Just keep the pointers for the beginning of a line uint8_t* and use uint16_t* to do the actual processing.
Unified · Split
| @@ -467,7 +467,6 @@ static av_always_inline void blend_plane_##depth##_##nbits##bits(AVFilterContext · 466 unmodified lines | |||
| 467 | int dst_hp = AV_CEIL_RSHIFT(dst_h, vsub); \ | 467 | int dst_hp = AV_CEIL_RSHIFT(dst_h, vsub); \ |
| 468 | int yp = y>>vsub; \ | 468 | int yp = y>>vsub; \ |
| 469 | int xp = x>>hsub; \ | 469 | int xp = x>>hsub; \ |
| 470 | uint##depth##_t *s, *sp, *d, *dp, *dap, *a, *ap; \ | ||
| 471 | int jmax, j, k, kmax; \ | 470 | int jmax, j, k, kmax; \ |
| 472 | int slice_start, slice_end; \ | 471 | int slice_start, slice_end; \ |
| 473 | const uint##depth##_t max = (1 << nbits) - 1; \ | 472 | const uint##depth##_t max = (1 << nbits) - 1; \ |
| @@ -481,20 +480,19 @@ static av_always_inline void blend_plane_##depth##_##nbits##bits(AVFilterContext · 7 unmodified lines | |||
| 481 | slice_start = j + (jmax * jobnr) / nb_jobs; \ | 480 | slice_start = j + (jmax * jobnr) / nb_jobs; \ |
| 482 | slice_end = j + (jmax * (jobnr+1)) / nb_jobs; \ | 481 | slice_end = j + (jmax * (jobnr+1)) / nb_jobs; \ |
| 483 | \ | 482 | \ |
| 484 | sp = (uint##depth##_t *)(src->data[i] + (slice_start) * src->linesize[i]); \ | 483 | const uint8_t *sp = src->data[i] + (slice_start) * src->linesize[i]; \ |
| 485 | dp = (uint##depth##_t *)(dst->data[dst_plane] \ | 484 | uint8_t *dp = dst->data[dst_plane] \ |
| 486 | + (yp + slice_start) * dst->linesize[dst_plane] \ | 485 | + (yp + slice_start) * dst->linesize[dst_plane] \ |
| 487 | + dst_offset); \ | 486 | + dst_offset; \ |
| 488 | ap = (uint##depth##_t *)(src->data[3] + (slice_start << vsub) * src->linesize[3]); \ | 487 | const uint8_t *ap = src->data[3] + (slice_start << vsub) * src->linesize[3]; \ |
| 489 | if (main_has_alpha) \ | 488 | const uint8_t *dap = main_has_alpha ? dst->data[3] + ((yp + slice_start) << vsub) * dst->linesize[3] : NULL; \ |
| 490 | dap = (uint##depth##_t *)(dst->data[3] + ((yp + slice_start) << vsub) * dst->linesize[3]); \ | ||
| 491 | \ | 489 | \ |
| 492 | for (j = slice_start; j < slice_end; j++) { \ | 490 | for (j = slice_start; j < slice_end; j++) { \ |
| 493 | k = FFMAX(-xp, 0); \ | 491 | k = FFMAX(-xp, 0); \ |
| 494 | d = dp + (xp+k) * dst_step; \ | 492 | const uint##depth##_t *s = (const uint##depth##_t *)sp + k; \ |
| 495 | s = sp + k; \ | 493 | const uint##depth##_t *a = (const uint##depth##_t *)ap + (k<<hsub); \ |
| 496 | a = ap + (k<<hsub); \ | 494 | const uint##depth##_t *da = main_has_alpha ? (uint##depth##_t *)dap + ((xp + k) << hsub) : NULL; \ |
| 497 | uint##depth##_t *da = main_has_alpha ? dap + ((xp+k) << hsub) : NULL; \ | 495 | uint##depth##_t *d = (uint##depth##_t *)dp + (xp + k) * dst_step; \ |
| 498 | kmax = FFMIN(-xp + dst_wp, src_wp); \ | 496 | kmax = FFMIN(-xp + dst_wp, src_wp); \ |
| 499 | \ | 497 | \ |
| 500 | if (nbits == 8 && ((vsub && j+1 < src_hp) || !vsub) && octx->blend_row[i]) { \ | 498 | if (nbits == 8 && ((vsub && j+1 < src_hp) || !vsub) && octx->blend_row[i]) { \ |
| @@ -566,11 +564,11 @@ static av_always_inline void blend_plane_##depth##_##nbits##bits(AVFilterContext · 65 unmodified lines | |||
| 566 | da += 1 << hsub; \ | 564 | da += 1 << hsub; \ |
| 567 | a += 1 << hsub; \ | 565 | a += 1 << hsub; \ |
| 568 | } \ | 566 | } \ |
| 569 | dp += dst->linesize[dst_plane] / bytes; \ | 567 | dp += dst->linesize[dst_plane]; \ |
| 570 | sp += src->linesize[i] / bytes; \ | 568 | sp += src->linesize[i]; \ |
| 571 | ap += (1 << vsub) * src->linesize[3] / bytes; \ | 569 | ap += (1 << vsub) * src->linesize[3]; \ |
| 572 | if (main_has_alpha) \ | 570 | if (main_has_alpha) \ |
| 573 | dap += (1 << vsub) * dst->linesize[3] / bytes; \ | 571 | dap += (1 << vsub) * dst->linesize[3]; \ |
| 574 | } \ | 572 | } \ |
| 575 | } | 573 | } |
| 576 | DEFINE_BLEND_PLANE(8, 8) | 574 | DEFINE_BLEND_PLANE(8, 8) |
| @@ -584,11 +582,9 @@ static inline void alpha_composite_##depth##_##nbits##bits(const AVFrame *src, c · 7 unmodified lines | |||
| 584 | int jobnr, int nb_jobs) \ | 582 | int jobnr, int nb_jobs) \ |
| 585 | { \ | 583 | { \ |
| 586 | uint##depth##_t alpha; /* the amount of overlay to blend on to main */ \ | 584 | uint##depth##_t alpha; /* the amount of overlay to blend on to main */ \ |
| 587 | uint##depth##_t *s, *sa, *d, *da; \ | ||
| 588 | int i, imax, j, jmax; \ | 585 | int i, imax, j, jmax; \ |
| 589 | int slice_start, slice_end; \ | 586 | int slice_start, slice_end; \ |
| 590 | const uint##depth##_t max = (1 << nbits) - 1; \ | 587 | const uint##depth##_t max = (1 << nbits) - 1; \ |
| 591 | int bytes = depth / 8; \ | ||
| 592 | \ | 588 | \ |
| 593 | imax = FFMIN3(-y + dst_h, FFMIN(src_h, dst_h), y + src_h); \ | 589 | imax = FFMIN3(-y + dst_h, FFMIN(src_h, dst_h), y + src_h); \ |
| 594 | i = FFMAX(-y, 0); \ | 590 | i = FFMAX(-y, 0); \ |
| @@ -596,13 +592,13 @@ static inline void alpha_composite_##depth##_##nbits##bits(const AVFrame *src, c · 1 unmodified line | |||
| 596 | slice_start = i + (imax * jobnr) / nb_jobs; \ | 592 | slice_start = i + (imax * jobnr) / nb_jobs; \ |
| 597 | slice_end = i + ((imax * (jobnr+1)) / nb_jobs); \ | 593 | slice_end = i + ((imax * (jobnr+1)) / nb_jobs); \ |
| 598 | \ | 594 | \ |
| 599 | sa = (uint##depth##_t *)(src->data[3] + (slice_start) * src->linesize[3]); \ | 595 | const uint8_t *sa = src->data[3] + (slice_start) * src->linesize[3]; \ |
| 600 | da = (uint##depth##_t *)(dst->data[3] + (y + slice_start) * dst->linesize[3]); \ | 596 | uint8_t *da = dst->data[3] + (y + slice_start) * dst->linesize[3]; \ |
| 601 | \ | 597 | \ |
| 602 | for (i = slice_start; i < slice_end; i++) { \ | 598 | for (i = slice_start; i < slice_end; i++) { \ |
| 603 | j = FFMAX(-x, 0); \ | 599 | j = FFMAX(-x, 0); \ |
| 604 | s = sa + j; \ | 600 | const uint##depth##_t *s = (const uint##depth##_t *)sa + j; \ |
| 605 | d = da + x+j; \ | 601 | uint##depth##_t *d = (uint##depth##_t *)da + x+j; \ |
| 606 | \ | 602 | \ |
| 607 | for (jmax = FFMIN(-x + dst_w, src_w); j < jmax; j++) { \ | 603 | for (jmax = FFMIN(-x + dst_w, src_w); j < jmax; j++) { \ |
| 608 | alpha = *s; \ | 604 | alpha = *s; \ |
| @@ -622,8 +618,8 @@ static inline void alpha_composite_##depth##_##nbits##bits(const AVFrame *src, c · 13 unmodified lines | |||
| 622 | d += 1; \ | 618 | d += 1; \ |
| 623 | s += 1; \ | 619 | s += 1; \ |
| 624 | } \ | 620 | } \ |
| 625 | da += dst->linesize[3] / bytes; \ | 621 | da += dst->linesize[3]; \ |
| 626 | sa += src->linesize[3] / bytes; \ | 622 | sa += src->linesize[3]; \ |
| 627 | } \ | 623 | } \ |
| 628 | } | 624 | } |
| 629 | DEFINE_ALPHA_COMPOSITE(8, 8) | 625 | DEFINE_ALPHA_COMPOSITE(8, 8) |