falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
76672 Commits
32 Branches
1.0GB
Tree: 2d8c2f1a28
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '2d8c2f1a28'
${ noResults }
FFmpeg/libswscale/swscale.c

1281 lines
49KB
Raw Blame History 

  1. /*

  2.  * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

  6.  * FFmpeg is free software; you can redistribute it and/or

  7.  * modify it under the terms of the GNU Lesser General Public

  8.  * License as published by the Free Software Foundation; either

  9.  * version 2.1 of the License, or (at your option) any later version.

  10.  *

  11.  * FFmpeg is distributed in the hope that it will be useful,

  12.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  13.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  14.  * Lesser General Public License for more details.

  15.  *

  16.  * You should have received a copy of the GNU Lesser General Public

  17.  * License along with FFmpeg; if not, write to the Free Software

  18.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  19.  */

  20. 

  21. #include <inttypes.h>

  22. #include <math.h>

  23. #include <stdio.h>

  24. #include <string.h>

  25. 

  26. #include "libavutil/avassert.h"

  27. #include "libavutil/avutil.h"

  28. #include "libavutil/bswap.h"

  29. #include "libavutil/cpu.h"

  30. #include "libavutil/imgutils.h"

  31. #include "libavutil/intreadwrite.h"

  32. #include "libavutil/mathematics.h"

  33. #include "libavutil/pixdesc.h"

  34. #include "config.h"

  35. #include "rgb2rgb.h"

  36. #include "swscale_internal.h"

  37. #include "swscale.h"

  38. 

  39. DECLARE_ALIGNED(8, const uint8_t, ff_dither_8x8_128)[9][8] = {

  40.     {  36, 68,  60, 92,  34, 66,  58, 90, },

  41.     { 100,  4, 124, 28,  98,  2, 122, 26, },

  42.     {  52, 84,  44, 76,  50, 82,  42, 74, },

  43.     { 116, 20, 108, 12, 114, 18, 106, 10, },

  44.     {  32, 64,  56, 88,  38, 70,  62, 94, },

  45.     {  96,  0, 120, 24, 102,  6, 126, 30, },

  46.     {  48, 80,  40, 72,  54, 86,  46, 78, },

  47.     { 112, 16, 104,  8, 118, 22, 110, 14, },

  48.     {  36, 68,  60, 92,  34, 66,  58, 90, },

  49. };

  50. 

  51. DECLARE_ALIGNED(8, static const uint8_t, sws_pb_64)[8] = {

  52.     64, 64, 64, 64, 64, 64, 64, 64

  53. };

  54. 

  55. #ifndef NEW_FILTER

  56. static void gamma_convert(uint8_t * src[], int width, uint16_t *gamma)

  57. {

  58.     int i;

  59.     uint16_t *src1 = (uint16_t*)src[0];

  60. 

  61.     for (i = 0; i < width; ++i) {

  62.         uint16_t r = AV_RL16(src1 + i*4 + 0);

  63.         uint16_t g = AV_RL16(src1 + i*4 + 1);

  64.         uint16_t b = AV_RL16(src1 + i*4 + 2);

  65. 

  66.         AV_WL16(src1 + i*4 + 0, gamma[r]);

  67.         AV_WL16(src1 + i*4 + 1, gamma[g]);

  68.         AV_WL16(src1 + i*4 + 2, gamma[b]);

  69.     }

  70. }

  71. #endif

  72. 

  73. static av_always_inline void fillPlane(uint8_t *plane, int stride, int width,

  74.                                        int height, int y, uint8_t val)

  75. {

  76.     int i;

  77.     uint8_t *ptr = plane + stride * y;

  78.     for (i = 0; i < height; i++) {

  79.         memset(ptr, val, width);

  80.         ptr += stride;

  81.     }

  82. }

  83. 

  84. static void hScale16To19_c(SwsContext *c, int16_t *_dst, int dstW,

  85.                            const uint8_t *_src, const int16_t *filter,

  86.                            const int32_t *filterPos, int filterSize)

  87. {

  88.     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->srcFormat);

  89.     int i;

  90.     int32_t *dst        = (int32_t *) _dst;

  91.     const uint16_t *src = (const uint16_t *) _src;

  92.     int bits            = desc->comp[0].depth - 1;

  93.     int sh              = bits - 4;

  94. 

  95.     if((isAnyRGB(c->srcFormat) || c->srcFormat==AV_PIX_FMT_PAL8) && desc->comp[0].depth<16)

  96.         sh= 9;

  97. 

  98.     for (i = 0; i < dstW; i++) {

  99.         int j;

  100.         int srcPos = filterPos[i];

  101.         int val    = 0;

  102. 

  103.         for (j = 0; j < filterSize; j++) {

  104.             val += src[srcPos + j] * filter[filterSize * i + j];

  105.         }

  106.         // filter=14 bit, input=16 bit, output=30 bit, >> 11 makes 19 bit

  107.         dst[i] = FFMIN(val >> sh, (1 << 19) - 1);

  108.     }

  109. }

  110. 

  111. static void hScale16To15_c(SwsContext *c, int16_t *dst, int dstW,

  112.                            const uint8_t *_src, const int16_t *filter,

  113.                            const int32_t *filterPos, int filterSize)

  114. {

  115.     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->srcFormat);

  116.     int i;

  117.     const uint16_t *src = (const uint16_t *) _src;

  118.     int sh              = desc->comp[0].depth - 1;

  119. 

  120.     if(sh<15)

  121.         sh= isAnyRGB(c->srcFormat) || c->srcFormat==AV_PIX_FMT_PAL8 ? 13 : (desc->comp[0].depth - 1);

  122. 

  123.     for (i = 0; i < dstW; i++) {

  124.         int j;

  125.         int srcPos = filterPos[i];

  126.         int val    = 0;

  127. 

  128.         for (j = 0; j < filterSize; j++) {

  129.             val += src[srcPos + j] * filter[filterSize * i + j];

  130.         }

  131.         // filter=14 bit, input=16 bit, output=30 bit, >> 15 makes 15 bit

  132.         dst[i] = FFMIN(val >> sh, (1 << 15) - 1);

  133.     }

  134. }

  135. 

  136. // bilinear / bicubic scaling

  137. static void hScale8To15_c(SwsContext *c, int16_t *dst, int dstW,

  138.                           const uint8_t *src, const int16_t *filter,

  139.                           const int32_t *filterPos, int filterSize)

  140. {

  141.     int i;

  142.     for (i = 0; i < dstW; i++) {

  143.         int j;

  144.         int srcPos = filterPos[i];

  145.         int val    = 0;

  146.         for (j = 0; j < filterSize; j++) {

  147.             val += ((int)src[srcPos + j]) * filter[filterSize * i + j];

  148.         }

  149.         dst[i] = FFMIN(val >> 7, (1 << 15) - 1); // the cubic equation does overflow ...

  150.     }

  151. }

  152. 

  153. static void hScale8To19_c(SwsContext *c, int16_t *_dst, int dstW,

  154.                           const uint8_t *src, const int16_t *filter,

  155.                           const int32_t *filterPos, int filterSize)

  156. {

  157.     int i;

  158.     int32_t *dst = (int32_t *) _dst;

  159.     for (i = 0; i < dstW; i++) {

  160.         int j;

  161.         int srcPos = filterPos[i];

  162.         int val    = 0;

  163.         for (j = 0; j < filterSize; j++) {

  164.             val += ((int)src[srcPos + j]) * filter[filterSize * i + j];

  165.         }

  166.         dst[i] = FFMIN(val >> 3, (1 << 19) - 1); // the cubic equation does overflow ...

  167.     }

  168. }

  169. 

  170. // FIXME all pal and rgb srcFormats could do this conversion as well

  171. // FIXME all scalers more complex than bilinear could do half of this transform

  172. static void chrRangeToJpeg_c(int16_t *dstU, int16_t *dstV, int width)

  173. {

  174.     int i;

  175.     for (i = 0; i < width; i++) {

  176.         dstU[i] = (FFMIN(dstU[i], 30775) * 4663 - 9289992) >> 12; // -264

  177.         dstV[i] = (FFMIN(dstV[i], 30775) * 4663 - 9289992) >> 12; // -264

  178.     }

  179. }

  180. 

  181. static void chrRangeFromJpeg_c(int16_t *dstU, int16_t *dstV, int width)

  182. {

  183.     int i;

  184.     for (i = 0; i < width; i++) {

  185.         dstU[i] = (dstU[i] * 1799 + 4081085) >> 11; // 1469

  186.         dstV[i] = (dstV[i] * 1799 + 4081085) >> 11; // 1469

  187.     }

  188. }

  189. 

  190. static void lumRangeToJpeg_c(int16_t *dst, int width)

  191. {

  192.     int i;

  193.     for (i = 0; i < width; i++)

  194.         dst[i] = (FFMIN(dst[i], 30189) * 19077 - 39057361) >> 14;

  195. }

  196. 

  197. static void lumRangeFromJpeg_c(int16_t *dst, int width)

  198. {

  199.     int i;

  200.     for (i = 0; i < width; i++)

  201.         dst[i] = (dst[i] * 14071 + 33561947) >> 14;

  202. }

  203. 

  204. static void chrRangeToJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)

  205. {

  206.     int i;

  207.     int32_t *dstU = (int32_t *) _dstU;

  208.     int32_t *dstV = (int32_t *) _dstV;

  209.     for (i = 0; i < width; i++) {

  210.         dstU[i] = (FFMIN(dstU[i], 30775 << 4) * 4663 - (9289992 << 4)) >> 12; // -264

  211.         dstV[i] = (FFMIN(dstV[i], 30775 << 4) * 4663 - (9289992 << 4)) >> 12; // -264

  212.     }

  213. }

  214. 

  215. static void chrRangeFromJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)

  216. {

  217.     int i;

  218.     int32_t *dstU = (int32_t *) _dstU;

  219.     int32_t *dstV = (int32_t *) _dstV;

  220.     for (i = 0; i < width; i++) {

  221.         dstU[i] = (dstU[i] * 1799 + (4081085 << 4)) >> 11; // 1469

  222.         dstV[i] = (dstV[i] * 1799 + (4081085 << 4)) >> 11; // 1469

  223.     }

  224. }

  225. 

  226. static void lumRangeToJpeg16_c(int16_t *_dst, int width)

  227. {

  228.     int i;

  229.     int32_t *dst = (int32_t *) _dst;

  230.     for (i = 0; i < width; i++) {

  231.         dst[i] = ((int)(FFMIN(dst[i], 30189 << 4) * 4769U - (39057361 << 2))) >> 12;

  232.     }

  233. }

  234. 

  235. static void lumRangeFromJpeg16_c(int16_t *_dst, int width)

  236. {

  237.     int i;

  238.     int32_t *dst = (int32_t *) _dst;

  239.     for (i = 0; i < width; i++)

  240.         dst[i] = (dst[i]*(14071/4) + (33561947<<4)/4)>>12;

  241. }

  242. 

  243. #ifndef NEW_FILTER

  244. // *** horizontal scale Y line to temp buffer

  245. static av_always_inline void hyscale(SwsContext *c, int16_t *dst, int dstWidth,

  246.                                      const uint8_t *src_in[4],

  247.                                      int srcW, int xInc,

  248.                                      const int16_t *hLumFilter,

  249.                                      const int32_t *hLumFilterPos,

  250.                                      int hLumFilterSize,

  251.                                      uint8_t *formatConvBuffer,

  252.                                      uint32_t *pal, int isAlpha)

  253. {

  254.     void (*toYV12)(uint8_t *, const uint8_t *, const uint8_t *, const uint8_t *, int, uint32_t *) =

  255.         isAlpha ? c->alpToYV12 : c->lumToYV12;

  256.     void (*convertRange)(int16_t *, int) = isAlpha ? NULL : c->lumConvertRange;

  257.     const uint8_t *src = src_in[isAlpha ? 3 : 0];

  258. 

  259.     if (toYV12) {

  260.         toYV12(formatConvBuffer, src, src_in[1], src_in[2], srcW, pal);

  261.         src = formatConvBuffer;

  262.     } else if (c->readLumPlanar && !isAlpha) {

  263.         c->readLumPlanar(formatConvBuffer, src_in, srcW, c->input_rgb2yuv_table);

  264.         src = formatConvBuffer;

  265.     } else if (c->readAlpPlanar && isAlpha) {

  266.         c->readAlpPlanar(formatConvBuffer, src_in, srcW, NULL);

  267.         src = formatConvBuffer;

  268.     }

  269. 

  270.     if (!c->hyscale_fast) {

  271.         c->hyScale(c, dst, dstWidth, src, hLumFilter,

  272.                    hLumFilterPos, hLumFilterSize);

  273.     } else { // fast bilinear upscale / crap downscale

  274.         c->hyscale_fast(c, dst, dstWidth, src, srcW, xInc);

  275.     }

  276. 

  277.     if (convertRange)

  278.         convertRange(dst, dstWidth);

  279. }

  280. 

  281. static av_always_inline void hcscale(SwsContext *c, int16_t *dst1,

  282.                                      int16_t *dst2, int dstWidth,

  283.                                      const uint8_t *src_in[4],

  284.                                      int srcW, int xInc,

  285.                                      const int16_t *hChrFilter,

  286.                                      const int32_t *hChrFilterPos,

  287.                                      int hChrFilterSize,

  288.                                      uint8_t *formatConvBuffer, uint32_t *pal)

  289. {

  290.     const uint8_t *src1 = src_in[1], *src2 = src_in[2];

  291.     if (c->chrToYV12) {

  292.         uint8_t *buf2 = formatConvBuffer +

  293.                         FFALIGN(srcW*2+78, 16);

  294.         c->chrToYV12(formatConvBuffer, buf2, src_in[0], src1, src2, srcW, pal);

  295.         src1= formatConvBuffer;

  296.         src2= buf2;

  297.     } else if (c->readChrPlanar) {

  298.         uint8_t *buf2 = formatConvBuffer +

  299.                         FFALIGN(srcW*2+78, 16);

  300.         c->readChrPlanar(formatConvBuffer, buf2, src_in, srcW, c->input_rgb2yuv_table);

  301.         src1 = formatConvBuffer;

  302.         src2 = buf2;

  303.     }

  304. 

  305.     if (!c->hcscale_fast) {

  306.         c->hcScale(c, dst1, dstWidth, src1, hChrFilter, hChrFilterPos, hChrFilterSize);

  307.         c->hcScale(c, dst2, dstWidth, src2, hChrFilter, hChrFilterPos, hChrFilterSize);

  308.     } else { // fast bilinear upscale / crap downscale

  309.         c->hcscale_fast(c, dst1, dst2, dstWidth, src1, src2, srcW, xInc);

  310.     }

  311. 

  312.     if (c->chrConvertRange)

  313.         c->chrConvertRange(dst1, dst2, dstWidth);

  314. }

  315. #endif /* NEW_FILTER */

  316. 

  317. #define DEBUG_SWSCALE_BUFFERS 0

  318. #define DEBUG_BUFFERS(...)                      \

  319.     if (DEBUG_SWSCALE_BUFFERS)                  \

  320.         av_log(c, AV_LOG_DEBUG, __VA_ARGS__)

  321. 

  322. static int swscale(SwsContext *c, const uint8_t *src[],

  323.                    int srcStride[], int srcSliceY,

  324.                    int srcSliceH, uint8_t *dst[], int dstStride[])

  325. {

  326.     /* load a few things into local vars to make the code more readable?

  327.      * and faster */

  328. #ifndef NEW_FILTER

  329.     const int srcW                   = c->srcW;

  330. #endif

  331.     const int dstW                   = c->dstW;

  332.     const int dstH                   = c->dstH;

  333. #ifndef NEW_FILTER

  334.     const int chrDstW                = c->chrDstW;

  335.     const int chrSrcW                = c->chrSrcW;

  336.     const int lumXInc                = c->lumXInc;

  337.     const int chrXInc                = c->chrXInc;

  338. #endif

  339.     const enum AVPixelFormat dstFormat = c->dstFormat;

  340.     const int flags                  = c->flags;

  341.     int32_t *vLumFilterPos           = c->vLumFilterPos;

  342.     int32_t *vChrFilterPos           = c->vChrFilterPos;

  343. #ifndef NEW_FILTER

  344.     int32_t *hLumFilterPos           = c->hLumFilterPos;

  345.     int32_t *hChrFilterPos           = c->hChrFilterPos;

  346.     int16_t *hLumFilter              = c->hLumFilter;

  347.     int16_t *hChrFilter              = c->hChrFilter;

  348.     int32_t *lumMmxFilter            = c->lumMmxFilter;

  349.     int32_t *chrMmxFilter            = c->chrMmxFilter;

  350. #endif

  351.     const int vLumFilterSize         = c->vLumFilterSize;

  352.     const int vChrFilterSize         = c->vChrFilterSize;

  353. #ifndef NEW_FILTER

  354.     const int hLumFilterSize         = c->hLumFilterSize;

  355.     const int hChrFilterSize         = c->hChrFilterSize;

  356.     int16_t **lumPixBuf              = c->lumPixBuf;

  357.     int16_t **chrUPixBuf             = c->chrUPixBuf;

  358.     int16_t **chrVPixBuf             = c->chrVPixBuf;

  359. #endif

  360.     int16_t **alpPixBuf              = c->alpPixBuf;

  361.     const int vLumBufSize            = c->vLumBufSize;

  362.     const int vChrBufSize            = c->vChrBufSize;

  363. #ifndef NEW_FILTER

  364.     uint8_t *formatConvBuffer        = c->formatConvBuffer;

  365.     uint32_t *pal                    = c->pal_yuv;

  366.     int perform_gamma = c->is_internal_gamma;

  367. #endif

  368.     yuv2planar1_fn yuv2plane1        = c->yuv2plane1;

  369.     yuv2planarX_fn yuv2planeX        = c->yuv2planeX;

  370.     yuv2interleavedX_fn yuv2nv12cX   = c->yuv2nv12cX;

  371.     yuv2packed1_fn yuv2packed1       = c->yuv2packed1;

  372.     yuv2packed2_fn yuv2packed2       = c->yuv2packed2;

  373.     yuv2packedX_fn yuv2packedX       = c->yuv2packedX;

  374.     yuv2anyX_fn yuv2anyX             = c->yuv2anyX;

  375.     const int chrSrcSliceY           =                srcSliceY >> c->chrSrcVSubSample;

  376.     const int chrSrcSliceH           = FF_CEIL_RSHIFT(srcSliceH,   c->chrSrcVSubSample);

  377.     int should_dither                = is9_OR_10BPS(c->srcFormat) ||

  378.                                        is16BPS(c->srcFormat);

  379.     int lastDstY;

  380. 

  381.     /* vars which will change and which we need to store back in the context */

  382.     int dstY         = c->dstY;

  383.     int lumBufIndex  = c->lumBufIndex;

  384.     int chrBufIndex  = c->chrBufIndex;

  385.     int lastInLumBuf = c->lastInLumBuf;

  386.     int lastInChrBuf = c->lastInChrBuf;

  387. 

  388. #ifdef NEW_FILTER

  389.     int lumStart = 0;

  390.     int lumEnd = c->descIndex[0];

  391.     int chrStart = lumEnd;

  392.     int chrEnd = c->descIndex[1];

  393.     int vStart = chrEnd;

  394.     int vEnd = c->numDesc;

  395.     SwsSlice *src_slice = &c->slice[lumStart];

  396.     SwsSlice *hout_slice = &c->slice[c->numSlice-2];

  397.     SwsSlice *vout_slice = &c->slice[c->numSlice-1];

  398.     SwsFilterDescriptor *desc = c->desc;

  399. 

  400.     int hasLumHoles = 1;

  401.     int hasChrHoles = 1;

  402. #endif

  403. 

  404. #ifndef NEW_FILTER

  405.     if (!usePal(c->srcFormat)) {

  406.         pal = c->input_rgb2yuv_table;

  407.     }

  408. #endif

  409. 

  410.     if (isPacked(c->srcFormat)) {

  411.         src[0] =

  412.         src[1] =

  413.         src[2] =

  414.         src[3] = src[0];

  415.         srcStride[0] =

  416.         srcStride[1] =

  417.         srcStride[2] =

  418.         srcStride[3] = srcStride[0];

  419.     }

  420.     srcStride[1] <<= c->vChrDrop;

  421.     srcStride[2] <<= c->vChrDrop;

  422. 

  423.     DEBUG_BUFFERS("swscale() %p[%d] %p[%d] %p[%d] %p[%d] -> %p[%d] %p[%d] %p[%d] %p[%d]\n",

  424.                   src[0], srcStride[0], src[1], srcStride[1],

  425.                   src[2], srcStride[2], src[3], srcStride[3],

  426.                   dst[0], dstStride[0], dst[1], dstStride[1],

  427.                   dst[2], dstStride[2], dst[3], dstStride[3]);

  428.     DEBUG_BUFFERS("srcSliceY: %d srcSliceH: %d dstY: %d dstH: %d\n",

  429.                   srcSliceY, srcSliceH, dstY, dstH);

  430.     DEBUG_BUFFERS("vLumFilterSize: %d vLumBufSize: %d vChrFilterSize: %d vChrBufSize: %d\n",

  431.                   vLumFilterSize, vLumBufSize, vChrFilterSize, vChrBufSize);

  432. 

  433.     if (dstStride[0]&15 || dstStride[1]&15 ||

  434.         dstStride[2]&15 || dstStride[3]&15) {

  435.         static int warnedAlready = 0; // FIXME maybe move this into the context

  436.         if (flags & SWS_PRINT_INFO && !warnedAlready) {

  437.             av_log(c, AV_LOG_WARNING,

  438.                    "Warning: dstStride is not aligned!\n"

  439.                    "         ->cannot do aligned memory accesses anymore\n");

  440.             warnedAlready = 1;

  441.         }

  442.     }

  443. 

  444.     if (   (uintptr_t)dst[0]&15 || (uintptr_t)dst[1]&15 || (uintptr_t)dst[2]&15

  445.         || (uintptr_t)src[0]&15 || (uintptr_t)src[1]&15 || (uintptr_t)src[2]&15

  446.         || dstStride[0]&15 || dstStride[1]&15 || dstStride[2]&15 || dstStride[3]&15

  447.         || srcStride[0]&15 || srcStride[1]&15 || srcStride[2]&15 || srcStride[3]&15

  448.     ) {

  449.         static int warnedAlready=0;

  450.         int cpu_flags = av_get_cpu_flags();

  451.         if (HAVE_MMXEXT && (cpu_flags & AV_CPU_FLAG_SSE2) && !warnedAlready){

  452.             av_log(c, AV_LOG_WARNING, "Warning: data is not aligned! This can lead to a speedloss\n");

  453.             warnedAlready=1;

  454.         }

  455.     }

  456. 

  457.     /* Note the user might start scaling the picture in the middle so this

  458.      * will not get executed. This is not really intended but works

  459.      * currently, so people might do it. */

  460.     if (srcSliceY == 0) {

  461.         lumBufIndex  = -1;

  462.         chrBufIndex  = -1;

  463.         dstY         = 0;

  464.         lastInLumBuf = -1;

  465.         lastInChrBuf = -1;

  466.     }

  467. 

  468.     if (!should_dither) {

  469.         c->chrDither8 = c->lumDither8 = sws_pb_64;

  470.     }

  471.     lastDstY = dstY;

  472. 

  473. #ifdef NEW_FILTER

  474.     ff_init_vscale_pfn(c, yuv2plane1, yuv2planeX, yuv2nv12cX,

  475.                    yuv2packed1, yuv2packed2, yuv2packedX, yuv2anyX, c->use_mmx_vfilter);

  476. 

  477.     ff_init_slice_from_src(src_slice, (uint8_t**)src, srcStride, c->srcW,

  478.             srcSliceY, srcSliceH, chrSrcSliceY, chrSrcSliceH, 1);

  479. 

  480.     ff_init_slice_from_src(vout_slice, (uint8_t**)dst, dstStride, c->dstW,

  481.             dstY, dstH, dstY >> c->chrDstVSubSample,

  482.             FF_CEIL_RSHIFT(dstH, c->chrDstVSubSample), 0);

  483.     if (srcSliceY == 0) {

  484.         hout_slice->plane[0].sliceY = lastInLumBuf + 1;

  485.         hout_slice->plane[1].sliceY = lastInChrBuf + 1;

  486.         hout_slice->plane[2].sliceY = lastInChrBuf + 1;

  487.         hout_slice->plane[3].sliceY = lastInLumBuf + 1;

  488. 

  489.         hout_slice->plane[0].sliceH =

  490.         hout_slice->plane[1].sliceH =

  491.         hout_slice->plane[2].sliceH =

  492.         hout_slice->plane[3].sliceH = 0;

  493.         hout_slice->width = dstW;

  494.     }

  495. #endif

  496. 

  497.     for (; dstY < dstH; dstY++) {

  498.         const int chrDstY = dstY >> c->chrDstVSubSample;

  499. #ifndef NEW_FILTER

  500.         uint8_t *dest[4]  = {

  501.             dst[0] + dstStride[0] * dstY,

  502.             dst[1] + dstStride[1] * chrDstY,

  503.             dst[2] + dstStride[2] * chrDstY,

  504.             (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? dst[3] + dstStride[3] * dstY : NULL,

  505.         };

  506. #endif

  507.         int use_mmx_vfilter= c->use_mmx_vfilter;

  508. 

  509.         // First line needed as input

  510.         const int firstLumSrcY  = FFMAX(1 - vLumFilterSize, vLumFilterPos[dstY]);

  511.         const int firstLumSrcY2 = FFMAX(1 - vLumFilterSize, vLumFilterPos[FFMIN(dstY | ((1 << c->chrDstVSubSample) - 1), dstH - 1)]);

  512.         // First line needed as input

  513.         const int firstChrSrcY  = FFMAX(1 - vChrFilterSize, vChrFilterPos[chrDstY]);

  514. 

  515.         // Last line needed as input

  516.         int lastLumSrcY  = FFMIN(c->srcH,    firstLumSrcY  + vLumFilterSize) - 1;

  517.         int lastLumSrcY2 = FFMIN(c->srcH,    firstLumSrcY2 + vLumFilterSize) - 1;

  518.         int lastChrSrcY  = FFMIN(c->chrSrcH, firstChrSrcY  + vChrFilterSize) - 1;

  519.         int enough_lines;

  520. #ifdef NEW_FILTER

  521.         int i;

  522.         int posY, cPosY, firstPosY, lastPosY, firstCPosY, lastCPosY;

  523. #endif

  524. 

  525.         // handle holes (FAST_BILINEAR & weird filters)

  526.         if (firstLumSrcY > lastInLumBuf) {

  527. #ifdef NEW_FILTER

  528.             hasLumHoles = lastInLumBuf != firstLumSrcY - 1;

  529.             if (hasLumHoles) {

  530.                 hout_slice->plane[0].sliceY = firstLumSrcY;

  531.                 hout_slice->plane[3].sliceY = firstLumSrcY;

  532.                 hout_slice->plane[0].sliceH =

  533.                 hout_slice->plane[3].sliceH = 0;

  534.             }

  535. #endif

  536.             lastInLumBuf = firstLumSrcY - 1;

  537.         }

  538.         if (firstChrSrcY > lastInChrBuf) {

  539. #ifdef NEW_FILTER

  540.             hasChrHoles = lastInChrBuf != firstChrSrcY - 1;

  541.             if (hasChrHoles) {

  542.                 hout_slice->plane[1].sliceY = firstChrSrcY;

  543.                 hout_slice->plane[2].sliceY = firstChrSrcY;

  544.                 hout_slice->plane[1].sliceH =

  545.                 hout_slice->plane[2].sliceH = 0;

  546.             }

  547. #endif

  548.             lastInChrBuf = firstChrSrcY - 1;

  549.         }

  550.         av_assert0(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1);

  551.         av_assert0(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1);

  552. 

  553.         DEBUG_BUFFERS("dstY: %d\n", dstY);

  554.         DEBUG_BUFFERS("\tfirstLumSrcY: %d lastLumSrcY: %d lastInLumBuf: %d\n",

  555.                       firstLumSrcY, lastLumSrcY, lastInLumBuf);

  556.         DEBUG_BUFFERS("\tfirstChrSrcY: %d lastChrSrcY: %d lastInChrBuf: %d\n",

  557.                       firstChrSrcY, lastChrSrcY, lastInChrBuf);

  558. 

  559.         // Do we have enough lines in this slice to output the dstY line

  560.         enough_lines = lastLumSrcY2 < srcSliceY + srcSliceH &&

  561.                        lastChrSrcY < FF_CEIL_RSHIFT(srcSliceY + srcSliceH, c->chrSrcVSubSample);

  562. 

  563.         if (!enough_lines) {

  564.             lastLumSrcY = srcSliceY + srcSliceH - 1;

  565.             lastChrSrcY = chrSrcSliceY + chrSrcSliceH - 1;

  566.             DEBUG_BUFFERS("buffering slice: lastLumSrcY %d lastChrSrcY %d\n",

  567.                           lastLumSrcY, lastChrSrcY);

  568.         }

  569. 

  570. #ifdef NEW_FILTER

  571.         posY = hout_slice->plane[0].sliceY + hout_slice->plane[0].sliceH;

  572.         if (posY <= lastLumSrcY && !hasLumHoles) {

  573.             firstPosY = FFMAX(firstLumSrcY, posY);

  574.             lastPosY = FFMIN(lastLumSrcY + MAX_LINES_AHEAD, srcSliceY + srcSliceH - 1);

  575.         } else {

  576.             firstPosY = lastInLumBuf + 1;

  577.             lastPosY = lastLumSrcY;

  578.         }

  579. 

  580.         cPosY = hout_slice->plane[1].sliceY + hout_slice->plane[1].sliceH;

  581.         if (cPosY <= lastChrSrcY && !hasChrHoles) {

  582.             firstCPosY = FFMAX(firstChrSrcY, cPosY);

  583.             lastCPosY = FFMIN(lastChrSrcY + MAX_LINES_AHEAD, FF_CEIL_RSHIFT(srcSliceY + srcSliceH, c->chrSrcVSubSample) - 1);

  584.         } else {

  585.             firstCPosY = lastInChrBuf + 1;

  586.             lastCPosY = lastChrSrcY;

  587.         }

  588. 

  589.         ff_rotate_slice(hout_slice, lastPosY, lastCPosY);

  590. 

  591.         if (posY < lastLumSrcY + 1) {

  592.             for (i = lumStart; i < lumEnd; ++i)

  593.                 desc[i].process(c, &desc[i], firstPosY, lastPosY - firstPosY + 1);

  594.         }

  595. 

  596.         lumBufIndex += lastLumSrcY - lastInLumBuf;

  597.         lastInLumBuf = lastLumSrcY;

  598. 

  599.         if (cPosY < lastChrSrcY + 1) {

  600.             for (i = chrStart; i < chrEnd; ++i)

  601.                 desc[i].process(c, &desc[i], firstCPosY, lastCPosY - firstCPosY + 1);

  602.         }

  603. 

  604.         chrBufIndex += lastChrSrcY - lastInChrBuf;

  605.         lastInChrBuf = lastChrSrcY;

  606. 

  607. #else

  608.         // Do horizontal scaling

  609.         while (lastInLumBuf < lastLumSrcY) {

  610.             const uint8_t *src1[4] = {

  611.                 src[0] + (lastInLumBuf + 1 - srcSliceY) * srcStride[0],

  612.                 src[1] + (lastInLumBuf + 1 - srcSliceY) * srcStride[1],

  613.                 src[2] + (lastInLumBuf + 1 - srcSliceY) * srcStride[2],

  614.                 src[3] + (lastInLumBuf + 1 - srcSliceY) * srcStride[3],

  615.             };

  616.             lumBufIndex++;

  617.             av_assert0(lumBufIndex < 2 * vLumBufSize);

  618.             av_assert0(lastInLumBuf + 1 - srcSliceY < srcSliceH);

  619.             av_assert0(lastInLumBuf + 1 - srcSliceY >= 0);

  620. 

  621.             if (perform_gamma)

  622.                 gamma_convert((uint8_t **)src1, srcW, c->inv_gamma);

  623. 

  624.             hyscale(c, lumPixBuf[lumBufIndex], dstW, src1, srcW, lumXInc,

  625.                     hLumFilter, hLumFilterPos, hLumFilterSize,

  626.                     formatConvBuffer, pal, 0);

  627.             if (CONFIG_SWSCALE_ALPHA && alpPixBuf)

  628.                 hyscale(c, alpPixBuf[lumBufIndex], dstW, src1, srcW,

  629.                         lumXInc, hLumFilter, hLumFilterPos, hLumFilterSize,

  630.                         formatConvBuffer, pal, 1);

  631.             lastInLumBuf++;

  632.             DEBUG_BUFFERS("\t\tlumBufIndex %d: lastInLumBuf: %d\n",

  633.                           lumBufIndex, lastInLumBuf);

  634.         }

  635.         while (lastInChrBuf < lastChrSrcY) {

  636.             const uint8_t *src1[4] = {

  637.                 src[0] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[0],

  638.                 src[1] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[1],

  639.                 src[2] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[2],

  640.                 src[3] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[3],

  641.             };

  642.             chrBufIndex++;

  643.             av_assert0(chrBufIndex < 2 * vChrBufSize);

  644.             av_assert0(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH));

  645.             av_assert0(lastInChrBuf + 1 - chrSrcSliceY >= 0);

  646.             // FIXME replace parameters through context struct (some at least)

  647. 

  648.             if (c->needs_hcscale)

  649.                 hcscale(c, chrUPixBuf[chrBufIndex], chrVPixBuf[chrBufIndex],

  650.                         chrDstW, src1, chrSrcW, chrXInc,

  651.                         hChrFilter, hChrFilterPos, hChrFilterSize,

  652.                         formatConvBuffer, pal);

  653.             lastInChrBuf++;

  654.             DEBUG_BUFFERS("\t\tchrBufIndex %d: lastInChrBuf: %d\n",

  655.                           chrBufIndex, lastInChrBuf);

  656.         }

  657. #endif

  658.         // wrap buf index around to stay inside the ring buffer

  659.         if (lumBufIndex >= vLumBufSize)

  660.             lumBufIndex -= vLumBufSize;

  661.         if (chrBufIndex >= vChrBufSize)

  662.             chrBufIndex -= vChrBufSize;

  663.         if (!enough_lines)

  664.             break;  // we can't output a dstY line so let's try with the next slice

  665. 

  666. #if HAVE_MMX_INLINE

  667.         ff_updateMMXDitherTables(c, dstY, lumBufIndex, chrBufIndex,

  668.                               lastInLumBuf, lastInChrBuf);

  669. #endif

  670.         if (should_dither) {

  671.             c->chrDither8 = ff_dither_8x8_128[chrDstY & 7];

  672.             c->lumDither8 = ff_dither_8x8_128[dstY    & 7];

  673.         }

  674.         if (dstY >= dstH - 2) {

  675.             /* hmm looks like we can't use MMX here without overwriting

  676.              * this array's tail */

  677.             ff_sws_init_output_funcs(c, &yuv2plane1, &yuv2planeX, &yuv2nv12cX,

  678.                                      &yuv2packed1, &yuv2packed2, &yuv2packedX, &yuv2anyX);

  679.             use_mmx_vfilter= 0;

  680.             ff_init_vscale_pfn(c, yuv2plane1, yuv2planeX, yuv2nv12cX,

  681.                            yuv2packed1, yuv2packed2, yuv2packedX, yuv2anyX, use_mmx_vfilter);

  682.         }

  683. 

  684.         {

  685. #ifdef NEW_FILTER

  686.             for (i = vStart; i < vEnd; ++i)

  687.                 desc[i].process(c, &desc[i], dstY, 1);

  688. #else

  689.             const int16_t **lumSrcPtr  = (const int16_t **)(void*) lumPixBuf  + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;

  690.             const int16_t **chrUSrcPtr = (const int16_t **)(void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;

  691.             const int16_t **chrVSrcPtr = (const int16_t **)(void*) chrVPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;

  692.             const int16_t **alpSrcPtr  = (CONFIG_SWSCALE_ALPHA && alpPixBuf) ?

  693.                                          (const int16_t **)(void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;

  694.             int16_t *vLumFilter = c->vLumFilter;

  695.             int16_t *vChrFilter = c->vChrFilter;

  696. 

  697.             if (isPlanarYUV(dstFormat) ||

  698.                 (isGray(dstFormat) && !isALPHA(dstFormat))) { // YV12 like

  699.                 const int chrSkipMask = (1 << c->chrDstVSubSample) - 1;

  700. 

  701.                 vLumFilter +=    dstY * vLumFilterSize;

  702.                 vChrFilter += chrDstY * vChrFilterSize;

  703. 

  704. //                 av_assert0(use_mmx_vfilter != (

  705. //                                yuv2planeX == yuv2planeX_10BE_c

  706. //                             || yuv2planeX == yuv2planeX_10LE_c

  707. //                             || yuv2planeX == yuv2planeX_9BE_c

  708. //                             || yuv2planeX == yuv2planeX_9LE_c

  709. //                             || yuv2planeX == yuv2planeX_16BE_c

  710. //                             || yuv2planeX == yuv2planeX_16LE_c

  711. //                             || yuv2planeX == yuv2planeX_8_c) || !ARCH_X86);

  712. 

  713.                 if(use_mmx_vfilter){

  714.                     vLumFilter= (int16_t *)c->lumMmxFilter;

  715.                     vChrFilter= (int16_t *)c->chrMmxFilter;

  716.                 }

  717. 

  718.                 if (vLumFilterSize == 1) {

  719.                     yuv2plane1(lumSrcPtr[0], dest[0], dstW, c->lumDither8, 0);

  720.                 } else {

  721.                     yuv2planeX(vLumFilter, vLumFilterSize,

  722.                                lumSrcPtr, dest[0],

  723.                                dstW, c->lumDither8, 0);

  724.                 }

  725. 

  726.                 if (!((dstY & chrSkipMask) || isGray(dstFormat))) {

  727.                     if (yuv2nv12cX) {

  728.                         yuv2nv12cX(c, vChrFilter,

  729.                                    vChrFilterSize, chrUSrcPtr, chrVSrcPtr,

  730.                                    dest[1], chrDstW);

  731.                     } else if (vChrFilterSize == 1) {

  732.                         yuv2plane1(chrUSrcPtr[0], dest[1], chrDstW, c->chrDither8, 0);

  733.                         yuv2plane1(chrVSrcPtr[0], dest[2], chrDstW, c->chrDither8, 3);

  734.                     } else {

  735.                         yuv2planeX(vChrFilter,

  736.                                    vChrFilterSize, chrUSrcPtr, dest[1],

  737.                                    chrDstW, c->chrDither8, 0);

  738.                         yuv2planeX(vChrFilter,

  739.                                    vChrFilterSize, chrVSrcPtr, dest[2],

  740.                                    chrDstW, c->chrDither8, use_mmx_vfilter ? (c->uv_offx2 >> 1) : 3);

  741.                     }

  742.                 }

  743. 

  744.                 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {

  745.                     if(use_mmx_vfilter){

  746.                         vLumFilter= (int16_t *)c->alpMmxFilter;

  747.                     }

  748.                     if (vLumFilterSize == 1) {

  749.                         yuv2plane1(alpSrcPtr[0], dest[3], dstW,

  750.                                    c->lumDither8, 0);

  751.                     } else {

  752.                         yuv2planeX(vLumFilter,

  753.                                    vLumFilterSize, alpSrcPtr, dest[3],

  754.                                    dstW, c->lumDither8, 0);

  755.                     }

  756.                 }

  757.             } else if (yuv2packedX) {

  758.                 av_assert1(lumSrcPtr  + vLumFilterSize - 1 < (const int16_t **)lumPixBuf  + vLumBufSize * 2);

  759.                 av_assert1(chrUSrcPtr + vChrFilterSize - 1 < (const int16_t **)chrUPixBuf + vChrBufSize * 2);

  760.                 if (c->yuv2packed1 && vLumFilterSize == 1 &&

  761.                     vChrFilterSize <= 2) { // unscaled RGB

  762.                     int chrAlpha = vChrFilterSize == 1 ? 0 : vChrFilter[2 * dstY + 1];

  763.                     yuv2packed1(c, *lumSrcPtr, chrUSrcPtr, chrVSrcPtr,

  764.                                 alpPixBuf ? *alpSrcPtr : NULL,

  765.                                 dest[0], dstW, chrAlpha, dstY);

  766.                 } else if (c->yuv2packed2 && vLumFilterSize == 2 &&

  767.                            vChrFilterSize == 2) { // bilinear upscale RGB

  768.                     int lumAlpha = vLumFilter[2 * dstY + 1];

  769.                     int chrAlpha = vChrFilter[2 * dstY + 1];

  770.                     lumMmxFilter[2] =

  771.                     lumMmxFilter[3] = vLumFilter[2 * dstY]    * 0x10001;

  772.                     chrMmxFilter[2] =

  773.                     chrMmxFilter[3] = vChrFilter[2 * chrDstY] * 0x10001;

  774.                     yuv2packed2(c, lumSrcPtr, chrUSrcPtr, chrVSrcPtr,

  775.                                 alpPixBuf ? alpSrcPtr : NULL,

  776.                                 dest[0], dstW, lumAlpha, chrAlpha, dstY);

  777.                 } else { // general RGB

  778.                     yuv2packedX(c, vLumFilter + dstY * vLumFilterSize,

  779.                                 lumSrcPtr, vLumFilterSize,

  780.                                 vChrFilter + dstY * vChrFilterSize,

  781.                                 chrUSrcPtr, chrVSrcPtr, vChrFilterSize,

  782.                                 alpSrcPtr, dest[0], dstW, dstY);

  783.                 }

  784.             } else {

  785.                 av_assert1(!yuv2packed1 && !yuv2packed2);

  786.                 yuv2anyX(c, vLumFilter + dstY * vLumFilterSize,

  787.                          lumSrcPtr, vLumFilterSize,

  788.                          vChrFilter + dstY * vChrFilterSize,

  789.                          chrUSrcPtr, chrVSrcPtr, vChrFilterSize,

  790.                          alpSrcPtr, dest, dstW, dstY);

  791.             }

  792.             if (perform_gamma)

  793.                 gamma_convert(dest, dstW, c->gamma);

  794. #endif

  795.         }

  796.     }

  797.     if (isPlanar(dstFormat) && isALPHA(dstFormat) && !alpPixBuf) {

  798.         int length = dstW;

  799.         int height = dstY - lastDstY;

  800. 

  801.         if (is16BPS(dstFormat) || isNBPS(dstFormat)) {

  802.             const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(dstFormat);

  803.             fillPlane16(dst[3], dstStride[3], length, height, lastDstY,

  804.                     1, desc->comp[3].depth,

  805.                     isBE(dstFormat));

  806.         } else

  807.             fillPlane(dst[3], dstStride[3], length, height, lastDstY, 255);

  808.     }

  809. 

  810. #if HAVE_MMXEXT_INLINE

  811.     if (av_get_cpu_flags() & AV_CPU_FLAG_MMXEXT)

  812.         __asm__ volatile ("sfence" ::: "memory");

  813. #endif

  814.     emms_c();

  815. 

  816.     /* store changed local vars back in the context */

  817.     c->dstY         = dstY;

  818.     c->lumBufIndex  = lumBufIndex;

  819.     c->chrBufIndex  = chrBufIndex;

  820.     c->lastInLumBuf = lastInLumBuf;

  821.     c->lastInChrBuf = lastInChrBuf;

  822. 

  823.     return dstY - lastDstY;

  824. }

  825. 

  826. av_cold void ff_sws_init_range_convert(SwsContext *c)

  827. {

  828.     c->lumConvertRange = NULL;

  829.     c->chrConvertRange = NULL;

  830.     if (c->srcRange != c->dstRange && !isAnyRGB(c->dstFormat)) {

  831.         if (c->dstBpc <= 14) {

  832.             if (c->srcRange) {

  833.                 c->lumConvertRange = lumRangeFromJpeg_c;

  834.                 c->chrConvertRange = chrRangeFromJpeg_c;

  835.             } else {

  836.                 c->lumConvertRange = lumRangeToJpeg_c;

  837.                 c->chrConvertRange = chrRangeToJpeg_c;

  838.             }

  839.         } else {

  840.             if (c->srcRange) {

  841.                 c->lumConvertRange = lumRangeFromJpeg16_c;

  842.                 c->chrConvertRange = chrRangeFromJpeg16_c;

  843.             } else {

  844.                 c->lumConvertRange = lumRangeToJpeg16_c;

  845.                 c->chrConvertRange = chrRangeToJpeg16_c;

  846.             }

  847.         }

  848.     }

  849. }

  850. 

  851. static av_cold void sws_init_swscale(SwsContext *c)

  852. {

  853.     enum AVPixelFormat srcFormat = c->srcFormat;

  854. 

  855.     ff_sws_init_output_funcs(c, &c->yuv2plane1, &c->yuv2planeX,

  856.                              &c->yuv2nv12cX, &c->yuv2packed1,

  857.                              &c->yuv2packed2, &c->yuv2packedX, &c->yuv2anyX);

  858. 

  859.     ff_sws_init_input_funcs(c);

  860. 

  861. 

  862.     if (c->srcBpc == 8) {

  863.         if (c->dstBpc <= 14) {

  864.             c->hyScale = c->hcScale = hScale8To15_c;

  865.             if (c->flags & SWS_FAST_BILINEAR) {

  866.                 c->hyscale_fast = ff_hyscale_fast_c;

  867.                 c->hcscale_fast = ff_hcscale_fast_c;

  868.             }

  869.         } else {

  870.             c->hyScale = c->hcScale = hScale8To19_c;

  871.         }

  872.     } else {

  873.         c->hyScale = c->hcScale = c->dstBpc > 14 ? hScale16To19_c

  874.                                                  : hScale16To15_c;

  875.     }

  876. 

  877.     ff_sws_init_range_convert(c);

  878. 

  879.     if (!(isGray(srcFormat) || isGray(c->dstFormat) ||

  880.           srcFormat == AV_PIX_FMT_MONOBLACK || srcFormat == AV_PIX_FMT_MONOWHITE))

  881.         c->needs_hcscale = 1;

  882. }

  883. 

  884. SwsFunc ff_getSwsFunc(SwsContext *c)

  885. {

  886.     sws_init_swscale(c);

  887. 

  888.     if (ARCH_PPC)

  889.         ff_sws_init_swscale_ppc(c);

  890.     if (ARCH_X86)

  891.         ff_sws_init_swscale_x86(c);

  892. 

  893.     return swscale;

  894. }

  895. 

  896. static void reset_ptr(const uint8_t *src[], enum AVPixelFormat format)

  897. {

  898.     if (!isALPHA(format))

  899.         src[3] = NULL;

  900.     if (!isPlanar(format)) {

  901.         src[3] = src[2] = NULL;

  902. 

  903.         if (!usePal(format))

  904.             src[1] = NULL;

  905.     }

  906. }

  907. 

  908. static int check_image_pointers(const uint8_t * const data[4], enum AVPixelFormat pix_fmt,

  909.                                 const int linesizes[4])

  910. {

  911.     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);

  912.     int i;

  913. 

  914.     av_assert2(desc);

  915. 

  916.     for (i = 0; i < 4; i++) {

  917.         int plane = desc->comp[i].plane;

  918.         if (!data[plane] || !linesizes[plane])

  919.             return 0;

  920.     }

  921. 

  922.     return 1;

  923. }

  924. 

  925. static void xyz12Torgb48(struct SwsContext *c, uint16_t *dst,

  926.                          const uint16_t *src, int stride, int h)

  927. {

  928.     int xp,yp;

  929.     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->srcFormat);

  930. 

  931.     for (yp=0; yp<h; yp++) {

  932.         for (xp=0; xp+2<stride; xp+=3) {

  933.             int x, y, z, r, g, b;

  934. 

  935.             if (desc->flags & AV_PIX_FMT_FLAG_BE) {

  936.                 x = AV_RB16(src + xp + 0);

  937.                 y = AV_RB16(src + xp + 1);

  938.                 z = AV_RB16(src + xp + 2);

  939.             } else {

  940.                 x = AV_RL16(src + xp + 0);

  941.                 y = AV_RL16(src + xp + 1);

  942.                 z = AV_RL16(src + xp + 2);

  943.             }

  944. 

  945.             x = c->xyzgamma[x>>4];

  946.             y = c->xyzgamma[y>>4];

  947.             z = c->xyzgamma[z>>4];

  948. 

  949.             // convert from XYZlinear to sRGBlinear

  950.             r = c->xyz2rgb_matrix[0][0] * x +

  951.                 c->xyz2rgb_matrix[0][1] * y +

  952.                 c->xyz2rgb_matrix[0][2] * z >> 12;

  953.             g = c->xyz2rgb_matrix[1][0] * x +

  954.                 c->xyz2rgb_matrix[1][1] * y +

  955.                 c->xyz2rgb_matrix[1][2] * z >> 12;

  956.             b = c->xyz2rgb_matrix[2][0] * x +

  957.                 c->xyz2rgb_matrix[2][1] * y +

  958.                 c->xyz2rgb_matrix[2][2] * z >> 12;

  959. 

  960.             // limit values to 12-bit depth

  961.             r = av_clip_uintp2(r, 12);

  962.             g = av_clip_uintp2(g, 12);

  963.             b = av_clip_uintp2(b, 12);

  964. 

  965.             // convert from sRGBlinear to RGB and scale from 12bit to 16bit

  966.             if (desc->flags & AV_PIX_FMT_FLAG_BE) {

  967.                 AV_WB16(dst + xp + 0, c->rgbgamma[r] << 4);

  968.                 AV_WB16(dst + xp + 1, c->rgbgamma[g] << 4);

  969.                 AV_WB16(dst + xp + 2, c->rgbgamma[b] << 4);

  970.             } else {

  971.                 AV_WL16(dst + xp + 0, c->rgbgamma[r] << 4);

  972.                 AV_WL16(dst + xp + 1, c->rgbgamma[g] << 4);

  973.                 AV_WL16(dst + xp + 2, c->rgbgamma[b] << 4);

  974.             }

  975.         }

  976.         src += stride;

  977.         dst += stride;

  978.     }

  979. }

  980. 

  981. static void rgb48Toxyz12(struct SwsContext *c, uint16_t *dst,

  982.                          const uint16_t *src, int stride, int h)

  983. {

  984.     int xp,yp;

  985.     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->dstFormat);

  986. 

  987.     for (yp=0; yp<h; yp++) {

  988.         for (xp=0; xp+2<stride; xp+=3) {

  989.             int x, y, z, r, g, b;

  990. 

  991.             if (desc->flags & AV_PIX_FMT_FLAG_BE) {

  992.                 r = AV_RB16(src + xp + 0);

  993.                 g = AV_RB16(src + xp + 1);

  994.                 b = AV_RB16(src + xp + 2);

  995.             } else {

  996.                 r = AV_RL16(src + xp + 0);

  997.                 g = AV_RL16(src + xp + 1);

  998.                 b = AV_RL16(src + xp + 2);

  999.             }

  1000. 

  1001.             r = c->rgbgammainv[r>>4];

  1002.             g = c->rgbgammainv[g>>4];

  1003.             b = c->rgbgammainv[b>>4];

  1004. 

  1005.             // convert from sRGBlinear to XYZlinear

  1006.             x = c->rgb2xyz_matrix[0][0] * r +

  1007.                 c->rgb2xyz_matrix[0][1] * g +

  1008.                 c->rgb2xyz_matrix[0][2] * b >> 12;

  1009.             y = c->rgb2xyz_matrix[1][0] * r +

  1010.                 c->rgb2xyz_matrix[1][1] * g +

  1011.                 c->rgb2xyz_matrix[1][2] * b >> 12;

  1012.             z = c->rgb2xyz_matrix[2][0] * r +

  1013.                 c->rgb2xyz_matrix[2][1] * g +

  1014.                 c->rgb2xyz_matrix[2][2] * b >> 12;

  1015. 

  1016.             // limit values to 12-bit depth

  1017.             x = av_clip_uintp2(x, 12);

  1018.             y = av_clip_uintp2(y, 12);

  1019.             z = av_clip_uintp2(z, 12);

  1020. 

  1021.             // convert from XYZlinear to X'Y'Z' and scale from 12bit to 16bit

  1022.             if (desc->flags & AV_PIX_FMT_FLAG_BE) {

  1023.                 AV_WB16(dst + xp + 0, c->xyzgammainv[x] << 4);

  1024.                 AV_WB16(dst + xp + 1, c->xyzgammainv[y] << 4);

  1025.                 AV_WB16(dst + xp + 2, c->xyzgammainv[z] << 4);

  1026.             } else {

  1027.                 AV_WL16(dst + xp + 0, c->xyzgammainv[x] << 4);

  1028.                 AV_WL16(dst + xp + 1, c->xyzgammainv[y] << 4);

  1029.                 AV_WL16(dst + xp + 2, c->xyzgammainv[z] << 4);

  1030.             }

  1031.         }

  1032.         src += stride;

  1033.         dst += stride;

  1034.     }

  1035. }

  1036. 

  1037. /**

  1038.  * swscale wrapper, so we don't need to export the SwsContext.

  1039.  * Assumes planar YUV to be in YUV order instead of YVU.

  1040.  */

  1041. int attribute_align_arg sws_scale(struct SwsContext *c,

  1042.                                   const uint8_t * const srcSlice[],

  1043.                                   const int srcStride[], int srcSliceY,

  1044.                                   int srcSliceH, uint8_t *const dst[],

  1045.                                   const int dstStride[])

  1046. {

  1047.     int i, ret;

  1048.     const uint8_t *src2[4];

  1049.     uint8_t *dst2[4];

  1050.     uint8_t *rgb0_tmp = NULL;

  1051. 

  1052.     if (!srcStride || !dstStride || !dst || !srcSlice) {

  1053.         av_log(c, AV_LOG_ERROR, "One of the input parameters to sws_scale() is NULL, please check the calling code\n");

  1054.         return 0;

  1055.     }

  1056. 

  1057.     if (c->gamma_flag && c->cascaded_context[0]) {

  1058. 

  1059. 

  1060.         ret = sws_scale(c->cascaded_context[0],

  1061.                     srcSlice, srcStride, srcSliceY, srcSliceH,

  1062.                     c->cascaded_tmp, c->cascaded_tmpStride);

  1063. 

  1064.         if (ret < 0)

  1065.             return ret;

  1066. 

  1067.         if (c->cascaded_context[2])

  1068.             ret = sws_scale(c->cascaded_context[1], (const uint8_t * const *)c->cascaded_tmp, c->cascaded_tmpStride, srcSliceY, srcSliceH, c->cascaded1_tmp, c->cascaded1_tmpStride);

  1069.         else

  1070.             ret = sws_scale(c->cascaded_context[1], (const uint8_t * const *)c->cascaded_tmp, c->cascaded_tmpStride, srcSliceY, srcSliceH, dst, dstStride);

  1071. 

  1072.         if (ret < 0)

  1073.             return ret;

  1074. 

  1075.         if (c->cascaded_context[2]) {

  1076.             ret = sws_scale(c->cascaded_context[2],

  1077.                         (const uint8_t * const *)c->cascaded1_tmp, c->cascaded1_tmpStride, c->cascaded_context[1]->dstY - ret, c->cascaded_context[1]->dstY,

  1078.                         dst, dstStride);

  1079.         }

  1080.         return ret;

  1081.     }

  1082. 

  1083.     if (c->cascaded_context[0] && srcSliceY == 0 && srcSliceH == c->cascaded_context[0]->srcH) {

  1084.         ret = sws_scale(c->cascaded_context[0],

  1085.                         srcSlice, srcStride, srcSliceY, srcSliceH,

  1086.                         c->cascaded_tmp, c->cascaded_tmpStride);

  1087.         if (ret < 0)

  1088.             return ret;

  1089.         ret = sws_scale(c->cascaded_context[1],

  1090.                         (const uint8_t * const * )c->cascaded_tmp, c->cascaded_tmpStride, 0, c->cascaded_context[0]->dstH,

  1091.                         dst, dstStride);

  1092.         return ret;

  1093.     }

  1094. 

  1095.     memcpy(src2, srcSlice, sizeof(src2));

  1096.     memcpy(dst2, dst, sizeof(dst2));

  1097. 

  1098.     // do not mess up sliceDir if we have a "trailing" 0-size slice

  1099.     if (srcSliceH == 0)

  1100.         return 0;

  1101. 

  1102.     if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) {

  1103.         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");

  1104.         return 0;

  1105.     }

  1106.     if (!check_image_pointers((const uint8_t* const*)dst, c->dstFormat, dstStride)) {

  1107.         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");

  1108.         return 0;

  1109.     }

  1110. 

  1111.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  1112.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  1113.         return 0;

  1114.     }

  1115.     if (c->sliceDir == 0) {

  1116.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  1117.     }

  1118. 

  1119.     if (usePal(c->srcFormat)) {

  1120.         for (i = 0; i < 256; i++) {

  1121.             int r, g, b, y, u, v, a = 0xff;

  1122.             if (c->srcFormat == AV_PIX_FMT_PAL8) {

  1123.                 uint32_t p = ((const uint32_t *)(srcSlice[1]))[i];

  1124.                 a = (p >> 24) & 0xFF;

  1125.                 r = (p >> 16) & 0xFF;

  1126.                 g = (p >>  8) & 0xFF;

  1127.                 b =  p        & 0xFF;

  1128.             } else if (c->srcFormat == AV_PIX_FMT_RGB8) {

  1129.                 r = ( i >> 5     ) * 36;

  1130.                 g = ((i >> 2) & 7) * 36;

  1131.                 b = ( i       & 3) * 85;

  1132.             } else if (c->srcFormat == AV_PIX_FMT_BGR8) {

  1133.                 b = ( i >> 6     ) * 85;

  1134.                 g = ((i >> 3) & 7) * 36;

  1135.                 r = ( i       & 7) * 36;

  1136.             } else if (c->srcFormat == AV_PIX_FMT_RGB4_BYTE) {

  1137.                 r = ( i >> 3     ) * 255;

  1138.                 g = ((i >> 1) & 3) * 85;

  1139.                 b = ( i       & 1) * 255;

  1140.             } else if (c->srcFormat == AV_PIX_FMT_GRAY8 || c->srcFormat == AV_PIX_FMT_GRAY8A) {

  1141.                 r = g = b = i;

  1142.             } else {

  1143.                 av_assert1(c->srcFormat == AV_PIX_FMT_BGR4_BYTE);

  1144.                 b = ( i >> 3     ) * 255;

  1145.                 g = ((i >> 1) & 3) * 85;

  1146.                 r = ( i       & 1) * 255;

  1147.             }

  1148. #define RGB2YUV_SHIFT 15

  1149. #define BY ( (int) (0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1150. #define BV (-(int) (0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1151. #define BU ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1152. #define GY ( (int) (0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1153. #define GV (-(int) (0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1154. #define GU (-(int) (0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1155. #define RY ( (int) (0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1156. #define RV ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1157. #define RU (-(int) (0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1158. 

  1159.             y = av_clip_uint8((RY * r + GY * g + BY * b + ( 33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1160.             u = av_clip_uint8((RU * r + GU * g + BU * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1161.             v = av_clip_uint8((RV * r + GV * g + BV * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1162.             c->pal_yuv[i]= y + (u<<8) + (v<<16) + ((unsigned)a<<24);

  1163. 

  1164.             switch (c->dstFormat) {

  1165.             case AV_PIX_FMT_BGR32:

  1166. #if !HAVE_BIGENDIAN

  1167.             case AV_PIX_FMT_RGB24:

  1168. #endif

  1169.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16) + ((unsigned)a<<24);

  1170.                 break;

  1171.             case AV_PIX_FMT_BGR32_1:

  1172. #if HAVE_BIGENDIAN

  1173.             case AV_PIX_FMT_BGR24:

  1174. #endif

  1175.                 c->pal_rgb[i]= a + (r<<8) + (g<<16) + ((unsigned)b<<24);

  1176.                 break;

  1177.             case AV_PIX_FMT_RGB32_1:

  1178. #if HAVE_BIGENDIAN

  1179.             case AV_PIX_FMT_RGB24:

  1180. #endif

  1181.                 c->pal_rgb[i]= a + (b<<8) + (g<<16) + ((unsigned)r<<24);

  1182.                 break;

  1183.             case AV_PIX_FMT_RGB32:

  1184. #if !HAVE_BIGENDIAN

  1185.             case AV_PIX_FMT_BGR24:

  1186. #endif

  1187.             default:

  1188.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16) + ((unsigned)a<<24);

  1189.             }

  1190.         }

  1191.     }

  1192. 

  1193.     if (c->src0Alpha && !c->dst0Alpha && isALPHA(c->dstFormat)) {

  1194.         uint8_t *base;

  1195.         int x,y;

  1196.         rgb0_tmp = av_malloc(FFABS(srcStride[0]) * srcSliceH + 32);

  1197.         if (!rgb0_tmp)

  1198.             return AVERROR(ENOMEM);

  1199. 

  1200.         base = srcStride[0] < 0 ? rgb0_tmp - srcStride[0] * (srcSliceH-1) : rgb0_tmp;

  1201.         for (y=0; y<srcSliceH; y++){

  1202.             memcpy(base + srcStride[0]*y, src2[0] + srcStride[0]*y, 4*c->srcW);

  1203.             for (x=c->src0Alpha-1; x<4*c->srcW; x+=4) {

  1204.                 base[ srcStride[0]*y + x] = 0xFF;

  1205.             }

  1206.         }

  1207.         src2[0] = base;

  1208.     }

  1209. 

  1210.     if (c->srcXYZ && !(c->dstXYZ && c->srcW==c->dstW && c->srcH==c->dstH)) {

  1211.         uint8_t *base;

  1212.         rgb0_tmp = av_malloc(FFABS(srcStride[0]) * srcSliceH + 32);

  1213.         if (!rgb0_tmp)

  1214.             return AVERROR(ENOMEM);

  1215. 

  1216.         base = srcStride[0] < 0 ? rgb0_tmp - srcStride[0] * (srcSliceH-1) : rgb0_tmp;

  1217. 

  1218.         xyz12Torgb48(c, (uint16_t*)base, (const uint16_t*)src2[0], srcStride[0]/2, srcSliceH);

  1219.         src2[0] = base;

  1220.     }

  1221. 

  1222.     if (!srcSliceY && (c->flags & SWS_BITEXACT) && c->dither == SWS_DITHER_ED && c->dither_error[0])

  1223.         for (i = 0; i < 4; i++)

  1224.             memset(c->dither_error[i], 0, sizeof(c->dither_error[0][0]) * (c->dstW+2));

  1225. 

  1226. 

  1227.     // copy strides, so they can safely be modified

  1228.     if (c->sliceDir == 1) {

  1229.         // slices go from top to bottom

  1230.         int srcStride2[4] = { srcStride[0], srcStride[1], srcStride[2],

  1231.                               srcStride[3] };

  1232.         int dstStride2[4] = { dstStride[0], dstStride[1], dstStride[2],

  1233.                               dstStride[3] };

  1234. 

  1235.         reset_ptr(src2, c->srcFormat);

  1236.         reset_ptr((void*)dst2, c->dstFormat);

  1237. 

  1238.         /* reset slice direction at end of frame */

  1239.         if (srcSliceY + srcSliceH == c->srcH)

  1240.             c->sliceDir = 0;

  1241. 

  1242.         ret = c->swscale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2,

  1243.                           dstStride2);

  1244.     } else {

  1245.         // slices go from bottom to top => we flip the image internally

  1246.         int srcStride2[4] = { -srcStride[0], -srcStride[1], -srcStride[2],

  1247.                               -srcStride[3] };

  1248.         int dstStride2[4] = { -dstStride[0], -dstStride[1], -dstStride[2],

  1249.                               -dstStride[3] };

  1250. 

  1251.         src2[0] += (srcSliceH - 1) * srcStride[0];

  1252.         if (!usePal(c->srcFormat))

  1253.             src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];

  1254.         src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];

  1255.         src2[3] += (srcSliceH - 1) * srcStride[3];

  1256.         dst2[0] += ( c->dstH                         - 1) * dstStride[0];

  1257.         dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1];

  1258.         dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2];

  1259.         dst2[3] += ( c->dstH                         - 1) * dstStride[3];

  1260. 

  1261.         reset_ptr(src2, c->srcFormat);

  1262.         reset_ptr((void*)dst2, c->dstFormat);

  1263. 

  1264.         /* reset slice direction at end of frame */

  1265.         if (!srcSliceY)

  1266.             c->sliceDir = 0;

  1267. 

  1268.         ret = c->swscale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH,

  1269.                           srcSliceH, dst2, dstStride2);

  1270.     }

  1271. 

  1272. 

  1273.     if (c->dstXYZ && !(c->srcXYZ && c->srcW==c->dstW && c->srcH==c->dstH)) {

  1274.         /* replace on the same data */

  1275.         rgb48Toxyz12(c, (uint16_t*)dst2[0], (const uint16_t*)dst2[0], dstStride[0]/2, ret);

  1276.     }

  1277. 

  1278.     av_free(rgb0_tmp);

  1279.     return ret;

  1280. }