utils.c

    cpu_flags = av_get_cpu_flags();
    flags     = c->flags;
    emms_c();
    if (!rgb15to16)
        sws_rgb2rgb_init();

    unscaled = (srcW == dstW && srcH == dstH);

    c->srcRange |= handle_jpeg(&c->srcFormat);
    c->dstRange |= handle_jpeg(&c->dstFormat);

    if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat)
        av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");

    if (!c->contrast && !c->saturation && !c->dstFormatBpp)
        sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange,
                                 ff_yuv2rgb_coeffs[SWS_CS_DEFAULT],
                                 c->dstRange, 0, 1 << 16, 1 << 16);

    handle_formats(c);
    srcFormat = c->srcFormat;
    dstFormat = c->dstFormat;
    desc_src = av_pix_fmt_desc_get(srcFormat);
    desc_dst = av_pix_fmt_desc_get(dstFormat);

    if (!(unscaled && sws_isSupportedEndiannessConversion(srcFormat) &&
          av_pix_fmt_swap_endianness(srcFormat) == dstFormat)) {
    if (!sws_isSupportedInput(srcFormat)) {
        av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
               av_get_pix_fmt_name(srcFormat));
        return AVERROR(EINVAL);
    }
    if (!sws_isSupportedOutput(dstFormat)) {
        av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n",
               av_get_pix_fmt_name(dstFormat));
        return AVERROR(EINVAL);
    }
    }
    av_assert2(desc_src && desc_dst);

    i = flags & (SWS_POINT         |
                 SWS_AREA          |
                 SWS_BILINEAR      |
                 SWS_FAST_BILINEAR |
                 SWS_BICUBIC       |
                 SWS_X             |
                 SWS_GAUSS         |
                 SWS_LANCZOS       |
                 SWS_SINC          |
                 SWS_SPLINE        |
                 SWS_BICUBLIN);

    /* provide a default scaler if not set by caller */
    if (!i) {
        if (dstW < srcW && dstH < srcH)
            flags |= SWS_BICUBIC;
        else if (dstW > srcW && dstH > srcH)
            flags |= SWS_BICUBIC;
        else
            flags |= SWS_BICUBIC;
        c->flags = flags;
    } else if (i & (i - 1)) {
        av_log(c, AV_LOG_ERROR,
               "Exactly one scaler algorithm must be chosen, got %X\n", i);
        return AVERROR(EINVAL);
    }
    /* sanity check */
    if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) {
        /* FIXME check if these are enough and try to lower them after
         * fixing the relevant parts of the code */
        av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
               srcW, srcH, dstW, dstH);
        return AVERROR(EINVAL);
    }

    if (!dstFilter)
        dstFilter = &dummyFilter;
    if (!srcFilter)
        srcFilter = &dummyFilter;

    c->lumXInc      = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
    c->lumYInc      = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH;
    c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
    c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
    c->vRounder     = 4 * 0x0001000100010001ULL;

    usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) ||
                  (srcFilter->chrV && srcFilter->chrV->length > 1) ||
                  (dstFilter->lumV && dstFilter->lumV->length > 1) ||
                  (dstFilter->chrV && dstFilter->chrV->length > 1);
    usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) ||
                  (srcFilter->chrH && srcFilter->chrH->length > 1) ||
                  (dstFilter->lumH && dstFilter->lumH->length > 1) ||
                  (dstFilter->chrH && dstFilter->chrH->length > 1);

    av_pix_fmt_get_chroma_sub_sample(srcFormat, &c->chrSrcHSubSample, &c->chrSrcVSubSample);
    av_pix_fmt_get_chroma_sub_sample(dstFormat, &c->chrDstHSubSample, &c->chrDstVSubSample);

    if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) {
        if (dstW&1) {
            av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
            flags |= SWS_FULL_CHR_H_INT;
            c->flags = flags;
        }

        if (   c->chrSrcHSubSample == 0
            && c->chrSrcVSubSample == 0
            && c->dither != SWS_DITHER_BAYER //SWS_FULL_CHR_H_INT is currently not supported with SWS_DITHER_BAYER
            && !(c->flags & SWS_FAST_BILINEAR)
        ) {
            av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to input having non subsampled chroma\n");
            flags |= SWS_FULL_CHR_H_INT;
            c->flags = flags;
        }
    }

    if (c->dither == SWS_DITHER_AUTO) {
        if (flags & SWS_ERROR_DIFFUSION)
            c->dither = SWS_DITHER_ED;
    }

    if(dstFormat == AV_PIX_FMT_BGR4_BYTE ||
       dstFormat == AV_PIX_FMT_RGB4_BYTE ||
       dstFormat == AV_PIX_FMT_BGR8 ||
       dstFormat == AV_PIX_FMT_RGB8) {
        if (c->dither == SWS_DITHER_AUTO)
            c->dither = (flags & SWS_FULL_CHR_H_INT) ? SWS_DITHER_ED : SWS_DITHER_BAYER;
        if (!(flags & SWS_FULL_CHR_H_INT)) {
            if (c->dither == SWS_DITHER_ED || c->dither == SWS_DITHER_A_DITHER || c->dither == SWS_DITHER_X_DITHER) {
                av_log(c, AV_LOG_DEBUG,
                    "Desired dithering only supported in full chroma interpolation for destination format '%s'\n",
                    av_get_pix_fmt_name(dstFormat));
                flags   |= SWS_FULL_CHR_H_INT;
                c->flags = flags;
            }
        }
        if (flags & SWS_FULL_CHR_H_INT) {
            if (c->dither == SWS_DITHER_BAYER) {
                av_log(c, AV_LOG_DEBUG,
                    "Ordered dither is not supported in full chroma interpolation for destination format '%s'\n",
                    av_get_pix_fmt_name(dstFormat));
                c->dither = SWS_DITHER_ED;
            }
        }
    }
    if (isPlanarRGB(dstFormat)) {
        if (!(flags & SWS_FULL_CHR_H_INT)) {
            av_log(c, AV_LOG_DEBUG,
                   "%s output is not supported with half chroma resolution, switching to full\n",
                   av_get_pix_fmt_name(dstFormat));
            flags   |= SWS_FULL_CHR_H_INT;
            c->flags = flags;
        }
    }

    /* reuse chroma for 2 pixels RGB/BGR unless user wants full
     * chroma interpolation */
    if (flags & SWS_FULL_CHR_H_INT &&
        isAnyRGB(dstFormat)        &&
        !isPlanarRGB(dstFormat)    &&
        dstFormat != AV_PIX_FMT_RGBA64LE &&
        dstFormat != AV_PIX_FMT_RGBA64BE &&
        dstFormat != AV_PIX_FMT_BGRA64LE &&
        dstFormat != AV_PIX_FMT_BGRA64BE &&
        dstFormat != AV_PIX_FMT_RGB48LE &&
        dstFormat != AV_PIX_FMT_RGB48BE &&
        dstFormat != AV_PIX_FMT_BGR48LE &&
        dstFormat != AV_PIX_FMT_BGR48BE &&
        dstFormat != AV_PIX_FMT_RGBA  &&
        dstFormat != AV_PIX_FMT_ARGB  &&
        dstFormat != AV_PIX_FMT_BGRA  &&
        dstFormat != AV_PIX_FMT_ABGR  &&
        dstFormat != AV_PIX_FMT_RGB24 &&
        dstFormat != AV_PIX_FMT_BGR24 &&
        dstFormat != AV_PIX_FMT_BGR4_BYTE &&
        dstFormat != AV_PIX_FMT_RGB4_BYTE &&
        dstFormat != AV_PIX_FMT_BGR8 &&
        dstFormat != AV_PIX_FMT_RGB8
    ) {
        av_log(c, AV_LOG_WARNING,
               "full chroma interpolation for destination format '%s' not yet implemented\n",
               av_get_pix_fmt_name(dstFormat));
        flags   &= ~SWS_FULL_CHR_H_INT;
        c->flags = flags;
    }
    if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
        c->chrDstHSubSample = 1;

    // drop some chroma lines if the user wants it
    c->vChrDrop          = (flags & SWS_SRC_V_CHR_DROP_MASK) >>
                           SWS_SRC_V_CHR_DROP_SHIFT;
    c->chrSrcVSubSample += c->vChrDrop;

    /* drop every other pixel for chroma calculation unless user
     * wants full chroma */
    if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP)   &&
        srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 &&
        srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 &&
        srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE &&
        srcFormat != AV_PIX_FMT_GBRP9BE   && srcFormat != AV_PIX_FMT_GBRP9LE  &&
        srcFormat != AV_PIX_FMT_GBRP10BE  && srcFormat != AV_PIX_FMT_GBRP10LE &&
        srcFormat != AV_PIX_FMT_GBRP12BE  && srcFormat != AV_PIX_FMT_GBRP12LE &&
        srcFormat != AV_PIX_FMT_GBRP14BE  && srcFormat != AV_PIX_FMT_GBRP14LE &&
        srcFormat != AV_PIX_FMT_GBRP16BE  && srcFormat != AV_PIX_FMT_GBRP16LE &&
        ((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
         (flags & SWS_FAST_BILINEAR)))
        c->chrSrcHSubSample = 1;

    // Note the FF_CEIL_RSHIFT is so that we always round toward +inf.
    c->chrSrcW = FF_CEIL_RSHIFT(srcW, c->chrSrcHSubSample);
    c->chrSrcH = FF_CEIL_RSHIFT(srcH, c->chrSrcVSubSample);
    c->chrDstW = FF_CEIL_RSHIFT(dstW, c->chrDstHSubSample);
    c->chrDstH = FF_CEIL_RSHIFT(dstH, c->chrDstVSubSample);

    FF_ALLOCZ_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);

    c->srcBpc = 1 + desc_src->comp[0].depth_minus1;
    if (c->srcBpc < 8)
        c->srcBpc = 8;
    c->dstBpc = 1 + desc_dst->comp[0].depth_minus1;
    if (c->dstBpc < 8)
        c->dstBpc = 8;
    if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8)
        c->srcBpc = 16;
    if (c->dstBpc == 16)
        dst_stride <<= 1;

    if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) {
        c->canMMXEXTBeUsed = dstW >= srcW && (dstW & 31) == 0 &&
                             c->chrDstW >= c->chrSrcW &&
                             (srcW & 15) == 0;
        if (!c->canMMXEXTBeUsed && dstW >= srcW && c->chrDstW >= c->chrSrcW && (srcW & 15) == 0

            && (flags & SWS_FAST_BILINEAR)) {
            if (flags & SWS_PRINT_INFO)
                av_log(c, AV_LOG_INFO,
                       "output width is not a multiple of 32 -> no MMXEXT scaler\n");
        }
        if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat))
            c->canMMXEXTBeUsed = 0;
    } else
        c->canMMXEXTBeUsed = 0;

    c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW;
    c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH;

    /* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src
     * to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do
     * correct scaling.
     * n-2 is the last chrominance sample available.
     * This is not perfect, but no one should notice the difference, the more
     * correct variant would be like the vertical one, but that would require
     * some special code for the first and last pixel */
    if (flags & SWS_FAST_BILINEAR) {
        if (c->canMMXEXTBeUsed) {
            c->lumXInc += 20;
            c->chrXInc += 20;
        }
        // we don't use the x86 asm scaler if MMX is available
        else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) {
            c->lumXInc = ((int64_t)(srcW       - 2) << 16) / (dstW       - 2) - 20;
            c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
        }
    }

    // hardcoded for now
    c->gamma_value = 2.2;
    tmpFmt = AV_PIX_FMT_RGBA64LE;


    if (!unscaled && c->gamma_flag && (srcFormat != tmpFmt || dstFormat != tmpFmt)) {
        SwsContext *c2;
        c->cascaded_context[0] = NULL;

        ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
                            srcW, srcH, tmpFmt, 64);
        if (ret < 0)
            return ret;

        c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
                                                srcW, srcH, tmpFmt,
                                                flags, NULL, NULL, c->param);
        if (!c->cascaded_context[0]) {
            return -1;
        }

        c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFmt,
                                                dstW, dstH, tmpFmt,
                                                flags, srcFilter, dstFilter, c->param);

        if (!c->cascaded_context[1])
            return -1;

        c2 = c->cascaded_context[1];
        c2->is_internal_gamma = 1;
        c2->gamma     = alloc_gamma_tbl(    c->gamma_value);
        c2->inv_gamma = alloc_gamma_tbl(1.f/c->gamma_value);
        if (!c2->gamma || !c2->inv_gamma)
            return AVERROR(ENOMEM);

        c->cascaded_context[2] = NULL;
        if (dstFormat != tmpFmt) {
            ret = av_image_alloc(c->cascaded1_tmp, c->cascaded1_tmpStride,
                                dstW, dstH, tmpFmt, 64);
            if (ret < 0)
                return ret;

            c->cascaded_context[2] = sws_getContext(dstW, dstH, tmpFmt,
                                                dstW, dstH, dstFormat,
                                                flags, NULL, NULL, c->param);
            if (!c->cascaded_context[2])
                return -1;
        }
        return 0;
    }

    if (isBayer(srcFormat)) {
        if (!unscaled ||
            (dstFormat != AV_PIX_FMT_RGB24 && dstFormat != AV_PIX_FMT_YUV420P)) {
            enum AVPixelFormat tmpFormat = AV_PIX_FMT_RGB24;

            ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
                                srcW, srcH, tmpFormat, 64);
            if (ret < 0)
                return ret;

            c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
                                                    srcW, srcH, tmpFormat,
                                                    flags, srcFilter, NULL, c->param);
            if (!c->cascaded_context[0])
                return -1;

            c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFormat,
                                                    dstW, dstH, dstFormat,
                                                    flags, NULL, dstFilter, c->param);
            if (!c->cascaded_context[1])
                return -1;
            return 0;
        }
    }

#define USE_MMAP (HAVE_MMAP && HAVE_MPROTECT && defined MAP_ANONYMOUS)

    /* precalculate horizontal scaler filter coefficients */
    {
#if HAVE_MMXEXT_INLINE
// can't downscale !!!
        if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) {
            c->lumMmxextFilterCodeSize = ff_init_hscaler_mmxext(dstW, c->lumXInc, NULL,
                                                             NULL, NULL, 8);
            c->chrMmxextFilterCodeSize = ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc,
                                                             NULL, NULL, NULL, 4);

#if USE_MMAP
            c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize,
                                          PROT_READ | PROT_WRITE,
                                          MAP_PRIVATE | MAP_ANONYMOUS,
                                          -1, 0);
            c->chrMmxextFilterCode = mmap(NULL, c->chrMmxextFilterCodeSize,
                                          PROT_READ | PROT_WRITE,
                                          MAP_PRIVATE | MAP_ANONYMOUS,
                                          -1, 0);
#elif HAVE_VIRTUALALLOC
            c->lumMmxextFilterCode = VirtualAlloc(NULL,
                                                  c->lumMmxextFilterCodeSize,
                                                  MEM_COMMIT,
                                                  PAGE_EXECUTE_READWRITE);
            c->chrMmxextFilterCode = VirtualAlloc(NULL,
                                                  c->chrMmxextFilterCodeSize,
                                                  MEM_COMMIT,
                                                  PAGE_EXECUTE_READWRITE);
#else
            c->lumMmxextFilterCode = av_malloc(c->lumMmxextFilterCodeSize);
            c->chrMmxextFilterCode = av_malloc(c->chrMmxextFilterCodeSize);
#endif

#ifdef MAP_ANONYMOUS
            if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED)
#else
            if (!c->lumMmxextFilterCode || !c->chrMmxextFilterCode)
#endif
            {
                av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n");
                return AVERROR(ENOMEM);
            }

            FF_ALLOCZ_OR_GOTO(c, c->hLumFilter,    (dstW           / 8 + 8) * sizeof(int16_t), fail);
            FF_ALLOCZ_OR_GOTO(c, c->hChrFilter,    (c->chrDstW     / 4 + 8) * sizeof(int16_t), fail);
            FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW       / 2 / 8 + 8) * sizeof(int32_t), fail);
            FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail);

            ff_init_hscaler_mmxext(      dstW, c->lumXInc, c->lumMmxextFilterCode,
                                c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
            ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
                                c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);

#if USE_MMAP
            if (   mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1
                || mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1) {
                av_log(c, AV_LOG_ERROR, "mprotect failed, cannot use fast bilinear scaler\n");
                goto fail;
            }
#endif
        } else
#endif /* HAVE_MMXEXT_INLINE */
        {
            const int filterAlign = X86_MMX(cpu_flags)     ? 4 :
                                    PPC_ALTIVEC(cpu_flags) ? 8 : 1;

            if ((ret = initFilter(&c->hLumFilter, &c->hLumFilterPos,
                           &c->hLumFilterSize, c->lumXInc,
                           srcW, dstW, filterAlign, 1 << 14,
                           (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
                           cpu_flags, srcFilter->lumH, dstFilter->lumH,
                           c->param,
                           get_local_pos(c, 0, 0, 0),
                           get_local_pos(c, 0, 0, 0))) < 0)
                goto fail;
            if ((ret = initFilter(&c->hChrFilter, &c->hChrFilterPos,
                           &c->hChrFilterSize, c->chrXInc,
                           c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
                           (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
                           cpu_flags, srcFilter->chrH, dstFilter->chrH,
                           c->param,
                           get_local_pos(c, c->chrSrcHSubSample, c->src_h_chr_pos, 0),
                           get_local_pos(c, c->chrDstHSubSample, c->dst_h_chr_pos, 0))) < 0)
                goto fail;
        }
    } // initialize horizontal stuff

    /* precalculate vertical scaler filter coefficients */
    {
        const int filterAlign = X86_MMX(cpu_flags)     ? 2 :
                                PPC_ALTIVEC(cpu_flags) ? 8 : 1;

        if ((ret = initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize,
                       c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
                       (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
                       cpu_flags, srcFilter->lumV, dstFilter->lumV,
                       c->param,
                       get_local_pos(c, 0, 0, 1),
                       get_local_pos(c, 0, 0, 1))) < 0)
            goto fail;
        if ((ret = initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize,
                       c->chrYInc, c->chrSrcH, c->chrDstH,
                       filterAlign, (1 << 12),
                       (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
                       cpu_flags, srcFilter->chrV, dstFilter->chrV,
                       c->param,
                       get_local_pos(c, c->chrSrcVSubSample, c->src_v_chr_pos, 1),
                       get_local_pos(c, c->chrDstVSubSample, c->dst_v_chr_pos, 1))) < 0)

            goto fail;

#if HAVE_ALTIVEC
        FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof(vector signed short) * c->vLumFilterSize * c->dstH,    fail);
        FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof(vector signed short) * c->vChrFilterSize * c->chrDstH, fail);

        for (i = 0; i < c->vLumFilterSize * c->dstH; i++) {
            int j;
            short *p = (short *)&c->vYCoeffsBank[i];
            for (j = 0; j < 8; j++)
                p[j] = c->vLumFilter[i];
        }

        for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) {
            int j;
            short *p = (short *)&c->vCCoeffsBank[i];
            for (j = 0; j < 8; j++)
                p[j] = c->vChrFilter[i];
        }
#endif
    }

    // calculate buffer sizes so that they won't run out while handling these damn slices
    c->vLumBufSize = c->vLumFilterSize;
    c->vChrBufSize = c->vChrFilterSize;
    for (i = 0; i < dstH; i++) {
        int chrI      = (int64_t)i * c->chrDstH / dstH;
        int nextSlice = FFMAX(c->vLumFilterPos[i] + c->vLumFilterSize - 1,
                              ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)
                               << c->chrSrcVSubSample));

        nextSlice >>= c->chrSrcVSubSample;
        nextSlice <<= c->chrSrcVSubSample;
        if (c->vLumFilterPos[i] + c->vLumBufSize < nextSlice)
            c->vLumBufSize = nextSlice - c->vLumFilterPos[i];
        if (c->vChrFilterPos[chrI] + c->vChrBufSize <
            (nextSlice >> c->chrSrcVSubSample))
            c->vChrBufSize = (nextSlice >> c->chrSrcVSubSample) -
                             c->vChrFilterPos[chrI];
    }

    for (i = 0; i < 4; i++)
        FF_ALLOCZ_OR_GOTO(c, c->dither_error[i], (c->dstW+2) * sizeof(int), fail);

    /* Allocate pixbufs (we use dynamic allocation because otherwise we would
     * need to allocate several megabytes to handle all possible cases) */
    FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf,  c->vLumBufSize * 3 * sizeof(int16_t *), fail);
    FF_ALLOCZ_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
    FF_ALLOCZ_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
    if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
        FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
    /* Note we need at least one pixel more at the end because of the MMX code
     * (just in case someone wants to replace the 4000/8000). */
    /* align at 16 bytes for AltiVec */
    for (i = 0; i < c->vLumBufSize; i++) {
        FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i + c->vLumBufSize],
                          dst_stride + 16, fail);
        c->lumPixBuf[i] = c->lumPixBuf[i + c->vLumBufSize];
    }
    // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
    c->uv_off   = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
    c->uv_offx2 = dst_stride + 16;
    for (i = 0; i < c->vChrBufSize; i++) {
        FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i + c->vChrBufSize],
                         dst_stride * 2 + 32, fail);
        c->chrUPixBuf[i] = c->chrUPixBuf[i + c->vChrBufSize];
        c->chrVPixBuf[i] = c->chrVPixBuf[i + c->vChrBufSize]
                         = c->chrUPixBuf[i] + (dst_stride >> 1) + 8;
    }
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
        for (i = 0; i < c->vLumBufSize; i++) {
            FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i + c->vLumBufSize],
                              dst_stride + 16, fail);
            c->alpPixBuf[i] = c->alpPixBuf[i + c->vLumBufSize];
        }

    // try to avoid drawing green stuff between the right end and the stride end
    for (i = 0; i < c->vChrBufSize; i++)
        if(desc_dst->comp[0].depth_minus1 == 15){
            av_assert0(c->dstBpc > 14);
            for(j=0; j<dst_stride/2+1; j++)
                ((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
        } else
            for(j=0; j<dst_stride+1; j++)
                ((int16_t*)(c->chrUPixBuf[i]))[j] = 1<<14;

    av_assert0(c->chrDstH <= dstH);

    if (flags & SWS_PRINT_INFO) {
        const char *scaler = NULL, *cpucaps;

        for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
            if (flags & scale_algorithms[i].flag) {
                scaler = scale_algorithms[i].description;
                break;
            }
        }
        if (!scaler)
            scaler =  "ehh flags invalid?!";
        av_log(c, AV_LOG_INFO, "%s scaler, from %s to %s%s ",
               scaler,
               av_get_pix_fmt_name(srcFormat),
#ifdef DITHER1XBPP
               dstFormat == AV_PIX_FMT_BGR555   || dstFormat == AV_PIX_FMT_BGR565   ||
               dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE ||
               dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ?
                                                             "dithered " : "",
#else
               "",
#endif
               av_get_pix_fmt_name(dstFormat));

        if (INLINE_MMXEXT(cpu_flags))
            cpucaps = "MMXEXT";
        else if (INLINE_AMD3DNOW(cpu_flags))
            cpucaps = "3DNOW";
        else if (INLINE_MMX(cpu_flags))
            cpucaps = "MMX";
        else if (PPC_ALTIVEC(cpu_flags))
            cpucaps = "AltiVec";
        else
            cpucaps = "C";

        av_log(c, AV_LOG_INFO, "using %s\n", cpucaps);

        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
        av_log(c, AV_LOG_DEBUG,
               "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
               c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
        av_log(c, AV_LOG_DEBUG,
               "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
               c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH,
               c->chrXInc, c->chrYInc);
    }

    /* unscaled special cases */
    if (unscaled && !usesHFilter && !usesVFilter &&
        (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
        ff_get_unscaled_swscale(c);

        if (c->swscale) {
            if (flags & SWS_PRINT_INFO)
                av_log(c, AV_LOG_INFO,
                       "using unscaled %s -> %s special converter\n",
                       av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
            return 0;
        }
    }

    c->swscale = ff_getSwsFunc(c);
    return 0;
fail: // FIXME replace things by appropriate error codes
    if (ret == RETCODE_USE_CASCADE)  {
        int tmpW = sqrt(srcW * (int64_t)dstW);
        int tmpH = sqrt(srcH * (int64_t)dstH);
        enum AVPixelFormat tmpFormat = AV_PIX_FMT_YUV420P;

        if (srcW*(int64_t)srcH <= 4LL*dstW*dstH)
            return AVERROR(EINVAL);

        ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
                             tmpW, tmpH, tmpFormat, 64);
        if (ret < 0)
            return ret;

        c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
                                                tmpW, tmpH, tmpFormat,
                                                flags, srcFilter, NULL, c->param);
        if (!c->cascaded_context[0])
            return -1;

        c->cascaded_context[1] = sws_getContext(tmpW, tmpH, tmpFormat,
                                                dstW, dstH, dstFormat,
                                                flags, NULL, dstFilter, c->param);
        if (!c->cascaded_context[1])
            return -1;
        return 0;
    }
    return -1;
}

SwsContext *sws_alloc_set_opts(int srcW, int srcH, enum AVPixelFormat srcFormat,
                               int dstW, int dstH, enum AVPixelFormat dstFormat,
                               int flags, const double *param)
{
    SwsContext *c;

    if (!(c = sws_alloc_context()))
        return NULL;

    c->flags     = flags;
    c->srcW      = srcW;
    c->srcH      = srcH;
    c->dstW      = dstW;
    c->dstH      = dstH;
    c->srcFormat = srcFormat;
    c->dstFormat = dstFormat;

    if (param) {
        c->param[0] = param[0];
        c->param[1] = param[1];
    }

    return c;
}

SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
                           int dstW, int dstH, enum AVPixelFormat dstFormat,
                           int flags, SwsFilter *srcFilter,
                           SwsFilter *dstFilter, const double *param)
{
    SwsContext *c;

    c = sws_alloc_set_opts(srcW, srcH, srcFormat,
                           dstW, dstH, dstFormat,
                           flags, param);
    if (!c)
        return NULL;

    if (sws_init_context(c, srcFilter, dstFilter) < 0) {
        sws_freeContext(c);
        return NULL;
    }

    return c;
}

static int isnan_vec(SwsVector *a)
{
    int i;
    for (i=0; i<a->length; i++)
        if (isnan(a->coeff[i]))
            return 1;
    return 0;
}

static void makenan_vec(SwsVector *a)
{
    int i;
    for (i=0; i<a->length; i++)
        a->coeff[i] = NAN;
}

SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
                                float lumaSharpen, float chromaSharpen,
                                float chromaHShift, float chromaVShift,
                                int verbose)
{
    SwsFilter *filter = av_malloc(sizeof(SwsFilter));
    if (!filter)
        return NULL;

    if (lumaGBlur != 0.0) {
        filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
        filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0);
    } else {
        filter->lumH = sws_getIdentityVec();
        filter->lumV = sws_getIdentityVec();
    }

    if (chromaGBlur != 0.0) {
        filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
        filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0);
    } else {
        filter->chrH = sws_getIdentityVec();
        filter->chrV = sws_getIdentityVec();
    }

    if (!filter->lumH || !filter->lumV || !filter->chrH || !filter->chrV)
        goto fail;

    if (chromaSharpen != 0.0) {
        SwsVector *id = sws_getIdentityVec();
        if (!id)
            goto fail;
        sws_scaleVec(filter->chrH, -chromaSharpen);
        sws_scaleVec(filter->chrV, -chromaSharpen);
        sws_addVec(filter->chrH, id);
        sws_addVec(filter->chrV, id);
        sws_freeVec(id);
    }

    if (lumaSharpen != 0.0) {
        SwsVector *id = sws_getIdentityVec();
        if (!id)
            goto fail;
        sws_scaleVec(filter->lumH, -lumaSharpen);
        sws_scaleVec(filter->lumV, -lumaSharpen);
        sws_addVec(filter->lumH, id);
        sws_addVec(filter->lumV, id);
        sws_freeVec(id);
    }

    if (chromaHShift != 0.0)
        sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5));

    if (chromaVShift != 0.0)
        sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5));

    sws_normalizeVec(filter->chrH, 1.0);
    sws_normalizeVec(filter->chrV, 1.0);
    sws_normalizeVec(filter->lumH, 1.0);
    sws_normalizeVec(filter->lumV, 1.0);

    if (isnan_vec(filter->chrH) ||
        isnan_vec(filter->chrV) ||
        isnan_vec(filter->lumH) ||
        isnan_vec(filter->lumV))
        goto fail;

    if (verbose)
        sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
    if (verbose)
        sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);

    return filter;

fail:
    sws_freeVec(filter->lumH);
    sws_freeVec(filter->lumV);
    sws_freeVec(filter->chrH);
    sws_freeVec(filter->chrV);
    av_freep(&filter);
    return NULL;
}

SwsVector *sws_allocVec(int length)
{
    SwsVector *vec;

    if(length <= 0 || length > INT_MAX/ sizeof(double))
        return NULL;

    vec = av_malloc(sizeof(SwsVector));
    if (!vec)
        return NULL;
    vec->length = length;
    vec->coeff  = av_malloc(sizeof(double) * length);
    if (!vec->coeff)
        av_freep(&vec);
    return vec;
}

SwsVector *sws_getGaussianVec(double variance, double quality)
{
    const int length = (int)(variance * quality + 0.5) | 1;
    int i;
    double middle  = (length - 1) * 0.5;
    SwsVector *vec;

    if(variance < 0 || quality < 0)
        return NULL;

    vec = sws_allocVec(length);

    if (!vec)
        return NULL;

    for (i = 0; i < length; i++) {
        double dist = i - middle;
        vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
                        sqrt(2 * variance * M_PI);
    }

    sws_normalizeVec(vec, 1.0);

    return vec;
}

SwsVector *sws_getConstVec(double c, int length)
{
    int i;
    SwsVector *vec = sws_allocVec(length);

    if (!vec)
        return NULL;

    for (i = 0; i < length; i++)
        vec->coeff[i] = c;

    return vec;
}

SwsVector *sws_getIdentityVec(void)
{
    return sws_getConstVec(1.0, 1);
}

static double sws_dcVec(SwsVector *a)
{
    int i;
    double sum = 0;

    for (i = 0; i < a->length; i++)
        sum += a->coeff[i];

    return sum;
}

void sws_scaleVec(SwsVector *a, double scalar)
{
    int i;

    for (i = 0; i < a->length; i++)
        a->coeff[i] *= scalar;
}

void sws_normalizeVec(SwsVector *a, double height)
{
    sws_scaleVec(a, height / sws_dcVec(a));
}

static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
{
    int length = a->length + b->length - 1;
    int i, j;
    SwsVector *vec = sws_getConstVec(0.0, length);

    if (!vec)
        return NULL;

    for (i = 0; i < a->length; i++) {
        for (j = 0; j < b->length; j++) {
            vec->coeff[i + j] += a->coeff[i] * b->coeff[j];
        }
    }

    return vec;
}

static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
{
    int length = FFMAX(a->length, b->length);
    int i;
    SwsVector *vec = sws_getConstVec(0.0, length);

    if (!vec)
        return NULL;

    for (i = 0; i < a->length; i++)
        vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
    for (i = 0; i < b->length; i++)
        vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i];

    return vec;
}

static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
{
    int length = FFMAX(a->length, b->length);
    int i;
    SwsVector *vec = sws_getConstVec(0.0, length);

    if (!vec)
        return NULL;

    for (i = 0; i < a->length; i++)
        vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
    for (i = 0; i < b->length; i++)
        vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] -= b->coeff[i];

    return vec;
}

/* shift left / or right if "shift" is negative */
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
{
    int length = a->length + FFABS(shift) * 2;
    int i;
    SwsVector *vec = sws_getConstVec(0.0, length);

    if (!vec)
        return NULL;

    for (i = 0; i < a->length; i++) {
        vec->coeff[i + (length    - 1) / 2 -
                       (a->length - 1) / 2 - shift] = a->coeff[i];
    }

    return vec;
}

void sws_shiftVec(SwsVector *a, int shift)
{
    SwsVector *shifted = sws_getShiftedVec(a, shift);
    if (!shifted) {
        makenan_vec(a);
        return;
    }
    av_free(a->coeff);
    a->coeff  = shifted->coeff;
    a->length = shifted->length;
    av_free(shifted);
}

void sws_addVec(SwsVector *a, SwsVector *b)
{
    SwsVector *sum = sws_sumVec(a, b);
    if (!sum) {
        makenan_vec(a);
        return;
    }
    av_free(a->coeff);
    a->coeff  = sum->coeff;
    a->length = sum->length;
    av_free(sum);
}

void sws_subVec(SwsVector *a, SwsVector *b)
{
    SwsVector *diff = sws_diffVec(a, b);
    if (!diff) {
        makenan_vec(a);
        return;
    }
    av_free(a->coeff);
    a->coeff  = diff->coeff;
    a->length = diff->length;
    av_free(diff);
}

void sws_convVec(SwsVector *a, SwsVector *b)
{
    SwsVector *conv = sws_getConvVec(a, b);
    if (!conv) {
        makenan_vec(a);
        return;
    }
    av_free(a->coeff);
    a->coeff  = conv->coeff;
    a->length = conv->length;
    av_free(conv);
}

SwsVector *sws_cloneVec(SwsVector *a)
{
    SwsVector *vec = sws_allocVec(a->length);

    if (!vec)
        return NULL;

    memcpy(vec->coeff, a->coeff, a->length * sizeof(*a->coeff));

    return vec;
}

void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)