Split swscale.c into scaler code (swscale.c) and utility code (utils.c).

Originally committed as revision 30411 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale

4 files changed, 1620 insertions, 1539 deletions

diff --git a/libswscale/Makefile b/libswscale/Makefile
index 247ae8d239..085027789e 100644
--- a/libswscale/Makefile
+++ b/libswscale/Makefile
@@ -5,7 +5,7 @@ FFLIBS = avutil
 
 HEADERS = swscale.h
 
-OBJS = options.o rgb2rgb.o swscale.o yuv2rgb.o
+OBJS = options.o rgb2rgb.o swscale.o utils.o yuv2rgb.o
 
 OBJS-$(ARCH_BFIN)          +=  bfin/internal_bfin.o     \
                                bfin/swscale_bfin.o      \
diff --git a/libswscale/swscale.c b/libswscale/swscale.c
index 67a9364c97..fb46dcd5fe 100644
--- a/libswscale/swscale.c
+++ b/libswscale/swscale.c
@@ -54,23 +54,12 @@ untested special converters
   BGR24 -> YV12
 */
 
-#define _SVID_SOURCE //needed for MAP_ANONYMOUS
 #include <inttypes.h>
 #include <string.h>
 #include <math.h>
 #include <stdio.h>
 #include "config.h"
 #include <assert.h>
-#if HAVE_SYS_MMAN_H
-#include <sys/mman.h>
-#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
-#define MAP_ANONYMOUS MAP_ANON
-#endif
-#endif
-#if HAVE_VIRTUALALLOC
-#define WIN32_LEAN_AND_MEAN
-#include <windows.h>
-#endif
 #include "swscale.h"
 #include "swscale_internal.h"
 #include "rgb2rgb.h"
@@ -80,22 +69,6 @@ untested special converters
 #include "libavutil/bswap.h"
 #include "libavutil/pixdesc.h"
 
-unsigned swscale_version(void)
-{
-    return LIBSWSCALE_VERSION_INT;
-}
-
-const char *swscale_configuration(void)
-{
-    return FFMPEG_CONFIGURATION;
-}
-
-const char *swscale_license(void)
-{
-#define LICENSE_PREFIX "libswscale license: "
-    return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
-}
-
 #undef MOVNTQ
 #undef PAVGB
 
@@ -107,92 +80,12 @@ const char *swscale_license(void)
 
 #define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
 
-#define RET 0xC3 //near return opcode for x86
-
 #ifdef M_PI
 #define PI M_PI
 #else
 #define PI 3.14159265358979323846
 #endif
 
-#define isSupportedIn(x)    (       \
-           (x)==PIX_FMT_YUV420P     \
-        || (x)==PIX_FMT_YUVA420P    \
-        || (x)==PIX_FMT_YUYV422     \
-        || (x)==PIX_FMT_UYVY422     \
-        || (x)==PIX_FMT_RGB48BE     \
-        || (x)==PIX_FMT_RGB48LE     \
-        || (x)==PIX_FMT_RGB32       \
-        || (x)==PIX_FMT_RGB32_1     \
-        || (x)==PIX_FMT_BGR24       \
-        || (x)==PIX_FMT_BGR565      \
-        || (x)==PIX_FMT_BGR555      \
-        || (x)==PIX_FMT_BGR32       \
-        || (x)==PIX_FMT_BGR32_1     \
-        || (x)==PIX_FMT_RGB24       \
-        || (x)==PIX_FMT_RGB565      \
-        || (x)==PIX_FMT_RGB555      \
-        || (x)==PIX_FMT_GRAY8       \
-        || (x)==PIX_FMT_YUV410P     \
-        || (x)==PIX_FMT_YUV440P     \
-        || (x)==PIX_FMT_NV12        \
-        || (x)==PIX_FMT_NV21        \
-        || (x)==PIX_FMT_GRAY16BE    \
-        || (x)==PIX_FMT_GRAY16LE    \
-        || (x)==PIX_FMT_YUV444P     \
-        || (x)==PIX_FMT_YUV422P     \
-        || (x)==PIX_FMT_YUV411P     \
-        || (x)==PIX_FMT_PAL8        \
-        || (x)==PIX_FMT_BGR8        \
-        || (x)==PIX_FMT_RGB8        \
-        || (x)==PIX_FMT_BGR4_BYTE   \
-        || (x)==PIX_FMT_RGB4_BYTE   \
-        || (x)==PIX_FMT_YUV440P     \
-        || (x)==PIX_FMT_MONOWHITE   \
-        || (x)==PIX_FMT_MONOBLACK   \
-        || (x)==PIX_FMT_YUV420P16LE   \
-        || (x)==PIX_FMT_YUV422P16LE   \
-        || (x)==PIX_FMT_YUV444P16LE   \
-        || (x)==PIX_FMT_YUV420P16BE   \
-        || (x)==PIX_FMT_YUV422P16BE   \
-        || (x)==PIX_FMT_YUV444P16BE   \
-    )
-
-int sws_isSupportedInput(enum PixelFormat pix_fmt)
-{
-    return isSupportedIn(pix_fmt);
-}
-
-#define isSupportedOut(x)   (       \
-           (x)==PIX_FMT_YUV420P     \
-        || (x)==PIX_FMT_YUVA420P    \
-        || (x)==PIX_FMT_YUYV422     \
-        || (x)==PIX_FMT_UYVY422     \
-        || (x)==PIX_FMT_YUV444P     \
-        || (x)==PIX_FMT_YUV422P     \
-        || (x)==PIX_FMT_YUV411P     \
-        || isRGB(x)                 \
-        || isBGR(x)                 \
-        || (x)==PIX_FMT_NV12        \
-        || (x)==PIX_FMT_NV21        \
-        || (x)==PIX_FMT_GRAY16BE    \
-        || (x)==PIX_FMT_GRAY16LE    \
-        || (x)==PIX_FMT_GRAY8       \
-        || (x)==PIX_FMT_YUV410P     \
-        || (x)==PIX_FMT_YUV440P     \
-        || (x)==PIX_FMT_YUV420P16LE   \
-        || (x)==PIX_FMT_YUV422P16LE   \
-        || (x)==PIX_FMT_YUV444P16LE   \
-        || (x)==PIX_FMT_YUV420P16BE   \
-        || (x)==PIX_FMT_YUV422P16BE   \
-        || (x)==PIX_FMT_YUV444P16BE   \
-    )
-
-int sws_isSupportedOutput(enum PixelFormat pix_fmt)
-{
-    return isSupportedOut(pix_fmt);
-}
-
 #define isPacked(x)         (       \
            (x)==PIX_FMT_PAL8        \
         || (x)==PIX_FMT_YUYV422     \
@@ -213,8 +106,6 @@ int sws_isSupportedOutput(enum PixelFormat pix_fmt)
 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
 
-extern const int32_t ff_yuv2rgb_coeffs[8][4];
-
 static const double rgb2yuv_table[8][9]={
     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
@@ -379,14 +270,6 @@ DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 };
 #endif
 
-const char *sws_format_name(enum PixelFormat format)
-{
-    if ((unsigned)format < PIX_FMT_NB && av_pix_fmt_descriptors[format].name)
-        return av_pix_fmt_descriptors[format].name;
-    else
-        return "Unknown format";
-}
-
 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
                                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
                                                     const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
@@ -1331,515 +1214,7 @@ static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uin
 
 #endif //ARCH_X86
 
-static double getSplineCoeff(double a, double b, double c, double d, double dist)
-{
-//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
-    if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
-    else           return getSplineCoeff(        0.0,
-                                          b+ 2.0*c + 3.0*d,
-                                                 c + 3.0*d,
-                                         -b- 3.0*c - 6.0*d,
-                                         dist-1.0);
-}
-
-static int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
-                      int srcW, int dstW, int filterAlign, int one, int flags,
-                      SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
-{
-    int i;
-    int filterSize;
-    int filter2Size;
-    int minFilterSize;
-    int64_t *filter=NULL;
-    int64_t *filter2=NULL;
-    const int64_t fone= 1LL<<54;
-    int ret= -1;
-#if ARCH_X86
-    if (flags & SWS_CPU_CAPS_MMX)
-        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
-#endif
-
-    // NOTE: the +1 is for the MMX scaler which reads over the end
-    FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW+1)*sizeof(int16_t), fail);
-
-    if (FFABS(xInc - 0x10000) <10) { // unscaled
-        int i;
-        filterSize= 1;
-        FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
-
-        for (i=0; i<dstW; i++) {
-            filter[i*filterSize]= fone;
-            (*filterPos)[i]=i;
-        }
-
-    } else if (flags&SWS_POINT) { // lame looking point sampling mode
-        int i;
-        int xDstInSrc;
-        filterSize= 1;
-        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
-
-        xDstInSrc= xInc/2 - 0x8000;
-        for (i=0; i<dstW; i++) {
-            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
-
-            (*filterPos)[i]= xx;
-            filter[i]= fone;
-            xDstInSrc+= xInc;
-        }
-    } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale
-        int i;
-        int xDstInSrc;
-        filterSize= 2;
-        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
-
-        xDstInSrc= xInc/2 - 0x8000;
-        for (i=0; i<dstW; i++) {
-            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
-            int j;
-
-            (*filterPos)[i]= xx;
-            //bilinear upscale / linear interpolate / area averaging
-            for (j=0; j<filterSize; j++) {
-                int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
-                if (coeff<0) coeff=0;
-                filter[i*filterSize + j]= coeff;
-                xx++;
-            }
-            xDstInSrc+= xInc;
-        }
-    } else {
-        int xDstInSrc;
-        int sizeFactor;
-
-        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
-        else if (flags&SWS_X)            sizeFactor=  8;
-        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
-        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
-        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
-        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
-        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
-        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
-        else {
-            sizeFactor= 0; //GCC warning killer
-            assert(0);
-        }
-
-        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
-        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
-
-        if (filterSize > srcW-2) filterSize=srcW-2;
-
-        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
-
-        xDstInSrc= xInc - 0x10000;
-        for (i=0; i<dstW; i++) {
-            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
-            int j;
-            (*filterPos)[i]= xx;
-            for (j=0; j<filterSize; j++) {
-                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
-                double floatd;
-                int64_t coeff;
-
-                if (xInc > 1<<16)
-                    d= d*dstW/srcW;
-                floatd= d * (1.0/(1<<30));
-
-                if (flags & SWS_BICUBIC) {
-                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
-                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
-                    int64_t dd = ( d*d)>>30;
-                    int64_t ddd= (dd*d)>>30;
-
-                    if      (d < 1LL<<30)
-                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
-                    else if (d < 1LL<<31)
-                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
-                    else
-                        coeff=0.0;
-                    coeff *= fone>>(30+24);
-                }
-/*                else if (flags & SWS_X) {
-                    double p= param ? param*0.01 : 0.3;
-                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
-                    coeff*= pow(2.0, - p*d*d);
-                }*/
-                else if (flags & SWS_X) {
-                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
-                    double c;
-
-                    if (floatd<1.0)
-                        c = cos(floatd*PI);
-                    else
-                        c=-1.0;
-                    if (c<0.0)      c= -pow(-c, A);
-                    else            c=  pow( c, A);
-                    coeff= (c*0.5 + 0.5)*fone;
-                } else if (flags & SWS_AREA) {
-                    int64_t d2= d - (1<<29);
-                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
-                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
-                    else coeff=0.0;
-                    coeff *= fone>>(30+16);
-                } else if (flags & SWS_GAUSS) {
-                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
-                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
-                } else if (flags & SWS_SINC) {
-                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
-                } else if (flags & SWS_LANCZOS) {
-                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
-                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
-                    if (floatd>p) coeff=0;
-                } else if (flags & SWS_BILINEAR) {
-                    coeff= (1<<30) - d;
-                    if (coeff<0) coeff=0;
-                    coeff *= fone >> 30;
-                } else if (flags & SWS_SPLINE) {
-                    double p=-2.196152422706632;
-                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
-                } else {
-                    coeff= 0.0; //GCC warning killer
-                    assert(0);
-                }
-
-                filter[i*filterSize + j]= coeff;
-                xx++;
-            }
-            xDstInSrc+= 2*xInc;
-        }
-    }
-
-    /* apply src & dst Filter to filter -> filter2
-       av_free(filter);
-    */
-    assert(filterSize>0);
-    filter2Size= filterSize;
-    if (srcFilter) filter2Size+= srcFilter->length - 1;
-    if (dstFilter) filter2Size+= dstFilter->length - 1;
-    assert(filter2Size>0);
-    FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail);
-
-    for (i=0; i<dstW; i++) {
-        int j, k;
-
-        if(srcFilter) {
-            for (k=0; k<srcFilter->length; k++) {
-                for (j=0; j<filterSize; j++)
-                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
-            }
-        } else {
-            for (j=0; j<filterSize; j++)
-                filter2[i*filter2Size + j]= filter[i*filterSize + j];
-        }
-        //FIXME dstFilter
-
-        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
-    }
-    av_freep(&filter);
-
-    /* try to reduce the filter-size (step1 find size and shift left) */
-    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
-    minFilterSize= 0;
-    for (i=dstW-1; i>=0; i--) {
-        int min= filter2Size;
-        int j;
-        int64_t cutOff=0.0;
-
-        /* get rid of near zero elements on the left by shifting left */
-        for (j=0; j<filter2Size; j++) {
-            int k;
-            cutOff += FFABS(filter2[i*filter2Size]);
-
-            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
-
-            /* preserve monotonicity because the core can't handle the filter otherwise */
-            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
-
-            // move filter coefficients left
-            for (k=1; k<filter2Size; k++)
-                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
-            filter2[i*filter2Size + k - 1]= 0;
-            (*filterPos)[i]++;
-        }
-
-        cutOff=0;
-        /* count near zeros on the right */
-        for (j=filter2Size-1; j>0; j--) {
-            cutOff += FFABS(filter2[i*filter2Size + j]);
-
-            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
-            min--;
-        }
-
-        if (min>minFilterSize) minFilterSize= min;
-    }
-
-    if (flags & SWS_CPU_CAPS_ALTIVEC) {
-        // we can handle the special case 4,
-        // so we don't want to go to the full 8
-        if (minFilterSize < 5)
-            filterAlign = 4;
-
-        // We really don't want to waste our time
-        // doing useless computation, so fall back on
-        // the scalar C code for very small filters.
-        // Vectorizing is worth it only if you have a
-        // decent-sized vector.
-        if (minFilterSize < 3)
-            filterAlign = 1;
-    }
-
-    if (flags & SWS_CPU_CAPS_MMX) {
-        // special case for unscaled vertical filtering
-        if (minFilterSize == 1 && filterAlign == 2)
-            filterAlign= 1;
-    }
-
-    assert(minFilterSize > 0);
-    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
-    assert(filterSize > 0);
-    filter= av_malloc(filterSize*dstW*sizeof(*filter));
-    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
-        goto fail;
-    *outFilterSize= filterSize;
-
-    if (flags&SWS_PRINT_INFO)
-        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
-    /* try to reduce the filter-size (step2 reduce it) */
-    for (i=0; i<dstW; i++) {
-        int j;
-
-        for (j=0; j<filterSize; j++) {
-            if (j>=filter2Size) filter[i*filterSize + j]= 0;
-            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
-            if((flags & SWS_BITEXACT) && j>=minFilterSize)
-                filter[i*filterSize + j]= 0;
-        }
-    }
-
-    //FIXME try to align filterPos if possible
-
-    //fix borders
-    for (i=0; i<dstW; i++) {
-        int j;
-        if ((*filterPos)[i] < 0) {
-            // move filter coefficients left to compensate for filterPos
-            for (j=1; j<filterSize; j++) {
-                int left= FFMAX(j + (*filterPos)[i], 0);
-                filter[i*filterSize + left] += filter[i*filterSize + j];
-                filter[i*filterSize + j]=0;
-            }
-            (*filterPos)[i]= 0;
-        }
-
-        if ((*filterPos)[i] + filterSize > srcW) {
-            int shift= (*filterPos)[i] + filterSize - srcW;
-            // move filter coefficients right to compensate for filterPos
-            for (j=filterSize-2; j>=0; j--) {
-                int right= FFMIN(j + shift, filterSize-1);
-                filter[i*filterSize +right] += filter[i*filterSize +j];
-                filter[i*filterSize +j]=0;
-            }
-            (*filterPos)[i]= srcW - filterSize;
-        }
-    }
-
-    // Note the +1 is for the MMX scaler which reads over the end
-    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
-    FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+1)*sizeof(int16_t), fail);
-
-    /* normalize & store in outFilter */
-    for (i=0; i<dstW; i++) {
-        int j;
-        int64_t error=0;
-        int64_t sum=0;
-
-        for (j=0; j<filterSize; j++) {
-            sum+= filter[i*filterSize + j];
-        }
-        sum= (sum + one/2)/ one;
-        for (j=0; j<*outFilterSize; j++) {
-            int64_t v= filter[i*filterSize + j] + error;
-            int intV= ROUNDED_DIV(v, sum);
-            (*outFilter)[i*(*outFilterSize) + j]= intV;
-            error= v - intV*sum;
-        }
-    }
-
-    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
-    for (i=0; i<*outFilterSize; i++) {
-        int j= dstW*(*outFilterSize);
-        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
-    }
-
-    ret=0;
-fail:
-    av_free(filter);
-    av_free(filter2);
-    return ret;
-}
-
-#ifdef COMPILE_MMX2
-static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)
-{
-    uint8_t *fragmentA;
-    x86_reg imm8OfPShufW1A;
-    x86_reg imm8OfPShufW2A;
-    x86_reg fragmentLengthA;
-    uint8_t *fragmentB;
-    x86_reg imm8OfPShufW1B;
-    x86_reg imm8OfPShufW2B;
-    x86_reg fragmentLengthB;
-    int fragmentPos;
-
-    int xpos, i;
-
-    // create an optimized horizontal scaling routine
-    /* This scaler is made of runtime-generated MMX2 code using specially
-     * tuned pshufw instructions. For every four output pixels, if four
-     * input pixels are enough for the fast bilinear scaling, then a chunk
-     * of fragmentB is used. If five input pixels are needed, then a chunk
-     * of fragmentA is used.
-     */
-
-    //code fragment
-
-    __asm__ volatile(
-        "jmp                         9f                 \n\t"
-    // Begin
-        "0:                                             \n\t"
-        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
-        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
-        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
-        "punpcklbw                %%mm7, %%mm1          \n\t"
-        "punpcklbw                %%mm7, %%mm0          \n\t"
-        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
-        "1:                                             \n\t"
-        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
-        "2:                                             \n\t"
-        "psubw                    %%mm1, %%mm0          \n\t"
-        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
-        "pmullw                   %%mm3, %%mm0          \n\t"
-        "psllw                       $7, %%mm1          \n\t"
-        "paddw                    %%mm1, %%mm0          \n\t"
-
-        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
-
-        "add                         $8, %%"REG_a"      \n\t"
-    // End
-        "9:                                             \n\t"
-//        "int $3                                         \n\t"
-        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
-        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
-        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
-        "dec                         %1                 \n\t"
-        "dec                         %2                 \n\t"
-        "sub                         %0, %1             \n\t"
-        "sub                         %0, %2             \n\t"
-        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
-        "sub                         %0, %3             \n\t"
-
-
-        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
-        "=r" (fragmentLengthA)
-    );
-
-    __asm__ volatile(
-        "jmp                         9f                 \n\t"
-    // Begin
-        "0:                                             \n\t"
-        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
-        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
-        "punpcklbw                %%mm7, %%mm0          \n\t"
-        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
-        "1:                                             \n\t"
-        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
-        "2:                                             \n\t"
-        "psubw                    %%mm1, %%mm0          \n\t"
-        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
-        "pmullw                   %%mm3, %%mm0          \n\t"
-        "psllw                       $7, %%mm1          \n\t"
-        "paddw                    %%mm1, %%mm0          \n\t"
-
-        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
-
-        "add                         $8, %%"REG_a"      \n\t"
-    // End
-        "9:                                             \n\t"
-//        "int                       $3                   \n\t"
-        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
-        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
-        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
-        "dec                         %1                 \n\t"
-        "dec                         %2                 \n\t"
-        "sub                         %0, %1             \n\t"
-        "sub                         %0, %2             \n\t"
-        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
-        "sub                         %0, %3             \n\t"
-
-
-        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
-        "=r" (fragmentLengthB)
-    );
-
-    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
-    fragmentPos=0;
-
-    for (i=0; i<dstW/numSplits; i++) {
-        int xx=xpos>>16;
-
-        if ((i&3) == 0) {
-            int a=0;
-            int b=((xpos+xInc)>>16) - xx;
-            int c=((xpos+xInc*2)>>16) - xx;
-            int d=((xpos+xInc*3)>>16) - xx;
-            int inc                = (d+1<4);
-            uint8_t *fragment      = (d+1<4) ? fragmentB       : fragmentA;
-            x86_reg imm8OfPShufW1  = (d+1<4) ? imm8OfPShufW1B  : imm8OfPShufW1A;
-            x86_reg imm8OfPShufW2  = (d+1<4) ? imm8OfPShufW2B  : imm8OfPShufW2A;
-            x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;
-            int maxShift= 3-(d+inc);
-            int shift=0;
-
-            if (filterCode) {
-                filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
-                filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
-                filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
-                filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
-                filterPos[i/2]= xx;
-
-                memcpy(filterCode + fragmentPos, fragment, fragmentLength);
-
-                filterCode[fragmentPos + imm8OfPShufW1]=
-                    (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);
-                filterCode[fragmentPos + imm8OfPShufW2]=
-                    a | (b<<2) | (c<<4) | (d<<6);
-
-                if (i+4-inc>=dstW) shift=maxShift; //avoid overread
-                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
-
-                if (shift && i>=shift) {
-                    filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;
-                    filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;
-                    filterPos[i/2]-=shift;
-                }
-            }
-
-            fragmentPos+= fragmentLength;
-
-            if (filterCode)
-                filterCode[fragmentPos]= RET;
-        }
-        xpos+=xInc;
-    }
-    if (filterCode)
-        filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
-
-    return fragmentPos + 1;
-}
-#endif /* COMPILE_MMX2 */
-
-static SwsFunc getSwsFunc(SwsContext *c)
+SwsFunc ff_getSwsFunc(SwsContext *c)
 {
 #if CONFIG_RUNTIME_CPUDETECT
     int flags = c->flags;
@@ -2260,137 +1635,9 @@ static int planarCopy(SwsContext *c, const uint8_t* src[], int srcStride[], int
     return srcSliceH;
 }
 
-static void getSubSampleFactors(int *h, int *v, enum PixelFormat format)
-{
-    *h = av_pix_fmt_descriptors[format].log2_chroma_w;
-    *v = av_pix_fmt_descriptors[format].log2_chroma_h;
-}
-
-static uint16_t roundToInt16(int64_t f)
-{
-    int r= (f + (1<<15))>>16;
-         if (r<-0x7FFF) return 0x8000;
-    else if (r> 0x7FFF) return 0x7FFF;
-    else                return r;
-}
-
-int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation)
+int ff_hardcodedcpuflags(void)
 {
-    int64_t crv =  inv_table[0];
-    int64_t cbu =  inv_table[1];
-    int64_t cgu = -inv_table[2];
-    int64_t cgv = -inv_table[3];
-    int64_t cy  = 1<<16;
-    int64_t oy  = 0;
-
-    memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
-    memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
-
-    c->brightness= brightness;
-    c->contrast  = contrast;
-    c->saturation= saturation;
-    c->srcRange  = srcRange;
-    c->dstRange  = dstRange;
-    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
-
-    c->uOffset=   0x0400040004000400LL;
-    c->vOffset=   0x0400040004000400LL;
-
-    if (!srcRange) {
-        cy= (cy*255) / 219;
-        oy= 16<<16;
-    } else {
-        crv= (crv*224) / 255;
-        cbu= (cbu*224) / 255;
-        cgu= (cgu*224) / 255;
-        cgv= (cgv*224) / 255;
-    }
-
-    cy = (cy *contrast             )>>16;
-    crv= (crv*contrast * saturation)>>32;
-    cbu= (cbu*contrast * saturation)>>32;
-    cgu= (cgu*contrast * saturation)>>32;
-    cgv= (cgv*contrast * saturation)>>32;
-
-    oy -= 256*brightness;
-
-    c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
-    c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
-    c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
-    c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
-    c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
-    c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
-
-    c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);
-    c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
-    c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
-    c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
-    c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
-    c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
-
-    ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
-    //FIXME factorize
-
-#ifdef COMPILE_ALTIVEC
-    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
-        ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
-#endif
-    return 0;
-}
-
-int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation)
-{
-    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
-
-    *inv_table = c->srcColorspaceTable;
-    *table     = c->dstColorspaceTable;
-    *srcRange  = c->srcRange;
-    *dstRange  = c->dstRange;
-    *brightness= c->brightness;
-    *contrast  = c->contrast;
-    *saturation= c->saturation;
-
-    return 0;
-}
-
-static int handle_jpeg(enum PixelFormat *format)
-{
-    switch (*format) {
-    case PIX_FMT_YUVJ420P:
-        *format = PIX_FMT_YUV420P;
-        return 1;
-    case PIX_FMT_YUVJ422P:
-        *format = PIX_FMT_YUV422P;
-        return 1;
-    case PIX_FMT_YUVJ444P:
-        *format = PIX_FMT_YUV444P;
-        return 1;
-    case PIX_FMT_YUVJ440P:
-        *format = PIX_FMT_YUV440P;
-        return 1;
-    default:
-        return 0;
-    }
-}
-
-SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat,
-                           int dstW, int dstH, enum PixelFormat dstFormat, int flags,
-                           SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
-{
-
-    SwsContext *c;
-    int i;
-    int usesVFilter, usesHFilter;
-    int unscaled, needsDither;
-    int srcRange, dstRange;
-    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
-#if ARCH_X86
-    if (flags & SWS_CPU_CAPS_MMX)
-        __asm__ volatile("emms\n\t"::: "memory");
-#endif
-
-#if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
-    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
+    int flags = 0;
 #if   COMPILE_TEMPLATE_MMX2
     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
 #elif COMPILE_TEMPLATE_AMD3DNOW
@@ -2402,116 +1649,21 @@ SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat,
 #elif ARCH_BFIN
     flags |= SWS_CPU_CAPS_BFIN;
 #endif
-#endif /* CONFIG_RUNTIME_CPUDETECT */
-    if (!rgb15to16) sws_rgb2rgb_init(flags);
+    return flags;
+}
+
+void ff_get_unscaled_swscale(SwsContext *c)
+{
+    const enum PixelFormat srcFormat = c->srcFormat;
+    const enum PixelFormat dstFormat = c->dstFormat;
+    const int flags = c->flags;
+    const int dstH = c->dstH;
+    int needsDither;
 
-    unscaled = (srcW == dstW && srcH == dstH);
     needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
         && (fmt_depth(dstFormat))<24
         && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
 
-    srcRange = handle_jpeg(&srcFormat);
-    dstRange = handle_jpeg(&dstFormat);
-
-    if (!isSupportedIn(srcFormat)) {
-        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
-        return NULL;
-    }
-    if (!isSupportedOut(dstFormat)) {
-        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
-        return NULL;
-    }
-
-    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);
-    if(!i || (i & (i-1))) {
-        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
-        return NULL;
-    }
-
-    /* sanity check */
-    if (srcW<4 || srcH<1 || dstW<8 || dstH<1) { //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code
-        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
-               srcW, srcH, dstW, dstH);
-        return NULL;
-    }
-    if(srcW > VOFW || dstW > VOFW) {
-        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
-        return NULL;
-    }
-
-    if (!dstFilter) dstFilter= &dummyFilter;
-    if (!srcFilter) srcFilter= &dummyFilter;
-
-    FF_ALLOCZ_OR_GOTO(NULL, c, sizeof(SwsContext), fail);
-
-    c->av_class = &sws_context_class;
-    c->srcW= srcW;
-    c->srcH= srcH;
-    c->dstW= dstW;
-    c->dstH= dstH;
-    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
-    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
-    c->flags= flags;
-    c->dstFormat= dstFormat;
-    c->srcFormat= srcFormat;
-    c->vRounder= 4* 0x0001000100010001ULL;
-
-    usesHFilter= usesVFilter= 0;
-    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
-    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
-    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
-    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
-    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
-    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
-    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
-    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
-
-    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
-    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
-
-    // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
-    if ((isBGR(dstFormat) || isRGB(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 ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
-      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
-      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
-      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
-      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
-        c->chrSrcHSubSample=1;
-
-    if (param) {
-        c->param[0] = param[0];
-        c->param[1] = param[1];
-    } else {
-        c->param[0] =
-        c->param[1] = SWS_PARAM_DEFAULT;
-    }
-
-    // Note the -((-x)>>y) is so that we always round toward +inf.
-    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
-    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
-    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
-    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
-
-    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
-
-    /* unscaled special cases */
-    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat))) {
         /* yv12_to_nv12 */
         if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
             c->swScale= PlanarToNV12Wrapper;
@@ -2617,298 +1769,6 @@ SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat,
         if (flags & SWS_CPU_CAPS_BFIN)
             ff_bfin_get_unscaled_swscale (c);
 #endif
-
-        if (c->swScale) {
-            if (flags&SWS_PRINT_INFO)
-                av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
-                       sws_format_name(srcFormat), sws_format_name(dstFormat));
-            return c;
-        }
-    }
-
-    if (flags & SWS_CPU_CAPS_MMX2) {
-        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
-        if (!c->canMMX2BeUsed && dstW >=srcW && (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 MMX2 scaler\n");
-        }
-        if (usesHFilter) c->canMMX2BeUsed=0;
-    }
-    else
-        c->canMMX2BeUsed=0;
-
-    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
-    c->chrYInc= ((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->canMMX2BeUsed) {
-            c->lumXInc+= 20;
-            c->chrXInc+= 20;
-        }
-        //we don't use the x86 asm scaler if MMX is available
-        else if (flags & SWS_CPU_CAPS_MMX) {
-            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
-            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
-        }
-    }
-
-    /* precalculate horizontal scaler filter coefficients */
-    {
-#if defined(COMPILE_MMX2)
-// can't downscale !!!
-        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {
-            c->lumMmx2FilterCodeSize = initMMX2HScaler(      dstW, c->lumXInc, NULL, NULL, NULL, 8);
-            c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);
-
-#ifdef MAP_ANONYMOUS
-            c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
-            c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
-#elif HAVE_VIRTUALALLOC
-            c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
-            c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
-#else
-            c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize);
-            c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);
-#endif
-
-            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);
-
-            initMMX2HScaler(      dstW, c->lumXInc, c->lumMmx2FilterCode, c->hLumFilter, c->hLumFilterPos, 8);
-            initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->hChrFilter, c->hChrFilterPos, 4);
-
-#ifdef MAP_ANONYMOUS
-            mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
-            mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
-#endif
-        } else
-#endif /* defined(COMPILE_MMX2) */
-        {
-            const int filterAlign=
-                (flags & SWS_CPU_CAPS_MMX) ? 4 :
-                (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
-                1;
-
-            if (initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
-                           srcW      ,       dstW, filterAlign, 1<<14,
-                           (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
-                           srcFilter->lumH, dstFilter->lumH, c->param) < 0)
-                goto fail;
-            if (initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
-                           c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
-                           (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
-                           srcFilter->chrH, dstFilter->chrH, c->param) < 0)
-                goto fail;
-        }
-    } // initialize horizontal stuff
-
-    /* precalculate vertical scaler filter coefficients */
-    {
-        const int filterAlign=
-            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
-            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
-            1;
-
-        if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
-                       srcH      ,        dstH, filterAlign, (1<<12),
-                       (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
-                       srcFilter->lumV, dstFilter->lumV, c->param) < 0)
-            goto fail;
-        if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
-                       c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
-                       (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
-                       srcFilter->chrV, dstFilter->chrV, c->param) < 0)
-            goto fail;
-
-#ifdef COMPILE_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= 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];
-    }
-
-    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
-    // allocate several megabytes to handle all possible cases)
-    FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);
-    FF_ALLOC_OR_GOTO(c, c->chrPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail);
-    if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
-        FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);
-    //Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna 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], VOF+1, fail);
-        c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize];
-    }
-    for (i=0; i<c->vChrBufSize; i++) {
-        FF_ALLOC_OR_GOTO(c, c->chrPixBuf[i+c->vChrBufSize], (VOF+1)*2, fail);
-        c->chrPixBuf[i] = c->chrPixBuf[i+c->vChrBufSize];
-    }
-    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
-        for (i=0; i<c->vLumBufSize; i++) {
-            FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i+c->vLumBufSize], VOF+1, 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++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
-
-    assert(2*VOFW == VOF);
-
-    assert(c->chrDstH <= dstH);
-
-    if (flags&SWS_PRINT_INFO) {
-        if (flags&SWS_FAST_BILINEAR)
-            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
-        else if (flags&SWS_BILINEAR)
-            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
-        else if (flags&SWS_BICUBIC)
-            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
-        else if (flags&SWS_X)
-            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
-        else if (flags&SWS_POINT)
-            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
-        else if (flags&SWS_AREA)
-            av_log(c, AV_LOG_INFO, "Area Averaging scaler, ");
-        else if (flags&SWS_BICUBLIN)
-            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
-        else if (flags&SWS_GAUSS)
-            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
-        else if (flags&SWS_SINC)
-            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
-        else if (flags&SWS_LANCZOS)
-            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
-        else if (flags&SWS_SPLINE)
-            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
-        else
-            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
-
-        av_log(c, AV_LOG_INFO, "from %s to %s%s ",
-               sws_format_name(srcFormat),
-#ifdef DITHER1XBPP
-               dstFormat == PIX_FMT_BGR555 || dstFormat == PIX_FMT_BGR565 ? "dithered " : "",
-#else
-               "",
-#endif
-               sws_format_name(dstFormat));
-
-        if (flags & SWS_CPU_CAPS_MMX2)
-            av_log(c, AV_LOG_INFO, "using MMX2\n");
-        else if (flags & SWS_CPU_CAPS_3DNOW)
-            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
-        else if (flags & SWS_CPU_CAPS_MMX)
-            av_log(c, AV_LOG_INFO, "using MMX\n");
-        else if (flags & SWS_CPU_CAPS_ALTIVEC)
-            av_log(c, AV_LOG_INFO, "using AltiVec\n");
-        else
-            av_log(c, AV_LOG_INFO, "using C\n");
-    }
-
-    if (flags & SWS_PRINT_INFO) {
-        if (flags & SWS_CPU_CAPS_MMX) {
-            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
-                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
-            else {
-                if (c->hLumFilterSize==4)
-                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
-                else if (c->hLumFilterSize==8)
-                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
-                else
-                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
-
-                if (c->hChrFilterSize==4)
-                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
-                else if (c->hChrFilterSize==8)
-                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
-                else
-                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
-            }
-        } else {
-#if ARCH_X86
-            av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
-#else
-            if (flags & SWS_FAST_BILINEAR)
-                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
-            else
-                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
-#endif
-        }
-        if (isPlanarYUV(dstFormat)) {
-            if (c->vLumFilterSize==1)
-                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
-            else
-                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
-        } else {
-            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
-                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
-                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
-            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
-                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
-            else
-                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
-        }
-
-        if (dstFormat==PIX_FMT_BGR24)
-            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
-                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
-        else if (dstFormat==PIX_FMT_RGB32)
-            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
-        else if (dstFormat==PIX_FMT_BGR565)
-            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
-        else if (dstFormat==PIX_FMT_BGR555)
-            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
-
-        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
-    }
-    if (flags & SWS_PRINT_INFO) {
-        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);
-    }
-
-    c->swScale= getSwsFunc(c);
-    return c;
-
-fail:
-    sws_freeContext(c);
-    return NULL;
 }
 
 static void reset_ptr(const uint8_t* src[], int format)
@@ -3053,388 +1913,3 @@ int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[]
     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
 }
 #endif
-
-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 (chromaSharpen!=0.0) {
-        SwsVector *id= sws_getIdentityVec();
-        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();
-        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 (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
-    if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
-
-    return filter;
-}
-
-SwsVector *sws_allocVec(int length)
-{
-    SwsVector *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= 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*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);
-}
-
-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);
-    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);
-    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);
-    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);
-    av_free(a->coeff);
-    a->coeff= conv->coeff;
-    a->length= conv->length;
-    av_free(conv);
-}
-
-SwsVector *sws_cloneVec(SwsVector *a)
-{
-    int i;
-    SwsVector *vec= sws_allocVec(a->length);
-
-    if (!vec)
-        return NULL;
-
-    for (i=0; i<a->length; i++) vec->coeff[i]= a->coeff[i];
-
-    return vec;
-}
-
-void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
-{
-    int i;
-    double max=0;
-    double min=0;
-    double range;
-
-    for (i=0; i<a->length; i++)
-        if (a->coeff[i]>max) max= a->coeff[i];
-
-    for (i=0; i<a->length; i++)
-        if (a->coeff[i]<min) min= a->coeff[i];
-
-    range= max - min;
-
-    for (i=0; i<a->length; i++) {
-        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
-        av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
-        for (;x>0; x--) av_log(log_ctx, log_level, " ");
-        av_log(log_ctx, log_level, "|\n");
-    }
-}
-
-#if LIBSWSCALE_VERSION_MAJOR < 1
-void sws_printVec(SwsVector *a)
-{
-    sws_printVec2(a, NULL, AV_LOG_DEBUG);
-}
-#endif
-
-void sws_freeVec(SwsVector *a)
-{
-    if (!a) return;
-    av_freep(&a->coeff);
-    a->length=0;
-    av_free(a);
-}
-
-void sws_freeFilter(SwsFilter *filter)
-{
-    if (!filter) return;
-
-    if (filter->lumH) sws_freeVec(filter->lumH);
-    if (filter->lumV) sws_freeVec(filter->lumV);
-    if (filter->chrH) sws_freeVec(filter->chrH);
-    if (filter->chrV) sws_freeVec(filter->chrV);
-    av_free(filter);
-}
-
-void sws_freeContext(SwsContext *c)
-{
-    int i;
-    if (!c) return;
-
-    if (c->lumPixBuf) {
-        for (i=0; i<c->vLumBufSize; i++)
-            av_freep(&c->lumPixBuf[i]);
-        av_freep(&c->lumPixBuf);
-    }
-
-    if (c->chrPixBuf) {
-        for (i=0; i<c->vChrBufSize; i++)
-            av_freep(&c->chrPixBuf[i]);
-        av_freep(&c->chrPixBuf);
-    }
-
-    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
-        for (i=0; i<c->vLumBufSize; i++)
-            av_freep(&c->alpPixBuf[i]);
-        av_freep(&c->alpPixBuf);
-    }
-
-    av_freep(&c->vLumFilter);
-    av_freep(&c->vChrFilter);
-    av_freep(&c->hLumFilter);
-    av_freep(&c->hChrFilter);
-#ifdef COMPILE_ALTIVEC
-    av_freep(&c->vYCoeffsBank);
-    av_freep(&c->vCCoeffsBank);
-#endif
-
-    av_freep(&c->vLumFilterPos);
-    av_freep(&c->vChrFilterPos);
-    av_freep(&c->hLumFilterPos);
-    av_freep(&c->hChrFilterPos);
-
-#if ARCH_X86 && CONFIG_GPL
-#ifdef MAP_ANONYMOUS
-    if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);
-    if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);
-#elif HAVE_VIRTUALALLOC
-    if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, MEM_RELEASE);
-    if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, MEM_RELEASE);
-#else
-    av_free(c->lumMmx2FilterCode);
-    av_free(c->chrMmx2FilterCode);
-#endif
-    c->lumMmx2FilterCode=NULL;
-    c->chrMmx2FilterCode=NULL;
-#endif /* ARCH_X86 && CONFIG_GPL */
-
-    av_freep(&c->yuvTable);
-
-    av_free(c);
-}
-
-struct SwsContext *sws_getCachedContext(struct SwsContext *context,
-                                        int srcW, int srcH, enum PixelFormat srcFormat,
-                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
-                                        SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
-{
-    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
-
-    if (!param)
-        param = default_param;
-
-    if (context) {
-        if (context->srcW != srcW || context->srcH != srcH ||
-            context->srcFormat != srcFormat ||
-            context->dstW != dstW || context->dstH != dstH ||
-            context->dstFormat != dstFormat || context->flags != flags ||
-            context->param[0] != param[0] || context->param[1] != param[1])
-        {
-            sws_freeContext(context);
-            context = NULL;
-        }
-    }
-    if (!context) {
-        return sws_getContext(srcW, srcH, srcFormat,
-                              dstW, dstH, dstFormat, flags,
-                              srcFilter, dstFilter, param);
-    }
-    return context;
-}
-
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index 0ae6dbc0f2..658c4aef72 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -460,4 +460,21 @@ extern const uint64_t ff_dither8[2];
 
 extern const AVClass sws_context_class;
 
+/**
+ * Sets c->swScale to an unscaled converter if one exists for the specific
+ * source and destination formats, bit depths, flags, etc.
+ */
+void ff_get_unscaled_swscale(SwsContext *c);
+
+/**
+ * Returns the SWS_CPU_CAPS for the optimized code compiled into swscale.
+ */
+int ff_hardcodedcpuflags(void);
+
+/**
+ * Returns function pointer to fastest main scaler path function depending
+ * on architecture and available optimizations.
+ */
+SwsFunc ff_getSwsFunc(SwsContext *c);
+
 #endif /* SWSCALE_SWSCALE_INTERNAL_H */
diff --git a/libswscale/utils.c b/libswscale/utils.c
new file mode 100644
index 0000000000..1a7a36454b
--- /dev/null
+++ b/libswscale/utils.c
@@ -0,0 +1,1589 @@
+/*
+ * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ *
+ * the C code (not assembly, mmx, ...) of this file can be used
+ * under the LGPL license too
+ */
+
+#define _SVID_SOURCE //needed for MAP_ANONYMOUS
+#include <inttypes.h>
+#include <string.h>
+#include <math.h>
+#include <stdio.h>
+#include "config.h"
+#include <assert.h>
+#if HAVE_SYS_MMAN_H
+#include <sys/mman.h>
+#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
+#define MAP_ANONYMOUS MAP_ANON
+#endif
+#endif
+#if HAVE_VIRTUALALLOC
+#define WIN32_LEAN_AND_MEAN
+#include <windows.h>
+#endif
+#include "swscale.h"
+#include "swscale_internal.h"
+#include "rgb2rgb.h"
+#include "libavutil/intreadwrite.h"
+#include "libavutil/x86_cpu.h"
+#include "libavutil/avutil.h"
+#include "libavutil/bswap.h"
+#include "libavutil/pixdesc.h"
+
+unsigned swscale_version(void)
+{
+    return LIBSWSCALE_VERSION_INT;
+}
+
+const char *swscale_configuration(void)
+{
+    return FFMPEG_CONFIGURATION;
+}
+
+const char *swscale_license(void)
+{
+#define LICENSE_PREFIX "libswscale license: "
+    return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
+}
+
+#define RET 0xC3 //near return opcode for x86
+
+#ifdef M_PI
+#define PI M_PI
+#else
+#define PI 3.14159265358979323846
+#endif
+
+#define isSupportedIn(x)    (       \
+           (x)==PIX_FMT_YUV420P     \
+        || (x)==PIX_FMT_YUVA420P    \
+        || (x)==PIX_FMT_YUYV422     \
+        || (x)==PIX_FMT_UYVY422     \
+        || (x)==PIX_FMT_RGB48BE     \
+        || (x)==PIX_FMT_RGB48LE     \
+        || (x)==PIX_FMT_RGB32       \
+        || (x)==PIX_FMT_RGB32_1     \
+        || (x)==PIX_FMT_BGR24       \
+        || (x)==PIX_FMT_BGR565      \
+        || (x)==PIX_FMT_BGR555      \
+        || (x)==PIX_FMT_BGR32       \
+        || (x)==PIX_FMT_BGR32_1     \
+        || (x)==PIX_FMT_RGB24       \
+        || (x)==PIX_FMT_RGB565      \
+        || (x)==PIX_FMT_RGB555      \
+        || (x)==PIX_FMT_GRAY8       \
+        || (x)==PIX_FMT_YUV410P     \
+        || (x)==PIX_FMT_YUV440P     \
+        || (x)==PIX_FMT_NV12        \
+        || (x)==PIX_FMT_NV21        \
+        || (x)==PIX_FMT_GRAY16BE    \
+        || (x)==PIX_FMT_GRAY16LE    \
+        || (x)==PIX_FMT_YUV444P     \
+        || (x)==PIX_FMT_YUV422P     \
+        || (x)==PIX_FMT_YUV411P     \
+        || (x)==PIX_FMT_PAL8        \
+        || (x)==PIX_FMT_BGR8        \
+        || (x)==PIX_FMT_RGB8        \
+        || (x)==PIX_FMT_BGR4_BYTE   \
+        || (x)==PIX_FMT_RGB4_BYTE   \
+        || (x)==PIX_FMT_YUV440P     \
+        || (x)==PIX_FMT_MONOWHITE   \
+        || (x)==PIX_FMT_MONOBLACK   \
+        || (x)==PIX_FMT_YUV420P16LE   \
+        || (x)==PIX_FMT_YUV422P16LE   \
+        || (x)==PIX_FMT_YUV444P16LE   \
+        || (x)==PIX_FMT_YUV420P16BE   \
+        || (x)==PIX_FMT_YUV422P16BE   \
+        || (x)==PIX_FMT_YUV444P16BE   \
+    )
+
+int sws_isSupportedInput(enum PixelFormat pix_fmt)
+{
+    return isSupportedIn(pix_fmt);
+}
+
+#define isSupportedOut(x)   (       \
+           (x)==PIX_FMT_YUV420P     \
+        || (x)==PIX_FMT_YUVA420P    \
+        || (x)==PIX_FMT_YUYV422     \
+        || (x)==PIX_FMT_UYVY422     \
+        || (x)==PIX_FMT_YUV444P     \
+        || (x)==PIX_FMT_YUV422P     \
+        || (x)==PIX_FMT_YUV411P     \
+        || isRGB(x)                 \
+        || isBGR(x)                 \
+        || (x)==PIX_FMT_NV12        \
+        || (x)==PIX_FMT_NV21        \
+        || (x)==PIX_FMT_GRAY16BE    \
+        || (x)==PIX_FMT_GRAY16LE    \
+        || (x)==PIX_FMT_GRAY8       \
+        || (x)==PIX_FMT_YUV410P     \
+        || (x)==PIX_FMT_YUV440P     \
+        || (x)==PIX_FMT_YUV420P16LE   \
+        || (x)==PIX_FMT_YUV422P16LE   \
+        || (x)==PIX_FMT_YUV444P16LE   \
+        || (x)==PIX_FMT_YUV420P16BE   \
+        || (x)==PIX_FMT_YUV422P16BE   \
+        || (x)==PIX_FMT_YUV444P16BE   \
+    )
+
+int sws_isSupportedOutput(enum PixelFormat pix_fmt)
+{
+    return isSupportedOut(pix_fmt);
+}
+
+#define usePal(x) (av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL)
+
+extern const int32_t ff_yuv2rgb_coeffs[8][4];
+
+const char *sws_format_name(enum PixelFormat format)
+{
+    if ((unsigned)format < PIX_FMT_NB && av_pix_fmt_descriptors[format].name)
+        return av_pix_fmt_descriptors[format].name;
+    else
+        return "Unknown format";
+}
+
+static double getSplineCoeff(double a, double b, double c, double d, double dist)
+{
+//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
+    if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
+    else           return getSplineCoeff(        0.0,
+                                          b+ 2.0*c + 3.0*d,
+                                                 c + 3.0*d,
+                                         -b- 3.0*c - 6.0*d,
+                                         dist-1.0);
+}
+
+static int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
+                      int srcW, int dstW, int filterAlign, int one, int flags,
+                      SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
+{
+    int i;
+    int filterSize;
+    int filter2Size;
+    int minFilterSize;
+    int64_t *filter=NULL;
+    int64_t *filter2=NULL;
+    const int64_t fone= 1LL<<54;
+    int ret= -1;
+#if ARCH_X86
+    if (flags & SWS_CPU_CAPS_MMX)
+        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
+#endif
+
+    // NOTE: the +1 is for the MMX scaler which reads over the end
+    FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW+1)*sizeof(int16_t), fail);
+
+    if (FFABS(xInc - 0x10000) <10) { // unscaled
+        int i;
+        filterSize= 1;
+        FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
+
+        for (i=0; i<dstW; i++) {
+            filter[i*filterSize]= fone;
+            (*filterPos)[i]=i;
+        }
+
+    } else if (flags&SWS_POINT) { // lame looking point sampling mode
+        int i;
+        int xDstInSrc;
+        filterSize= 1;
+        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
+
+        xDstInSrc= xInc/2 - 0x8000;
+        for (i=0; i<dstW; i++) {
+            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
+
+            (*filterPos)[i]= xx;
+            filter[i]= fone;
+            xDstInSrc+= xInc;
+        }
+    } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale
+        int i;
+        int xDstInSrc;
+        filterSize= 2;
+        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
+
+        xDstInSrc= xInc/2 - 0x8000;
+        for (i=0; i<dstW; i++) {
+            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
+            int j;
+
+            (*filterPos)[i]= xx;
+            //bilinear upscale / linear interpolate / area averaging
+            for (j=0; j<filterSize; j++) {
+                int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
+                if (coeff<0) coeff=0;
+                filter[i*filterSize + j]= coeff;
+                xx++;
+            }
+            xDstInSrc+= xInc;
+        }
+    } else {
+        int xDstInSrc;
+        int sizeFactor;
+
+        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
+        else if (flags&SWS_X)            sizeFactor=  8;
+        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
+        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
+        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
+        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
+        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
+        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
+        else {
+            sizeFactor= 0; //GCC warning killer
+            assert(0);
+        }
+
+        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
+        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
+
+        if (filterSize > srcW-2) filterSize=srcW-2;
+
+        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
+
+        xDstInSrc= xInc - 0x10000;
+        for (i=0; i<dstW; i++) {
+            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
+            int j;
+            (*filterPos)[i]= xx;
+            for (j=0; j<filterSize; j++) {
+                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
+                double floatd;
+                int64_t coeff;
+
+                if (xInc > 1<<16)
+                    d= d*dstW/srcW;
+                floatd= d * (1.0/(1<<30));
+
+                if (flags & SWS_BICUBIC) {
+                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
+                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
+                    int64_t dd = ( d*d)>>30;
+                    int64_t ddd= (dd*d)>>30;
+
+                    if      (d < 1LL<<30)
+                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
+                    else if (d < 1LL<<31)
+                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
+                    else
+                        coeff=0.0;
+                    coeff *= fone>>(30+24);
+                }
+/*                else if (flags & SWS_X) {
+                    double p= param ? param*0.01 : 0.3;
+                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
+                    coeff*= pow(2.0, - p*d*d);
+                }*/
+                else if (flags & SWS_X) {
+                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
+                    double c;
+
+                    if (floatd<1.0)
+                        c = cos(floatd*PI);
+                    else
+                        c=-1.0;
+                    if (c<0.0)      c= -pow(-c, A);
+                    else            c=  pow( c, A);
+                    coeff= (c*0.5 + 0.5)*fone;
+                } else if (flags & SWS_AREA) {
+                    int64_t d2= d - (1<<29);
+                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
+                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
+                    else coeff=0.0;
+                    coeff *= fone>>(30+16);
+                } else if (flags & SWS_GAUSS) {
+                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
+                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
+                } else if (flags & SWS_SINC) {
+                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
+                } else if (flags & SWS_LANCZOS) {
+                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
+                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
+                    if (floatd>p) coeff=0;
+                } else if (flags & SWS_BILINEAR) {
+                    coeff= (1<<30) - d;
+                    if (coeff<0) coeff=0;
+                    coeff *= fone >> 30;
+                } else if (flags & SWS_SPLINE) {
+                    double p=-2.196152422706632;
+                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
+                } else {
+                    coeff= 0.0; //GCC warning killer
+                    assert(0);
+                }
+
+                filter[i*filterSize + j]= coeff;
+                xx++;
+            }
+            xDstInSrc+= 2*xInc;
+        }
+    }
+
+    /* apply src & dst Filter to filter -> filter2
+       av_free(filter);
+    */
+    assert(filterSize>0);
+    filter2Size= filterSize;
+    if (srcFilter) filter2Size+= srcFilter->length - 1;
+    if (dstFilter) filter2Size+= dstFilter->length - 1;
+    assert(filter2Size>0);
+    FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail);
+
+    for (i=0; i<dstW; i++) {
+        int j, k;
+
+        if(srcFilter) {
+            for (k=0; k<srcFilter->length; k++) {
+                for (j=0; j<filterSize; j++)
+                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
+            }
+        } else {
+            for (j=0; j<filterSize; j++)
+                filter2[i*filter2Size + j]= filter[i*filterSize + j];
+        }
+        //FIXME dstFilter
+
+        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
+    }
+    av_freep(&filter);
+
+    /* try to reduce the filter-size (step1 find size and shift left) */
+    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
+    minFilterSize= 0;
+    for (i=dstW-1; i>=0; i--) {
+        int min= filter2Size;
+        int j;
+        int64_t cutOff=0.0;
+
+        /* get rid of near zero elements on the left by shifting left */
+        for (j=0; j<filter2Size; j++) {
+            int k;
+            cutOff += FFABS(filter2[i*filter2Size]);
+
+            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
+
+            /* preserve monotonicity because the core can't handle the filter otherwise */
+            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
+
+            // move filter coefficients left
+            for (k=1; k<filter2Size; k++)
+                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
+            filter2[i*filter2Size + k - 1]= 0;
+            (*filterPos)[i]++;
+        }
+
+        cutOff=0;
+        /* count near zeros on the right */
+        for (j=filter2Size-1; j>0; j--) {
+            cutOff += FFABS(filter2[i*filter2Size + j]);
+
+            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
+            min--;
+        }
+
+        if (min>minFilterSize) minFilterSize= min;
+    }
+
+    if (flags & SWS_CPU_CAPS_ALTIVEC) {
+        // we can handle the special case 4,
+        // so we don't want to go to the full 8
+        if (minFilterSize < 5)
+            filterAlign = 4;
+
+        // We really don't want to waste our time
+        // doing useless computation, so fall back on
+        // the scalar C code for very small filters.
+        // Vectorizing is worth it only if you have a
+        // decent-sized vector.
+        if (minFilterSize < 3)
+            filterAlign = 1;
+    }
+
+    if (flags & SWS_CPU_CAPS_MMX) {
+        // special case for unscaled vertical filtering
+        if (minFilterSize == 1 && filterAlign == 2)
+            filterAlign= 1;
+    }
+
+    assert(minFilterSize > 0);
+    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
+    assert(filterSize > 0);
+    filter= av_malloc(filterSize*dstW*sizeof(*filter));
+    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
+        goto fail;
+    *outFilterSize= filterSize;
+
+    if (flags&SWS_PRINT_INFO)
+        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
+    /* try to reduce the filter-size (step2 reduce it) */
+    for (i=0; i<dstW; i++) {
+        int j;
+
+        for (j=0; j<filterSize; j++) {
+            if (j>=filter2Size) filter[i*filterSize + j]= 0;
+            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
+            if((flags & SWS_BITEXACT) && j>=minFilterSize)
+                filter[i*filterSize + j]= 0;
+        }
+    }
+
+    //FIXME try to align filterPos if possible
+
+    //fix borders
+    for (i=0; i<dstW; i++) {
+        int j;
+        if ((*filterPos)[i] < 0) {
+            // move filter coefficients left to compensate for filterPos
+            for (j=1; j<filterSize; j++) {
+                int left= FFMAX(j + (*filterPos)[i], 0);
+                filter[i*filterSize + left] += filter[i*filterSize + j];
+                filter[i*filterSize + j]=0;
+            }
+            (*filterPos)[i]= 0;
+        }
+
+        if ((*filterPos)[i] + filterSize > srcW) {
+            int shift= (*filterPos)[i] + filterSize - srcW;
+            // move filter coefficients right to compensate for filterPos
+            for (j=filterSize-2; j>=0; j--) {
+                int right= FFMIN(j + shift, filterSize-1);
+                filter[i*filterSize +right] += filter[i*filterSize +j];
+                filter[i*filterSize +j]=0;
+            }
+            (*filterPos)[i]= srcW - filterSize;
+        }
+    }
+
+    // Note the +1 is for the MMX scaler which reads over the end
+    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
+    FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+1)*sizeof(int16_t), fail);
+
+    /* normalize & store in outFilter */
+    for (i=0; i<dstW; i++) {
+        int j;
+        int64_t error=0;
+        int64_t sum=0;
+
+        for (j=0; j<filterSize; j++) {
+            sum+= filter[i*filterSize + j];
+        }
+        sum= (sum + one/2)/ one;
+        for (j=0; j<*outFilterSize; j++) {
+            int64_t v= filter[i*filterSize + j] + error;
+            int intV= ROUNDED_DIV(v, sum);
+            (*outFilter)[i*(*outFilterSize) + j]= intV;
+            error= v - intV*sum;
+        }
+    }
+
+    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
+    for (i=0; i<*outFilterSize; i++) {
+        int j= dstW*(*outFilterSize);
+        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
+    }
+
+    ret=0;
+fail:
+    av_free(filter);
+    av_free(filter2);
+    return ret;
+}
+
+#if ARCH_X86 && (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
+static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)
+{
+    uint8_t *fragmentA;
+    x86_reg imm8OfPShufW1A;
+    x86_reg imm8OfPShufW2A;
+    x86_reg fragmentLengthA;
+    uint8_t *fragmentB;
+    x86_reg imm8OfPShufW1B;
+    x86_reg imm8OfPShufW2B;
+    x86_reg fragmentLengthB;
+    int fragmentPos;
+
+    int xpos, i;
+
+    // create an optimized horizontal scaling routine
+    /* This scaler is made of runtime-generated MMX2 code using specially
+     * tuned pshufw instructions. For every four output pixels, if four
+     * input pixels are enough for the fast bilinear scaling, then a chunk
+     * of fragmentB is used. If five input pixels are needed, then a chunk
+     * of fragmentA is used.
+     */
+
+    //code fragment
+
+    __asm__ volatile(
+        "jmp                         9f                 \n\t"
+    // Begin
+        "0:                                             \n\t"
+        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
+        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
+        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
+        "punpcklbw                %%mm7, %%mm1          \n\t"
+        "punpcklbw                %%mm7, %%mm0          \n\t"
+        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
+        "1:                                             \n\t"
+        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
+        "2:                                             \n\t"
+        "psubw                    %%mm1, %%mm0          \n\t"
+        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
+        "pmullw                   %%mm3, %%mm0          \n\t"
+        "psllw                       $7, %%mm1          \n\t"
+        "paddw                    %%mm1, %%mm0          \n\t"
+
+        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
+
+        "add                         $8, %%"REG_a"      \n\t"
+    // End
+        "9:                                             \n\t"
+//        "int $3                                         \n\t"
+        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
+        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
+        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
+        "dec                         %1                 \n\t"
+        "dec                         %2                 \n\t"
+        "sub                         %0, %1             \n\t"
+        "sub                         %0, %2             \n\t"
+        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
+        "sub                         %0, %3             \n\t"
+
+
+        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
+        "=r" (fragmentLengthA)
+    );
+
+    __asm__ volatile(
+        "jmp                         9f                 \n\t"
+    // Begin
+        "0:                                             \n\t"
+        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
+        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
+        "punpcklbw                %%mm7, %%mm0          \n\t"
+        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
+        "1:                                             \n\t"
+        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
+        "2:                                             \n\t"
+        "psubw                    %%mm1, %%mm0          \n\t"
+        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
+        "pmullw                   %%mm3, %%mm0          \n\t"
+        "psllw                       $7, %%mm1          \n\t"
+        "paddw                    %%mm1, %%mm0          \n\t"
+
+        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
+
+        "add                         $8, %%"REG_a"      \n\t"
+    // End
+        "9:                                             \n\t"
+//        "int                       $3                   \n\t"
+        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
+        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
+        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
+        "dec                         %1                 \n\t"
+        "dec                         %2                 \n\t"
+        "sub                         %0, %1             \n\t"
+        "sub                         %0, %2             \n\t"
+        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
+        "sub                         %0, %3             \n\t"
+
+
+        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
+        "=r" (fragmentLengthB)
+    );
+
+    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
+    fragmentPos=0;
+
+    for (i=0; i<dstW/numSplits; i++) {
+        int xx=xpos>>16;
+
+        if ((i&3) == 0) {
+            int a=0;
+            int b=((xpos+xInc)>>16) - xx;
+            int c=((xpos+xInc*2)>>16) - xx;
+            int d=((xpos+xInc*3)>>16) - xx;
+            int inc                = (d+1<4);
+            uint8_t *fragment      = (d+1<4) ? fragmentB       : fragmentA;
+            x86_reg imm8OfPShufW1  = (d+1<4) ? imm8OfPShufW1B  : imm8OfPShufW1A;
+            x86_reg imm8OfPShufW2  = (d+1<4) ? imm8OfPShufW2B  : imm8OfPShufW2A;
+            x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;
+            int maxShift= 3-(d+inc);
+            int shift=0;
+
+            if (filterCode) {
+                filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
+                filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
+                filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
+                filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
+                filterPos[i/2]= xx;
+
+                memcpy(filterCode + fragmentPos, fragment, fragmentLength);
+
+                filterCode[fragmentPos + imm8OfPShufW1]=
+                    (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);
+                filterCode[fragmentPos + imm8OfPShufW2]=
+                    a | (b<<2) | (c<<4) | (d<<6);
+
+                if (i+4-inc>=dstW) shift=maxShift; //avoid overread
+                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
+
+                if (shift && i>=shift) {
+                    filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;
+                    filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;
+                    filterPos[i/2]-=shift;
+                }
+            }
+
+            fragmentPos+= fragmentLength;
+
+            if (filterCode)
+                filterCode[fragmentPos]= RET;
+        }
+        xpos+=xInc;
+    }
+    if (filterCode)
+        filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
+
+    return fragmentPos + 1;
+}
+#endif /* ARCH_X86 && (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL */
+
+static void getSubSampleFactors(int *h, int *v, enum PixelFormat format)
+{
+    *h = av_pix_fmt_descriptors[format].log2_chroma_w;
+    *v = av_pix_fmt_descriptors[format].log2_chroma_h;
+}
+
+static uint16_t roundToInt16(int64_t f)
+{
+    int r= (f + (1<<15))>>16;
+         if (r<-0x7FFF) return 0x8000;
+    else if (r> 0x7FFF) return 0x7FFF;
+    else                return r;
+}
+
+int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation)
+{
+    int64_t crv =  inv_table[0];
+    int64_t cbu =  inv_table[1];
+    int64_t cgu = -inv_table[2];
+    int64_t cgv = -inv_table[3];
+    int64_t cy  = 1<<16;
+    int64_t oy  = 0;
+
+    memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
+    memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
+
+    c->brightness= brightness;
+    c->contrast  = contrast;
+    c->saturation= saturation;
+    c->srcRange  = srcRange;
+    c->dstRange  = dstRange;
+    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
+
+    c->uOffset=   0x0400040004000400LL;
+    c->vOffset=   0x0400040004000400LL;
+
+    if (!srcRange) {
+        cy= (cy*255) / 219;
+        oy= 16<<16;
+    } else {
+        crv= (crv*224) / 255;
+        cbu= (cbu*224) / 255;
+        cgu= (cgu*224) / 255;
+        cgv= (cgv*224) / 255;
+    }
+
+    cy = (cy *contrast             )>>16;
+    crv= (crv*contrast * saturation)>>32;
+    cbu= (cbu*contrast * saturation)>>32;
+    cgu= (cgu*contrast * saturation)>>32;
+    cgv= (cgv*contrast * saturation)>>32;
+
+    oy -= 256*brightness;
+
+    c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
+    c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
+    c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
+    c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
+    c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
+    c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
+
+    c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);
+    c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
+    c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
+    c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
+    c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
+    c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
+
+    ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
+    //FIXME factorize
+
+#if ARCH_PPC && (HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT)
+    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
+        ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
+#endif
+    return 0;
+}
+
+int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation)
+{
+    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
+
+    *inv_table = c->srcColorspaceTable;
+    *table     = c->dstColorspaceTable;
+    *srcRange  = c->srcRange;
+    *dstRange  = c->dstRange;
+    *brightness= c->brightness;
+    *contrast  = c->contrast;
+    *saturation= c->saturation;
+
+    return 0;
+}
+
+static int handle_jpeg(enum PixelFormat *format)
+{
+    switch (*format) {
+    case PIX_FMT_YUVJ420P:
+        *format = PIX_FMT_YUV420P;
+        return 1;
+    case PIX_FMT_YUVJ422P:
+        *format = PIX_FMT_YUV422P;
+        return 1;
+    case PIX_FMT_YUVJ444P:
+        *format = PIX_FMT_YUV444P;
+        return 1;
+    case PIX_FMT_YUVJ440P:
+        *format = PIX_FMT_YUV440P;
+        return 1;
+    default:
+        return 0;
+    }
+}
+
+SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat,
+                           int dstW, int dstH, enum PixelFormat dstFormat, int flags,
+                           SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
+{
+
+    SwsContext *c;
+    int i;
+    int usesVFilter, usesHFilter;
+    int unscaled;
+    int srcRange, dstRange;
+    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
+#if ARCH_X86
+    if (flags & SWS_CPU_CAPS_MMX)
+        __asm__ volatile("emms\n\t"::: "memory");
+#endif
+
+#if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
+    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
+    flags |= ff_hardcodedcpuflags();
+#endif /* CONFIG_RUNTIME_CPUDETECT */
+    if (!rgb15to16) sws_rgb2rgb_init(flags);
+
+    unscaled = (srcW == dstW && srcH == dstH);
+
+    srcRange = handle_jpeg(&srcFormat);
+    dstRange = handle_jpeg(&dstFormat);
+
+    if (!isSupportedIn(srcFormat)) {
+        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
+        return NULL;
+    }
+    if (!isSupportedOut(dstFormat)) {
+        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
+        return NULL;
+    }
+
+    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);
+    if(!i || (i & (i-1))) {
+        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
+        return NULL;
+    }
+
+    /* sanity check */
+    if (srcW<4 || srcH<1 || dstW<8 || dstH<1) { //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code
+        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
+               srcW, srcH, dstW, dstH);
+        return NULL;
+    }
+    if(srcW > VOFW || dstW > VOFW) {
+        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
+        return NULL;
+    }
+
+    if (!dstFilter) dstFilter= &dummyFilter;
+    if (!srcFilter) srcFilter= &dummyFilter;
+
+    FF_ALLOCZ_OR_GOTO(NULL, c, sizeof(SwsContext), fail);
+
+    c->av_class = &sws_context_class;
+    c->srcW= srcW;
+    c->srcH= srcH;
+    c->dstW= dstW;
+    c->dstH= dstH;
+    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
+    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
+    c->flags= flags;
+    c->dstFormat= dstFormat;
+    c->srcFormat= srcFormat;
+    c->vRounder= 4* 0x0001000100010001ULL;
+
+    usesHFilter= usesVFilter= 0;
+    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
+    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
+    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
+    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
+    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
+    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
+    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
+    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
+
+    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
+    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
+
+    // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
+    if ((isBGR(dstFormat) || isRGB(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 ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
+      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
+      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
+      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
+      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
+        c->chrSrcHSubSample=1;
+
+    if (param) {
+        c->param[0] = param[0];
+        c->param[1] = param[1];
+    } else {
+        c->param[0] =
+        c->param[1] = SWS_PARAM_DEFAULT;
+    }
+
+    // Note the -((-x)>>y) is so that we always round toward +inf.
+    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
+    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
+    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
+    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
+
+    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
+
+    /* unscaled special cases */
+    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(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",
+                       sws_format_name(srcFormat), sws_format_name(dstFormat));
+            return c;
+        }
+    }
+
+    if (flags & SWS_CPU_CAPS_MMX2) {
+        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
+        if (!c->canMMX2BeUsed && dstW >=srcW && (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 MMX2 scaler\n");
+        }
+        if (usesHFilter) c->canMMX2BeUsed=0;
+    }
+    else
+        c->canMMX2BeUsed=0;
+
+    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
+    c->chrYInc= ((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->canMMX2BeUsed) {
+            c->lumXInc+= 20;
+            c->chrXInc+= 20;
+        }
+        //we don't use the x86 asm scaler if MMX is available
+        else if (flags & SWS_CPU_CAPS_MMX) {
+            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
+            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
+        }
+    }
+
+    /* precalculate horizontal scaler filter coefficients */
+    {
+#if ARCH_X86 && (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
+// can't downscale !!!
+        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {
+            c->lumMmx2FilterCodeSize = initMMX2HScaler(      dstW, c->lumXInc, NULL, NULL, NULL, 8);
+            c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);
+
+#ifdef MAP_ANONYMOUS
+            c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
+            c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
+#elif HAVE_VIRTUALALLOC
+            c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
+            c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
+#else
+            c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize);
+            c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);
+#endif
+
+            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);
+
+            initMMX2HScaler(      dstW, c->lumXInc, c->lumMmx2FilterCode, c->hLumFilter, c->hLumFilterPos, 8);
+            initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->hChrFilter, c->hChrFilterPos, 4);
+
+#ifdef MAP_ANONYMOUS
+            mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
+            mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
+#endif
+        } else
+#endif /* ARCH_X86 && (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL */
+        {
+            const int filterAlign=
+                (flags & SWS_CPU_CAPS_MMX) ? 4 :
+                (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
+                1;
+
+            if (initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
+                           srcW      ,       dstW, filterAlign, 1<<14,
+                           (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
+                           srcFilter->lumH, dstFilter->lumH, c->param) < 0)
+                goto fail;
+            if (initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
+                           c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
+                           (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
+                           srcFilter->chrH, dstFilter->chrH, c->param) < 0)
+                goto fail;
+        }
+    } // initialize horizontal stuff
+
+    /* precalculate vertical scaler filter coefficients */
+    {
+        const int filterAlign=
+            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
+            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
+            1;
+
+        if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
+                       srcH      ,        dstH, filterAlign, (1<<12),
+                       (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
+                       srcFilter->lumV, dstFilter->lumV, c->param) < 0)
+            goto fail;
+        if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
+                       c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
+                       (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
+                       srcFilter->chrV, dstFilter->chrV, c->param) < 0)
+            goto fail;
+
+#if ARCH_PPC && (HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT)
+        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= 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];
+    }
+
+    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
+    // allocate several megabytes to handle all possible cases)
+    FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);
+    FF_ALLOC_OR_GOTO(c, c->chrPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail);
+    if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
+        FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);
+    //Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna 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], VOF+1, fail);
+        c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize];
+    }
+    for (i=0; i<c->vChrBufSize; i++) {
+        FF_ALLOC_OR_GOTO(c, c->chrPixBuf[i+c->vChrBufSize], (VOF+1)*2, fail);
+        c->chrPixBuf[i] = c->chrPixBuf[i+c->vChrBufSize];
+    }
+    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
+        for (i=0; i<c->vLumBufSize; i++) {
+            FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i+c->vLumBufSize], VOF+1, 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++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
+
+    assert(2*VOFW == VOF);
+
+    assert(c->chrDstH <= dstH);
+
+    if (flags&SWS_PRINT_INFO) {
+        if (flags&SWS_FAST_BILINEAR)
+            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
+        else if (flags&SWS_BILINEAR)
+            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
+        else if (flags&SWS_BICUBIC)
+            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
+        else if (flags&SWS_X)
+            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
+        else if (flags&SWS_POINT)
+            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
+        else if (flags&SWS_AREA)
+            av_log(c, AV_LOG_INFO, "Area Averaging scaler, ");
+        else if (flags&SWS_BICUBLIN)
+            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
+        else if (flags&SWS_GAUSS)
+            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
+        else if (flags&SWS_SINC)
+            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
+        else if (flags&SWS_LANCZOS)
+            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
+        else if (flags&SWS_SPLINE)
+            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
+        else
+            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
+
+        av_log(c, AV_LOG_INFO, "from %s to %s%s ",
+               sws_format_name(srcFormat),
+#ifdef DITHER1XBPP
+               dstFormat == PIX_FMT_BGR555 || dstFormat == PIX_FMT_BGR565 ? "dithered " : "",
+#else
+               "",
+#endif
+               sws_format_name(dstFormat));
+
+        if (flags & SWS_CPU_CAPS_MMX2)
+            av_log(c, AV_LOG_INFO, "using MMX2\n");
+        else if (flags & SWS_CPU_CAPS_3DNOW)
+            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
+        else if (flags & SWS_CPU_CAPS_MMX)
+            av_log(c, AV_LOG_INFO, "using MMX\n");
+        else if (flags & SWS_CPU_CAPS_ALTIVEC)
+            av_log(c, AV_LOG_INFO, "using AltiVec\n");
+        else
+            av_log(c, AV_LOG_INFO, "using C\n");
+    }
+
+    if (flags & SWS_PRINT_INFO) {
+        if (flags & SWS_CPU_CAPS_MMX) {
+            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
+                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
+            else {
+                if (c->hLumFilterSize==4)
+                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
+                else if (c->hLumFilterSize==8)
+                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
+                else
+                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
+
+                if (c->hChrFilterSize==4)
+                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
+                else if (c->hChrFilterSize==8)
+                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
+                else
+                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
+            }
+        } else {
+#if ARCH_X86
+            av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
+#else
+            if (flags & SWS_FAST_BILINEAR)
+                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
+            else
+                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
+#endif
+        }
+        if (isPlanarYUV(dstFormat)) {
+            if (c->vLumFilterSize==1)
+                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
+            else
+                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
+        } else {
+            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
+                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
+                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
+            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
+                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
+            else
+                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
+        }
+
+        if (dstFormat==PIX_FMT_BGR24)
+            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
+                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
+        else if (dstFormat==PIX_FMT_RGB32)
+            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
+        else if (dstFormat==PIX_FMT_BGR565)
+            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
+        else if (dstFormat==PIX_FMT_BGR555)
+            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
+
+        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
+    }
+    if (flags & SWS_PRINT_INFO) {
+        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);
+    }
+
+    c->swScale= ff_getSwsFunc(c);
+    return c;
+
+fail:
+    sws_freeContext(c);
+    return NULL;
+}
+
+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 (chromaSharpen!=0.0) {
+        SwsVector *id= sws_getIdentityVec();
+        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();
+        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 (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
+    if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
+
+    return filter;
+}
+
+SwsVector *sws_allocVec(int length)
+{
+    SwsVector *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= 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*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);
+}
+
+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);
+    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);
+    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);
+    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);
+    av_free(a->coeff);
+    a->coeff= conv->coeff;
+    a->length= conv->length;
+    av_free(conv);
+}
+
+SwsVector *sws_cloneVec(SwsVector *a)
+{
+    int i;
+    SwsVector *vec= sws_allocVec(a->length);
+
+    if (!vec)
+        return NULL;
+
+    for (i=0; i<a->length; i++) vec->coeff[i]= a->coeff[i];
+
+    return vec;
+}
+
+void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
+{
+    int i;
+    double max=0;
+    double min=0;
+    double range;
+
+    for (i=0; i<a->length; i++)
+        if (a->coeff[i]>max) max= a->coeff[i];
+
+    for (i=0; i<a->length; i++)
+        if (a->coeff[i]<min) min= a->coeff[i];
+
+    range= max - min;
+
+    for (i=0; i<a->length; i++) {
+        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
+        av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
+        for (;x>0; x--) av_log(log_ctx, log_level, " ");
+        av_log(log_ctx, log_level, "|\n");
+    }
+}
+
+#if LIBSWSCALE_VERSION_MAJOR < 1
+void sws_printVec(SwsVector *a)
+{
+    sws_printVec2(a, NULL, AV_LOG_DEBUG);
+}
+#endif
+
+void sws_freeVec(SwsVector *a)
+{
+    if (!a) return;
+    av_freep(&a->coeff);
+    a->length=0;
+    av_free(a);
+}
+
+void sws_freeFilter(SwsFilter *filter)
+{
+    if (!filter) return;
+
+    if (filter->lumH) sws_freeVec(filter->lumH);
+    if (filter->lumV) sws_freeVec(filter->lumV);
+    if (filter->chrH) sws_freeVec(filter->chrH);
+    if (filter->chrV) sws_freeVec(filter->chrV);
+    av_free(filter);
+}
+
+void sws_freeContext(SwsContext *c)
+{
+    int i;
+    if (!c) return;
+
+    if (c->lumPixBuf) {
+        for (i=0; i<c->vLumBufSize; i++)
+            av_freep(&c->lumPixBuf[i]);
+        av_freep(&c->lumPixBuf);
+    }
+
+    if (c->chrPixBuf) {
+        for (i=0; i<c->vChrBufSize; i++)
+            av_freep(&c->chrPixBuf[i]);
+        av_freep(&c->chrPixBuf);
+    }
+
+    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
+        for (i=0; i<c->vLumBufSize; i++)
+            av_freep(&c->alpPixBuf[i]);
+        av_freep(&c->alpPixBuf);
+    }
+
+    av_freep(&c->vLumFilter);
+    av_freep(&c->vChrFilter);
+    av_freep(&c->hLumFilter);
+    av_freep(&c->hChrFilter);
+#if ARCH_PPC && (HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT)
+    av_freep(&c->vYCoeffsBank);
+    av_freep(&c->vCCoeffsBank);
+#endif
+
+    av_freep(&c->vLumFilterPos);
+    av_freep(&c->vChrFilterPos);
+    av_freep(&c->hLumFilterPos);
+    av_freep(&c->hChrFilterPos);
+
+#if ARCH_X86 && CONFIG_GPL
+#ifdef MAP_ANONYMOUS
+    if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);
+    if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);
+#elif HAVE_VIRTUALALLOC
+    if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, MEM_RELEASE);
+    if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, MEM_RELEASE);
+#else
+    av_free(c->lumMmx2FilterCode);
+    av_free(c->chrMmx2FilterCode);
+#endif
+    c->lumMmx2FilterCode=NULL;
+    c->chrMmx2FilterCode=NULL;
+#endif /* ARCH_X86 && CONFIG_GPL */
+
+    av_freep(&c->yuvTable);
+
+    av_free(c);
+}
+
+struct SwsContext *sws_getCachedContext(struct SwsContext *context,
+                                        int srcW, int srcH, enum PixelFormat srcFormat,
+                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
+                                        SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
+{
+    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
+
+    if (!param)
+        param = default_param;
+
+    if (context) {
+        if (context->srcW != srcW || context->srcH != srcH ||
+            context->srcFormat != srcFormat ||
+            context->dstW != dstW || context->dstH != dstH ||
+            context->dstFormat != dstFormat || context->flags != flags ||
+            context->param[0] != param[0] || context->param[1] != param[1])
+        {
+            sws_freeContext(context);
+            context = NULL;
+        }
+    }
+    if (!context) {
+        return sws_getContext(srcW, srcH, srcFormat,
+                              dstW, dstH, dstFormat, flags,
+                              srcFilter, dstFilter, param);
+    }
+    return context;
+}
+