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.
75375 Commits
32 Branches
293MB
Tree: fa5358f07d
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'fa5358f07d'
${ noResults }
FFmpeg/libswscale/swscale.c

1278 lines
48KB
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. // *** horizontal scale Y line to temp buffer

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

  245.                                      const uint8_t *src_in[4],

  246.                                      int srcW, int xInc,

  247.                                      const int16_t *hLumFilter,

  248.                                      const int32_t *hLumFilterPos,

  249.                                      int hLumFilterSize,

  250.                                      uint8_t *formatConvBuffer,

  251.                                      uint32_t *pal, int isAlpha)

  252. {

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

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

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

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

  257. 

  258.     if (toYV12) {

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

  260.         src = formatConvBuffer;

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

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

  263.         src = formatConvBuffer;

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

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

  266.         src = formatConvBuffer;

  267.     }

  268. 

  269.     if (!c->hyscale_fast) {

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

  271.                    hLumFilterPos, hLumFilterSize);

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

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

  274.     }

  275. 

  276.     if (convertRange)

  277.         convertRange(dst, dstWidth);

  278. }

  279. 

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

  281.                                      int16_t *dst2, int dstWidth,

  282.                                      const uint8_t *src_in[4],

  283.                                      int srcW, int xInc,

  284.                                      const int16_t *hChrFilter,

  285.                                      const int32_t *hChrFilterPos,

  286.                                      int hChrFilterSize,

  287.                                      uint8_t *formatConvBuffer, uint32_t *pal)

  288. {

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

  290.     if (c->chrToYV12) {

  291.         uint8_t *buf2 = formatConvBuffer +

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

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

  294.         src1= formatConvBuffer;

  295.         src2= buf2;

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

  297.         uint8_t *buf2 = formatConvBuffer +

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

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

  300.         src1 = formatConvBuffer;

  301.         src2 = buf2;

  302.     }

  303. 

  304.     if (!c->hcscale_fast) {

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

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

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

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

  309.     }

  310. 

  311.     if (c->chrConvertRange)

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

  313. }

  314. 

  315. #define DEBUG_SWSCALE_BUFFERS 0

  316. #define DEBUG_BUFFERS(...)                      \

  317.     if (DEBUG_SWSCALE_BUFFERS)                  \

  318.         av_log(c, AV_LOG_DEBUG, __VA_ARGS__)

  319. 

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

  321.                    int srcStride[], int srcSliceY,

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

  323. {

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

  325.      * and faster */

  326. #ifndef NEW_FILTER

  327.     const int srcW                   = c->srcW;

  328. #endif

  329.     const int dstW                   = c->dstW;

  330.     const int dstH                   = c->dstH;

  331. #ifndef NEW_FILTER

  332.     const int chrDstW                = c->chrDstW;

  333.     const int chrSrcW                = c->chrSrcW;

  334.     const int lumXInc                = c->lumXInc;

  335.     const int chrXInc                = c->chrXInc;

  336. #endif

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

  338.     const int flags                  = c->flags;

  339.     int32_t *vLumFilterPos           = c->vLumFilterPos;

  340.     int32_t *vChrFilterPos           = c->vChrFilterPos;

  341. #ifndef NEW_FILTER

  342.     int32_t *hLumFilterPos           = c->hLumFilterPos;

  343.     int32_t *hChrFilterPos           = c->hChrFilterPos;

  344.     int16_t *hLumFilter              = c->hLumFilter;

  345.     int16_t *hChrFilter              = c->hChrFilter;

  346.     int32_t *lumMmxFilter            = c->lumMmxFilter;

  347.     int32_t *chrMmxFilter            = c->chrMmxFilter;

  348. #endif

  349.     const int vLumFilterSize         = c->vLumFilterSize;

  350.     const int vChrFilterSize         = c->vChrFilterSize;

  351. #ifndef NEW_FILTER

  352.     const int hLumFilterSize         = c->hLumFilterSize;

  353.     const int hChrFilterSize         = c->hChrFilterSize;

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

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

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

  357. #endif

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

  359.     const int vLumBufSize            = c->vLumBufSize;

  360.     const int vChrBufSize            = c->vChrBufSize;

  361. #ifndef NEW_FILTER

  362.     uint8_t *formatConvBuffer        = c->formatConvBuffer;

  363.     uint32_t *pal                    = c->pal_yuv;

  364.     int perform_gamma = c->is_internal_gamma;

  365. #endif

  366.     yuv2planar1_fn yuv2plane1        = c->yuv2plane1;

  367.     yuv2planarX_fn yuv2planeX        = c->yuv2planeX;

  368.     yuv2interleavedX_fn yuv2nv12cX   = c->yuv2nv12cX;

  369.     yuv2packed1_fn yuv2packed1       = c->yuv2packed1;

  370.     yuv2packed2_fn yuv2packed2       = c->yuv2packed2;

  371.     yuv2packedX_fn yuv2packedX       = c->yuv2packedX;

  372.     yuv2anyX_fn yuv2anyX             = c->yuv2anyX;

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

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

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

  376.                                        is16BPS(c->srcFormat);

  377.     int lastDstY;

  378. 

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

  380.     int dstY         = c->dstY;

  381.     int lumBufIndex  = c->lumBufIndex;

  382.     int chrBufIndex  = c->chrBufIndex;

  383.     int lastInLumBuf = c->lastInLumBuf;

  384.     int lastInChrBuf = c->lastInChrBuf;

  385. 

  386. #ifdef NEW_FILTER

  387.     int lumStart = 0;

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

  389.     int chrStart = lumEnd;

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

  391.     int vStart = chrEnd;

  392.     int vEnd = c->numDesc;

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

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

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

  396.     SwsFilterDescriptor *desc = c->desc;

  397. 

  398.     int hasLumHoles = 1;

  399.     int hasChrHoles = 1;

  400. #endif

  401. 

  402. #ifndef NEW_FILTER

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

  404.         pal = c->input_rgb2yuv_table;

  405.     }

  406. #endif

  407. 

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

  409.         src[0] =

  410.         src[1] =

  411.         src[2] =

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

  413.         srcStride[0] =

  414.         srcStride[1] =

  415.         srcStride[2] =

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

  417.     }

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

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

  420. 

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

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

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

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

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

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

  427.                   srcSliceY, srcSliceH, dstY, dstH);

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

  429.                   vLumFilterSize, vLumBufSize, vChrFilterSize, vChrBufSize);

  430. 

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

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

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

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

  435.             av_log(c, AV_LOG_WARNING,

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

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

  438.             warnedAlready = 1;

  439.         }

  440.     }

  441. 

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

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

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

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

  446.     ) {

  447.         static int warnedAlready=0;

  448.         int cpu_flags = av_get_cpu_flags();

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

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

  451.             warnedAlready=1;

  452.         }

  453.     }

  454. 

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

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

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

  458.     if (srcSliceY == 0) {

  459.         lumBufIndex  = -1;

  460.         chrBufIndex  = -1;

  461.         dstY         = 0;

  462.         lastInLumBuf = -1;

  463.         lastInChrBuf = -1;

  464.     }

  465. 

  466.     if (!should_dither) {

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

  468.     }

  469.     lastDstY = dstY;

  470. 

  471. #ifdef NEW_FILTER

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

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

  474. 

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

  476.             srcSliceY, srcSliceH, chrSrcSliceY, chrSrcSliceH);

  477. 

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

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

  480.             FF_CEIL_RSHIFT(dstH, c->chrDstVSubSample));

  481. 

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

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

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

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

  486. 

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

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

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

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

  491.     hout_slice->width = dstW;

  492. #endif

  493. 

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

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

  496. #ifndef NEW_FILTER

  497.         uint8_t *dest[4]  = {

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

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

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

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

  502.         };

  503. #endif

  504.         int use_mmx_vfilter= c->use_mmx_vfilter;

  505. 

  506.         // First line needed as input

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

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

  509.         // First line needed as input

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

  511. 

  512.         // Last line needed as input

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

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

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

  516.         int enough_lines;

  517. #ifdef NEW_FILTER

  518.         int i;

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

  520. #endif

  521. 

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

  523.         if (firstLumSrcY > lastInLumBuf) {

  524. #ifdef NEW_FILTER

  525.             hasLumHoles = lastInLumBuf != firstLumSrcY - 1;

  526.             if (hasLumHoles) {

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

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

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

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

  531.             }

  532. #endif

  533.             lastInLumBuf = firstLumSrcY - 1;

  534.         }

  535.         if (firstChrSrcY > lastInChrBuf) {

  536. #ifdef NEW_FILTER

  537.             hasChrHoles = lastInChrBuf != firstChrSrcY - 1;

  538.             if (hasChrHoles) {

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

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

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

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

  543.             }

  544. #endif

  545.             lastInChrBuf = firstChrSrcY - 1;

  546.         }

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

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

  549. 

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

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

  552.                       firstLumSrcY, lastLumSrcY, lastInLumBuf);

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

  554.                       firstChrSrcY, lastChrSrcY, lastInChrBuf);

  555. 

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

  557.         enough_lines = lastLumSrcY2 < srcSliceY + srcSliceH &&

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

  559. 

  560.         if (!enough_lines) {

  561.             lastLumSrcY = srcSliceY + srcSliceH - 1;

  562.             lastChrSrcY = chrSrcSliceY + chrSrcSliceH - 1;

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

  564.                           lastLumSrcY, lastChrSrcY);

  565.         }

  566. 

  567. #ifdef NEW_FILTER

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

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

  570.             firstPosY = FFMAX(firstLumSrcY, posY);

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

  572.         } else {

  573.             firstPosY = lastInLumBuf + 1;

  574.             lastPosY = lastLumSrcY;

  575.         }

  576. 

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

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

  579.             firstCPosY = FFMAX(firstChrSrcY, cPosY);

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

  581.         } else {

  582.             firstCPosY = lastInChrBuf + 1;

  583.             lastCPosY = lastChrSrcY;

  584.         }

  585. 

  586.         ff_rotate_slice(hout_slice, lastPosY, lastCPosY);

  587. 

  588.         if (posY < lastLumSrcY + 1) {

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

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

  591.         }

  592. 

  593.         lumBufIndex += lastLumSrcY - lastInLumBuf;

  594.         lastInLumBuf = lastLumSrcY;

  595. 

  596.         if (cPosY < lastChrSrcY + 1) {

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

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

  599.         }

  600. 

  601.         chrBufIndex += lastChrSrcY - lastInChrBuf;

  602.         lastInChrBuf = lastChrSrcY;

  603. 

  604. #else

  605.         // Do horizontal scaling

  606.         while (lastInLumBuf < lastLumSrcY) {

  607.             const uint8_t *src1[4] = {

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

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

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

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

  612.             };

  613.             lumBufIndex++;

  614.             av_assert0(lumBufIndex < 2 * vLumBufSize);

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

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

  617. 

  618.             if (perform_gamma)

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

  620. 

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

  622.                     hLumFilter, hLumFilterPos, hLumFilterSize,

  623.                     formatConvBuffer, pal, 0);

  624.             if (CONFIG_SWSCALE_ALPHA && alpPixBuf)

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

  626.                         lumXInc, hLumFilter, hLumFilterPos, hLumFilterSize,

  627.                         formatConvBuffer, pal, 1);

  628.             lastInLumBuf++;

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

  630.                           lumBufIndex, lastInLumBuf);

  631.         }

  632.         while (lastInChrBuf < lastChrSrcY) {

  633.             const uint8_t *src1[4] = {

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

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

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

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

  638.             };

  639.             chrBufIndex++;

  640.             av_assert0(chrBufIndex < 2 * vChrBufSize);

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

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

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

  644. 

  645.             if (c->needs_hcscale)

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

  647.                         chrDstW, src1, chrSrcW, chrXInc,

  648.                         hChrFilter, hChrFilterPos, hChrFilterSize,

  649.                         formatConvBuffer, pal);

  650.             lastInChrBuf++;

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

  652.                           chrBufIndex, lastInChrBuf);

  653.         }

  654. #endif

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

  656.         if (lumBufIndex >= vLumBufSize)

  657.             lumBufIndex -= vLumBufSize;

  658.         if (chrBufIndex >= vChrBufSize)

  659.             chrBufIndex -= vChrBufSize;

  660.         if (!enough_lines)

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

  662. 

  663. #if HAVE_MMX_INLINE

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

  665.                               lastInLumBuf, lastInChrBuf);

  666. #endif

  667.         if (should_dither) {

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

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

  670.         }

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

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

  673.              * this array's tail */

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

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

  676.             use_mmx_vfilter= 0;

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

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

  679.         }

  680. 

  681.         {

  682. #ifdef NEW_FILTER

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

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

  685. #else

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

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

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

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

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

  691.             int16_t *vLumFilter = c->vLumFilter;

  692.             int16_t *vChrFilter = c->vChrFilter;

  693. 

  694.             if (isPlanarYUV(dstFormat) ||

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

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

  697. 

  698.                 vLumFilter +=    dstY * vLumFilterSize;

  699.                 vChrFilter += chrDstY * vChrFilterSize;

  700. 

  701. //                 av_assert0(use_mmx_vfilter != (

  702. //                                yuv2planeX == yuv2planeX_10BE_c

  703. //                             || yuv2planeX == yuv2planeX_10LE_c

  704. //                             || yuv2planeX == yuv2planeX_9BE_c

  705. //                             || yuv2planeX == yuv2planeX_9LE_c

  706. //                             || yuv2planeX == yuv2planeX_16BE_c

  707. //                             || yuv2planeX == yuv2planeX_16LE_c

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

  709. 

  710.                 if(use_mmx_vfilter){

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

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

  713.                 }

  714. 

  715.                 if (vLumFilterSize == 1) {

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

  717.                 } else {

  718.                     yuv2planeX(vLumFilter, vLumFilterSize,

  719.                                lumSrcPtr, dest[0],

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

  721.                 }

  722. 

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

  724.                     if (yuv2nv12cX) {

  725.                         yuv2nv12cX(c, vChrFilter,

  726.                                    vChrFilterSize, chrUSrcPtr, chrVSrcPtr,

  727.                                    dest[1], chrDstW);

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

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

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

  731.                     } else {

  732.                         yuv2planeX(vChrFilter,

  733.                                    vChrFilterSize, chrUSrcPtr, dest[1],

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

  735.                         yuv2planeX(vChrFilter,

  736.                                    vChrFilterSize, chrVSrcPtr, dest[2],

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

  738.                     }

  739.                 }

  740. 

  741.                 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {

  742.                     if(use_mmx_vfilter){

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

  744.                     }

  745.                     if (vLumFilterSize == 1) {

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

  747.                                    c->lumDither8, 0);

  748.                     } else {

  749.                         yuv2planeX(vLumFilter,

  750.                                    vLumFilterSize, alpSrcPtr, dest[3],

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

  752.                     }

  753.                 }

  754.             } else if (yuv2packedX) {

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

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

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

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

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

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

  761.                                 alpPixBuf ? *alpSrcPtr : NULL,

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

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

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

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

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

  767.                     lumMmxFilter[2] =

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

  769.                     chrMmxFilter[2] =

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

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

  772.                                 alpPixBuf ? alpSrcPtr : NULL,

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

  774.                 } else { // general RGB

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

  776.                                 lumSrcPtr, vLumFilterSize,

  777.                                 vChrFilter + dstY * vChrFilterSize,

  778.                                 chrUSrcPtr, chrVSrcPtr, vChrFilterSize,

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

  780.                 }

  781.             } else {

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

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

  784.                          lumSrcPtr, vLumFilterSize,

  785.                          vChrFilter + dstY * vChrFilterSize,

  786.                          chrUSrcPtr, chrVSrcPtr, vChrFilterSize,

  787.                          alpSrcPtr, dest, dstW, dstY);

  788.             }

  789.             if (perform_gamma)

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

  791. #endif

  792.         }

  793.     }

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

  795.         int length = dstW;

  796.         int height = dstY - lastDstY;

  797. 

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

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

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

  801.                     1, desc->comp[3].depth,

  802.                     isBE(dstFormat));

  803.         } else

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

  805.     }

  806. 

  807. #if HAVE_MMXEXT_INLINE

  808.     if (av_get_cpu_flags() & AV_CPU_FLAG_MMXEXT)

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

  810. #endif

  811.     emms_c();

  812. 

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

  814.     c->dstY         = dstY;

  815.     c->lumBufIndex  = lumBufIndex;

  816.     c->chrBufIndex  = chrBufIndex;

  817.     c->lastInLumBuf = lastInLumBuf;

  818.     c->lastInChrBuf = lastInChrBuf;

  819. 

  820.     return dstY - lastDstY;

  821. }

  822. 

  823. av_cold void ff_sws_init_range_convert(SwsContext *c)

  824. {

  825.     c->lumConvertRange = NULL;

  826.     c->chrConvertRange = NULL;

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

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

  829.             if (c->srcRange) {

  830.                 c->lumConvertRange = lumRangeFromJpeg_c;

  831.                 c->chrConvertRange = chrRangeFromJpeg_c;

  832.             } else {

  833.                 c->lumConvertRange = lumRangeToJpeg_c;

  834.                 c->chrConvertRange = chrRangeToJpeg_c;

  835.             }

  836.         } else {

  837.             if (c->srcRange) {

  838.                 c->lumConvertRange = lumRangeFromJpeg16_c;

  839.                 c->chrConvertRange = chrRangeFromJpeg16_c;

  840.             } else {

  841.                 c->lumConvertRange = lumRangeToJpeg16_c;

  842.                 c->chrConvertRange = chrRangeToJpeg16_c;

  843.             }

  844.         }

  845.     }

  846. }

  847. 

  848. static av_cold void sws_init_swscale(SwsContext *c)

  849. {

  850.     enum AVPixelFormat srcFormat = c->srcFormat;

  851. 

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

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

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

  855. 

  856.     ff_sws_init_input_funcs(c);

  857. 

  858. 

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

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

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

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

  863.                 c->hyscale_fast = ff_hyscale_fast_c;

  864.                 c->hcscale_fast = ff_hcscale_fast_c;

  865.             }

  866.         } else {

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

  868.         }

  869.     } else {

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

  871.                                                  : hScale16To15_c;

  872.     }

  873. 

  874.     ff_sws_init_range_convert(c);

  875. 

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

  877.           srcFormat == AV_PIX_FMT_MONOBLACK || srcFormat == AV_PIX_FMT_MONOWHITE))

  878.         c->needs_hcscale = 1;

  879. }

  880. 

  881. SwsFunc ff_getSwsFunc(SwsContext *c)

  882. {

  883.     sws_init_swscale(c);

  884. 

  885.     if (ARCH_PPC)

  886.         ff_sws_init_swscale_ppc(c);

  887.     if (ARCH_X86)

  888.         ff_sws_init_swscale_x86(c);

  889. 

  890.     return swscale;

  891. }

  892. 

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

  894. {

  895.     if (!isALPHA(format))

  896.         src[3] = NULL;

  897.     if (!isPlanar(format)) {

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

  899. 

  900.         if (!usePal(format))

  901.             src[1] = NULL;

  902.     }

  903. }

  904. 

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

  906.                                 const int linesizes[4])

  907. {

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

  909.     int i;

  910. 

  911.     av_assert2(desc);

  912. 

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

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

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

  916.             return 0;

  917.     }

  918. 

  919.     return 1;

  920. }

  921. 

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

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

  924. {

  925.     int xp,yp;

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

  927. 

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

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

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

  931. 

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

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

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

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

  936.             } else {

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

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

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

  940.             }

  941. 

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

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

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

  945. 

  946.             // convert from XYZlinear to sRGBlinear

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

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

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

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

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

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

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

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

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

  956. 

  957.             // limit values to 12-bit depth

  958.             r = av_clip_uintp2(r, 12);

  959.             g = av_clip_uintp2(g, 12);

  960.             b = av_clip_uintp2(b, 12);

  961. 

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

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

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

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

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

  967.             } else {

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

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

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

  971.             }

  972.         }

  973.         src += stride;

  974.         dst += stride;

  975.     }

  976. }

  977. 

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

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

  980. {

  981.     int xp,yp;

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

  983. 

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

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

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

  987. 

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

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

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

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

  992.             } else {

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

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

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

  996.             }

  997. 

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

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

  1000.             b = c->rgbgammainv[b>>4];

  1001. 

  1002.             // convert from sRGBlinear to XYZlinear

  1003.             x = c->rgb2xyz_matrix[0][0] * r +

  1004.                 c->rgb2xyz_matrix[0][1] * g +

  1005.                 c->rgb2xyz_matrix[0][2] * b >> 12;

  1006.             y = c->rgb2xyz_matrix[1][0] * r +

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

  1008.                 c->rgb2xyz_matrix[1][2] * b >> 12;

  1009.             z = c->rgb2xyz_matrix[2][0] * r +

  1010.                 c->rgb2xyz_matrix[2][1] * g +

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

  1012. 

  1013.             // limit values to 12-bit depth

  1014.             x = av_clip_uintp2(x, 12);

  1015.             y = av_clip_uintp2(y, 12);

  1016.             z = av_clip_uintp2(z, 12);

  1017. 

  1018.             // convert from XYZlinear to X'Y'Z' and scale from 12bit to 16bit

  1019.             if (desc->flags & AV_PIX_FMT_FLAG_BE) {

  1020.                 AV_WB16(dst + xp + 0, c->xyzgammainv[x] << 4);

  1021.                 AV_WB16(dst + xp + 1, c->xyzgammainv[y] << 4);

  1022.                 AV_WB16(dst + xp + 2, c->xyzgammainv[z] << 4);

  1023.             } else {

  1024.                 AV_WL16(dst + xp + 0, c->xyzgammainv[x] << 4);

  1025.                 AV_WL16(dst + xp + 1, c->xyzgammainv[y] << 4);

  1026.                 AV_WL16(dst + xp + 2, c->xyzgammainv[z] << 4);

  1027.             }

  1028.         }

  1029.         src += stride;

  1030.         dst += stride;

  1031.     }

  1032. }

  1033. 

  1034. /**

  1035.  * swscale wrapper, so we don't need to export the SwsContext.

  1036.  * Assumes planar YUV to be in YUV order instead of YVU.

  1037.  */

  1038. int attribute_align_arg sws_scale(struct SwsContext *c,

  1039.                                   const uint8_t * const srcSlice[],

  1040.                                   const int srcStride[], int srcSliceY,

  1041.                                   int srcSliceH, uint8_t *const dst[],

  1042.                                   const int dstStride[])

  1043. {

  1044.     int i, ret;

  1045.     const uint8_t *src2[4];

  1046.     uint8_t *dst2[4];

  1047.     uint8_t *rgb0_tmp = NULL;

  1048. 

  1049.     if (!srcStride || !dstStride || !dst || !srcSlice) {

  1050.         av_log(c, AV_LOG_ERROR, "One of the input parameters to sws_scale() is NULL, please check the calling code\n");

  1051.         return 0;

  1052.     }

  1053. 

  1054.     if (c->gamma_flag && c->cascaded_context[0]) {

  1055. 

  1056. 

  1057.         ret = sws_scale(c->cascaded_context[0],

  1058.                     srcSlice, srcStride, srcSliceY, srcSliceH,

  1059.                     c->cascaded_tmp, c->cascaded_tmpStride);

  1060. 

  1061.         if (ret < 0)

  1062.             return ret;

  1063. 

  1064.         if (c->cascaded_context[2])

  1065.             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);

  1066.         else

  1067.             ret = sws_scale(c->cascaded_context[1], (const uint8_t * const *)c->cascaded_tmp, c->cascaded_tmpStride, srcSliceY, srcSliceH, dst, dstStride);

  1068. 

  1069.         if (ret < 0)

  1070.             return ret;

  1071. 

  1072.         if (c->cascaded_context[2]) {

  1073.             ret = sws_scale(c->cascaded_context[2],

  1074.                         (const uint8_t * const *)c->cascaded1_tmp, c->cascaded1_tmpStride, c->cascaded_context[1]->dstY - ret, c->cascaded_context[1]->dstY,

  1075.                         dst, dstStride);

  1076.         }

  1077.         return ret;

  1078.     }

  1079. 

  1080.     if (c->cascaded_context[0] && srcSliceY == 0 && srcSliceH == c->cascaded_context[0]->srcH) {

  1081.         ret = sws_scale(c->cascaded_context[0],

  1082.                         srcSlice, srcStride, srcSliceY, srcSliceH,

  1083.                         c->cascaded_tmp, c->cascaded_tmpStride);

  1084.         if (ret < 0)

  1085.             return ret;

  1086.         ret = sws_scale(c->cascaded_context[1],

  1087.                         (const uint8_t * const * )c->cascaded_tmp, c->cascaded_tmpStride, 0, c->cascaded_context[0]->dstH,

  1088.                         dst, dstStride);

  1089.         return ret;

  1090.     }

  1091. 

  1092.     memcpy(src2, srcSlice, sizeof(src2));

  1093.     memcpy(dst2, dst, sizeof(dst2));

  1094. 

  1095.     // do not mess up sliceDir if we have a "trailing" 0-size slice

  1096.     if (srcSliceH == 0)

  1097.         return 0;

  1098. 

  1099.     if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) {

  1100.         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");

  1101.         return 0;

  1102.     }

  1103.     if (!check_image_pointers((const uint8_t* const*)dst, c->dstFormat, dstStride)) {

  1104.         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");

  1105.         return 0;

  1106.     }

  1107. 

  1108.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  1109.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  1110.         return 0;

  1111.     }

  1112.     if (c->sliceDir == 0) {

  1113.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  1114.     }

  1115. 

  1116.     if (usePal(c->srcFormat)) {

  1117.         for (i = 0; i < 256; i++) {

  1118.             int r, g, b, y, u, v, a = 0xff;

  1119.             if (c->srcFormat == AV_PIX_FMT_PAL8) {

  1120.                 uint32_t p = ((const uint32_t *)(srcSlice[1]))[i];

  1121.                 a = (p >> 24) & 0xFF;

  1122.                 r = (p >> 16) & 0xFF;

  1123.                 g = (p >>  8) & 0xFF;

  1124.                 b =  p        & 0xFF;

  1125.             } else if (c->srcFormat == AV_PIX_FMT_RGB8) {

  1126.                 r = ( i >> 5     ) * 36;

  1127.                 g = ((i >> 2) & 7) * 36;

  1128.                 b = ( i       & 3) * 85;

  1129.             } else if (c->srcFormat == AV_PIX_FMT_BGR8) {

  1130.                 b = ( i >> 6     ) * 85;

  1131.                 g = ((i >> 3) & 7) * 36;

  1132.                 r = ( i       & 7) * 36;

  1133.             } else if (c->srcFormat == AV_PIX_FMT_RGB4_BYTE) {

  1134.                 r = ( i >> 3     ) * 255;

  1135.                 g = ((i >> 1) & 3) * 85;

  1136.                 b = ( i       & 1) * 255;

  1137.             } else if (c->srcFormat == AV_PIX_FMT_GRAY8 || c->srcFormat == AV_PIX_FMT_GRAY8A) {

  1138.                 r = g = b = i;

  1139.             } else {

  1140.                 av_assert1(c->srcFormat == AV_PIX_FMT_BGR4_BYTE);

  1141.                 b = ( i >> 3     ) * 255;

  1142.                 g = ((i >> 1) & 3) * 85;

  1143.                 r = ( i       & 1) * 255;

  1144.             }

  1145. #define RGB2YUV_SHIFT 15

  1146. #define BY ( (int) (0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1147. #define BV (-(int) (0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1148. #define BU ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1149. #define GY ( (int) (0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1150. #define GV (-(int) (0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1151. #define GU (-(int) (0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1152. #define RY ( (int) (0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1153. #define RV ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1154. #define RU (-(int) (0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  1155. 

  1156.             y = av_clip_uint8((RY * r + GY * g + BY * b + ( 33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1157.             u = av_clip_uint8((RU * r + GU * g + BU * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1158.             v = av_clip_uint8((RV * r + GV * g + BV * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1159.             c->pal_yuv[i]= y + (u<<8) + (v<<16) + ((unsigned)a<<24);

  1160. 

  1161.             switch (c->dstFormat) {

  1162.             case AV_PIX_FMT_BGR32:

  1163. #if !HAVE_BIGENDIAN

  1164.             case AV_PIX_FMT_RGB24:

  1165. #endif

  1166.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16) + ((unsigned)a<<24);

  1167.                 break;

  1168.             case AV_PIX_FMT_BGR32_1:

  1169. #if HAVE_BIGENDIAN

  1170.             case AV_PIX_FMT_BGR24:

  1171. #endif

  1172.                 c->pal_rgb[i]= a + (r<<8) + (g<<16) + ((unsigned)b<<24);

  1173.                 break;

  1174.             case AV_PIX_FMT_RGB32_1:

  1175. #if HAVE_BIGENDIAN

  1176.             case AV_PIX_FMT_RGB24:

  1177. #endif

  1178.                 c->pal_rgb[i]= a + (b<<8) + (g<<16) + ((unsigned)r<<24);

  1179.                 break;

  1180.             case AV_PIX_FMT_RGB32:

  1181. #if !HAVE_BIGENDIAN

  1182.             case AV_PIX_FMT_BGR24:

  1183. #endif

  1184.             default:

  1185.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16) + ((unsigned)a<<24);

  1186.             }

  1187.         }

  1188.     }

  1189. 

  1190.     if (c->src0Alpha && !c->dst0Alpha && isALPHA(c->dstFormat)) {

  1191.         uint8_t *base;

  1192.         int x,y;

  1193.         rgb0_tmp = av_malloc(FFABS(srcStride[0]) * srcSliceH + 32);

  1194.         if (!rgb0_tmp)

  1195.             return AVERROR(ENOMEM);

  1196. 

  1197.         base = srcStride[0] < 0 ? rgb0_tmp - srcStride[0] * (srcSliceH-1) : rgb0_tmp;

  1198.         for (y=0; y<srcSliceH; y++){

  1199.             memcpy(base + srcStride[0]*y, src2[0] + srcStride[0]*y, 4*c->srcW);

  1200.             for (x=c->src0Alpha-1; x<4*c->srcW; x+=4) {

  1201.                 base[ srcStride[0]*y + x] = 0xFF;

  1202.             }

  1203.         }

  1204.         src2[0] = base;

  1205.     }

  1206. 

  1207.     if (c->srcXYZ && !(c->dstXYZ && c->srcW==c->dstW && c->srcH==c->dstH)) {

  1208.         uint8_t *base;

  1209.         rgb0_tmp = av_malloc(FFABS(srcStride[0]) * srcSliceH + 32);

  1210.         if (!rgb0_tmp)

  1211.             return AVERROR(ENOMEM);

  1212. 

  1213.         base = srcStride[0] < 0 ? rgb0_tmp - srcStride[0] * (srcSliceH-1) : rgb0_tmp;

  1214. 

  1215.         xyz12Torgb48(c, (uint16_t*)base, (const uint16_t*)src2[0], srcStride[0]/2, srcSliceH);

  1216.         src2[0] = base;

  1217.     }

  1218. 

  1219.     if (!srcSliceY && (c->flags & SWS_BITEXACT) && c->dither == SWS_DITHER_ED && c->dither_error[0])

  1220.         for (i = 0; i < 4; i++)

  1221.             memset(c->dither_error[i], 0, sizeof(c->dither_error[0][0]) * (c->dstW+2));

  1222. 

  1223. 

  1224.     // copy strides, so they can safely be modified

  1225.     if (c->sliceDir == 1) {

  1226.         // slices go from top to bottom

  1227.         int srcStride2[4] = { srcStride[0], srcStride[1], srcStride[2],

  1228.                               srcStride[3] };

  1229.         int dstStride2[4] = { dstStride[0], dstStride[1], dstStride[2],

  1230.                               dstStride[3] };

  1231. 

  1232.         reset_ptr(src2, c->srcFormat);

  1233.         reset_ptr((void*)dst2, c->dstFormat);

  1234. 

  1235.         /* reset slice direction at end of frame */

  1236.         if (srcSliceY + srcSliceH == c->srcH)

  1237.             c->sliceDir = 0;

  1238. 

  1239.         ret = c->swscale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2,

  1240.                           dstStride2);

  1241.     } else {

  1242.         // slices go from bottom to top => we flip the image internally

  1243.         int srcStride2[4] = { -srcStride[0], -srcStride[1], -srcStride[2],

  1244.                               -srcStride[3] };

  1245.         int dstStride2[4] = { -dstStride[0], -dstStride[1], -dstStride[2],

  1246.                               -dstStride[3] };

  1247. 

  1248.         src2[0] += (srcSliceH - 1) * srcStride[0];

  1249.         if (!usePal(c->srcFormat))

  1250.             src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];

  1251.         src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];

  1252.         src2[3] += (srcSliceH - 1) * srcStride[3];

  1253.         dst2[0] += ( c->dstH                         - 1) * dstStride[0];

  1254.         dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1];

  1255.         dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2];

  1256.         dst2[3] += ( c->dstH                         - 1) * dstStride[3];

  1257. 

  1258.         reset_ptr(src2, c->srcFormat);

  1259.         reset_ptr((void*)dst2, c->dstFormat);

  1260. 

  1261.         /* reset slice direction at end of frame */

  1262.         if (!srcSliceY)

  1263.             c->sliceDir = 0;

  1264. 

  1265.         ret = c->swscale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH,

  1266.                           srcSliceH, dst2, dstStride2);

  1267.     }

  1268. 

  1269. 

  1270.     if (c->dstXYZ && !(c->srcXYZ && c->srcW==c->dstW && c->srcH==c->dstH)) {

  1271.         /* replace on the same data */

  1272.         rgb48Toxyz12(c, (uint16_t*)dst2[0], (const uint16_t*)dst2[0], dstStride[0]/2, ret);

  1273.     }

  1274. 

  1275.     av_free(rgb0_tmp);

  1276.     return ret;

  1277. }