cleanup
authorMichael Niedermayer <michaelni@gmx.at>
Thu, 13 Feb 2003 14:19:05 +0000 (14:19 +0000)
committerMichael Niedermayer <michaelni@gmx.at>
Thu, 13 Feb 2003 14:19:05 +0000 (14:19 +0000)
Originally committed as revision 9414 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc

postproc/swscale.c
postproc/swscale.h
postproc/swscale_template.c

index 6936f71..044085b 100644 (file)
@@ -2019,6 +2019,15 @@ SwsContext *getSwsContext(int srcW, int srcH, int srcFormat, int dstW, int dstH,
        c->dstFormat= dstFormat;
        c->srcFormat= srcFormat;
 
+       c->yCoeff=    0x2568256825682568LL;
+       c->vrCoeff=   0x3343334333433343LL;
+       c->ubCoeff=   0x40cf40cf40cf40cfLL;
+       c->vgCoeff=   0xE5E2E5E2E5E2E5E2LL;
+       c->ugCoeff=   0xF36EF36EF36EF36ELL;
+       c->yOffset=   0x0080008000800080LL;
+       c->uOffset=   0x0400040004000400LL;
+       c->vOffset=   0x0400040004000400LL;
+
        usesFilter=0;
        if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesFilter=1;
        if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesFilter=1;
@@ -2261,19 +2270,6 @@ SwsContext *getSwsContext(int srcW, int srcH, int srcFormat, int dstW, int dstH,
 
        ASSERT(c->chrDstH <= dstH)
 
-       // pack filter data for mmx code
-       if(cpuCaps.hasMMX)
-       {
-               c->lumMmxFilter= (int16_t*)memalign(8, c->vLumFilterSize*      dstH*4*sizeof(int16_t));
-               c->chrMmxFilter= (int16_t*)memalign(8, c->vChrFilterSize*c->chrDstH*4*sizeof(int16_t));
-               for(i=0; i<c->vLumFilterSize*dstH; i++)
-                       c->lumMmxFilter[4*i]=c->lumMmxFilter[4*i+1]=c->lumMmxFilter[4*i+2]=c->lumMmxFilter[4*i+3]=
-                               c->vLumFilter[i];
-               for(i=0; i<c->vChrFilterSize*c->chrDstH; i++)
-                       c->chrMmxFilter[4*i]=c->chrMmxFilter[4*i+1]=c->chrMmxFilter[4*i+2]=c->chrMmxFilter[4*i+3]=
-                               c->vChrFilter[i];
-       }
-
        if(flags&SWS_PRINT_INFO)
        {
 #ifdef DITHER1XBPP
@@ -2668,11 +2664,6 @@ void freeSwsContext(SwsContext *c){
        if(c->hChrFilterPos) free(c->hChrFilterPos);
        c->hChrFilterPos = NULL;
 
-       if(c->lumMmxFilter) free(c->lumMmxFilter);
-       c->lumMmxFilter = NULL;
-       if(c->chrMmxFilter) free(c->chrMmxFilter);
-       c->chrMmxFilter = NULL;
-
        if(c->lumMmx2Filter) free(c->lumMmx2Filter);
        c->lumMmx2Filter=NULL;
        if(c->chrMmx2Filter) free(c->chrMmx2Filter);
index 54f8c71..a03927c 100644 (file)
@@ -44,6 +44,7 @@
 #define SWS_FULL_CHR_H_INP     0x4000
 #define SWS_DIRECT_BGR         0x8000
 
+#define MAX_FILTER_SIZE 256
 
 #define SWS_MAX_REDUCE_CUTOFF 0.002
 
@@ -70,9 +71,6 @@ typedef struct SwsContext{
        int16_t *vChrFilter;
        int16_t *vChrFilterPos;
 
-// Contain simply the values from v(Lum|Chr)Filter just nicely packed for mmx
-       int16_t  *lumMmxFilter;
-       int16_t  *chrMmxFilter;
        uint8_t formatConvBuffer[4000]; //FIXME dynamic alloc, but we have to change alot of code for this to be usefull
 
        int hLumFilterSize;
@@ -105,8 +103,40 @@ typedef struct SwsContext{
 
        void (*swScale)(struct SwsContext *context, uint8_t* src[], int srcStride[], int srcSliceY,
              int srcSliceH, uint8_t* dst[], int dstStride[]);
+
+#define RED_DITHER   "0*8"
+#define GREEN_DITHER "1*8"
+#define BLUE_DITHER  "2*8"
+#define Y_COEFF      "3*8"
+#define VR_COEFF     "4*8"
+#define UB_COEFF     "5*8"
+#define VG_COEFF     "6*8"
+#define UG_COEFF     "7*8"
+#define Y_OFFSET     "8*8"
+#define U_OFFSET     "9*8"
+#define V_OFFSET     "10*8"
+#define LUM_MMX_FILTER_OFFSET "11*8"
+#define CHR_MMX_FILTER_OFFSET "11*8+4*4*256"
+                  
+       uint64_t redDither   __attribute__((aligned(8)));
+       uint64_t greenDither __attribute__((aligned(8)));
+       uint64_t blueDither  __attribute__((aligned(8)));
+
+       uint64_t yCoeff      __attribute__((aligned(8)));
+       uint64_t vrCoeff     __attribute__((aligned(8)));
+       uint64_t ubCoeff     __attribute__((aligned(8)));
+       uint64_t vgCoeff     __attribute__((aligned(8)));
+       uint64_t ugCoeff     __attribute__((aligned(8)));
+       uint64_t yOffset     __attribute__((aligned(8)));
+       uint64_t uOffset     __attribute__((aligned(8)));
+       uint64_t vOffset     __attribute__((aligned(8)));
+       int32_t  lumMmxFilter[4*MAX_FILTER_SIZE];
+       int32_t  chrMmxFilter[4*MAX_FILTER_SIZE];
+       
 } SwsContext;
 //FIXME check init (where 0)
+//FIXME split private & public
+
 
 // when used for filters they must have an odd number of elements
 // coeffs cannot be shared between vectors
index 092f11b..dc87558 100644 (file)
 #define MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
 #endif
 
-#define YSCALEYUV2YV12X(x) \
+#define YSCALEYUV2YV12X(x, offset) \
                        "xorl %%eax, %%eax              \n\t"\
                        "pxor %%mm3, %%mm3              \n\t"\
                        "pxor %%mm4, %%mm4              \n\t"\
-                       "movl %0, %%edx                 \n\t"\
+                       "leal " offset "(%0), %%edx     \n\t"\
+                       "movl (%%edx), %%esi            \n\t"\
                        ".balign 16                     \n\t" /* FIXME Unroll? */\
                        "1:                             \n\t"\
-                       "movl (%1, %%edx, 4), %%esi     \n\t"\
-                       "movq (%2, %%edx, 8), %%mm0     \n\t" /* filterCoeff */\
+                       "movq 8(%%edx), %%mm0           \n\t" /* filterCoeff */\
                        "movq " #x "(%%esi, %%eax, 2), %%mm2    \n\t" /* srcData */\
                        "movq 8+" #x "(%%esi, %%eax, 2), %%mm5  \n\t" /* srcData */\
+                       "addl $16, %%edx                \n\t"\
+                       "movl (%%edx), %%esi            \n\t"\
+                       "testl %%esi, %%esi             \n\t"\
                        "pmulhw %%mm0, %%mm2            \n\t"\
                        "pmulhw %%mm0, %%mm5            \n\t"\
                        "paddw %%mm2, %%mm3             \n\t"\
                        "paddw %%mm5, %%mm4             \n\t"\
-                       "addl $1, %%edx                 \n\t"\
                        " jnz 1b                        \n\t"\
                        "psraw $3, %%mm3                \n\t"\
                        "psraw $3, %%mm4                \n\t"\
                        "packuswb %%mm4, %%mm3          \n\t"\
-                       MOVNTQ(%%mm3, (%3, %%eax))\
+                       MOVNTQ(%%mm3, (%1, %%eax))\
                        "addl $8, %%eax                 \n\t"\
-                       "cmpl %4, %%eax                 \n\t"\
+                       "cmpl %2, %%eax                 \n\t"\
                        "pxor %%mm3, %%mm3              \n\t"\
                        "pxor %%mm4, %%mm4              \n\t"\
-                       "movl %0, %%edx                 \n\t"\
+                       "leal " offset "(%0), %%edx     \n\t"\
+                       "movl (%%edx), %%esi            \n\t"\
                        "jb 1b                          \n\t"
 
 #define YSCALEYUV2YV121 \
 #define YSCALEYUV2PACKEDX \
                "xorl %%eax, %%eax              \n\t"\
                ".balign 16                     \n\t"\
+               "nop                            \n\t"\
                "1:                             \n\t"\
-               "movl %1, %%edx                 \n\t" /* -chrFilterSize */\
-               "movl %3, %%ebx                 \n\t" /* chrMmxFilter+chrFilterSize */\
-               "movl %7, %%ecx                 \n\t" /* chrSrc+chrFilterSize */\
+               "leal "CHR_MMX_FILTER_OFFSET"(%0), %%edx        \n\t"\
+               "movl (%%edx), %%esi            \n\t"\
                "pxor %%mm3, %%mm3              \n\t"\
                "pxor %%mm4, %%mm4              \n\t"\
+               ".balign 16                     \n\t"\
                "2:                             \n\t"\
-               "movl (%%ecx, %%edx, 4), %%esi  \n\t"\
-               "movq (%%ebx, %%edx, 8), %%mm0  \n\t" /* filterCoeff */\
+               "movq 8(%%edx), %%mm0           \n\t" /* filterCoeff */\
                "movq (%%esi, %%eax), %%mm2     \n\t" /* UsrcData */\
                "movq 4096(%%esi, %%eax), %%mm5 \n\t" /* VsrcData */\
+               "addl $16, %%edx                \n\t"\
+               "movl (%%edx), %%esi            \n\t"\
                "pmulhw %%mm0, %%mm2            \n\t"\
                "pmulhw %%mm0, %%mm5            \n\t"\
                "paddw %%mm2, %%mm3             \n\t"\
                "paddw %%mm5, %%mm4             \n\t"\
-               "addl $1, %%edx                 \n\t"\
+               "testl %%esi, %%esi             \n\t"\
                " jnz 2b                        \n\t"\
 \
-               "movl %0, %%edx                 \n\t" /* -lumFilterSize */\
-               "movl %2, %%ebx                 \n\t" /* lumMmxFilter+lumFilterSize */\
-               "movl %6, %%ecx                 \n\t" /* lumSrc+lumFilterSize */\
+               "leal "LUM_MMX_FILTER_OFFSET"(%0), %%edx        \n\t"\
+               "movl (%%edx), %%esi            \n\t"\
                "pxor %%mm1, %%mm1              \n\t"\
                "pxor %%mm7, %%mm7              \n\t"\
+               ".balign 16                     \n\t"\
                "2:                             \n\t"\
-               "movl (%%ecx, %%edx, 4), %%esi  \n\t"\
-               "movq (%%ebx, %%edx, 8), %%mm0  \n\t" /* filterCoeff */\
+               "movq 8(%%edx), %%mm0           \n\t" /* filterCoeff */\
                "movq (%%esi, %%eax, 2), %%mm2  \n\t" /* Y1srcData */\
                "movq 8(%%esi, %%eax, 2), %%mm5 \n\t" /* Y2srcData */\
+               "addl $16, %%edx                \n\t"\
+               "movl (%%edx), %%esi            \n\t"\
                "pmulhw %%mm0, %%mm2            \n\t"\
                "pmulhw %%mm0, %%mm5            \n\t"\
                "paddw %%mm2, %%mm1             \n\t"\
                "paddw %%mm5, %%mm7             \n\t"\
-               "addl $1, %%edx                 \n\t"\
+               "testl %%esi, %%esi             \n\t"\
                " jnz 2b                        \n\t"\
 
 
 #define YSCALEYUV2RGBX \
                YSCALEYUV2PACKEDX\
-               "psubw "MANGLE(w400)", %%mm3    \n\t" /* (U-128)8*/\
-               "psubw "MANGLE(w400)", %%mm4    \n\t" /* (V-128)8*/\
+               "psubw "U_OFFSET"(%0), %%mm3    \n\t" /* (U-128)8*/\
+               "psubw "V_OFFSET"(%0), %%mm4    \n\t" /* (V-128)8*/\
                "movq %%mm3, %%mm2              \n\t" /* (U-128)8*/\
                "movq %%mm4, %%mm5              \n\t" /* (V-128)8*/\
-               "pmulhw "MANGLE(ugCoeff)", %%mm3\n\t"\
-               "pmulhw "MANGLE(vgCoeff)", %%mm4\n\t"\
+               "pmulhw "UG_COEFF"(%0), %%mm3   \n\t"\
+               "pmulhw "VG_COEFF"(%0), %%mm4   \n\t"\
        /* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
-               "pmulhw "MANGLE(ubCoeff)", %%mm2\n\t"\
-               "pmulhw "MANGLE(vrCoeff)", %%mm5\n\t"\
-               "psubw "MANGLE(w80)", %%mm1     \n\t" /* 8(Y-16)*/\
-               "psubw "MANGLE(w80)", %%mm7     \n\t" /* 8(Y-16)*/\
-               "pmulhw "MANGLE(yCoeff)", %%mm1 \n\t"\
-               "pmulhw "MANGLE(yCoeff)", %%mm7 \n\t"\
+               "pmulhw "UB_COEFF"(%0), %%mm2   \n\t"\
+               "pmulhw "VR_COEFF"(%0), %%mm5   \n\t"\
+               "psubw "Y_OFFSET"(%0), %%mm1    \n\t" /* 8(Y-16)*/\
+               "psubw "Y_OFFSET"(%0), %%mm7    \n\t" /* 8(Y-16)*/\
+               "pmulhw "Y_COEFF"(%0), %%mm1    \n\t"\
+               "pmulhw "Y_COEFF"(%0), %%mm7    \n\t"\
        /* mm1= Y1, mm2=ub, mm3=ug, mm4=vg mm5=vr, mm7=Y2 */\
                "paddw %%mm3, %%mm4             \n\t"\
                "movq %%mm2, %%mm0              \n\t"\
                "packuswb %%mm6, %%mm5          \n\t"\
                "packuswb %%mm3, %%mm4          \n\t"\
                "pxor %%mm7, %%mm7              \n\t"
-
+#if 0
 #define FULL_YSCALEYUV2RGB \
                "pxor %%mm7, %%mm7              \n\t"\
                "movd %6, %%mm6                 \n\t" /*yalpha1*/\
                "paddw %%mm2, %%mm1             \n\t" /* G*/\
 \
                "packuswb %%mm1, %%mm1          \n\t"
+#endif
 
 #define YSCALEYUV2PACKED \
                "movd %6, %%mm6                 \n\t" /*yalpha1*/\
                        " jb 1b                         \n\t"
 
 
-static inline void RENAME(yuv2yuvX)(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
+static inline void RENAME(yuv2yuvX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
                                    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
                                    uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW,
-                                   int16_t * lumMmxFilter, int16_t * chrMmxFilter)
+                                   int32_t * lumMmxFilter, int32_t * chrMmxFilter)
 {
+       int dummy=0;
 #ifdef HAVE_MMX
        if(uDest != NULL)
        {
                asm volatile(
-                               YSCALEYUV2YV12X(0)
-                               :: "m" (-chrFilterSize), "r" (chrSrc+chrFilterSize),
-                               "r" (chrMmxFilter+chrFilterSize*4), "r" (uDest), "m" (chrDstW)
+                               YSCALEYUV2YV12X(0, CHR_MMX_FILTER_OFFSET)
+                               :: "r" (&c->redDither),
+                               "r" (uDest), "m" (chrDstW)
                                : "%eax", "%edx", "%esi"
                        );
 
                asm volatile(
-                               YSCALEYUV2YV12X(4096)
-                               :: "m" (-chrFilterSize), "r" (chrSrc+chrFilterSize),
-                               "r" (chrMmxFilter+chrFilterSize*4), "r" (vDest), "m" (chrDstW)
+                               YSCALEYUV2YV12X(4096, CHR_MMX_FILTER_OFFSET)
+                               :: "r" (&c->redDither),
+                               "r" (vDest), "m" (chrDstW)
                                : "%eax", "%edx", "%esi"
                        );
        }
 
        asm volatile(
-                       YSCALEYUV2YV12X(0)
-                       :: "m" (-lumFilterSize), "r" (lumSrc+lumFilterSize),
-                          "r" (lumMmxFilter+lumFilterSize*4), "r" (dest), "m" (dstW)
+                       YSCALEYUV2YV12X(0, LUM_MMX_FILTER_OFFSET)
+                       :: "r" (&c->redDither),
+                          "r" (dest), "m" (dstW)
                        : "%eax", "%edx", "%esi"
                );
 #else
@@ -844,8 +852,9 @@ static inline void RENAME(yuv2yuv1)(int16_t *lumSrc, int16_t *chrSrc,
  */
 static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
                                    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
-                           uint8_t *dest, int dstW, int16_t * lumMmxFilter, int16_t * chrMmxFilter, int dstY)
+                           uint8_t *dest, int dstW, int dstY)
 {
+       int dummy=0;
        switch(c->dstFormat)
        {
 #ifdef HAVE_MMX
@@ -855,11 +864,10 @@ static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_
                                YSCALEYUV2RGBX
                                WRITEBGR32
 
-                       :: "m" (-lumFilterSize), "m" (-chrFilterSize),
-                          "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4),
-                          "r" (dest), "m" (dstW),
-                          "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize)
-                       : "%eax", "%ebx", "%ecx", "%edx", "%esi"
+                       :: "r" (&c->redDither), 
+                          "m" (dummy), "m" (dummy), "m" (dummy),
+                          "r" (dest), "m" (dstW)
+                       : "%eax", "%edx", "%esi"
                        );
                }
                break;
@@ -871,11 +879,10 @@ static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_
                                "addl %4, %%ebx                 \n\t"
                                WRITEBGR24
 
-                       :: "m" (-lumFilterSize), "m" (-chrFilterSize),
-                          "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4),
-                          "r" (dest), "m" (dstW),
-                          "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize)
-                       : "%eax", "%ebx", "%ecx", "%edx", "%esi"
+                       :: "r" (&c->redDither), 
+                          "m" (dummy), "m" (dummy), "m" (dummy),
+                          "r" (dest), "m" (dstW)
+                       : "%eax", "%ebx", "%edx", "%esi" //FIXME ebx
                        );
                }
                break;
@@ -892,11 +899,10 @@ static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_
 
                                WRITEBGR15
 
-                       :: "m" (-lumFilterSize), "m" (-chrFilterSize),
-                          "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4),
-                          "r" (dest), "m" (dstW),
-                          "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize)
-                       : "%eax", "%ebx", "%ecx", "%edx", "%esi"
+                       :: "r" (&c->redDither), 
+                          "m" (dummy), "m" (dummy), "m" (dummy),
+                          "r" (dest), "m" (dstW)
+                       : "%eax", "%edx", "%esi"
                        );
                }
                break;
@@ -913,11 +919,10 @@ static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_
 
                                WRITEBGR16
 
-                       :: "m" (-lumFilterSize), "m" (-chrFilterSize),
-                          "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4),
-                          "r" (dest), "m" (dstW),
-                          "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize)
-                       : "%eax", "%ebx", "%ecx", "%edx", "%esi"
+                       :: "r" (&c->redDither), 
+                          "m" (dummy), "m" (dummy), "m" (dummy),
+                          "r" (dest), "m" (dstW)
+                       : "%eax", "%edx", "%esi"
                        );
                }
                break;
@@ -933,11 +938,10 @@ static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_
                                "psraw $3, %%mm7                \n\t"
                                WRITEYUY2
 
-                       :: "m" (-lumFilterSize), "m" (-chrFilterSize),
-                          "m" (lumMmxFilter+lumFilterSize*4), "m" (chrMmxFilter+chrFilterSize*4),
-                          "r" (dest), "m" (dstW),
-                          "m" (lumSrc+lumFilterSize), "m" (chrSrc+chrFilterSize)
-                       : "%eax", "%ebx", "%ecx", "%edx", "%esi"
+                       :: "r" (&c->redDither), 
+                          "m" (dummy), "m" (dummy), "m" (dummy),
+                          "r" (dest), "m" (dstW)
+                       : "%eax", "%edx", "%esi"
                        );
                }
                break;
@@ -2528,8 +2532,8 @@ static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStridePar
        int16_t *vChrFilter= c->vChrFilter;
        int16_t *hLumFilter= c->hLumFilter;
        int16_t *hChrFilter= c->hChrFilter;
-       int16_t *lumMmxFilter= c->lumMmxFilter;
-       int16_t *chrMmxFilter= c->chrMmxFilter;
+       int32_t *lumMmxFilter= c->lumMmxFilter;
+       int32_t *chrMmxFilter= c->chrMmxFilter;
        const int vLumFilterSize= c->vLumFilterSize;
        const int vChrFilterSize= c->vChrFilterSize;
        const int hLumFilterSize= c->hLumFilterSize;
@@ -2729,11 +2733,28 @@ i--;
                        {
                                int16_t **lumSrcPtr= lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
                                int16_t **chrSrcPtr= chrPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
-                               RENAME(yuv2yuvX)(
+                               int i;
+#ifdef HAVE_MMX
+                               for(i=0; i<vLumFilterSize; i++)
+                               {
+                                       lumMmxFilter[4*i+0]= (int32_t)lumSrcPtr[i];
+                                       lumMmxFilter[4*i+2]= 
+                                       lumMmxFilter[4*i+3]= 
+                                               ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001;
+                               }
+                               for(i=0; i<vChrFilterSize; i++)
+                               {
+                                       chrMmxFilter[4*i+0]= (int32_t)chrSrcPtr[i];
+                                       chrMmxFilter[4*i+2]= 
+                                       chrMmxFilter[4*i+3]= 
+                                               ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001;
+                               }
+#endif
+                               RENAME(yuv2yuvX)(c,
                                        vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
                                        vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
                                        dest, uDest, vDest, dstW, chrDstW,
-                                       lumMmxFilter+dstY*vLumFilterSize*4, chrMmxFilter+chrDstY*vChrFilterSize*4);
+                                       lumMmxFilter, chrMmxFilter);
                        }
                }
                else
@@ -2760,11 +2781,27 @@ i--;
                        }
                        else //General RGB
                        {
+                               int i;
+#ifdef HAVE_MMX
+                               for(i=0; i<vLumFilterSize; i++)
+                               {
+                                       lumMmxFilter[4*i+0]= (int32_t)lumSrcPtr[i];
+                                       lumMmxFilter[4*i+2]= 
+                                       lumMmxFilter[4*i+3]= 
+                                               ((uint16_t)vLumFilter[dstY*vLumFilterSize + i])*0x10001;
+                               }
+                               for(i=0; i<vChrFilterSize; i++)
+                               {
+                                       chrMmxFilter[4*i+0]= (int32_t)chrSrcPtr[i];
+                                       chrMmxFilter[4*i+2]= 
+                                       chrMmxFilter[4*i+3]= 
+                                               ((uint16_t)vChrFilter[chrDstY*vChrFilterSize + i])*0x10001;
+                               }
+#endif
                                RENAME(yuv2packedX)(c,
                                        vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
                                        vChrFilter+dstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
-                                       dest, dstW,
-                                       lumMmxFilter+dstY*vLumFilterSize*4, chrMmxFilter+dstY*vChrFilterSize*4, dstY);
+                                       dest, dstW, dstY);
                        }
                }
             }