OLD | NEW |
1 /* | 1 /* |
2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved. | 2 * Copyright (c) 2010 The WebM project authors. All Rights Reserved. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license | 4 * Use of this source code is governed by a BSD-style license |
5 * that can be found in the LICENSE file in the root of the source | 5 * that can be found in the LICENSE file in the root of the source |
6 * tree. An additional intellectual property rights grant can be found | 6 * tree. An additional intellectual property rights grant can be found |
7 * in the file PATENTS. All contributing project authors may | 7 * in the file PATENTS. All contributing project authors may |
8 * be found in the AUTHORS file in the root of the source tree. | 8 * be found in the AUTHORS file in the root of the source tree. |
9 */ | 9 */ |
10 | 10 |
(...skipping 249 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
260 for (i = 0; i < height; i++) { | 260 for (i = 0; i < height; i++) { |
261 for (j = 0; j < width; j++) { | 261 for (j = 0; j < width; j++) { |
262 const int tmp = pred[j] + ref[j]; | 262 const int tmp = pred[j] + ref[j]; |
263 comp_pred[j] = ROUND_POWER_OF_TWO(tmp, 1); | 263 comp_pred[j] = ROUND_POWER_OF_TWO(tmp, 1); |
264 } | 264 } |
265 comp_pred += width; | 265 comp_pred += width; |
266 pred += width; | 266 pred += width; |
267 ref += ref_stride; | 267 ref += ref_stride; |
268 } | 268 } |
269 } | 269 } |
| 270 |
| 271 #if CONFIG_VP9_HIGHBITDEPTH |
| 272 void high_variance64(const uint8_t *a8, int a_stride, |
| 273 const uint8_t *b8, int b_stride, |
| 274 int w, int h, uint64_t *sse, |
| 275 uint64_t *sum) { |
| 276 int i, j; |
| 277 |
| 278 uint16_t *a = CONVERT_TO_SHORTPTR(a8); |
| 279 uint16_t *b = CONVERT_TO_SHORTPTR(b8); |
| 280 *sum = 0; |
| 281 *sse = 0; |
| 282 |
| 283 for (i = 0; i < h; i++) { |
| 284 for (j = 0; j < w; j++) { |
| 285 const int diff = a[j] - b[j]; |
| 286 *sum += diff; |
| 287 *sse += diff * diff; |
| 288 } |
| 289 a += a_stride; |
| 290 b += b_stride; |
| 291 } |
| 292 } |
| 293 |
| 294 void high_variance(const uint8_t *a8, int a_stride, |
| 295 const uint8_t *b8, int b_stride, |
| 296 int w, int h, unsigned int *sse, |
| 297 int *sum) { |
| 298 uint64_t sse_long = 0; |
| 299 uint64_t sum_long = 0; |
| 300 high_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long); |
| 301 *sse = sse_long; |
| 302 *sum = sum_long; |
| 303 } |
| 304 |
| 305 void high_10_variance(const uint8_t *a8, int a_stride, |
| 306 const uint8_t *b8, int b_stride, |
| 307 int w, int h, unsigned int *sse, |
| 308 int *sum) { |
| 309 uint64_t sse_long = 0; |
| 310 uint64_t sum_long = 0; |
| 311 high_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long); |
| 312 *sum = ROUND_POWER_OF_TWO(sum_long, 2); |
| 313 *sse = ROUND_POWER_OF_TWO(sse_long, 4); |
| 314 } |
| 315 |
| 316 void high_12_variance(const uint8_t *a8, int a_stride, |
| 317 const uint8_t *b8, int b_stride, |
| 318 int w, int h, unsigned int *sse, |
| 319 int *sum) { |
| 320 uint64_t sse_long = 0; |
| 321 uint64_t sum_long = 0; |
| 322 high_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long); |
| 323 *sum = ROUND_POWER_OF_TWO(sum_long, 4); |
| 324 *sse = ROUND_POWER_OF_TWO(sse_long, 8); |
| 325 } |
| 326 |
| 327 static void high_var_filter_block2d_bil_first_pass( |
| 328 const uint8_t *src_ptr8, |
| 329 uint16_t *output_ptr, |
| 330 unsigned int src_pixels_per_line, |
| 331 int pixel_step, |
| 332 unsigned int output_height, |
| 333 unsigned int output_width, |
| 334 const int16_t *vp9_filter) { |
| 335 unsigned int i, j; |
| 336 uint16_t *src_ptr = CONVERT_TO_SHORTPTR(src_ptr8); |
| 337 for (i = 0; i < output_height; i++) { |
| 338 for (j = 0; j < output_width; j++) { |
| 339 output_ptr[j] = |
| 340 ROUND_POWER_OF_TWO((int)src_ptr[0] * vp9_filter[0] + |
| 341 (int)src_ptr[pixel_step] * vp9_filter[1], |
| 342 FILTER_BITS); |
| 343 |
| 344 src_ptr++; |
| 345 } |
| 346 |
| 347 // Next row... |
| 348 src_ptr += src_pixels_per_line - output_width; |
| 349 output_ptr += output_width; |
| 350 } |
| 351 } |
| 352 |
| 353 static void high_var_filter_block2d_bil_second_pass( |
| 354 const uint16_t *src_ptr, |
| 355 uint16_t *output_ptr, |
| 356 unsigned int src_pixels_per_line, |
| 357 unsigned int pixel_step, |
| 358 unsigned int output_height, |
| 359 unsigned int output_width, |
| 360 const int16_t *vp9_filter) { |
| 361 unsigned int i, j; |
| 362 |
| 363 for (i = 0; i < output_height; i++) { |
| 364 for (j = 0; j < output_width; j++) { |
| 365 output_ptr[j] = |
| 366 ROUND_POWER_OF_TWO((int)src_ptr[0] * vp9_filter[0] + |
| 367 (int)src_ptr[pixel_step] * vp9_filter[1], |
| 368 FILTER_BITS); |
| 369 src_ptr++; |
| 370 } |
| 371 |
| 372 src_ptr += src_pixels_per_line - output_width; |
| 373 output_ptr += output_width; |
| 374 } |
| 375 } |
| 376 |
| 377 #define HIGH_VAR(W, H) \ |
| 378 unsigned int vp9_high_variance##W##x##H##_c(const uint8_t *a, int a_stride, \ |
| 379 const uint8_t *b, int b_stride, \ |
| 380 unsigned int *sse) { \ |
| 381 int sum; \ |
| 382 high_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \ |
| 383 return *sse - (((int64_t)sum * sum) / (W * H)); \ |
| 384 } \ |
| 385 \ |
| 386 unsigned int vp9_high_10_variance##W##x##H##_c(const uint8_t *a, int a_stride, \ |
| 387 const uint8_t *b, int b_stride, \ |
| 388 unsigned int *sse) { \ |
| 389 int sum; \ |
| 390 high_10_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \ |
| 391 return *sse - (((int64_t)sum * sum) / (W * H)); \ |
| 392 } \ |
| 393 \ |
| 394 unsigned int vp9_high_12_variance##W##x##H##_c(const uint8_t *a, int a_stride, \ |
| 395 const uint8_t *b, int b_stride, \ |
| 396 unsigned int *sse) { \ |
| 397 int sum; \ |
| 398 high_12_variance(a, a_stride, b, b_stride, W, H, sse, &sum); \ |
| 399 return *sse - (((int64_t)sum * sum) / (W * H)); \ |
| 400 } |
| 401 |
| 402 #define HIGH_SUBPIX_VAR(W, H) \ |
| 403 unsigned int vp9_high_sub_pixel_variance##W##x##H##_c( \ |
| 404 const uint8_t *src, int src_stride, \ |
| 405 int xoffset, int yoffset, \ |
| 406 const uint8_t *dst, int dst_stride, \ |
| 407 unsigned int *sse) { \ |
| 408 uint16_t fdata3[(H + 1) * W]; \ |
| 409 uint16_t temp2[H * W]; \ |
| 410 \ |
| 411 high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \ |
| 412 W, BILINEAR_FILTERS_2TAP(xoffset)); \ |
| 413 high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ |
| 414 BILINEAR_FILTERS_2TAP(yoffset)); \ |
| 415 \ |
| 416 return vp9_high_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), W, dst, \ |
| 417 dst_stride, sse); \ |
| 418 } \ |
| 419 \ |
| 420 unsigned int vp9_high_10_sub_pixel_variance##W##x##H##_c( \ |
| 421 const uint8_t *src, int src_stride, \ |
| 422 int xoffset, int yoffset, \ |
| 423 const uint8_t *dst, int dst_stride, \ |
| 424 unsigned int *sse) { \ |
| 425 uint16_t fdata3[(H + 1) * W]; \ |
| 426 uint16_t temp2[H * W]; \ |
| 427 \ |
| 428 high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \ |
| 429 W, BILINEAR_FILTERS_2TAP(xoffset)); \ |
| 430 high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ |
| 431 BILINEAR_FILTERS_2TAP(yoffset)); \ |
| 432 \ |
| 433 return vp9_high_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), W, dst, \ |
| 434 dst_stride, sse); \ |
| 435 } \ |
| 436 \ |
| 437 unsigned int vp9_high_12_sub_pixel_variance##W##x##H##_c( \ |
| 438 const uint8_t *src, int src_stride, \ |
| 439 int xoffset, int yoffset, \ |
| 440 const uint8_t *dst, int dst_stride, \ |
| 441 unsigned int *sse) { \ |
| 442 uint16_t fdata3[(H + 1) * W]; \ |
| 443 uint16_t temp2[H * W]; \ |
| 444 \ |
| 445 high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \ |
| 446 W, BILINEAR_FILTERS_2TAP(xoffset)); \ |
| 447 high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ |
| 448 BILINEAR_FILTERS_2TAP(yoffset)); \ |
| 449 \ |
| 450 return vp9_high_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp2), W, dst, \ |
| 451 dst_stride, sse); \ |
| 452 } |
| 453 |
| 454 #define HIGH_SUBPIX_AVG_VAR(W, H) \ |
| 455 unsigned int vp9_high_sub_pixel_avg_variance##W##x##H##_c( \ |
| 456 const uint8_t *src, int src_stride, \ |
| 457 int xoffset, int yoffset, \ |
| 458 const uint8_t *dst, int dst_stride, \ |
| 459 unsigned int *sse, \ |
| 460 const uint8_t *second_pred) { \ |
| 461 uint16_t fdata3[(H + 1) * W]; \ |
| 462 uint16_t temp2[H * W]; \ |
| 463 DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \ |
| 464 \ |
| 465 high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \ |
| 466 W, BILINEAR_FILTERS_2TAP(xoffset)); \ |
| 467 high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ |
| 468 BILINEAR_FILTERS_2TAP(yoffset)); \ |
| 469 \ |
| 470 vp9_high_comp_avg_pred(temp3, second_pred, W, H, CONVERT_TO_BYTEPTR(temp2), \ |
| 471 W); \ |
| 472 \ |
| 473 return vp9_high_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, dst, \ |
| 474 dst_stride, sse); \ |
| 475 } \ |
| 476 \ |
| 477 unsigned int vp9_high_10_sub_pixel_avg_variance##W##x##H##_c( \ |
| 478 const uint8_t *src, int src_stride, \ |
| 479 int xoffset, int yoffset, \ |
| 480 const uint8_t *dst, int dst_stride, \ |
| 481 unsigned int *sse, \ |
| 482 const uint8_t *second_pred) { \ |
| 483 uint16_t fdata3[(H + 1) * W]; \ |
| 484 uint16_t temp2[H * W]; \ |
| 485 DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \ |
| 486 \ |
| 487 high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \ |
| 488 W, BILINEAR_FILTERS_2TAP(xoffset)); \ |
| 489 high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ |
| 490 BILINEAR_FILTERS_2TAP(yoffset)); \ |
| 491 \ |
| 492 vp9_high_comp_avg_pred(temp3, second_pred, W, H, CONVERT_TO_BYTEPTR(temp2), \ |
| 493 W); \ |
| 494 \ |
| 495 return vp9_high_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, dst, \ |
| 496 dst_stride, sse); \ |
| 497 } \ |
| 498 \ |
| 499 unsigned int vp9_high_12_sub_pixel_avg_variance##W##x##H##_c( \ |
| 500 const uint8_t *src, int src_stride, \ |
| 501 int xoffset, int yoffset, \ |
| 502 const uint8_t *dst, int dst_stride, \ |
| 503 unsigned int *sse, \ |
| 504 const uint8_t *second_pred) { \ |
| 505 uint16_t fdata3[(H + 1) * W]; \ |
| 506 uint16_t temp2[H * W]; \ |
| 507 DECLARE_ALIGNED_ARRAY(16, uint16_t, temp3, H * W); \ |
| 508 \ |
| 509 high_var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, \ |
| 510 W, BILINEAR_FILTERS_2TAP(xoffset)); \ |
| 511 high_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ |
| 512 BILINEAR_FILTERS_2TAP(yoffset)); \ |
| 513 \ |
| 514 vp9_high_comp_avg_pred(temp3, second_pred, W, H, CONVERT_TO_BYTEPTR(temp2), \ |
| 515 W); \ |
| 516 \ |
| 517 return vp9_high_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, dst, \ |
| 518 dst_stride, sse); \ |
| 519 } |
| 520 |
| 521 #define HIGH_GET_VAR(S) \ |
| 522 void vp9_high_get##S##x##S##var_c(const uint8_t *src, int src_stride, \ |
| 523 const uint8_t *ref, int ref_stride, \ |
| 524 unsigned int *sse, int *sum) { \ |
| 525 high_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \ |
| 526 } \ |
| 527 \ |
| 528 void vp9_high_10_get##S##x##S##var_c(const uint8_t *src, int src_stride, \ |
| 529 const uint8_t *ref, int ref_stride, \ |
| 530 unsigned int *sse, int *sum) { \ |
| 531 high_10_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \ |
| 532 } \ |
| 533 \ |
| 534 void vp9_high_12_get##S##x##S##var_c(const uint8_t *src, int src_stride, \ |
| 535 const uint8_t *ref, int ref_stride, \ |
| 536 unsigned int *sse, int *sum) { \ |
| 537 high_12_variance(src, src_stride, ref, ref_stride, S, S, sse, sum); \ |
| 538 } |
| 539 |
| 540 #define HIGH_MSE(W, H) \ |
| 541 unsigned int vp9_high_mse##W##x##H##_c(const uint8_t *src, int src_stride, \ |
| 542 const uint8_t *ref, int ref_stride, \ |
| 543 unsigned int *sse) { \ |
| 544 int sum; \ |
| 545 high_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \ |
| 546 return *sse; \ |
| 547 } \ |
| 548 \ |
| 549 unsigned int vp9_high_10_mse##W##x##H##_c(const uint8_t *src, int src_stride, \ |
| 550 const uint8_t *ref, int ref_stride, \ |
| 551 unsigned int *sse) { \ |
| 552 int sum; \ |
| 553 high_10_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \ |
| 554 return *sse; \ |
| 555 } \ |
| 556 \ |
| 557 unsigned int vp9_high_12_mse##W##x##H##_c(const uint8_t *src, int src_stride, \ |
| 558 const uint8_t *ref, int ref_stride, \ |
| 559 unsigned int *sse) { \ |
| 560 int sum; \ |
| 561 high_12_variance(src, src_stride, ref, ref_stride, W, H, sse, &sum); \ |
| 562 return *sse; \ |
| 563 } |
| 564 |
| 565 HIGH_GET_VAR(8) |
| 566 HIGH_GET_VAR(16) |
| 567 |
| 568 HIGH_MSE(16, 16) |
| 569 HIGH_MSE(16, 8) |
| 570 HIGH_MSE(8, 16) |
| 571 HIGH_MSE(8, 8) |
| 572 |
| 573 HIGH_VAR(4, 4) |
| 574 HIGH_SUBPIX_VAR(4, 4) |
| 575 HIGH_SUBPIX_AVG_VAR(4, 4) |
| 576 |
| 577 HIGH_VAR(4, 8) |
| 578 HIGH_SUBPIX_VAR(4, 8) |
| 579 HIGH_SUBPIX_AVG_VAR(4, 8) |
| 580 |
| 581 HIGH_VAR(8, 4) |
| 582 HIGH_SUBPIX_VAR(8, 4) |
| 583 HIGH_SUBPIX_AVG_VAR(8, 4) |
| 584 |
| 585 HIGH_VAR(8, 8) |
| 586 HIGH_SUBPIX_VAR(8, 8) |
| 587 HIGH_SUBPIX_AVG_VAR(8, 8) |
| 588 |
| 589 HIGH_VAR(8, 16) |
| 590 HIGH_SUBPIX_VAR(8, 16) |
| 591 HIGH_SUBPIX_AVG_VAR(8, 16) |
| 592 |
| 593 HIGH_VAR(16, 8) |
| 594 HIGH_SUBPIX_VAR(16, 8) |
| 595 HIGH_SUBPIX_AVG_VAR(16, 8) |
| 596 |
| 597 HIGH_VAR(16, 16) |
| 598 HIGH_SUBPIX_VAR(16, 16) |
| 599 HIGH_SUBPIX_AVG_VAR(16, 16) |
| 600 |
| 601 HIGH_VAR(16, 32) |
| 602 HIGH_SUBPIX_VAR(16, 32) |
| 603 HIGH_SUBPIX_AVG_VAR(16, 32) |
| 604 |
| 605 HIGH_VAR(32, 16) |
| 606 HIGH_SUBPIX_VAR(32, 16) |
| 607 HIGH_SUBPIX_AVG_VAR(32, 16) |
| 608 |
| 609 HIGH_VAR(32, 32) |
| 610 HIGH_SUBPIX_VAR(32, 32) |
| 611 HIGH_SUBPIX_AVG_VAR(32, 32) |
| 612 |
| 613 HIGH_VAR(32, 64) |
| 614 HIGH_SUBPIX_VAR(32, 64) |
| 615 HIGH_SUBPIX_AVG_VAR(32, 64) |
| 616 |
| 617 HIGH_VAR(64, 32) |
| 618 HIGH_SUBPIX_VAR(64, 32) |
| 619 HIGH_SUBPIX_AVG_VAR(64, 32) |
| 620 |
| 621 HIGH_VAR(64, 64) |
| 622 HIGH_SUBPIX_VAR(64, 64) |
| 623 HIGH_SUBPIX_AVG_VAR(64, 64) |
| 624 |
| 625 void vp9_high_comp_avg_pred(uint16_t *comp_pred, const uint8_t *pred8, |
| 626 int width, int height, const uint8_t *ref8, |
| 627 int ref_stride) { |
| 628 int i, j; |
| 629 uint16_t *pred = CONVERT_TO_SHORTPTR(pred8); |
| 630 uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); |
| 631 for (i = 0; i < height; i++) { |
| 632 for (j = 0; j < width; j++) { |
| 633 const int tmp = pred[j] + ref[j]; |
| 634 comp_pred[j] = ROUND_POWER_OF_TWO(tmp, 1); |
| 635 } |
| 636 comp_pred += width; |
| 637 pred += width; |
| 638 ref += ref_stride; |
| 639 } |
| 640 } |
| 641 #endif // CONFIG_VP9_HIGHBITDEPTH |
OLD | NEW |