Mercurial > libavcodec.hg
comparison vp6.c @ 4921:9a2c08b939ee libavcodec
faster and simpler vp6 bilinear mc
| author | lorenm |
|---|---|
| date | Sun, 06 May 2007 06:43:33 +0000 |
| parents | 143b89ab8187 |
| children | 58f751086434 |
comparison
equal
deleted
inserted
replaced
| 4920:6258f497d42e | 4921:9a2c08b939ee |
|---|---|
| 378 src += 2*stride; | 378 src += 2*stride; |
| 379 } | 379 } |
| 380 return (16*square_sum - sum*sum) >> 8; | 380 return (16*square_sum - sum*sum) >> 8; |
| 381 } | 381 } |
| 382 | 382 |
| 383 static void vp6_filter_hv2(vp56_context_t *s, uint8_t *dst, uint8_t *src, | |
| 384 int stride, int delta, int16_t weight) | |
| 385 { | |
| 386 s->dsp.put_pixels_tab[1][0](dst, src, stride, 8); | |
| 387 s->dsp.biweight_h264_pixels_tab[3](dst, src+delta, stride, 2, | |
| 388 8-weight, weight, 0); | |
| 389 } | |
| 390 | |
| 391 static void vp6_filter_hv4(uint8_t *dst, uint8_t *src, int stride, | 383 static void vp6_filter_hv4(uint8_t *dst, uint8_t *src, int stride, |
| 392 int delta, const int16_t *weights) | 384 int delta, const int16_t *weights) |
| 393 { | 385 { |
| 394 int x, y; | 386 int x, y; |
| 395 | 387 |
| 407 | 399 |
| 408 static void vp6_filter_diag2(vp56_context_t *s, uint8_t *dst, uint8_t *src, | 400 static void vp6_filter_diag2(vp56_context_t *s, uint8_t *dst, uint8_t *src, |
| 409 int stride, int h_weight, int v_weight) | 401 int stride, int h_weight, int v_weight) |
| 410 { | 402 { |
| 411 uint8_t *tmp = s->edge_emu_buffer+16; | 403 uint8_t *tmp = s->edge_emu_buffer+16; |
| 412 int x, xmax; | 404 s->dsp.put_h264_chroma_pixels_tab[0](tmp, src, stride, 9, h_weight, 0); |
| 413 | 405 s->dsp.put_h264_chroma_pixels_tab[0](dst, tmp, stride, 8, 0, v_weight); |
| 414 s->dsp.put_pixels_tab[1][0](tmp, src, stride, 8); | |
| 415 s->dsp.biweight_h264_pixels_tab[3](tmp, src+1, stride, 2, | |
| 416 8-h_weight, h_weight, 0); | |
| 417 /* we need a 8x9 block to do vertical filter, so compute one more line */ | |
| 418 for (x=8*stride, xmax=x+8; x<xmax; x++) | |
| 419 tmp[x] = (src[x]*(8-h_weight) + src[x+1]*h_weight + 4) >> 3; | |
| 420 | |
| 421 s->dsp.put_pixels_tab[1][0](dst, tmp, stride, 8); | |
| 422 s->dsp.biweight_h264_pixels_tab[3](dst, tmp+stride, stride, 2, | |
| 423 8-v_weight, v_weight, 0); | |
| 424 } | 406 } |
| 425 | 407 |
| 426 static void vp6_filter_diag4(uint8_t *dst, uint8_t *src, int stride, | 408 static void vp6_filter_diag4(uint8_t *dst, uint8_t *src, int stride, |
| 427 const int16_t *h_weights,const int16_t *v_weights) | 409 const int16_t *h_weights,const int16_t *v_weights) |
| 428 { | 410 { |
| 500 vp6_filter_diag4(dst, src+offset1, stride, | 482 vp6_filter_diag4(dst, src+offset1, stride, |
| 501 vp6_block_copy_filter[select][x8], | 483 vp6_block_copy_filter[select][x8], |
| 502 vp6_block_copy_filter[select][y8]); | 484 vp6_block_copy_filter[select][y8]); |
| 503 } | 485 } |
| 504 } else { | 486 } else { |
| 505 if (!y8) { /* left or right combine */ | 487 if (!x8 || !y8) { |
| 506 vp6_filter_hv2(s, dst, src+offset1, stride, 1, x8); | 488 s->dsp.put_h264_chroma_pixels_tab[0](dst, src+offset1, stride, 8, x8, y8); |
| 507 } else if (!x8) { /* above or below combine */ | |
| 508 vp6_filter_hv2(s, dst, src+offset1, stride, stride, y8); | |
| 509 } else if ((mv.x^mv.y) >> 31) { /* lower-left or upper-right combine */ | 489 } else if ((mv.x^mv.y) >> 31) { /* lower-left or upper-right combine */ |
| 510 vp6_filter_diag2(s, dst, src+offset1-1, stride, x8, y8); | 490 vp6_filter_diag2(s, dst, src+offset1-1, stride, x8, y8); |
| 511 } else { /* lower-right or upper-left combine */ | 491 } else { /* lower-right or upper-left combine */ |
| 512 vp6_filter_diag2(s, dst, src+offset1, stride, x8, y8); | 492 vp6_filter_diag2(s, dst, src+offset1, stride, x8, y8); |
| 513 } | 493 } |