yading@11: /*
yading@11:  * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
yading@11:  *
yading@11:  * This file is part of FFmpeg.
yading@11:  *
yading@11:  * FFmpeg is free software; you can redistribute it and/or
yading@11:  * modify it under the terms of the GNU Lesser General Public
yading@11:  * License as published by the Free Software Foundation; either
yading@11:  * version 2.1 of the License, or (at your option) any later version.
yading@11:  *
yading@11:  * FFmpeg is distributed in the hope that it will be useful,
yading@11:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@11:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@11:  * Lesser General Public License for more details.
yading@11:  *
yading@11:  * You should have received a copy of the GNU Lesser General Public
yading@11:  * License along with FFmpeg; if not, write to the Free Software
yading@11:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@11:  */
yading@11: 
yading@11: #include <assert.h>
yading@11: #include <inttypes.h>
yading@11: #include <math.h>
yading@11: #include <stdio.h>
yading@11: #include <string.h>
yading@11: 
yading@11: #include "libavutil/avassert.h"
yading@11: #include "libavutil/avutil.h"
yading@11: #include "libavutil/bswap.h"
yading@11: #include "libavutil/cpu.h"
yading@11: #include "libavutil/intreadwrite.h"
yading@11: #include "libavutil/mathematics.h"
yading@11: #include "libavutil/pixdesc.h"
yading@11: #include "config.h"
yading@11: #include "rgb2rgb.h"
yading@11: #include "swscale_internal.h"
yading@11: #include "swscale.h"
yading@11: 
yading@11: DECLARE_ALIGNED(8, const uint8_t, dither_8x8_128)[8][8] = {
yading@11:     {  36, 68,  60, 92,  34, 66,  58, 90, },
yading@11:     { 100,  4, 124, 28,  98,  2, 122, 26, },
yading@11:     {  52, 84,  44, 76,  50, 82,  42, 74, },
yading@11:     { 116, 20, 108, 12, 114, 18, 106, 10, },
yading@11:     {  32, 64,  56, 88,  38, 70,  62, 94, },
yading@11:     {  96,  0, 120, 24, 102,  6, 126, 30, },
yading@11:     {  48, 80,  40, 72,  54, 86,  46, 78, },
yading@11:     { 112, 16, 104,  8, 118, 22, 110, 14, },
yading@11: };
yading@11: 
yading@11: DECLARE_ALIGNED(8, const uint8_t, ff_sws_pb_64)[8] = {
yading@11:     64, 64, 64, 64, 64, 64, 64, 64
yading@11: };
yading@11: 
yading@11: static av_always_inline void fillPlane(uint8_t *plane, int stride, int width,
yading@11:                                        int height, int y, uint8_t val)
yading@11: {
yading@11:     int i;
yading@11:     uint8_t *ptr = plane + stride * y;
yading@11:     for (i = 0; i < height; i++) {
yading@11:         memset(ptr, val, width);
yading@11:         ptr += stride;
yading@11:     }
yading@11: }
yading@11: 
yading@11: static void hScale16To19_c(SwsContext *c, int16_t *_dst, int dstW,
yading@11:                            const uint8_t *_src, const int16_t *filter,
yading@11:                            const int32_t *filterPos, int filterSize)
yading@11: {
yading@11:     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->srcFormat);
yading@11:     int i;
yading@11:     int32_t *dst        = (int32_t *) _dst;
yading@11:     const uint16_t *src = (const uint16_t *) _src;
yading@11:     int bits            = desc->comp[0].depth_minus1;
yading@11:     int sh              = bits - 4;
yading@11: 
yading@11:     if((isAnyRGB(c->srcFormat) || c->srcFormat==AV_PIX_FMT_PAL8) && desc->comp[0].depth_minus1<15)
yading@11:         sh= 9;
yading@11: 
yading@11:     for (i = 0; i < dstW; i++) {
yading@11:         int j;
yading@11:         int srcPos = filterPos[i];
yading@11:         int val    = 0;
yading@11: 
yading@11:         for (j = 0; j < filterSize; j++) {
yading@11:             val += src[srcPos + j] * filter[filterSize * i + j];
yading@11:         }
yading@11:         // filter=14 bit, input=16 bit, output=30 bit, >> 11 makes 19 bit
yading@11:         dst[i] = FFMIN(val >> sh, (1 << 19) - 1);
yading@11:     }
yading@11: }
yading@11: 
yading@11: static void hScale16To15_c(SwsContext *c, int16_t *dst, int dstW,
yading@11:                            const uint8_t *_src, const int16_t *filter,
yading@11:                            const int32_t *filterPos, int filterSize)
yading@11: {
yading@11:     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->srcFormat);
yading@11:     int i;
yading@11:     const uint16_t *src = (const uint16_t *) _src;
yading@11:     int sh              = desc->comp[0].depth_minus1;
yading@11: 
yading@11:     if(sh<15)
yading@11:         sh= isAnyRGB(c->srcFormat) || c->srcFormat==AV_PIX_FMT_PAL8 ? 13 : desc->comp[0].depth_minus1;
yading@11: 
yading@11:     for (i = 0; i < dstW; i++) {
yading@11:         int j;
yading@11:         int srcPos = filterPos[i];
yading@11:         int val    = 0;
yading@11: 
yading@11:         for (j = 0; j < filterSize; j++) {
yading@11:             val += src[srcPos + j] * filter[filterSize * i + j];
yading@11:         }
yading@11:         // filter=14 bit, input=16 bit, output=30 bit, >> 15 makes 15 bit
yading@11:         dst[i] = FFMIN(val >> sh, (1 << 15) - 1);
yading@11:     }
yading@11: }
yading@11: 
yading@11: // bilinear / bicubic scaling
yading@11: static void hScale8To15_c(SwsContext *c, int16_t *dst, int dstW,
yading@11:                           const uint8_t *src, const int16_t *filter,
yading@11:                           const int32_t *filterPos, int filterSize)
yading@11: {
yading@11:     int i;
yading@11:     for (i = 0; i < dstW; i++) {
yading@11:         int j;
yading@11:         int srcPos = filterPos[i];
yading@11:         int val    = 0;
yading@11:         for (j = 0; j < filterSize; j++) {
yading@11:             val += ((int)src[srcPos + j]) * filter[filterSize * i + j];
yading@11:         }
yading@11:         dst[i] = FFMIN(val >> 7, (1 << 15) - 1); // the cubic equation does overflow ...
yading@11:     }
yading@11: }
yading@11: 
yading@11: static void hScale8To19_c(SwsContext *c, int16_t *_dst, int dstW,
yading@11:                           const uint8_t *src, const int16_t *filter,
yading@11:                           const int32_t *filterPos, int filterSize)
yading@11: {
yading@11:     int i;
yading@11:     int32_t *dst = (int32_t *) _dst;
yading@11:     for (i = 0; i < dstW; i++) {
yading@11:         int j;
yading@11:         int srcPos = filterPos[i];
yading@11:         int val    = 0;
yading@11:         for (j = 0; j < filterSize; j++) {
yading@11:             val += ((int)src[srcPos + j]) * filter[filterSize * i + j];
yading@11:         }
yading@11:         dst[i] = FFMIN(val >> 3, (1 << 19) - 1); // the cubic equation does overflow ...
yading@11:     }
yading@11: }
yading@11: 
yading@11: // FIXME all pal and rgb srcFormats could do this conversion as well
yading@11: // FIXME all scalers more complex than bilinear could do half of this transform
yading@11: static void chrRangeToJpeg_c(int16_t *dstU, int16_t *dstV, int width)
yading@11: {
yading@11:     int i;
yading@11:     for (i = 0; i < width; i++) {
yading@11:         dstU[i] = (FFMIN(dstU[i], 30775) * 4663 - 9289992) >> 12; // -264
yading@11:         dstV[i] = (FFMIN(dstV[i], 30775) * 4663 - 9289992) >> 12; // -264
yading@11:     }
yading@11: }
yading@11: 
yading@11: static void chrRangeFromJpeg_c(int16_t *dstU, int16_t *dstV, int width)
yading@11: {
yading@11:     int i;
yading@11:     for (i = 0; i < width; i++) {
yading@11:         dstU[i] = (dstU[i] * 1799 + 4081085) >> 11; // 1469
yading@11:         dstV[i] = (dstV[i] * 1799 + 4081085) >> 11; // 1469
yading@11:     }
yading@11: }
yading@11: 
yading@11: static void lumRangeToJpeg_c(int16_t *dst, int width)
yading@11: {
yading@11:     int i;
yading@11:     for (i = 0; i < width; i++)
yading@11:         dst[i] = (FFMIN(dst[i], 30189) * 19077 - 39057361) >> 14;
yading@11: }
yading@11: 
yading@11: static void lumRangeFromJpeg_c(int16_t *dst, int width)
yading@11: {
yading@11:     int i;
yading@11:     for (i = 0; i < width; i++)
yading@11:         dst[i] = (dst[i] * 14071 + 33561947) >> 14;
yading@11: }
yading@11: 
yading@11: static void chrRangeToJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)
yading@11: {
yading@11:     int i;
yading@11:     int32_t *dstU = (int32_t *) _dstU;
yading@11:     int32_t *dstV = (int32_t *) _dstV;
yading@11:     for (i = 0; i < width; i++) {
yading@11:         dstU[i] = (FFMIN(dstU[i], 30775 << 4) * 4663 - (9289992 << 4)) >> 12; // -264
yading@11:         dstV[i] = (FFMIN(dstV[i], 30775 << 4) * 4663 - (9289992 << 4)) >> 12; // -264
yading@11:     }
yading@11: }
yading@11: 
yading@11: static void chrRangeFromJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)
yading@11: {
yading@11:     int i;
yading@11:     int32_t *dstU = (int32_t *) _dstU;
yading@11:     int32_t *dstV = (int32_t *) _dstV;
yading@11:     for (i = 0; i < width; i++) {
yading@11:         dstU[i] = (dstU[i] * 1799 + (4081085 << 4)) >> 11; // 1469
yading@11:         dstV[i] = (dstV[i] * 1799 + (4081085 << 4)) >> 11; // 1469
yading@11:     }
yading@11: }
yading@11: 
yading@11: static void lumRangeToJpeg16_c(int16_t *_dst, int width)
yading@11: {
yading@11:     int i;
yading@11:     int32_t *dst = (int32_t *) _dst;
yading@11:     for (i = 0; i < width; i++)
yading@11:         dst[i] = (FFMIN(dst[i], 30189 << 4) * 4769 - (39057361 << 2)) >> 12;
yading@11: }
yading@11: 
yading@11: static void lumRangeFromJpeg16_c(int16_t *_dst, int width)
yading@11: {
yading@11:     int i;
yading@11:     int32_t *dst = (int32_t *) _dst;
yading@11:     for (i = 0; i < width; i++)
yading@11:         dst[i] = (dst[i]*(14071/4) + (33561947<<4)/4)>>12;
yading@11: }
yading@11: 
yading@11: static void hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
yading@11:                            const uint8_t *src, int srcW, int xInc)
yading@11: {
yading@11:     int i;
yading@11:     unsigned int xpos = 0;
yading@11:     for (i = 0; i < dstWidth; i++) {
yading@11:         register unsigned int xx     = xpos >> 16;
yading@11:         register unsigned int xalpha = (xpos & 0xFFFF) >> 9;
yading@11:         dst[i] = (src[xx] << 7) + (src[xx + 1] - src[xx]) * xalpha;
yading@11:         xpos  += xInc;
yading@11:     }
yading@11:     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
yading@11:         dst[i] = src[srcW-1]*128;
yading@11: }
yading@11: 
yading@11: // *** horizontal scale Y line to temp buffer
yading@11: static av_always_inline void hyscale(SwsContext *c, int16_t *dst, int dstWidth,
yading@11:                                      const uint8_t *src_in[4],
yading@11:                                      int srcW, int xInc,
yading@11:                                      const int16_t *hLumFilter,
yading@11:                                      const int32_t *hLumFilterPos,
yading@11:                                      int hLumFilterSize,
yading@11:                                      uint8_t *formatConvBuffer,
yading@11:                                      uint32_t *pal, int isAlpha)
yading@11: {
yading@11:     void (*toYV12)(uint8_t *, const uint8_t *, const uint8_t *, const uint8_t *, int, uint32_t *) =
yading@11:         isAlpha ? c->alpToYV12 : c->lumToYV12;
yading@11:     void (*convertRange)(int16_t *, int) = isAlpha ? NULL : c->lumConvertRange;
yading@11:     const uint8_t *src = src_in[isAlpha ? 3 : 0];
yading@11: 
yading@11:     if (toYV12) {
yading@11:         toYV12(formatConvBuffer, src, src_in[1], src_in[2], srcW, pal);
yading@11:         src = formatConvBuffer;
yading@11:     } else if (c->readLumPlanar && !isAlpha) {
yading@11:         c->readLumPlanar(formatConvBuffer, src_in, srcW, c->input_rgb2yuv_table);
yading@11:         src = formatConvBuffer;
yading@11:     }
yading@11: 
yading@11:     if (!c->hyscale_fast) {
yading@11:         c->hyScale(c, dst, dstWidth, src, hLumFilter,
yading@11:                    hLumFilterPos, hLumFilterSize);
yading@11:     } else { // fast bilinear upscale / crap downscale
yading@11:         c->hyscale_fast(c, dst, dstWidth, src, srcW, xInc);
yading@11:     }
yading@11: 
yading@11:     if (convertRange)
yading@11:         convertRange(dst, dstWidth);
yading@11: }
yading@11: 
yading@11: static void hcscale_fast_c(SwsContext *c, int16_t *dst1, int16_t *dst2,
yading@11:                            int dstWidth, const uint8_t *src1,
yading@11:                            const uint8_t *src2, int srcW, int xInc)
yading@11: {
yading@11:     int i;
yading@11:     unsigned int xpos = 0;
yading@11:     for (i = 0; i < dstWidth; i++) {
yading@11:         register unsigned int xx     = xpos >> 16;
yading@11:         register unsigned int xalpha = (xpos & 0xFFFF) >> 9;
yading@11:         dst1[i] = (src1[xx] * (xalpha ^ 127) + src1[xx + 1] * xalpha);
yading@11:         dst2[i] = (src2[xx] * (xalpha ^ 127) + src2[xx + 1] * xalpha);
yading@11:         xpos   += xInc;
yading@11:     }
yading@11:     for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
yading@11:         dst1[i] = src1[srcW-1]*128;
yading@11:         dst2[i] = src2[srcW-1]*128;
yading@11:     }
yading@11: }
yading@11: 
yading@11: static av_always_inline void hcscale(SwsContext *c, int16_t *dst1,
yading@11:                                      int16_t *dst2, int dstWidth,
yading@11:                                      const uint8_t *src_in[4],
yading@11:                                      int srcW, int xInc,
yading@11:                                      const int16_t *hChrFilter,
yading@11:                                      const int32_t *hChrFilterPos,
yading@11:                                      int hChrFilterSize,
yading@11:                                      uint8_t *formatConvBuffer, uint32_t *pal)
yading@11: {
yading@11:     const uint8_t *src1 = src_in[1], *src2 = src_in[2];
yading@11:     if (c->chrToYV12) {
yading@11:         uint8_t *buf2 = formatConvBuffer +
yading@11:                         FFALIGN(srcW*2+78, 16);
yading@11:         c->chrToYV12(formatConvBuffer, buf2, src_in[0], src1, src2, srcW, pal);
yading@11:         src1= formatConvBuffer;
yading@11:         src2= buf2;
yading@11:     } else if (c->readChrPlanar) {
yading@11:         uint8_t *buf2 = formatConvBuffer +
yading@11:                         FFALIGN(srcW*2+78, 16);
yading@11:         c->readChrPlanar(formatConvBuffer, buf2, src_in, srcW, c->input_rgb2yuv_table);
yading@11:         src1 = formatConvBuffer;
yading@11:         src2 = buf2;
yading@11:     }
yading@11: 
yading@11:     if (!c->hcscale_fast) {
yading@11:         c->hcScale(c, dst1, dstWidth, src1, hChrFilter, hChrFilterPos, hChrFilterSize);
yading@11:         c->hcScale(c, dst2, dstWidth, src2, hChrFilter, hChrFilterPos, hChrFilterSize);
yading@11:     } else { // fast bilinear upscale / crap downscale
yading@11:         c->hcscale_fast(c, dst1, dst2, dstWidth, src1, src2, srcW, xInc);
yading@11:     }
yading@11: 
yading@11:     if (c->chrConvertRange)
yading@11:         c->chrConvertRange(dst1, dst2, dstWidth);
yading@11: }
yading@11: 
yading@11: #define DEBUG_SWSCALE_BUFFERS 0
yading@11: #define DEBUG_BUFFERS(...)                      \
yading@11:     if (DEBUG_SWSCALE_BUFFERS)                  \
yading@11:         av_log(c, AV_LOG_DEBUG, __VA_ARGS__)
yading@11: 
yading@11: static int swScale(SwsContext *c, const uint8_t *src[],
yading@11:                    int srcStride[], int srcSliceY,
yading@11:                    int srcSliceH, uint8_t *dst[], int dstStride[])
yading@11: {
yading@11:     /* load a few things into local vars to make the code more readable?
yading@11:      * and faster */
yading@11:     const int srcW                   = c->srcW;
yading@11:     const int dstW                   = c->dstW;
yading@11:     const int dstH                   = c->dstH;
yading@11:     const int chrDstW                = c->chrDstW;
yading@11:     const int chrSrcW                = c->chrSrcW;
yading@11:     const int lumXInc                = c->lumXInc;
yading@11:     const int chrXInc                = c->chrXInc;
yading@11:     const enum AVPixelFormat dstFormat = c->dstFormat;
yading@11:     const int flags                  = c->flags;
yading@11:     int32_t *vLumFilterPos           = c->vLumFilterPos;
yading@11:     int32_t *vChrFilterPos           = c->vChrFilterPos;
yading@11:     int32_t *hLumFilterPos           = c->hLumFilterPos;
yading@11:     int32_t *hChrFilterPos           = c->hChrFilterPos;
yading@11:     int16_t *hLumFilter              = c->hLumFilter;
yading@11:     int16_t *hChrFilter              = c->hChrFilter;
yading@11:     int32_t *lumMmxFilter            = c->lumMmxFilter;
yading@11:     int32_t *chrMmxFilter            = c->chrMmxFilter;
yading@11:     const int vLumFilterSize         = c->vLumFilterSize;
yading@11:     const int vChrFilterSize         = c->vChrFilterSize;
yading@11:     const int hLumFilterSize         = c->hLumFilterSize;
yading@11:     const int hChrFilterSize         = c->hChrFilterSize;
yading@11:     int16_t **lumPixBuf              = c->lumPixBuf;
yading@11:     int16_t **chrUPixBuf             = c->chrUPixBuf;
yading@11:     int16_t **chrVPixBuf             = c->chrVPixBuf;
yading@11:     int16_t **alpPixBuf              = c->alpPixBuf;
yading@11:     const int vLumBufSize            = c->vLumBufSize;
yading@11:     const int vChrBufSize            = c->vChrBufSize;
yading@11:     uint8_t *formatConvBuffer        = c->formatConvBuffer;
yading@11:     uint32_t *pal                    = c->pal_yuv;
yading@11:     yuv2planar1_fn yuv2plane1        = c->yuv2plane1;
yading@11:     yuv2planarX_fn yuv2planeX        = c->yuv2planeX;
yading@11:     yuv2interleavedX_fn yuv2nv12cX   = c->yuv2nv12cX;
yading@11:     yuv2packed1_fn yuv2packed1       = c->yuv2packed1;
yading@11:     yuv2packed2_fn yuv2packed2       = c->yuv2packed2;
yading@11:     yuv2packedX_fn yuv2packedX       = c->yuv2packedX;
yading@11:     yuv2anyX_fn yuv2anyX             = c->yuv2anyX;
yading@11:     const int chrSrcSliceY           =     srcSliceY  >> c->chrSrcVSubSample;
yading@11:     const int chrSrcSliceH           = -((-srcSliceH) >> c->chrSrcVSubSample);
yading@11:     int should_dither                = is9_OR_10BPS(c->srcFormat) ||
yading@11:                                        is16BPS(c->srcFormat);
yading@11:     int lastDstY;
yading@11: 
yading@11:     /* vars which will change and which we need to store back in the context */
yading@11:     int dstY         = c->dstY;
yading@11:     int lumBufIndex  = c->lumBufIndex;
yading@11:     int chrBufIndex  = c->chrBufIndex;
yading@11:     int lastInLumBuf = c->lastInLumBuf;
yading@11:     int lastInChrBuf = c->lastInChrBuf;
yading@11: 
yading@11:     if (!usePal(c->srcFormat)) {
yading@11:         pal = c->input_rgb2yuv_table;
yading@11:     }
yading@11: 
yading@11:     if (isPacked(c->srcFormat)) {
yading@11:         src[0] =
yading@11:         src[1] =
yading@11:         src[2] =
yading@11:         src[3] = src[0];
yading@11:         srcStride[0] =
yading@11:         srcStride[1] =
yading@11:         srcStride[2] =
yading@11:         srcStride[3] = srcStride[0];
yading@11:     }
yading@11:     srcStride[1] <<= c->vChrDrop;
yading@11:     srcStride[2] <<= c->vChrDrop;
yading@11: 
yading@11:     DEBUG_BUFFERS("swScale() %p[%d] %p[%d] %p[%d] %p[%d] -> %p[%d] %p[%d] %p[%d] %p[%d]\n",
yading@11:                   src[0], srcStride[0], src[1], srcStride[1],
yading@11:                   src[2], srcStride[2], src[3], srcStride[3],
yading@11:                   dst[0], dstStride[0], dst[1], dstStride[1],
yading@11:                   dst[2], dstStride[2], dst[3], dstStride[3]);
yading@11:     DEBUG_BUFFERS("srcSliceY: %d srcSliceH: %d dstY: %d dstH: %d\n",
yading@11:                   srcSliceY, srcSliceH, dstY, dstH);
yading@11:     DEBUG_BUFFERS("vLumFilterSize: %d vLumBufSize: %d vChrFilterSize: %d vChrBufSize: %d\n",
yading@11:                   vLumFilterSize, vLumBufSize, vChrFilterSize, vChrBufSize);
yading@11: 
yading@11:     if (dstStride[0]%16 !=0 || dstStride[1]%16 !=0 ||
yading@11:         dstStride[2]%16 !=0 || dstStride[3]%16 != 0) {
yading@11:         static int warnedAlready = 0; // FIXME maybe move this into the context
yading@11:         if (flags & SWS_PRINT_INFO && !warnedAlready) {
yading@11:             av_log(c, AV_LOG_WARNING,
yading@11:                    "Warning: dstStride is not aligned!\n"
yading@11:                    "         ->cannot do aligned memory accesses anymore\n");
yading@11:             warnedAlready = 1;
yading@11:         }
yading@11:     }
yading@11: 
yading@11:     if (   (uintptr_t)dst[0]%16 || (uintptr_t)dst[1]%16 || (uintptr_t)dst[2]%16
yading@11:         || (uintptr_t)src[0]%16 || (uintptr_t)src[1]%16 || (uintptr_t)src[2]%16
yading@11:         || dstStride[0]%16 || dstStride[1]%16 || dstStride[2]%16 || dstStride[3]%16
yading@11:         || srcStride[0]%16 || srcStride[1]%16 || srcStride[2]%16 || srcStride[3]%16
yading@11:     ) {
yading@11:         static int warnedAlready=0;
yading@11:         int cpu_flags = av_get_cpu_flags();
yading@11:         if (HAVE_MMXEXT && (cpu_flags & AV_CPU_FLAG_SSE2) && !warnedAlready){
yading@11:             av_log(c, AV_LOG_WARNING, "Warning: data is not aligned! This can lead to a speedloss\n");
yading@11:             warnedAlready=1;
yading@11:         }
yading@11:     }
yading@11: 
yading@11:     /* Note the user might start scaling the picture in the middle so this
yading@11:      * will not get executed. This is not really intended but works
yading@11:      * currently, so people might do it. */
yading@11:     if (srcSliceY == 0) {
yading@11:         lumBufIndex  = -1;
yading@11:         chrBufIndex  = -1;
yading@11:         dstY         = 0;
yading@11:         lastInLumBuf = -1;
yading@11:         lastInChrBuf = -1;
yading@11:     }
yading@11: 
yading@11:     if (!should_dither) {
yading@11:         c->chrDither8 = c->lumDither8 = ff_sws_pb_64;
yading@11:     }
yading@11:     lastDstY = dstY;
yading@11: 
yading@11:     for (; dstY < dstH; dstY++) {
yading@11:         const int chrDstY = dstY >> c->chrDstVSubSample;
yading@11:         uint8_t *dest[4]  = {
yading@11:             dst[0] + dstStride[0] * dstY,
yading@11:             dst[1] + dstStride[1] * chrDstY,
yading@11:             dst[2] + dstStride[2] * chrDstY,
yading@11:             (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? dst[3] + dstStride[3] * dstY : NULL,
yading@11:         };
yading@11:         int use_mmx_vfilter= c->use_mmx_vfilter;
yading@11: 
yading@11:         // First line needed as input
yading@11:         const int firstLumSrcY  = FFMAX(1 - vLumFilterSize, vLumFilterPos[dstY]);
yading@11:         const int firstLumSrcY2 = FFMAX(1 - vLumFilterSize, vLumFilterPos[FFMIN(dstY | ((1 << c->chrDstVSubSample) - 1), dstH - 1)]);
yading@11:         // First line needed as input
yading@11:         const int firstChrSrcY  = FFMAX(1 - vChrFilterSize, vChrFilterPos[chrDstY]);
yading@11: 
yading@11:         // Last line needed as input
yading@11:         int lastLumSrcY  = FFMIN(c->srcH,    firstLumSrcY  + vLumFilterSize) - 1;
yading@11:         int lastLumSrcY2 = FFMIN(c->srcH,    firstLumSrcY2 + vLumFilterSize) - 1;
yading@11:         int lastChrSrcY  = FFMIN(c->chrSrcH, firstChrSrcY  + vChrFilterSize) - 1;
yading@11:         int enough_lines;
yading@11: 
yading@11:         // handle holes (FAST_BILINEAR & weird filters)
yading@11:         if (firstLumSrcY > lastInLumBuf)
yading@11:             lastInLumBuf = firstLumSrcY - 1;
yading@11:         if (firstChrSrcY > lastInChrBuf)
yading@11:             lastInChrBuf = firstChrSrcY - 1;
yading@11:         av_assert0(firstLumSrcY >= lastInLumBuf - vLumBufSize + 1);
yading@11:         av_assert0(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1);
yading@11: 
yading@11:         DEBUG_BUFFERS("dstY: %d\n", dstY);
yading@11:         DEBUG_BUFFERS("\tfirstLumSrcY: %d lastLumSrcY: %d lastInLumBuf: %d\n",
yading@11:                       firstLumSrcY, lastLumSrcY, lastInLumBuf);
yading@11:         DEBUG_BUFFERS("\tfirstChrSrcY: %d lastChrSrcY: %d lastInChrBuf: %d\n",
yading@11:                       firstChrSrcY, lastChrSrcY, lastInChrBuf);
yading@11: 
yading@11:         // Do we have enough lines in this slice to output the dstY line
yading@11:         enough_lines = lastLumSrcY2 < srcSliceY + srcSliceH &&
yading@11:                        lastChrSrcY < -((-srcSliceY - srcSliceH) >> c->chrSrcVSubSample);
yading@11: 
yading@11:         if (!enough_lines) {
yading@11:             lastLumSrcY = srcSliceY + srcSliceH - 1;
yading@11:             lastChrSrcY = chrSrcSliceY + chrSrcSliceH - 1;
yading@11:             DEBUG_BUFFERS("buffering slice: lastLumSrcY %d lastChrSrcY %d\n",
yading@11:                           lastLumSrcY, lastChrSrcY);
yading@11:         }
yading@11: 
yading@11:         // Do horizontal scaling
yading@11:         while (lastInLumBuf < lastLumSrcY) {
yading@11:             const uint8_t *src1[4] = {
yading@11:                 src[0] + (lastInLumBuf + 1 - srcSliceY) * srcStride[0],
yading@11:                 src[1] + (lastInLumBuf + 1 - srcSliceY) * srcStride[1],
yading@11:                 src[2] + (lastInLumBuf + 1 - srcSliceY) * srcStride[2],
yading@11:                 src[3] + (lastInLumBuf + 1 - srcSliceY) * srcStride[3],
yading@11:             };
yading@11:             lumBufIndex++;
yading@11:             av_assert0(lumBufIndex < 2 * vLumBufSize);
yading@11:             av_assert0(lastInLumBuf + 1 - srcSliceY < srcSliceH);
yading@11:             av_assert0(lastInLumBuf + 1 - srcSliceY >= 0);
yading@11:             hyscale(c, lumPixBuf[lumBufIndex], dstW, src1, srcW, lumXInc,
yading@11:                     hLumFilter, hLumFilterPos, hLumFilterSize,
yading@11:                     formatConvBuffer, pal, 0);
yading@11:             if (CONFIG_SWSCALE_ALPHA && alpPixBuf)
yading@11:                 hyscale(c, alpPixBuf[lumBufIndex], dstW, src1, srcW,
yading@11:                         lumXInc, hLumFilter, hLumFilterPos, hLumFilterSize,
yading@11:                         formatConvBuffer, pal, 1);
yading@11:             lastInLumBuf++;
yading@11:             DEBUG_BUFFERS("\t\tlumBufIndex %d: lastInLumBuf: %d\n",
yading@11:                           lumBufIndex, lastInLumBuf);
yading@11:         }
yading@11:         while (lastInChrBuf < lastChrSrcY) {
yading@11:             const uint8_t *src1[4] = {
yading@11:                 src[0] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[0],
yading@11:                 src[1] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[1],
yading@11:                 src[2] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[2],
yading@11:                 src[3] + (lastInChrBuf + 1 - chrSrcSliceY) * srcStride[3],
yading@11:             };
yading@11:             chrBufIndex++;
yading@11:             av_assert0(chrBufIndex < 2 * vChrBufSize);
yading@11:             av_assert0(lastInChrBuf + 1 - chrSrcSliceY < (chrSrcSliceH));
yading@11:             av_assert0(lastInChrBuf + 1 - chrSrcSliceY >= 0);
yading@11:             // FIXME replace parameters through context struct (some at least)
yading@11: 
yading@11:             if (c->needs_hcscale)
yading@11:                 hcscale(c, chrUPixBuf[chrBufIndex], chrVPixBuf[chrBufIndex],
yading@11:                         chrDstW, src1, chrSrcW, chrXInc,
yading@11:                         hChrFilter, hChrFilterPos, hChrFilterSize,
yading@11:                         formatConvBuffer, pal);
yading@11:             lastInChrBuf++;
yading@11:             DEBUG_BUFFERS("\t\tchrBufIndex %d: lastInChrBuf: %d\n",
yading@11:                           chrBufIndex, lastInChrBuf);
yading@11:         }
yading@11:         // wrap buf index around to stay inside the ring buffer
yading@11:         if (lumBufIndex >= vLumBufSize)
yading@11:             lumBufIndex -= vLumBufSize;
yading@11:         if (chrBufIndex >= vChrBufSize)
yading@11:             chrBufIndex -= vChrBufSize;
yading@11:         if (!enough_lines)
yading@11:             break;  // we can't output a dstY line so let's try with the next slice
yading@11: 
yading@11: #if HAVE_MMX_INLINE
yading@11:         updateMMXDitherTables(c, dstY, lumBufIndex, chrBufIndex,
yading@11:                               lastInLumBuf, lastInChrBuf);
yading@11: #endif
yading@11:         if (should_dither) {
yading@11:             c->chrDither8 = dither_8x8_128[chrDstY & 7];
yading@11:             c->lumDither8 = dither_8x8_128[dstY    & 7];
yading@11:         }
yading@11:         if (dstY >= dstH - 2) {
yading@11:             /* hmm looks like we can't use MMX here without overwriting
yading@11:              * this array's tail */
yading@11:             ff_sws_init_output_funcs(c, &yuv2plane1, &yuv2planeX, &yuv2nv12cX,
yading@11:                                      &yuv2packed1, &yuv2packed2, &yuv2packedX, &yuv2anyX);
yading@11:             use_mmx_vfilter= 0;
yading@11:         }
yading@11: 
yading@11:         {
yading@11:             const int16_t **lumSrcPtr  = (const int16_t **)(void*) lumPixBuf  + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
yading@11:             const int16_t **chrUSrcPtr = (const int16_t **)(void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
yading@11:             const int16_t **chrVSrcPtr = (const int16_t **)(void*) chrVPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
yading@11:             const int16_t **alpSrcPtr  = (CONFIG_SWSCALE_ALPHA && alpPixBuf) ?
yading@11:                                          (const int16_t **)(void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
yading@11:             int16_t *vLumFilter = c->vLumFilter;
yading@11:             int16_t *vChrFilter = c->vChrFilter;
yading@11: 
yading@11:             if (isPlanarYUV(dstFormat) ||
yading@11:                 (isGray(dstFormat) && !isALPHA(dstFormat))) { // YV12 like
yading@11:                 const int chrSkipMask = (1 << c->chrDstVSubSample) - 1;
yading@11: 
yading@11:                 vLumFilter +=    dstY * vLumFilterSize;
yading@11:                 vChrFilter += chrDstY * vChrFilterSize;
yading@11: 
yading@11: //                 av_assert0(use_mmx_vfilter != (
yading@11: //                                yuv2planeX == yuv2planeX_10BE_c
yading@11: //                             || yuv2planeX == yuv2planeX_10LE_c
yading@11: //                             || yuv2planeX == yuv2planeX_9BE_c
yading@11: //                             || yuv2planeX == yuv2planeX_9LE_c
yading@11: //                             || yuv2planeX == yuv2planeX_16BE_c
yading@11: //                             || yuv2planeX == yuv2planeX_16LE_c
yading@11: //                             || yuv2planeX == yuv2planeX_8_c) || !ARCH_X86);
yading@11: 
yading@11:                 if(use_mmx_vfilter){
yading@11:                     vLumFilter= (int16_t *)c->lumMmxFilter;
yading@11:                     vChrFilter= (int16_t *)c->chrMmxFilter;
yading@11:                 }
yading@11: 
yading@11:                 if (vLumFilterSize == 1) {
yading@11:                     yuv2plane1(lumSrcPtr[0], dest[0], dstW, c->lumDither8, 0);
yading@11:                 } else {
yading@11:                     yuv2planeX(vLumFilter, vLumFilterSize,
yading@11:                                lumSrcPtr, dest[0],
yading@11:                                dstW, c->lumDither8, 0);
yading@11:                 }
yading@11: 
yading@11:                 if (!((dstY & chrSkipMask) || isGray(dstFormat))) {
yading@11:                     if (yuv2nv12cX) {
yading@11:                         yuv2nv12cX(c, vChrFilter,
yading@11:                                    vChrFilterSize, chrUSrcPtr, chrVSrcPtr,
yading@11:                                    dest[1], chrDstW);
yading@11:                     } else if (vChrFilterSize == 1) {
yading@11:                         yuv2plane1(chrUSrcPtr[0], dest[1], chrDstW, c->chrDither8, 0);
yading@11:                         yuv2plane1(chrVSrcPtr[0], dest[2], chrDstW, c->chrDither8, 3);
yading@11:                     } else {
yading@11:                         yuv2planeX(vChrFilter,
yading@11:                                    vChrFilterSize, chrUSrcPtr, dest[1],
yading@11:                                    chrDstW, c->chrDither8, 0);
yading@11:                         yuv2planeX(vChrFilter,
yading@11:                                    vChrFilterSize, chrVSrcPtr, dest[2],
yading@11:                                    chrDstW, c->chrDither8, use_mmx_vfilter ? (c->uv_offx2 >> 1) : 3);
yading@11:                     }
yading@11:                 }
yading@11: 
yading@11:                 if (CONFIG_SWSCALE_ALPHA && alpPixBuf) {
yading@11:                     if(use_mmx_vfilter){
yading@11:                         vLumFilter= (int16_t *)c->alpMmxFilter;
yading@11:                     }
yading@11:                     if (vLumFilterSize == 1) {
yading@11:                         yuv2plane1(alpSrcPtr[0], dest[3], dstW,
yading@11:                                    c->lumDither8, 0);
yading@11:                     } else {
yading@11:                         yuv2planeX(vLumFilter,
yading@11:                                    vLumFilterSize, alpSrcPtr, dest[3],
yading@11:                                    dstW, c->lumDither8, 0);
yading@11:                     }
yading@11:                 }
yading@11:             } else if (yuv2packedX) {
yading@11:                 av_assert1(lumSrcPtr  + vLumFilterSize - 1 < (const int16_t **)lumPixBuf  + vLumBufSize * 2);
yading@11:                 av_assert1(chrUSrcPtr + vChrFilterSize - 1 < (const int16_t **)chrUPixBuf + vChrBufSize * 2);
yading@11:                 if (c->yuv2packed1 && vLumFilterSize == 1 &&
yading@11:                     vChrFilterSize <= 2) { // unscaled RGB
yading@11:                     int chrAlpha = vChrFilterSize == 1 ? 0 : vChrFilter[2 * dstY + 1];
yading@11:                     yuv2packed1(c, *lumSrcPtr, chrUSrcPtr, chrVSrcPtr,
yading@11:                                 alpPixBuf ? *alpSrcPtr : NULL,
yading@11:                                 dest[0], dstW, chrAlpha, dstY);
yading@11:                 } else if (c->yuv2packed2 && vLumFilterSize == 2 &&
yading@11:                            vChrFilterSize == 2) { // bilinear upscale RGB
yading@11:                     int lumAlpha = vLumFilter[2 * dstY + 1];
yading@11:                     int chrAlpha = vChrFilter[2 * dstY + 1];
yading@11:                     lumMmxFilter[2] =
yading@11:                     lumMmxFilter[3] = vLumFilter[2 * dstY]    * 0x10001;
yading@11:                     chrMmxFilter[2] =
yading@11:                     chrMmxFilter[3] = vChrFilter[2 * chrDstY] * 0x10001;
yading@11:                     yuv2packed2(c, lumSrcPtr, chrUSrcPtr, chrVSrcPtr,
yading@11:                                 alpPixBuf ? alpSrcPtr : NULL,
yading@11:                                 dest[0], dstW, lumAlpha, chrAlpha, dstY);
yading@11:                 } else { // general RGB
yading@11:                     yuv2packedX(c, vLumFilter + dstY * vLumFilterSize,
yading@11:                                 lumSrcPtr, vLumFilterSize,
yading@11:                                 vChrFilter + dstY * vChrFilterSize,
yading@11:                                 chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
yading@11:                                 alpSrcPtr, dest[0], dstW, dstY);
yading@11:                 }
yading@11:             } else {
yading@11:                 av_assert1(!yuv2packed1 && !yuv2packed2);
yading@11:                 yuv2anyX(c, vLumFilter + dstY * vLumFilterSize,
yading@11:                          lumSrcPtr, vLumFilterSize,
yading@11:                          vChrFilter + dstY * vChrFilterSize,
yading@11:                          chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
yading@11:                          alpSrcPtr, dest, dstW, dstY);
yading@11:             }
yading@11:         }
yading@11:     }
yading@11:     if (isPlanar(dstFormat) && isALPHA(dstFormat) && !alpPixBuf) {
yading@11:         int length = dstW;
yading@11:         int height = dstY - lastDstY;
yading@11: 
yading@11:         if (is16BPS(dstFormat) || isNBPS(dstFormat)) {
yading@11:             const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(dstFormat);
yading@11:             fillPlane16(dst[3], dstStride[3], length, height, lastDstY,
yading@11:                     1, desc->comp[3].depth_minus1,
yading@11:                     isBE(dstFormat));
yading@11:         } else
yading@11:             fillPlane(dst[3], dstStride[3], length, height, lastDstY, 255);
yading@11:     }
yading@11: 
yading@11: #if HAVE_MMXEXT_INLINE
yading@11:     if (av_get_cpu_flags() & AV_CPU_FLAG_MMXEXT)
yading@11:         __asm__ volatile ("sfence" ::: "memory");
yading@11: #endif
yading@11:     emms_c();
yading@11: 
yading@11:     /* store changed local vars back in the context */
yading@11:     c->dstY         = dstY;
yading@11:     c->lumBufIndex  = lumBufIndex;
yading@11:     c->chrBufIndex  = chrBufIndex;
yading@11:     c->lastInLumBuf = lastInLumBuf;
yading@11:     c->lastInChrBuf = lastInChrBuf;
yading@11: 
yading@11:     return dstY - lastDstY;
yading@11: }
yading@11: 
yading@11: static av_cold void sws_init_swScale_c(SwsContext *c)
yading@11: {
yading@11:     enum AVPixelFormat srcFormat = c->srcFormat;
yading@11: 
yading@11:     ff_sws_init_output_funcs(c, &c->yuv2plane1, &c->yuv2planeX,
yading@11:                              &c->yuv2nv12cX, &c->yuv2packed1,
yading@11:                              &c->yuv2packed2, &c->yuv2packedX, &c->yuv2anyX);
yading@11: 
yading@11:     ff_sws_init_input_funcs(c);
yading@11: 
yading@11: 
yading@11:     if (c->srcBpc == 8) {
yading@11:         if (c->dstBpc <= 14) {
yading@11:             c->hyScale = c->hcScale = hScale8To15_c;
yading@11:             if (c->flags & SWS_FAST_BILINEAR) {
yading@11:                 c->hyscale_fast = hyscale_fast_c;
yading@11:                 c->hcscale_fast = hcscale_fast_c;
yading@11:             }
yading@11:         } else {
yading@11:             c->hyScale = c->hcScale = hScale8To19_c;
yading@11:         }
yading@11:     } else {
yading@11:         c->hyScale = c->hcScale = c->dstBpc > 14 ? hScale16To19_c
yading@11:                                                  : hScale16To15_c;
yading@11:     }
yading@11: 
yading@11:     if (c->srcRange != c->dstRange && !isAnyRGB(c->dstFormat)) {
yading@11:         if (c->dstBpc <= 14) {
yading@11:             if (c->srcRange) {
yading@11:                 c->lumConvertRange = lumRangeFromJpeg_c;
yading@11:                 c->chrConvertRange = chrRangeFromJpeg_c;
yading@11:             } else {
yading@11:                 c->lumConvertRange = lumRangeToJpeg_c;
yading@11:                 c->chrConvertRange = chrRangeToJpeg_c;
yading@11:             }
yading@11:         } else {
yading@11:             if (c->srcRange) {
yading@11:                 c->lumConvertRange = lumRangeFromJpeg16_c;
yading@11:                 c->chrConvertRange = chrRangeFromJpeg16_c;
yading@11:             } else {
yading@11:                 c->lumConvertRange = lumRangeToJpeg16_c;
yading@11:                 c->chrConvertRange = chrRangeToJpeg16_c;
yading@11:             }
yading@11:         }
yading@11:     }
yading@11: 
yading@11:     if (!(isGray(srcFormat) || isGray(c->dstFormat) ||
yading@11:           srcFormat == AV_PIX_FMT_MONOBLACK || srcFormat == AV_PIX_FMT_MONOWHITE))
yading@11:         c->needs_hcscale = 1;
yading@11: }
yading@11: 
yading@11: SwsFunc ff_getSwsFunc(SwsContext *c)
yading@11: {
yading@11:     sws_init_swScale_c(c);
yading@11: 
yading@11:     if (HAVE_MMX)
yading@11:         ff_sws_init_swScale_mmx(c);
yading@11:     if (HAVE_ALTIVEC)
yading@11:         ff_sws_init_swScale_altivec(c);
yading@11: 
yading@11:     return swScale;
yading@11: }
yading@11: 
yading@11: static void reset_ptr(const uint8_t *src[], int format)
yading@11: {
yading@11:     if (!isALPHA(format))
yading@11:         src[3] = NULL;
yading@11:     if (!isPlanar(format)) {
yading@11:         src[3] = src[2] = NULL;
yading@11: 
yading@11:         if (!usePal(format))
yading@11:             src[1] = NULL;
yading@11:     }
yading@11: }
yading@11: 
yading@11: static int check_image_pointers(const uint8_t * const data[4], enum AVPixelFormat pix_fmt,
yading@11:                                 const int linesizes[4])
yading@11: {
yading@11:     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
yading@11:     int i;
yading@11: 
yading@11:     for (i = 0; i < 4; i++) {
yading@11:         int plane = desc->comp[i].plane;
yading@11:         if (!data[plane] || !linesizes[plane])
yading@11:             return 0;
yading@11:     }
yading@11: 
yading@11:     return 1;
yading@11: }
yading@11: 
yading@11: /**
yading@11:  * swscale wrapper, so we don't need to export the SwsContext.
yading@11:  * Assumes planar YUV to be in YUV order instead of YVU.
yading@11:  */
yading@11: int attribute_align_arg sws_scale(struct SwsContext *c,
yading@11:                                   const uint8_t * const srcSlice[],
yading@11:                                   const int srcStride[], int srcSliceY,
yading@11:                                   int srcSliceH, uint8_t *const dst[],
yading@11:                                   const int dstStride[])
yading@11: {
yading@11:     int i, ret;
yading@11:     const uint8_t *src2[4];
yading@11:     uint8_t *dst2[4];
yading@11:     uint8_t *rgb0_tmp = NULL;
yading@11: 
yading@11:     if (!srcSlice || !dstStride || !dst || !srcSlice) {
yading@11:         av_log(c, AV_LOG_ERROR, "One of the input parameters to sws_scale() is NULL, please check the calling code\n");
yading@11:         return 0;
yading@11:     }
yading@11:     memcpy(src2, srcSlice, sizeof(src2));
yading@11:     memcpy(dst2, dst, sizeof(dst2));
yading@11: 
yading@11:     // do not mess up sliceDir if we have a "trailing" 0-size slice
yading@11:     if (srcSliceH == 0)
yading@11:         return 0;
yading@11: 
yading@11:     if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) {
yading@11:         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
yading@11:         return 0;
yading@11:     }
yading@11:     if (!check_image_pointers((const uint8_t* const*)dst, c->dstFormat, dstStride)) {
yading@11:         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
yading@11:         return 0;
yading@11:     }
yading@11: 
yading@11:     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
yading@11:         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
yading@11:         return 0;
yading@11:     }
yading@11:     if (c->sliceDir == 0) {
yading@11:         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
yading@11:     }
yading@11: 
yading@11:     if (usePal(c->srcFormat)) {
yading@11:         for (i = 0; i < 256; i++) {
yading@11:             int p, r, g, b, y, u, v, a = 0xff;
yading@11:             if (c->srcFormat == AV_PIX_FMT_PAL8) {
yading@11:                 p = ((const uint32_t *)(srcSlice[1]))[i];
yading@11:                 a = (p >> 24) & 0xFF;
yading@11:                 r = (p >> 16) & 0xFF;
yading@11:                 g = (p >>  8) & 0xFF;
yading@11:                 b =  p        & 0xFF;
yading@11:             } else if (c->srcFormat == AV_PIX_FMT_RGB8) {
yading@11:                 r = ( i >> 5     ) * 36;
yading@11:                 g = ((i >> 2) & 7) * 36;
yading@11:                 b = ( i       & 3) * 85;
yading@11:             } else if (c->srcFormat == AV_PIX_FMT_BGR8) {
yading@11:                 b = ( i >> 6     ) * 85;
yading@11:                 g = ((i >> 3) & 7) * 36;
yading@11:                 r = ( i       & 7) * 36;
yading@11:             } else if (c->srcFormat == AV_PIX_FMT_RGB4_BYTE) {
yading@11:                 r = ( i >> 3     ) * 255;
yading@11:                 g = ((i >> 1) & 3) * 85;
yading@11:                 b = ( i       & 1) * 255;
yading@11:             } else if (c->srcFormat == AV_PIX_FMT_GRAY8 || c->srcFormat == AV_PIX_FMT_GRAY8A) {
yading@11:                 r = g = b = i;
yading@11:             } else {
yading@11:                 av_assert1(c->srcFormat == AV_PIX_FMT_BGR4_BYTE);
yading@11:                 b = ( i >> 3     ) * 255;
yading@11:                 g = ((i >> 1) & 3) * 85;
yading@11:                 r = ( i       & 1) * 255;
yading@11:             }
yading@11: #define RGB2YUV_SHIFT 15
yading@11: #define BY ( (int) (0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: #define BV (-(int) (0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: #define BU ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: #define GY ( (int) (0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: #define GV (-(int) (0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: #define GU (-(int) (0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: #define RY ( (int) (0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: #define RV ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: #define RU (-(int) (0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
yading@11: 
yading@11:             y = av_clip_uint8((RY * r + GY * g + BY * b + ( 33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
yading@11:             u = av_clip_uint8((RU * r + GU * g + BU * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
yading@11:             v = av_clip_uint8((RV * r + GV * g + BV * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
yading@11:             c->pal_yuv[i]= y + (u<<8) + (v<<16) + ((unsigned)a<<24);
yading@11: 
yading@11:             switch (c->dstFormat) {
yading@11:             case AV_PIX_FMT_BGR32:
yading@11: #if !HAVE_BIGENDIAN
yading@11:             case AV_PIX_FMT_RGB24:
yading@11: #endif
yading@11:                 c->pal_rgb[i]=  r + (g<<8) + (b<<16) + ((unsigned)a<<24);
yading@11:                 break;
yading@11:             case AV_PIX_FMT_BGR32_1:
yading@11: #if HAVE_BIGENDIAN
yading@11:             case AV_PIX_FMT_BGR24:
yading@11: #endif
yading@11:                 c->pal_rgb[i]= a + (r<<8) + (g<<16) + ((unsigned)b<<24);
yading@11:                 break;
yading@11:             case AV_PIX_FMT_RGB32_1:
yading@11: #if HAVE_BIGENDIAN
yading@11:             case AV_PIX_FMT_RGB24:
yading@11: #endif
yading@11:                 c->pal_rgb[i]= a + (b<<8) + (g<<16) + ((unsigned)r<<24);
yading@11:                 break;
yading@11:             case AV_PIX_FMT_RGB32:
yading@11: #if !HAVE_BIGENDIAN
yading@11:             case AV_PIX_FMT_BGR24:
yading@11: #endif
yading@11:             default:
yading@11:                 c->pal_rgb[i]=  b + (g<<8) + (r<<16) + ((unsigned)a<<24);
yading@11:             }
yading@11:         }
yading@11:     }
yading@11: 
yading@11:     if (c->src0Alpha && !c->dst0Alpha && isALPHA(c->dstFormat)) {
yading@11:         uint8_t *base;
yading@11:         int x,y;
yading@11:         rgb0_tmp = av_malloc(FFABS(srcStride[0]) * srcSliceH + 32);
yading@11:         base = srcStride[0] < 0 ? rgb0_tmp - srcStride[0] * (srcSliceH-1) : rgb0_tmp;
yading@11:         for (y=0; y<srcSliceH; y++){
yading@11:             memcpy(base + srcStride[0]*y, src2[0] + srcStride[0]*y, 4*c->srcW);
yading@11:             for (x=c->src0Alpha-1; x<4*c->srcW; x+=4) {
yading@11:                 base[ srcStride[0]*y + x] = 0xFF;
yading@11:             }
yading@11:         }
yading@11:         src2[0] = base;
yading@11:     }
yading@11: 
yading@11:     if (!srcSliceY && (c->flags & SWS_BITEXACT) && (c->flags & SWS_ERROR_DIFFUSION) && c->dither_error[0])
yading@11:         for (i = 0; i < 4; i++)
yading@11:             memset(c->dither_error[i], 0, sizeof(c->dither_error[0][0]) * (c->dstW+2));
yading@11: 
yading@11: 
yading@11:     // copy strides, so they can safely be modified
yading@11:     if (c->sliceDir == 1) {
yading@11:         // slices go from top to bottom
yading@11:         int srcStride2[4] = { srcStride[0], srcStride[1], srcStride[2],
yading@11:                               srcStride[3] };
yading@11:         int dstStride2[4] = { dstStride[0], dstStride[1], dstStride[2],
yading@11:                               dstStride[3] };
yading@11: 
yading@11:         reset_ptr(src2, c->srcFormat);
yading@11:         reset_ptr((void*)dst2, c->dstFormat);
yading@11: 
yading@11:         /* reset slice direction at end of frame */
yading@11:         if (srcSliceY + srcSliceH == c->srcH)
yading@11:             c->sliceDir = 0;
yading@11: 
yading@11:         ret = c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2,
yading@11:                           dstStride2);
yading@11:     } else {
yading@11:         // slices go from bottom to top => we flip the image internally
yading@11:         int srcStride2[4] = { -srcStride[0], -srcStride[1], -srcStride[2],
yading@11:                               -srcStride[3] };
yading@11:         int dstStride2[4] = { -dstStride[0], -dstStride[1], -dstStride[2],
yading@11:                               -dstStride[3] };
yading@11: 
yading@11:         src2[0] += (srcSliceH - 1) * srcStride[0];
yading@11:         if (!usePal(c->srcFormat))
yading@11:             src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];
yading@11:         src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];
yading@11:         src2[3] += (srcSliceH - 1) * srcStride[3];
yading@11:         dst2[0] += ( c->dstH                         - 1) * dstStride[0];
yading@11:         dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1];
yading@11:         dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2];
yading@11:         dst2[3] += ( c->dstH                         - 1) * dstStride[3];
yading@11: 
yading@11:         reset_ptr(src2, c->srcFormat);
yading@11:         reset_ptr((void*)dst2, c->dstFormat);
yading@11: 
yading@11:         /* reset slice direction at end of frame */
yading@11:         if (!srcSliceY)
yading@11:             c->sliceDir = 0;
yading@11: 
yading@11:         ret = c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH,
yading@11:                           srcSliceH, dst2, dstStride2);
yading@11:     }
yading@11: 
yading@11:     av_free(rgb0_tmp);
yading@11:     return ret;
yading@11: }
yading@11: