| @@ -1369,6 +1369,57 @@ static av_always_inline void rgbaf16ToA_endian(uint16_t *dst, const uint16_t *sr · 1368 unmodified lines |
| 1369 | } | 1369 | } |
| 1370 | } | 1370 | } |
| 1371 | | 1371 | |
| | 1372 | static av_always_inline void rgbf16ToUV_half_endian(uint16_t *dstU, uint16_t *dstV, int is_be, |
| | 1373 | const uint16_t *src, int width, |
| | 1374 | int32_t *rgb2yuv, Half2FloatTables *h2f_tbl) |
| | 1375 | { |
| | 1376 | int32_t ru = rgb2yuv[RU_IDX], gu = rgb2yuv[GU_IDX], bu = rgb2yuv[BU_IDX]; |
| | 1377 | int32_t rv = rgb2yuv[RV_IDX], gv = rgb2yuv[GV_IDX], bv = rgb2yuv[BV_IDX]; |
| | 1378 | int i; |
| | 1379 | for (i = 0; i < width; i++) { |
| | 1380 | int r = (lrintf(av_clipf(65535.0f * rdpx(src[i*6+0]), 0.0f, 65535.0f)) + |
| | 1381 | lrintf(av_clipf(65535.0f * rdpx(src[i*6+3]), 0.0f, 65535.0f))) >> 1; |
| | 1382 | int g = (lrintf(av_clipf(65535.0f * rdpx(src[i*6+1]), 0.0f, 65535.0f)) + |
| | 1383 | lrintf(av_clipf(65535.0f * rdpx(src[i*6+4]), 0.0f, 65535.0f))) >> 1; |
| | 1384 | int b = (lrintf(av_clipf(65535.0f * rdpx(src[i*6+2]), 0.0f, 65535.0f)) + |
| | 1385 | lrintf(av_clipf(65535.0f * rdpx(src[i*6+5]), 0.0f, 65535.0f))) >> 1; |
| | 1386 | |
| | 1387 | dstU[i] = (ru*r + gu*g + bu*b + (0x10001<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT; |
| | 1388 | dstV[i] = (rv*r + gv*g + bv*b + (0x10001<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT; |
| | 1389 | } |
| | 1390 | } |
| | 1391 | |
| | 1392 | static av_always_inline void rgbf16ToUV_endian(uint16_t *dstU, uint16_t *dstV, int is_be, |
| | 1393 | const uint16_t *src, int width, |
| | 1394 | int32_t *rgb2yuv, Half2FloatTables *h2f_tbl) |
| | 1395 | { |
| | 1396 | int32_t ru = rgb2yuv[RU_IDX], gu = rgb2yuv[GU_IDX], bu = rgb2yuv[BU_IDX]; |
| | 1397 | int32_t rv = rgb2yuv[RV_IDX], gv = rgb2yuv[GV_IDX], bv = rgb2yuv[BV_IDX]; |
| | 1398 | int i; |
| | 1399 | for (i = 0; i < width; i++) { |
| | 1400 | int r = lrintf(av_clipf(65535.0f * rdpx(src[i*3+0]), 0.0f, 65535.0f)); |
| | 1401 | int g = lrintf(av_clipf(65535.0f * rdpx(src[i*3+1]), 0.0f, 65535.0f)); |
| | 1402 | int b = lrintf(av_clipf(65535.0f * rdpx(src[i*3+2]), 0.0f, 65535.0f)); |
| | 1403 | |
| | 1404 | dstU[i] = (ru*r + gu*g + bu*b + (0x10001<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT; |
| | 1405 | dstV[i] = (rv*r + gv*g + bv*b + (0x10001<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT; |
| | 1406 | } |
| | 1407 | } |
| | 1408 | |
| | 1409 | static av_always_inline void rgbf16ToY_endian(uint16_t *dst, const uint16_t *src, int is_be, |
| | 1410 | int width, int32_t *rgb2yuv, Half2FloatTables *h2f_tbl) |
| | 1411 | { |
| | 1412 | int32_t ry = rgb2yuv[RY_IDX], gy = rgb2yuv[GY_IDX], by = rgb2yuv[BY_IDX]; |
| | 1413 | int i; |
| | 1414 | for (i = 0; i < width; i++) { |
| | 1415 | int r = lrintf(av_clipf(65535.0f * rdpx(src[i*3+0]), 0.0f, 65535.0f)); |
| | 1416 | int g = lrintf(av_clipf(65535.0f * rdpx(src[i*3+1]), 0.0f, 65535.0f)); |
| | 1417 | int b = lrintf(av_clipf(65535.0f * rdpx(src[i*3+2]), 0.0f, 65535.0f)); |
| | 1418 | |
| | 1419 | dst[i] = (ry*r + gy*g + by*b + (0x2001<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT; |
| | 1420 | } |
| | 1421 | } |
| | 1422 | |
| 1372 | #undef rdpx | 1423 | #undef rdpx |
| 1373 | | 1424 | |
| 1374 | #define rgbaf16_funcs_endian(endian_name, endian) \ | 1425 | #define rgbaf16_funcs_endian(endian_name, endian) \ |
| @@ -1408,7 +1459,37 @@ static void rgbaf16##endian_name##ToA_c(uint8_t *_dst, const uint8_t *_src, cons · 33 unmodified lines |
| 1408 | const uint16_t *src = (const uint16_t*)_src; \ | 1459 | const uint16_t *src = (const uint16_t*)_src; \ |
| 1409 | uint16_t *dst = (uint16_t*)_dst; \ | 1460 | uint16_t *dst = (uint16_t*)_dst; \ |
| 1410 | rgbaf16ToA_endian(dst, src, endian, width, opq); \ | 1461 | rgbaf16ToA_endian(dst, src, endian, width, opq); \ |
| 1411 | } | 1462 | } \ |
| | 1463 | static void rgbf16##endian_name##ToUV_half_c(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *unused, \ |
| | 1464 | const uint8_t *src1, const uint8_t *src2, \ |
| | 1465 | int width, uint32_t *_rgb2yuv, void *opq) \ |
| | 1466 | { \ |
| | 1467 | const uint16_t *src = (const uint16_t*)src1; \ |
| | 1468 | uint16_t *dstU = (uint16_t*)_dstU; \ |
| | 1469 | uint16_t *dstV = (uint16_t*)_dstV; \ |
| | 1470 | int32_t *rgb2yuv = (int32_t*)_rgb2yuv; \ |
| | 1471 | av_assert1(src1==src2); \ |
| | 1472 | rgbf16ToUV_half_endian(dstU, dstV, endian, src, width, rgb2yuv, opq); \ |
| | 1473 | } \ |
| | 1474 | static void rgbf16##endian_name##ToUV_c(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *unused, \ |
| | 1475 | const uint8_t *src1, const uint8_t *src2, \ |
| | 1476 | int width, uint32_t *_rgb2yuv, void *opq) \ |
| | 1477 | { \ |
| | 1478 | const uint16_t *src = (const uint16_t*)src1; \ |
| | 1479 | uint16_t *dstU = (uint16_t*)_dstU; \ |
| | 1480 | uint16_t *dstV = (uint16_t*)_dstV; \ |
| | 1481 | int32_t *rgb2yuv = (int32_t*)_rgb2yuv; \ |
| | 1482 | av_assert1(src1==src2); \ |
| | 1483 | rgbf16ToUV_endian(dstU, dstV, endian, src, width, rgb2yuv, opq); \ |
| | 1484 | } \ |
| | 1485 | static void rgbf16##endian_name##ToY_c(uint8_t *_dst, const uint8_t *_src, const uint8_t *unused0, \ |
| | 1486 | const uint8_t *unused1, int width, uint32_t *_rgb2yuv, void *opq) \ |
| | 1487 | { \ |
| | 1488 | const uint16_t *src = (const uint16_t*)_src; \ |
| | 1489 | uint16_t *dst = (uint16_t*)_dst; \ |
| | 1490 | int32_t *rgb2yuv = (int32_t*)_rgb2yuv; \ |
| | 1491 | rgbf16ToY_endian(dst, src, endian, width, rgb2yuv, opq); \ |
| | 1492 | } \ |
| 1412 | | 1493 | |
| 1413 | rgbaf16_funcs_endian(le, 0) | 1494 | rgbaf16_funcs_endian(le, 0) |
| 1414 | rgbaf16_funcs_endian(be, 1) | 1495 | rgbaf16_funcs_endian(be, 1) |
| @@ -1729,6 +1810,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c, · 314 unmodified lines |
| 1729 | case AV_PIX_FMT_RGBAF16LE: | 1810 | case AV_PIX_FMT_RGBAF16LE: |
| 1730 | *chrToYV12 = rgbaf16leToUV_half_c; | 1811 | *chrToYV12 = rgbaf16leToUV_half_c; |
| 1731 | break; | 1812 | break; |
| | 1813 | case AV_PIX_FMT_RGBF16BE: |
| | 1814 | *chrToYV12 = rgbf16beToUV_half_c; |
| | 1815 | break; |
| | 1816 | case AV_PIX_FMT_RGBF16LE: |
| | 1817 | *chrToYV12 = rgbf16leToUV_half_c; |
| | 1818 | break; |
| 1732 | } | 1819 | } |
| 1733 | } else { | 1820 | } else { |
| 1734 | switch (srcFormat) { | 1821 | switch (srcFormat) { |
| @@ -1822,6 +1909,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c, · 87 unmodified lines |
| 1822 | case AV_PIX_FMT_RGBAF16LE: | 1909 | case AV_PIX_FMT_RGBAF16LE: |
| 1823 | *chrToYV12 = rgbaf16leToUV_c; | 1910 | *chrToYV12 = rgbaf16leToUV_c; |
| 1824 | break; | 1911 | break; |
| | 1912 | case AV_PIX_FMT_RGBF16BE: |
| | 1913 | *chrToYV12 = rgbf16beToUV_c; |
| | 1914 | break; |
| | 1915 | case AV_PIX_FMT_RGBF16LE: |
| | 1916 | *chrToYV12 = rgbf16leToUV_c; |
| | 1917 | break; |
| 1825 | } | 1918 | } |
| 1826 | } | 1919 | } |
| 1827 | | 1920 | |
| @@ -2150,6 +2243,12 @@ av_cold void ff_sws_init_input_funcs(SwsContext *c, · 322 unmodified lines |
| 2150 | case AV_PIX_FMT_RGBAF16LE: | 2243 | case AV_PIX_FMT_RGBAF16LE: |
| 2151 | *lumToYV12 = rgbaf16leToY_c; | 2244 | *lumToYV12 = rgbaf16leToY_c; |
| 2152 | break; | 2245 | break; |
| | 2246 | case AV_PIX_FMT_RGBF16BE: |
| | 2247 | *lumToYV12 = rgbf16beToY_c; |
| | 2248 | break; |
| | 2249 | case AV_PIX_FMT_RGBF16LE: |
| | 2250 | *lumToYV12 = rgbf16leToY_c; |
| | 2251 | break; |
| 2153 | case AV_PIX_FMT_RGBF32LE: | 2252 | case AV_PIX_FMT_RGBF32LE: |
| 2154 | *lumToYV12 = rgbf32le_to_y_c; | 2253 | *lumToYV12 = rgbf32le_to_y_c; |
| 2155 | break; | 2254 | break; |