++) {
        pixel_copy(dst, blk, bw);
        dst += PXSTRIDE(dst_stride);
    }

    // copy bottom
    dst += center_h * PXSTRIDE(dst_stride);
    for (int y = 0; y < bottom_ext; y++) {
        pixel_copy(dst, &dst[-PXSTRIDE(dst_stride)], bw);
        dst += PXSTRIDE(dst_stride);
    }
}

783 784
void bitfn(dav1d_mc_dsp_init)(Dav1dMCDSPContext *const c) {
#define init_mc_fns(type, name) do { \
785 786 787 788
    c->mc        [type] = put_##name##_c; \
    c->mc_scaled [type] = put_##name##_scaled_c; \
    c->mct       [type] = prep_##name##_c; \
    c->mct_scaled[type] = prep_##name##_scaled_c; \
789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
} while (0)

    init_mc_fns(FILTER_2D_8TAP_REGULAR,        8tap_regular);
    init_mc_fns(FILTER_2D_8TAP_REGULAR_SMOOTH, 8tap_regular_smooth);
    init_mc_fns(FILTER_2D_8TAP_REGULAR_SHARP,  8tap_regular_sharp);
    init_mc_fns(FILTER_2D_8TAP_SHARP_REGULAR,  8tap_sharp_regular);
    init_mc_fns(FILTER_2D_8TAP_SHARP_SMOOTH,   8tap_sharp_smooth);
    init_mc_fns(FILTER_2D_8TAP_SHARP,          8tap_sharp);
    init_mc_fns(FILTER_2D_8TAP_SMOOTH_REGULAR, 8tap_smooth_regular);
    init_mc_fns(FILTER_2D_8TAP_SMOOTH,         8tap_smooth);
    init_mc_fns(FILTER_2D_8TAP_SMOOTH_SHARP,   8tap_smooth_sharp);
    init_mc_fns(FILTER_2D_BILINEAR,            bilin);

    c->avg      = avg_c;
    c->w_avg    = w_avg_c;
    c->mask     = mask_c;
    c->blend    = blend_c;
Henrik Gramner's avatar
Henrik Gramner committed
806 807
    c->blend_v  = blend_v_c;
    c->blend_h  = blend_h_c;
808 809 810 811 812
    c->w_mask[0] = w_mask_444_c;
    c->w_mask[1] = w_mask_422_c;
    c->w_mask[2] = w_mask_420_c;
    c->warp8x8  = warp_affine_8x8_c;
    c->warp8x8t = warp_affine_8x8t_c;
813
    c->emu_edge = emu_edge_c;
Henrik Gramner's avatar
Henrik Gramner committed
814

815 816 817 818
#if HAVE_ASM
#if ARCH_AARCH64 || ARCH_ARM
    bitfn(dav1d_mc_dsp_init_arm)(c);
#elif ARCH_X86
Henrik Gramner's avatar
Henrik Gramner committed
819 820
    bitfn(dav1d_mc_dsp_init_x86)(c);
#endif
821
#endif
822
}