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.
80918 Commits
32 Branches
340MB
Tree: fcdbbbd85a
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'fcdbbbd85a'
${ noResults }
FFmpeg/libswscale/swscale.c

1004 lines
36KB
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. static av_always_inline void fillPlane(uint8_t *plane, int stride, int width,

  56.                                        int height, int y, uint8_t val)

  57. {

  58.     int i;

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

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

  61.         memset(ptr, val, width);

  62.         ptr += stride;

  63.     }

  64. }

  65. 

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

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

  68.                            const int32_t *filterPos, int filterSize)

  69. {

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

  71.     int i;

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

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

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

  75.     int sh              = bits - 4;

  76. 

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

  78.         sh= 9;

  79. 

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

  81.         int j;

  82.         int srcPos = filterPos[i];

  83.         int val    = 0;

  84. 

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

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

  87.         }

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

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

  90.     }

  91. }

  92. 

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

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

  95.                            const int32_t *filterPos, int filterSize)

  96. {

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

  98.     int i;

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

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

  101. 

  102.     if(sh<15)

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

  104. 

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

  106.         int j;

  107.         int srcPos = filterPos[i];

  108.         int val    = 0;

  109. 

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

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

  112.         }

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

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

  115.     }

  116. }

  117. 

  118. // bilinear / bicubic scaling

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

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

  121.                           const int32_t *filterPos, int filterSize)

  122. {

  123.     int i;

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

  125.         int j;

  126.         int srcPos = filterPos[i];

  127.         int val    = 0;

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

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

  130.         }

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

  132.     }

  133. }

  134. 

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

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

  137.                           const int32_t *filterPos, int filterSize)

  138. {

  139.     int i;

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

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

  142.         int j;

  143.         int srcPos = filterPos[i];

  144.         int val    = 0;

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

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

  147.         }

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

  149.     }

  150. }

  151. 

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

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

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

  155. {

  156.     int i;

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

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

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

  160.     }

  161. }

  162. 

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

  164. {

  165.     int i;

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

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

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

  169.     }

  170. }

  171. 

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

  173. {

  174.     int i;

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

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

  177. }

  178. 

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

  180. {

  181.     int i;

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

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

  184. }

  185. 

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

  187. {

  188.     int i;

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

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

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

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

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

  194.     }

  195. }

  196. 

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

  198. {

  199.     int i;

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

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

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

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

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

  205.     }

  206. }

  207. 

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

  209. {

  210.     int i;

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

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

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

  214.     }

  215. }

  216. 

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

  218. {

  219.     int i;

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

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

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

  223. }

  224. 

  225. 

  226. #define DEBUG_SWSCALE_BUFFERS 0

  227. #define DEBUG_BUFFERS(...)                      \

  228.     if (DEBUG_SWSCALE_BUFFERS)                  \

  229.         av_log(c, AV_LOG_DEBUG, __VA_ARGS__)

  230. 

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

  232.                    int srcStride[], int srcSliceY,

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

  234. {

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

  236.      * and faster */

  237.     const int dstW                   = c->dstW;

  238.     const int dstH                   = c->dstH;

  239. 

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

  241.     const int flags                  = c->flags;

  242.     int32_t *vLumFilterPos           = c->vLumFilterPos;

  243.     int32_t *vChrFilterPos           = c->vChrFilterPos;

  244. 

  245.     const int vLumFilterSize         = c->vLumFilterSize;

  246.     const int vChrFilterSize         = c->vChrFilterSize;

  247. 

  248.     yuv2planar1_fn yuv2plane1        = c->yuv2plane1;

  249.     yuv2planarX_fn yuv2planeX        = c->yuv2planeX;

  250.     yuv2interleavedX_fn yuv2nv12cX   = c->yuv2nv12cX;

  251.     yuv2packed1_fn yuv2packed1       = c->yuv2packed1;

  252.     yuv2packed2_fn yuv2packed2       = c->yuv2packed2;

  253.     yuv2packedX_fn yuv2packedX       = c->yuv2packedX;

  254.     yuv2anyX_fn yuv2anyX             = c->yuv2anyX;

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

  256.     const int chrSrcSliceH           = AV_CEIL_RSHIFT(srcSliceH,   c->chrSrcVSubSample);

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

  258.                                        is16BPS(c->srcFormat);

  259.     int lastDstY;

  260. 

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

  262.     int dstY         = c->dstY;

  263.     int lumBufIndex  = c->lumBufIndex;

  264.     int chrBufIndex  = c->chrBufIndex;

  265.     int lastInLumBuf = c->lastInLumBuf;

  266.     int lastInChrBuf = c->lastInChrBuf;

  267. 

  268. 

  269.     int lumStart = 0;

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

  271.     int chrStart = lumEnd;

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

  273.     int vStart = chrEnd;

  274.     int vEnd = c->numDesc;

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

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

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

  278.     SwsFilterDescriptor *desc = c->desc;

  279. 

  280. 

  281.     int needAlpha = c->needAlpha;

  282. 

  283.     int hasLumHoles = 1;

  284.     int hasChrHoles = 1;

  285. 

  286. 

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

  288.         src[0] =

  289.         src[1] =

  290.         src[2] =

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

  292.         srcStride[0] =

  293.         srcStride[1] =

  294.         srcStride[2] =

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

  296.     }

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

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

  299. 

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

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

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

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

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

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

  306.                   srcSliceY, srcSliceH, dstY, dstH);

  307.     DEBUG_BUFFERS("vLumFilterSize: %d vChrFilterSize: %d\n",

  308.                   vLumFilterSize, vChrFilterSize);

  309. 

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

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

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

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

  314.             av_log(c, AV_LOG_WARNING,

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

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

  317.             warnedAlready = 1;

  318.         }

  319.     }

  320. 

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

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

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

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

  325.     ) {

  326.         static int warnedAlready=0;

  327.         int cpu_flags = av_get_cpu_flags();

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

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

  330.             warnedAlready=1;

  331.         }

  332.     }

  333. 

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

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

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

  337.     if (srcSliceY == 0) {

  338.         lumBufIndex  = -1;

  339.         chrBufIndex  = -1;

  340.         dstY         = 0;

  341.         lastInLumBuf = -1;

  342.         lastInChrBuf = -1;

  343.     }

  344. 

  345.     if (!should_dither) {

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

  347.     }

  348.     lastDstY = dstY;

  349. 

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

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

  352. 

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

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

  355. 

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

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

  358.             AV_CEIL_RSHIFT(dstH, c->chrDstVSubSample), 0);

  359.     if (srcSliceY == 0) {

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

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

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

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

  364. 

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

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

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

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

  369.         hout_slice->width = dstW;

  370.     }

  371. 

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

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

  374.         int use_mmx_vfilter= c->use_mmx_vfilter;

  375. 

  376.         // First line needed as input

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

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

  379.         // First line needed as input

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

  381. 

  382.         // Last line needed as input

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

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

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

  386.         int enough_lines;

  387. 

  388.         int i;

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

  390. 

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

  392.         if (firstLumSrcY > lastInLumBuf) {

  393. 

  394.             hasLumHoles = lastInLumBuf != firstLumSrcY - 1;

  395.             if (hasLumHoles) {

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

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

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

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

  400.             }

  401. 

  402.             lastInLumBuf = firstLumSrcY - 1;

  403.         }

  404.         if (firstChrSrcY > lastInChrBuf) {

  405. 

  406.             hasChrHoles = lastInChrBuf != firstChrSrcY - 1;

  407.             if (hasChrHoles) {

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

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

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

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

  412.             }

  413. 

  414.             lastInChrBuf = firstChrSrcY - 1;

  415.         }

  416. 

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

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

  419.                       firstLumSrcY, lastLumSrcY, lastInLumBuf);

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

  421.                       firstChrSrcY, lastChrSrcY, lastInChrBuf);

  422. 

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

  424.         enough_lines = lastLumSrcY2 < srcSliceY + srcSliceH &&

  425.                        lastChrSrcY < AV_CEIL_RSHIFT(srcSliceY + srcSliceH, c->chrSrcVSubSample);

  426. 

  427.         if (!enough_lines) {

  428.             lastLumSrcY = srcSliceY + srcSliceH - 1;

  429.             lastChrSrcY = chrSrcSliceY + chrSrcSliceH - 1;

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

  431.                           lastLumSrcY, lastChrSrcY);

  432.         }

  433. 

  434.         av_assert0((lastLumSrcY - firstLumSrcY + 1) <= hout_slice->plane[0].available_lines);

  435.         av_assert0((lastChrSrcY - firstChrSrcY + 1) <= hout_slice->plane[1].available_lines);

  436. 

  437. 

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

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

  440.             firstPosY = FFMAX(firstLumSrcY, posY);

  441.             lastPosY = FFMIN(firstLumSrcY + hout_slice->plane[0].available_lines - 1, srcSliceY + srcSliceH - 1);

  442.         } else {

  443.             firstPosY = lastInLumBuf + 1;

  444.             lastPosY = lastLumSrcY;

  445.         }

  446. 

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

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

  449.             firstCPosY = FFMAX(firstChrSrcY, cPosY);

  450.             lastCPosY = FFMIN(firstChrSrcY + hout_slice->plane[1].available_lines - 1, AV_CEIL_RSHIFT(srcSliceY + srcSliceH, c->chrSrcVSubSample) - 1);

  451.         } else {

  452.             firstCPosY = lastInChrBuf + 1;

  453.             lastCPosY = lastChrSrcY;

  454.         }

  455. 

  456.         ff_rotate_slice(hout_slice, lastPosY, lastCPosY);

  457. 

  458.         if (posY < lastLumSrcY + 1) {

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

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

  461.         }

  462. 

  463.         lumBufIndex += lastLumSrcY - lastInLumBuf;

  464.         lastInLumBuf = lastLumSrcY;

  465. 

  466.         if (cPosY < lastChrSrcY + 1) {

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

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

  469.         }

  470. 

  471.         chrBufIndex += lastChrSrcY - lastInChrBuf;

  472.         lastInChrBuf = lastChrSrcY;

  473. 

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

  475.         if (lumBufIndex >= vLumFilterSize)

  476.             lumBufIndex -= vLumFilterSize;

  477.         if (chrBufIndex >= vChrFilterSize)

  478.             chrBufIndex -= vChrFilterSize;

  479.         if (!enough_lines)

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

  481. 

  482. #if HAVE_MMX_INLINE

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

  484.                               lastInLumBuf, lastInChrBuf);

  485. #endif

  486.         if (should_dither) {

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

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

  489.         }

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

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

  492.              * this array's tail */

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

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

  495.             use_mmx_vfilter= 0;

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

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

  498.         }

  499. 

  500.         {

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

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

  503.         }

  504.     }

  505.     if (isPlanar(dstFormat) && isALPHA(dstFormat) && !needAlpha) {

  506.         int length = dstW;

  507.         int height = dstY - lastDstY;

  508. 

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

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

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

  512.                     1, desc->comp[3].depth,

  513.                     isBE(dstFormat));

  514.         } else

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

  516.     }

  517. 

  518. #if HAVE_MMXEXT_INLINE

  519.     if (av_get_cpu_flags() & AV_CPU_FLAG_MMXEXT)

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

  521. #endif

  522.     emms_c();

  523. 

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

  525.     c->dstY         = dstY;

  526.     c->lumBufIndex  = lumBufIndex;

  527.     c->chrBufIndex  = chrBufIndex;

  528.     c->lastInLumBuf = lastInLumBuf;

  529.     c->lastInChrBuf = lastInChrBuf;

  530. 

  531.     return dstY - lastDstY;

  532. }

  533. 

  534. av_cold void ff_sws_init_range_convert(SwsContext *c)

  535. {

  536.     c->lumConvertRange = NULL;

  537.     c->chrConvertRange = NULL;

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

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

  540.             if (c->srcRange) {

  541.                 c->lumConvertRange = lumRangeFromJpeg_c;

  542.                 c->chrConvertRange = chrRangeFromJpeg_c;

  543.             } else {

  544.                 c->lumConvertRange = lumRangeToJpeg_c;

  545.                 c->chrConvertRange = chrRangeToJpeg_c;

  546.             }

  547.         } else {

  548.             if (c->srcRange) {

  549.                 c->lumConvertRange = lumRangeFromJpeg16_c;

  550.                 c->chrConvertRange = chrRangeFromJpeg16_c;

  551.             } else {

  552.                 c->lumConvertRange = lumRangeToJpeg16_c;

  553.                 c->chrConvertRange = chrRangeToJpeg16_c;

  554.             }

  555.         }

  556.     }

  557. }

  558. 

  559. static av_cold void sws_init_swscale(SwsContext *c)

  560. {

  561.     enum AVPixelFormat srcFormat = c->srcFormat;

  562. 

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

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

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

  566. 

  567.     ff_sws_init_input_funcs(c);

  568. 

  569.     if (HAVE_VSX && (!HAVE_BIGENDIAN)) {

  570.         ff_sws_init_input_funcs_vsx(c);

  571.     }

  572. 

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

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

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

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

  577.                 c->hyscale_fast = ff_hyscale_fast_c;

  578.                 c->hcscale_fast = ff_hcscale_fast_c;

  579.             }

  580.         } else {

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

  582.         }

  583.     } else {

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

  585.                                                  : hScale16To15_c;

  586.     }

  587. 

  588.     ff_sws_init_range_convert(c);

  589. 

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

  591.           srcFormat == AV_PIX_FMT_MONOBLACK || srcFormat == AV_PIX_FMT_MONOWHITE))

  592.         c->needs_hcscale = 1;

  593. }

  594. 

  595. SwsFunc ff_getSwsFunc(SwsContext *c)

  596. {

  597.     sws_init_swscale(c);

  598. 

  599.     if (ARCH_PPC)

  600.         ff_sws_init_swscale_ppc(c);

  601.     if (ARCH_X86)

  602.         ff_sws_init_swscale_x86(c);

  603.     if (ARCH_AARCH64)

  604.         ff_sws_init_swscale_aarch64(c);

  605.     if (ARCH_ARM)

  606.         ff_sws_init_swscale_arm(c);

  607. 

  608.     return swscale;

  609. }

  610. 

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

  612. {

  613.     if (!isALPHA(format))

  614.         src[3] = NULL;

  615.     if (!isPlanar(format)) {

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

  617. 

  618.         if (!usePal(format))

  619.             src[1] = NULL;

  620.     }

  621. }

  622. 

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

  624.                                 const int linesizes[4])

  625. {

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

  627.     int i;

  628. 

  629.     av_assert2(desc);

  630. 

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

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

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

  634.             return 0;

  635.     }

  636. 

  637.     return 1;

  638. }

  639. 

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

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

  642. {

  643.     int xp,yp;

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

  645. 

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

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

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

  649. 

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

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

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

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

  654.             } else {

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

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

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

  658.             }

  659. 

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

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

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

  663. 

  664.             // convert from XYZlinear to sRGBlinear

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

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

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

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

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

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

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

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

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

  674. 

  675.             // limit values to 12-bit depth

  676.             r = av_clip_uintp2(r, 12);

  677.             g = av_clip_uintp2(g, 12);

  678.             b = av_clip_uintp2(b, 12);

  679. 

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

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

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

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

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

  685.             } else {

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

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

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

  689.             }

  690.         }

  691.         src += stride;

  692.         dst += stride;

  693.     }

  694. }

  695. 

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

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

  698. {

  699.     int xp,yp;

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

  701. 

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

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

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

  705. 

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

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

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

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

  710.             } else {

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

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

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

  714.             }

  715. 

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

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

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

  719. 

  720.             // convert from sRGBlinear to XYZlinear

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

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

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

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

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

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

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

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

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

  730. 

  731.             // limit values to 12-bit depth

  732.             x = av_clip_uintp2(x, 12);

  733.             y = av_clip_uintp2(y, 12);

  734.             z = av_clip_uintp2(z, 12);

  735. 

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

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

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

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

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

  741.             } else {

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

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

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

  745.             }

  746.         }

  747.         src += stride;

  748.         dst += stride;

  749.     }

  750. }

  751. 

  752. /**

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

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

  755.  */

  756. int attribute_align_arg sws_scale(struct SwsContext *c,

  757.                                   const uint8_t * const srcSlice[],

  758.                                   const int srcStride[], int srcSliceY,

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

  760.                                   const int dstStride[])

  761. {

  762.     int i, ret;

  763.     const uint8_t *src2[4];

  764.     uint8_t *dst2[4];

  765.     uint8_t *rgb0_tmp = NULL;

  766.     int macro_height = isBayer(c->srcFormat) ? 2 : (1 << c->chrSrcVSubSample);

  767. 

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

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

  770.         return 0;

  771.     }

  772. 

  773.     if ((srcSliceY & (macro_height-1)) ||

  774.         ((srcSliceH& (macro_height-1)) && srcSliceY + srcSliceH != c->srcH) ||

  775.         srcSliceY + srcSliceH > c->srcH) {

  776.         av_log(c, AV_LOG_ERROR, "Slice parameters %d, %d are invalid\n", srcSliceY, srcSliceH);

  777.         return AVERROR(EINVAL);

  778.     }

  779. 

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

  781. 

  782. 

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

  784.                     srcSlice, srcStride, srcSliceY, srcSliceH,

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

  786. 

  787.         if (ret < 0)

  788.             return ret;

  789. 

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

  791.             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);

  792.         else

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

  794. 

  795.         if (ret < 0)

  796.             return ret;

  797. 

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

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

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

  801.                         dst, dstStride);

  802.         }

  803.         return ret;

  804.     }

  805. 

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

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

  808.                         srcSlice, srcStride, srcSliceY, srcSliceH,

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

  810.         if (ret < 0)

  811.             return ret;

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

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

  814.                         dst, dstStride);

  815.         return ret;

  816.     }

  817. 

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

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

  820. 

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

  822.     if (srcSliceH == 0)

  823.         return 0;

  824. 

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

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

  827.         return 0;

  828.     }

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

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

  831.         return 0;

  832.     }

  833. 

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

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

  836.         return 0;

  837.     }

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

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

  840.     }

  841. 

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

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

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

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

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

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

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

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

  850.                 b =  p        & 0xFF;

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

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

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

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

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

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

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

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

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

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

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

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

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

  864.                 r = g = b = i;

  865.             } else {

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

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

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

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

  870.             }

  871. #define RGB2YUV_SHIFT 15

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

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

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

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

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

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

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

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

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

  881. 

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

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

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

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

  886. 

  887.             switch (c->dstFormat) {

  888.             case AV_PIX_FMT_BGR32:

  889. #if !HAVE_BIGENDIAN

  890.             case AV_PIX_FMT_RGB24:

  891. #endif

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

  893.                 break;

  894.             case AV_PIX_FMT_BGR32_1:

  895. #if HAVE_BIGENDIAN

  896.             case AV_PIX_FMT_BGR24:

  897. #endif

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

  899.                 break;

  900.             case AV_PIX_FMT_RGB32_1:

  901. #if HAVE_BIGENDIAN

  902.             case AV_PIX_FMT_RGB24:

  903. #endif

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

  905.                 break;

  906.             case AV_PIX_FMT_RGB32:

  907. #if !HAVE_BIGENDIAN

  908.             case AV_PIX_FMT_BGR24:

  909. #endif

  910.             default:

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

  912.             }

  913.         }

  914.     }

  915. 

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

  917.         uint8_t *base;

  918.         int x,y;

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

  920.         if (!rgb0_tmp)

  921.             return AVERROR(ENOMEM);

  922. 

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

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

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

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

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

  928.             }

  929.         }

  930.         src2[0] = base;

  931.     }

  932. 

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

  934.         uint8_t *base;

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

  936.         if (!rgb0_tmp)

  937.             return AVERROR(ENOMEM);

  938. 

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

  940. 

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

  942.         src2[0] = base;

  943.     }

  944. 

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

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

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

  948. 

  949. 

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

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

  952.         // slices go from top to bottom

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

  954.                               srcStride[3] };

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

  956.                               dstStride[3] };

  957. 

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

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

  960. 

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

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

  963.             c->sliceDir = 0;

  964. 

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

  966.                           dstStride2);

  967.     } else {

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

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

  970.                               -srcStride[3] };

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

  972.                               -dstStride[3] };

  973. 

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

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

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

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

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

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

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

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

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

  983. 

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

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

  986. 

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

  988.         if (!srcSliceY)

  989.             c->sliceDir = 0;

  990. 

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

  992.                           srcSliceH, dst2, dstStride2);

  993.     }

  994. 

  995. 

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

  997.         /* replace on the same data */

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

  999.     }

  1000. 

  1001.     av_free(rgb0_tmp);

  1002.     return ret;

  1003. }