X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=postproc%2Fswscale_template.c;h=e76020eab7a48e0799e2afae012ac9805232ea98;hb=eec1c6b94cb2796faaebd5bbffb40bf61556625d;hp=ba8a8c1fe127aed492498fbbc2771a2624f355d1;hpb=6ff0ad6bfd0f00a3d54705811ee91a7ce3c22cda;p=ffmpeg

diff --git a/postproc/swscale_template.c b/postproc/swscale_template.c
index ba8a8c1fe12..e76020eab7a 100644
--- a/postproc/swscale_template.c
+++ b/postproc/swscale_template.c
@@ -993,6 +993,9 @@ FULL_YSCALEYUV2RGB
 		if(dstFormat==IMGFMT_BGR32)
 		{
 			int i;
+#ifdef WORDS_BIGENDIAN
+			dest++;
+#endif
 			for(i=0;i<dstW;i++){
 				// vertical linear interpolation && yuv2rgb in a single step:
 				int Y=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];
@@ -1116,6 +1119,9 @@ FULL_YSCALEYUV2RGB
 		if(dstFormat==IMGFMT_BGR32)
 		{
 			int i;
+#ifdef WORDS_BIGENDIAN
+			dest++;
+#endif
 			for(i=0; i<dstW-1; i+=2){
 				// vertical linear interpolation && yuv2rgb in a single step:
 				int Y1=yuvtab_2568[((buf0[i]*yalpha1+buf1[i]*yalpha)>>19)];
@@ -1394,6 +1400,9 @@ static inline void RENAME(yuv2rgb1)(uint16_t *buf0, uint16_t *uvbuf0, uint16_t *
 	if(dstFormat==IMGFMT_BGR32)
 	{
 		int i;
+#ifdef WORDS_BIGENDIAN
+		dest++;
+#endif
 		for(i=0; i<dstW-1; i+=2){
 			// vertical linear interpolation && yuv2rgb in a single step:
 			int Y1=yuvtab_2568[buf0[i]>>7];
@@ -1635,7 +1644,76 @@ static inline void RENAME(bgr32ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1
 
 static inline void RENAME(bgr24ToY)(uint8_t *dst, uint8_t *src, int width)
 {
-#ifdef HAVE_MMXFIXME
+#ifdef HAVE_MMX
+	asm volatile(
+		"movl %2, %%eax			\n\t"
+		"movq "MANGLE(bgr2YCoeff)", %%mm6		\n\t"
+		"movq "MANGLE(w1111)", %%mm5		\n\t"
+		"pxor %%mm7, %%mm7		\n\t"
+		"leal (%%eax, %%eax, 2), %%ebx	\n\t"
+		".balign 16			\n\t"
+		"1:				\n\t"
+		PREFETCH" 64(%0, %%ebx)		\n\t"
+		"movd (%0, %%ebx), %%mm0	\n\t"
+		"movd 3(%0, %%ebx), %%mm1	\n\t"
+		"punpcklbw %%mm7, %%mm0		\n\t"
+		"punpcklbw %%mm7, %%mm1		\n\t"
+		"movd 6(%0, %%ebx), %%mm2	\n\t"
+		"movd 9(%0, %%ebx), %%mm3	\n\t"
+		"punpcklbw %%mm7, %%mm2		\n\t"
+		"punpcklbw %%mm7, %%mm3		\n\t"
+		"pmaddwd %%mm6, %%mm0		\n\t"
+		"pmaddwd %%mm6, %%mm1		\n\t"
+		"pmaddwd %%mm6, %%mm2		\n\t"
+		"pmaddwd %%mm6, %%mm3		\n\t"
+#ifndef FAST_BGR2YV12
+		"psrad $8, %%mm0		\n\t"
+		"psrad $8, %%mm1		\n\t"
+		"psrad $8, %%mm2		\n\t"
+		"psrad $8, %%mm3		\n\t"
+#endif
+		"packssdw %%mm1, %%mm0		\n\t"
+		"packssdw %%mm3, %%mm2		\n\t"
+		"pmaddwd %%mm5, %%mm0		\n\t"
+		"pmaddwd %%mm5, %%mm2		\n\t"
+		"packssdw %%mm2, %%mm0		\n\t"
+		"psraw $7, %%mm0		\n\t"
+
+		"movd 12(%0, %%ebx), %%mm4	\n\t"
+		"movd 15(%0, %%ebx), %%mm1	\n\t"
+		"punpcklbw %%mm7, %%mm4		\n\t"
+		"punpcklbw %%mm7, %%mm1		\n\t"
+		"movd 18(%0, %%ebx), %%mm2	\n\t"
+		"movd 21(%0, %%ebx), %%mm3	\n\t"
+		"punpcklbw %%mm7, %%mm2		\n\t"
+		"punpcklbw %%mm7, %%mm3		\n\t"
+		"pmaddwd %%mm6, %%mm4		\n\t"
+		"pmaddwd %%mm6, %%mm1		\n\t"
+		"pmaddwd %%mm6, %%mm2		\n\t"
+		"pmaddwd %%mm6, %%mm3		\n\t"
+#ifndef FAST_BGR2YV12
+		"psrad $8, %%mm4		\n\t"
+		"psrad $8, %%mm1		\n\t"
+		"psrad $8, %%mm2		\n\t"
+		"psrad $8, %%mm3		\n\t"
+#endif
+		"packssdw %%mm1, %%mm4		\n\t"
+		"packssdw %%mm3, %%mm2		\n\t"
+		"pmaddwd %%mm5, %%mm4		\n\t"
+		"pmaddwd %%mm5, %%mm2		\n\t"
+		"addl $24, %%ebx		\n\t"
+		"packssdw %%mm2, %%mm4		\n\t"
+		"psraw $7, %%mm4		\n\t"
+
+		"packuswb %%mm4, %%mm0		\n\t"
+		"paddusb "MANGLE(bgr2YOffset)", %%mm0	\n\t"
+
+		"movq %%mm0, (%1, %%eax)	\n\t"
+		"addl $8, %%eax			\n\t"
+		" js 1b				\n\t"
+		: : "r" (src+width*3), "r" (dst+width), "g" (-width)
+		: "%eax", "%ebx"
+	);
 #else
 	int i;
 	for(i=0; i<width; i++)
@@ -1651,7 +1729,156 @@ static inline void RENAME(bgr24ToY)(uint8_t *dst, uint8_t *src, int width)
 
 static inline void RENAME(bgr24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width)
 {
-#ifdef HAVE_MMXFIXME
+#ifdef HAVE_MMX
+	asm volatile(
+		"movl %4, %%eax			\n\t"
+		"movq "MANGLE(w1111)", %%mm5		\n\t"
+		"movq "MANGLE(bgr2UCoeff)", %%mm6		\n\t"
+		"pxor %%mm7, %%mm7		\n\t"
+		"leal (%%eax, %%eax, 2), %%ebx	\n\t"
+		"addl %%ebx, %%ebx		\n\t"
+		".balign 16			\n\t"
+		"1:				\n\t"
+		PREFETCH" 64(%0, %%ebx)		\n\t"
+		PREFETCH" 64(%1, %%ebx)		\n\t"
+#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
+		"movq (%0, %%ebx), %%mm0	\n\t"
+		"movq (%1, %%ebx), %%mm1	\n\t"
+		"movq 6(%0, %%ebx), %%mm2	\n\t"
+		"movq 6(%1, %%ebx), %%mm3	\n\t"
+		PAVGB(%%mm1, %%mm0)
+		PAVGB(%%mm3, %%mm2)
+		"movq %%mm0, %%mm1		\n\t"
+		"movq %%mm2, %%mm3		\n\t"
+		"psrlq $24, %%mm0		\n\t"
+		"psrlq $24, %%mm2		\n\t"
+		PAVGB(%%mm1, %%mm0)
+		PAVGB(%%mm3, %%mm2)
+		"punpcklbw %%mm7, %%mm0		\n\t"
+		"punpcklbw %%mm7, %%mm2		\n\t"
+#else
+		"movd (%0, %%ebx), %%mm0	\n\t"
+		"movd (%1, %%ebx), %%mm1	\n\t"
+		"movd 3(%0, %%ebx), %%mm2	\n\t"
+		"movd 3(%1, %%ebx), %%mm3	\n\t"
+		"punpcklbw %%mm7, %%mm0		\n\t"
+		"punpcklbw %%mm7, %%mm1		\n\t"
+		"punpcklbw %%mm7, %%mm2		\n\t"
+		"punpcklbw %%mm7, %%mm3		\n\t"
+		"paddw %%mm1, %%mm0		\n\t"
+		"paddw %%mm3, %%mm2		\n\t"
+		"paddw %%mm2, %%mm0		\n\t"
+		"movd 6(%0, %%ebx), %%mm4	\n\t"
+		"movd 6(%1, %%ebx), %%mm1	\n\t"
+		"movd 9(%0, %%ebx), %%mm2	\n\t"
+		"movd 9(%1, %%ebx), %%mm3	\n\t"
+		"punpcklbw %%mm7, %%mm4		\n\t"
+		"punpcklbw %%mm7, %%mm1		\n\t"
+		"punpcklbw %%mm7, %%mm2		\n\t"
+		"punpcklbw %%mm7, %%mm3		\n\t"
+		"paddw %%mm1, %%mm4		\n\t"
+		"paddw %%mm3, %%mm2		\n\t"
+		"paddw %%mm4, %%mm2		\n\t"
+		"psrlw $2, %%mm0		\n\t"
+		"psrlw $2, %%mm2		\n\t"
+#endif
+		"movq "MANGLE(bgr2VCoeff)", %%mm1		\n\t"
+		"movq "MANGLE(bgr2VCoeff)", %%mm3		\n\t"
+		
+		"pmaddwd %%mm0, %%mm1		\n\t"
+		"pmaddwd %%mm2, %%mm3		\n\t"
+		"pmaddwd %%mm6, %%mm0		\n\t"
+		"pmaddwd %%mm6, %%mm2		\n\t"
+#ifndef FAST_BGR2YV12
+		"psrad $8, %%mm0		\n\t"
+		"psrad $8, %%mm1		\n\t"
+		"psrad $8, %%mm2		\n\t"
+		"psrad $8, %%mm3		\n\t"
+#endif
+		"packssdw %%mm2, %%mm0		\n\t"
+		"packssdw %%mm3, %%mm1		\n\t"
+		"pmaddwd %%mm5, %%mm0		\n\t"
+		"pmaddwd %%mm5, %%mm1		\n\t"
+		"packssdw %%mm1, %%mm0		\n\t" // V1 V0 U1 U0
+		"psraw $7, %%mm0		\n\t"
+
+#if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
+		"movq 12(%0, %%ebx), %%mm4	\n\t"
+		"movq 12(%1, %%ebx), %%mm1	\n\t"
+		"movq 18(%0, %%ebx), %%mm2	\n\t"
+		"movq 18(%1, %%ebx), %%mm3	\n\t"
+		PAVGB(%%mm1, %%mm4)
+		PAVGB(%%mm3, %%mm2)
+		"movq %%mm4, %%mm1		\n\t"
+		"movq %%mm2, %%mm3		\n\t"
+		"psrlq $24, %%mm4		\n\t"
+		"psrlq $24, %%mm2		\n\t"
+		PAVGB(%%mm1, %%mm4)
+		PAVGB(%%mm3, %%mm2)
+		"punpcklbw %%mm7, %%mm4		\n\t"
+		"punpcklbw %%mm7, %%mm2		\n\t"
+#else
+		"movd 12(%0, %%ebx), %%mm4	\n\t"
+		"movd 12(%1, %%ebx), %%mm1	\n\t"
+		"movd 15(%0, %%ebx), %%mm2	\n\t"
+		"movd 15(%1, %%ebx), %%mm3	\n\t"
+		"punpcklbw %%mm7, %%mm4		\n\t"
+		"punpcklbw %%mm7, %%mm1		\n\t"
+		"punpcklbw %%mm7, %%mm2		\n\t"
+		"punpcklbw %%mm7, %%mm3		\n\t"
+		"paddw %%mm1, %%mm4		\n\t"
+		"paddw %%mm3, %%mm2		\n\t"
+		"paddw %%mm2, %%mm4		\n\t"
+		"movd 18(%0, %%ebx), %%mm5	\n\t"
+		"movd 18(%1, %%ebx), %%mm1	\n\t"
+		"movd 21(%0, %%ebx), %%mm2	\n\t"
+		"movd 21(%1, %%ebx), %%mm3	\n\t"
+		"punpcklbw %%mm7, %%mm5		\n\t"
+		"punpcklbw %%mm7, %%mm1		\n\t"
+		"punpcklbw %%mm7, %%mm2		\n\t"
+		"punpcklbw %%mm7, %%mm3		\n\t"
+		"paddw %%mm1, %%mm5		\n\t"
+		"paddw %%mm3, %%mm2		\n\t"
+		"paddw %%mm5, %%mm2		\n\t"
+		"movq "MANGLE(w1111)", %%mm5		\n\t"
+		"psrlw $2, %%mm4		\n\t"
+		"psrlw $2, %%mm2		\n\t"
+#endif
+		"movq "MANGLE(bgr2VCoeff)", %%mm1		\n\t"
+		"movq "MANGLE(bgr2VCoeff)", %%mm3		\n\t"
+		
+		"pmaddwd %%mm4, %%mm1		\n\t"
+		"pmaddwd %%mm2, %%mm3		\n\t"
+		"pmaddwd %%mm6, %%mm4		\n\t"
+		"pmaddwd %%mm6, %%mm2		\n\t"
+#ifndef FAST_BGR2YV12
+		"psrad $8, %%mm4		\n\t"
+		"psrad $8, %%mm1		\n\t"
+		"psrad $8, %%mm2		\n\t"
+		"psrad $8, %%mm3		\n\t"
+#endif
+		"packssdw %%mm2, %%mm4		\n\t"
+		"packssdw %%mm3, %%mm1		\n\t"
+		"pmaddwd %%mm5, %%mm4		\n\t"
+		"pmaddwd %%mm5, %%mm1		\n\t"
+		"addl $24, %%ebx		\n\t"
+		"packssdw %%mm1, %%mm4		\n\t" // V3 V2 U3 U2
+		"psraw $7, %%mm4		\n\t"
+		
+		"movq %%mm0, %%mm1		\n\t"
+		"punpckldq %%mm4, %%mm0		\n\t"
+		"punpckhdq %%mm4, %%mm1		\n\t"
+		"packsswb %%mm1, %%mm0		\n\t"
+		"paddb "MANGLE(bgr2UVOffset)", %%mm0	\n\t"
+
+		"movd %%mm0, (%2, %%eax)	\n\t"
+		"punpckhdq %%mm0, %%mm0		\n\t"
+		"movd %%mm0, (%3, %%eax)	\n\t"
+		"addl $4, %%eax			\n\t"
+		" js 1b				\n\t"
+		: : "r" (src1+width*6), "r" (src2+width*6), "r" (dstU+width), "r" (dstV+width), "g" (-width)
+		: "%eax", "%ebx"
+	);
 #else
 	int i;
 	for(i=0; i<width; i++)
@@ -1666,6 +1893,185 @@ static inline void RENAME(bgr24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1
 #endif
 }
 
+static inline void RENAME(bgr16ToY)(uint8_t *dst, uint8_t *src, int width)
+{
+	int i;
+	for(i=0; i<width; i++)
+	{
+		int d= src[i*2] + (src[i*2+1]<<8);
+		int b= d&0x1F;
+		int g= (d>>5)&0x3F;
+		int r= (d>>11)&0x1F;
+
+		dst[i]= ((2*RY*r + GY*g + 2*BY*b)>>(RGB2YUV_SHIFT-2)) + 16;
+	}
+}
+
+static inline void RENAME(bgr16ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width)
+{
+	int i;
+	for(i=0; i<width; i++)
+	{
+#if 1
+		int d0= le2me_32( ((uint32_t*)src1)[i] );
+		int d1= le2me_32( ((uint32_t*)src2)[i] );
+		
+		int dl= (d0&0x07E0F81F) + (d1&0x07E0F81F);
+		int dh= ((d0>>5)&0x07C0F83F) + ((d1>>5)&0x07C0F83F);
+
+		int dh2= (dh>>11) + (dh<<21);
+		int d= dh2 + dl;
+
+		int b= d&0x7F;
+		int r= (d>>11)&0x7F;
+		int g= d>>21;
+#else
+		int d0= src1[i*4] + (src1[i*4+1]<<8);
+		int b0= d0&0x1F;
+		int g0= (d0>>5)&0x3F;
+		int r0= (d0>>11)&0x1F;
+
+		int d1= src1[i*4+2] + (src1[i*4+3]<<8);
+		int b1= d1&0x1F;
+		int g1= (d1>>5)&0x3F;
+		int r1= (d1>>11)&0x1F;
+
+		int d2= src2[i*4] + (src2[i*4+1]<<8);
+		int b2= d2&0x1F;
+		int g2= (d2>>5)&0x3F;
+		int r2= (d2>>11)&0x1F;
+
+		int d3= src2[i*4+2] + (src2[i*4+3]<<8);
+		int b3= d3&0x1F;
+		int g3= (d3>>5)&0x3F;
+		int r3= (d3>>11)&0x1F;
+
+		int b= b0 + b1 + b2 + b3;
+		int g= g0 + g1 + g2 + g3;
+		int r= r0 + r1 + r2 + r3;
+#endif
+		dstU[i]= ((2*RU*r + GU*g + 2*BU*b)>>(RGB2YUV_SHIFT+2-2)) + 128;
+		dstV[i]= ((2*RV*r + GV*g + 2*BV*b)>>(RGB2YUV_SHIFT+2-2)) + 128;
+	}
+}
+
+static inline void RENAME(bgr15ToY)(uint8_t *dst, uint8_t *src, int width)
+{
+	int i;
+	for(i=0; i<width; i++)
+	{
+		int d= src[i*2] + (src[i*2+1]<<8);
+		int b= d&0x1F;
+		int g= (d>>5)&0x1F;
+		int r= (d>>10)&0x1F;
+
+		dst[i]= ((RY*r + GY*g + BY*b)>>(RGB2YUV_SHIFT-3)) + 16;
+	}
+}
+
+static inline void RENAME(bgr15ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width)
+{
+	int i;
+	for(i=0; i<width; i++)
+	{
+#if 1
+		int d0= le2me_32( ((uint32_t*)src1)[i] );
+		int d1= le2me_32( ((uint32_t*)src2)[i] );
+		
+		int dl= (d0&0x03E07C1F) + (d1&0x03E07C1F);
+		int dh= ((d0>>5)&0x03E0F81F) + ((d1>>5)&0x03E0F81F);
+
+		int dh2= (dh>>11) + (dh<<21);
+		int d= dh2 + dl;
+
+		int b= d&0x7F;
+		int r= (d>>10)&0x7F;
+		int g= d>>21;
+#else
+		int d0= src1[i*4] + (src1[i*4+1]<<8);
+		int b0= d0&0x1F;
+		int g0= (d0>>5)&0x1F;
+		int r0= (d0>>10)&0x1F;
+
+		int d1= src1[i*4+2] + (src1[i*4+3]<<8);
+		int b1= d1&0x1F;
+		int g1= (d1>>5)&0x1F;
+		int r1= (d1>>10)&0x1F;
+
+		int d2= src2[i*4] + (src2[i*4+1]<<8);
+		int b2= d2&0x1F;
+		int g2= (d2>>5)&0x1F;
+		int r2= (d2>>10)&0x1F;
+
+		int d3= src2[i*4+2] + (src2[i*4+3]<<8);
+		int b3= d3&0x1F;
+		int g3= (d3>>5)&0x1F;
+		int r3= (d3>>10)&0x1F;
+
+		int b= b0 + b1 + b2 + b3;
+		int g= g0 + g1 + g2 + g3;
+		int r= r0 + r1 + r2 + r3;
+#endif
+		dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+2-3)) + 128;
+		dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+2-3)) + 128;
+	}
+}
+
+
+static inline void RENAME(rgb32ToY)(uint8_t *dst, uint8_t *src, int width)
+{
+	int i;
+	for(i=0; i<width; i++)
+	{
+		int r= src[i*4+0];
+		int g= src[i*4+1];
+		int b= src[i*4+2];
+
+		dst[i]= ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;
+	}
+}
+
+static inline void RENAME(rgb32ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width)
+{
+	int i;
+	for(i=0; i<width; i++)
+	{
+		int r= src1[8*i + 0] + src1[8*i + 4] + src2[8*i + 0] + src2[8*i + 4];
+		int g= src1[8*i + 1] + src1[8*i + 5] + src2[8*i + 1] + src2[8*i + 5];
+		int b= src1[8*i + 2] + src1[8*i + 6] + src2[8*i + 2] + src2[8*i + 6];
+
+		dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+2)) + 128;
+		dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+2)) + 128;
+	}
+}
+
+static inline void RENAME(rgb24ToY)(uint8_t *dst, uint8_t *src, int width)
+{
+	int i;
+	for(i=0; i<width; i++)
+	{
+		int r= src[i*3+0];
+		int g= src[i*3+1];
+		int b= src[i*3+2];
+
+		dst[i]= ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16;
+	}
+}
+
+static inline void RENAME(rgb24ToUV)(uint8_t *dstU, uint8_t *dstV, uint8_t *src1, uint8_t *src2, int width)
+{
+	int i;
+	for(i=0; i<width; i++)
+	{
+		int r= src1[6*i + 0] + src1[6*i + 3] + src2[6*i + 0] + src2[6*i + 3];
+		int g= src1[6*i + 1] + src1[6*i + 4] + src2[6*i + 1] + src2[6*i + 4];
+		int b= src1[6*i + 2] + src1[6*i + 5] + src2[6*i + 2] + src2[6*i + 5];
+
+		dstU[i]= ((RU*r + GU*g + BU*b)>>(RGB2YUV_SHIFT+2)) + 128;
+		dstV[i]= ((RV*r + GV*g + BV*b)>>(RGB2YUV_SHIFT+2)) + 128;
+	}
+}
+
 
 // Bilinear / Bicubic scaling
 static inline void RENAME(hScale)(int16_t *dst, int dstW, uint8_t *src, int srcW, int xInc,
@@ -1832,7 +2238,8 @@ static inline void RENAME(hScale)(int16_t *dst, int dstW, uint8_t *src, int srcW
 static inline void RENAME(hyscale)(uint16_t *dst, int dstWidth, uint8_t *src, int srcW, int xInc,
 				   int flags, int canMMX2BeUsed, int16_t *hLumFilter,
 				   int16_t *hLumFilterPos, int hLumFilterSize, void *funnyYCode, 
-				   int srcFormat, uint8_t *formatConvBuffer)
+				   int srcFormat, uint8_t *formatConvBuffer, int16_t *mmx2Filter,
+				   int32_t *mmx2FilterPos)
 {
     if(srcFormat==IMGFMT_YUY2)
     {
@@ -1849,6 +2256,26 @@ static inline void RENAME(hyscale)(uint16_t *dst, int dstWidth, uint8_t *src, in
 	RENAME(bgr24ToY)(formatConvBuffer, src, srcW);
 	src= formatConvBuffer;
     }
+    else if(srcFormat==IMGFMT_BGR16)
+    {
+	RENAME(bgr16ToY)(formatConvBuffer, src, srcW);
+	src= formatConvBuffer;
+    }
+    else if(srcFormat==IMGFMT_BGR15)
+    {
+	RENAME(bgr15ToY)(formatConvBuffer, src, srcW);
+	src= formatConvBuffer;
+    }
+    else if(srcFormat==IMGFMT_RGB32)
+    {
+	RENAME(rgb32ToY)(formatConvBuffer, src, srcW);
+	src= formatConvBuffer;
+    }
+    else if(srcFormat==IMGFMT_RGB24)
+    {
+	RENAME(rgb24ToY)(formatConvBuffer, src, srcW);
+	src= formatConvBuffer;
+    }
 
 #ifdef HAVE_MMX
 	// use the new MMX scaler if th mmx2 cant be used (its faster than the x86asm one)
@@ -1868,35 +2295,21 @@ static inline void RENAME(hyscale)(uint16_t *dst, int dstWidth, uint8_t *src, in
 	{
 		asm volatile(
 			"pxor %%mm7, %%mm7		\n\t"
-			"pxor %%mm2, %%mm2		\n\t" // 2*xalpha
-			"movd %5, %%mm6			\n\t" // xInc&0xFFFF
-			"punpcklwd %%mm6, %%mm6		\n\t"
-			"punpcklwd %%mm6, %%mm6		\n\t"
-			"movq %%mm6, %%mm2		\n\t"
-			"psllq $16, %%mm2		\n\t"
-			"paddw %%mm6, %%mm2		\n\t"
-			"psllq $16, %%mm2		\n\t"
-			"paddw %%mm6, %%mm2		\n\t"
-			"psllq $16, %%mm2		\n\t" //0,t,2t,3t		t=xInc&0xFF
-			"movq %%mm2, %%mm4		\n\t"
-			"movd %4, %%mm6			\n\t" //(xInc*4)&0xFFFF
-			"punpcklwd %%mm6, %%mm6		\n\t"
-			"punpcklwd %%mm6, %%mm6		\n\t"
+			"movl %0, %%ecx			\n\t"
+			"movl %1, %%edi			\n\t"
+			"movl %2, %%edx			\n\t"
+			"movl %3, %%ebx			\n\t"
 			"xorl %%eax, %%eax		\n\t" // i
-			"movl %0, %%esi			\n\t" // src
-			"movl %1, %%edi			\n\t" // buf1
-			"movl %3, %%edx			\n\t" // (xInc*4)>>16
-			"xorl %%ecx, %%ecx		\n\t"
-			"xorl %%ebx, %%ebx		\n\t"
-			"movw %4, %%bx			\n\t" // (xInc*4)&0xFFFF
+			PREFETCH" (%%ecx)		\n\t"
+			PREFETCH" 32(%%ecx)		\n\t"
+			PREFETCH" 64(%%ecx)		\n\t"
 
 #define FUNNY_Y_CODE \
-			PREFETCH" 1024(%%esi)		\n\t"\
-			PREFETCH" 1056(%%esi)		\n\t"\
-			PREFETCH" 1088(%%esi)		\n\t"\
-			"call *%6			\n\t"\
-			"movq %%mm4, %%mm2		\n\t"\
-			"xorl %%ecx, %%ecx		\n\t"
+			"movl (%%ebx), %%esi		\n\t"\
+			"call *%4			\n\t"\
+			"addl (%%ebx, %%eax), %%ecx	\n\t"\
+			"addl %%eax, %%edi		\n\t"\
+			"xorl %%eax, %%eax		\n\t"\
 
 FUNNY_Y_CODE
 FUNNY_Y_CODE
@@ -1907,8 +2320,8 @@ FUNNY_Y_CODE
 FUNNY_Y_CODE
 FUNNY_Y_CODE
 
-			:: "m" (src), "m" (dst), "m" (dstWidth), "m" ((xInc*4)>>16),
-			"m" ((xInc*4)&0xFFFF), "m" (xInc&0xFFFF), "m" (funnyYCode)
+			:: "m" (src), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos),
+			"m" (funnyYCode)
 			: "%eax", "%ebx", "%ecx", "%edx", "%esi", "%edi"
 		);
 		for(i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) dst[i] = src[srcW-1]*128;
@@ -1976,7 +2389,8 @@ FUNNY_Y_CODE
 inline static void RENAME(hcscale)(uint16_t *dst, int dstWidth, uint8_t *src1, uint8_t *src2,
 				   int srcW, int xInc, int flags, int canMMX2BeUsed, int16_t *hChrFilter,
 				   int16_t *hChrFilterPos, int hChrFilterSize, void *funnyUVCode,
-				   int srcFormat, uint8_t *formatConvBuffer)
+				   int srcFormat, uint8_t *formatConvBuffer, int16_t *mmx2Filter,
+				   int32_t *mmx2FilterPos)
 {
     if(srcFormat==IMGFMT_YUY2)
     {
@@ -1996,6 +2410,30 @@ inline static void RENAME(hcscale)(uint16_t *dst, int dstWidth, uint8_t *src1, u
 	src1= formatConvBuffer;
 	src2= formatConvBuffer+2048;
     }
+    else if(srcFormat==IMGFMT_BGR16)
+    {
+	RENAME(bgr16ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW);
+	src1= formatConvBuffer;
+	src2= formatConvBuffer+2048;
+    }
+    else if(srcFormat==IMGFMT_BGR15)
+    {
+	RENAME(bgr15ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW);
+	src1= formatConvBuffer;
+	src2= formatConvBuffer+2048;
+    }
+    else if(srcFormat==IMGFMT_RGB32)
+    {
+	RENAME(rgb32ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW);
+	src1= formatConvBuffer;
+	src2= formatConvBuffer+2048;
+    }
+    else if(srcFormat==IMGFMT_RGB24)
+    {
+	RENAME(rgb24ToUV)(formatConvBuffer, formatConvBuffer+2048, src1, src2, srcW);
+	src1= formatConvBuffer;
+	src2= formatConvBuffer+2048;
+    }
     else if(isGray(srcFormat))
     {
     	return;
@@ -2019,65 +2457,44 @@ inline static void RENAME(hcscale)(uint16_t *dst, int dstWidth, uint8_t *src1, u
 	if(canMMX2BeUsed)
 	{
 		asm volatile(
-		"pxor %%mm7, %%mm7		\n\t"
-		"pxor %%mm2, %%mm2		\n\t" // 2*xalpha
-		"movd %5, %%mm6			\n\t" // xInc&0xFFFF
-		"punpcklwd %%mm6, %%mm6		\n\t"
-		"punpcklwd %%mm6, %%mm6		\n\t"
-		"movq %%mm6, %%mm2		\n\t"
-		"psllq $16, %%mm2		\n\t"
-		"paddw %%mm6, %%mm2		\n\t"
-		"psllq $16, %%mm2		\n\t"
-		"paddw %%mm6, %%mm2		\n\t"
-		"psllq $16, %%mm2		\n\t" //0,t,2t,3t		t=xInc&0xFFFF
-		"movq %%mm2, %%mm4		\n\t"
-		"movd %4, %%mm6			\n\t" //(xInc*4)&0xFFFF
-		"punpcklwd %%mm6, %%mm6		\n\t"
-		"punpcklwd %%mm6, %%mm6		\n\t"
-		"xorl %%eax, %%eax		\n\t" // i
-		"movl %0, %%esi			\n\t" // src
-		"movl %1, %%edi			\n\t" // buf1
-		"movl %3, %%edx			\n\t" // (xInc*4)>>16
-		"xorl %%ecx, %%ecx		\n\t"
-		"xorl %%ebx, %%ebx		\n\t"
-		"movw %4, %%bx			\n\t" // (xInc*4)&0xFFFF
-
-#define FUNNYUVCODE \
-			PREFETCH" 1024(%%esi)		\n\t"\
-			PREFETCH" 1056(%%esi)		\n\t"\
-			PREFETCH" 1088(%%esi)		\n\t"\
-			"call *%7			\n\t"\
-			"movq %%mm4, %%mm2	\n\t"\
-			"xorl %%ecx, %%ecx		\n\t"
-
-FUNNYUVCODE
-FUNNYUVCODE
-FUNNYUVCODE
-FUNNYUVCODE
-
-FUNNYUVCODE
-FUNNYUVCODE
-FUNNYUVCODE
-FUNNYUVCODE
-		"xorl %%eax, %%eax		\n\t" // i
-		"movl %6, %%esi			\n\t" // src
-		"movl %1, %%edi			\n\t" // buf1
-		"addl $4096, %%edi		\n\t"
-
-FUNNYUVCODE
-FUNNYUVCODE
-FUNNYUVCODE
-FUNNYUVCODE
-
-FUNNYUVCODE
-FUNNYUVCODE
-FUNNYUVCODE
-FUNNYUVCODE
-
-		:: "m" (src1), "m" (dst), "m" (dstWidth), "m" ((xInc*4)>>16),
-		  "m" ((xInc*4)&0xFFFF), "m" (xInc&0xFFFF), "m" (src2), "m" (funnyUVCode)
-		: "%eax", "%ebx", "%ecx", "%edx", "%esi", "%edi"
-	);
+			"pxor %%mm7, %%mm7		\n\t"
+			"movl %0, %%ecx			\n\t"
+			"movl %1, %%edi			\n\t"
+			"movl %2, %%edx			\n\t"
+			"movl %3, %%ebx			\n\t"
+			"xorl %%eax, %%eax		\n\t" // i
+			PREFETCH" (%%ecx)		\n\t"
+			PREFETCH" 32(%%ecx)		\n\t"
+			PREFETCH" 64(%%ecx)		\n\t"
+
+#define FUNNY_UV_CODE \
+			"movl (%%ebx), %%esi		\n\t"\
+			"call *%4			\n\t"\
+			"addl (%%ebx, %%eax), %%ecx	\n\t"\
+			"addl %%eax, %%edi		\n\t"\
+			"xorl %%eax, %%eax		\n\t"\
+
+FUNNY_UV_CODE
+FUNNY_UV_CODE
+FUNNY_UV_CODE
+FUNNY_UV_CODE
+			"xorl %%eax, %%eax		\n\t" // i
+			"movl %5, %%ecx			\n\t" // src
+			"movl %1, %%edi			\n\t" // buf1
+			"addl $4096, %%edi		\n\t"
+			PREFETCH" (%%ecx)		\n\t"
+			PREFETCH" 32(%%ecx)		\n\t"
+			PREFETCH" 64(%%ecx)		\n\t"
+
+FUNNY_UV_CODE
+FUNNY_UV_CODE
+FUNNY_UV_CODE
+FUNNY_UV_CODE
+
+			:: "m" (src1), "m" (dst), "m" (mmx2Filter), "m" (mmx2FilterPos),
+			"m" (funnyUVCode), "m" (src2)
+			: "%eax", "%ebx", "%ecx", "%edx", "%esi", "%edi"
+		);
 		for(i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
 		{
 //			printf("%d %d %d\n", dstWidth, i, srcW);
@@ -2148,7 +2565,7 @@ FUNNYUVCODE
 }
 
 static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStrideParam[], int srcSliceY,
-             int srcSliceH, uint8_t* dstParam[], int dstStride[]){
+             int srcSliceH, uint8_t* dstParam[], int dstStrideParam[]){
 
 	/* load a few things into local vars to make the code more readable? and faster */
 	const int srcW= c->srcW;
@@ -2189,10 +2606,11 @@ static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStridePar
 	int lastInLumBuf= c->lastInLumBuf;
 	int lastInChrBuf= c->lastInChrBuf;
 	int srcStride[3];
+	int dstStride[3];
 	uint8_t *src[3];
 	uint8_t *dst[3];
 	
-	if((c->srcFormat == IMGFMT_IYUV) || (c->srcFormat == IMGFMT_I420)){
+	if(c->srcFormat == IMGFMT_I420){
 		src[0]= srcParam[0];
 		src[1]= srcParam[2];
 		src[2]= srcParam[1];
@@ -2225,17 +2643,24 @@ static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStridePar
 		srcStride[2]= 0;
 	}
 
-	if((c->dstFormat == IMGFMT_IYUV) || (c->dstFormat == IMGFMT_I420)){
+	if(dstFormat == IMGFMT_I420){
 		dst[0]= dstParam[0];
 		dst[1]= dstParam[2];
 		dst[2]= dstParam[1];
-		
+		dstStride[0]= dstStrideParam[0];
+		dstStride[1]= dstStrideParam[2];
+		dstStride[2]= dstStrideParam[1];
 	}else{
 		dst[0]= dstParam[0];
 		dst[1]= dstParam[1];
 		dst[2]= dstParam[2];
+		dstStride[0]= dstStrideParam[0];
+		dstStride[1]= dstStrideParam[1];
+		dstStride[2]= dstStrideParam[2];
 	}
-	
+
+//printf("sws Strides:%d %d %d -> %d %d %d\n", srcStride[0],srcStride[1],srcStride[2],
+//dstStride[0],dstStride[1],dstStride[2]);
 
 	if(dstStride[0]%8 !=0 || dstStride[1]%8 !=0 || dstStride[2]%8 !=0)
 	{
@@ -2277,7 +2702,7 @@ static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStridePar
 		ASSERT(firstChrSrcY >= lastInChrBuf - vChrBufSize + 1)
 
 		// Do we have enough lines in this slice to output the dstY line
-		if(lastLumSrcY < srcSliceY + srcSliceH && lastChrSrcY < ((srcSliceY + srcSliceH)>>1))
+		if(lastLumSrcY < srcSliceY + srcSliceH && lastChrSrcY < ((srcSliceY + srcSliceH + 1)>>1))
 		{
 			//Do horizontal scaling
 			while(lastInLumBuf < lastLumSrcY)
@@ -2291,7 +2716,8 @@ static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStridePar
 //				printf("%d %d\n", lumBufIndex, vLumBufSize);
 				RENAME(hyscale)(lumPixBuf[ lumBufIndex ], dstW, s, srcW, lumXInc,
 						flags, canMMX2BeUsed, hLumFilter, hLumFilterPos, hLumFilterSize,
-						funnyYCode, c->srcFormat, formatConvBuffer);
+						funnyYCode, c->srcFormat, formatConvBuffer, 
+						c->lumMmx2Filter, c->lumMmx2FilterPos);
 				lastInLumBuf++;
 			}
 			while(lastInChrBuf < lastChrSrcY)
@@ -2300,12 +2726,13 @@ static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStridePar
 				uint8_t *src2= src[2]+(lastInChrBuf + 1 - (srcSliceY>>1))*srcStride[2];
 				chrBufIndex++;
 				ASSERT(chrBufIndex < 2*vChrBufSize)
-				ASSERT(lastInChrBuf + 1 - (srcSliceY>>1) < (srcSliceH>>1))
+				ASSERT(lastInChrBuf + 1 - (srcSliceY>>1) < ((srcSliceH+1)>>1))
 				ASSERT(lastInChrBuf + 1 - (srcSliceY>>1) >= 0)
 				//FIXME replace parameters through context struct (some at least)
 				RENAME(hcscale)(chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, (srcW+1)>>1, chrXInc,
 						flags, canMMX2BeUsed, hChrFilter, hChrFilterPos, hChrFilterSize,
-						funnyUVCode, c->srcFormat, formatConvBuffer);
+						funnyUVCode, c->srcFormat, formatConvBuffer, 
+						c->chrMmx2Filter, c->chrMmx2FilterPos);
 				lastInChrBuf++;
 			}
 			//wrap buf index around to stay inside the ring buffer
@@ -2329,7 +2756,8 @@ static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStridePar
 				ASSERT(lastInLumBuf + 1 - srcSliceY >= 0)
 				RENAME(hyscale)(lumPixBuf[ lumBufIndex ], dstW, s, srcW, lumXInc,
 						flags, canMMX2BeUsed, hLumFilter, hLumFilterPos, hLumFilterSize,
-						funnyYCode, c->srcFormat, formatConvBuffer);
+						funnyYCode, c->srcFormat, formatConvBuffer, 
+						c->lumMmx2Filter, c->lumMmx2FilterPos);
 				lastInLumBuf++;
 			}
 			while(lastInChrBuf+1 < ((srcSliceY + srcSliceH)>>1))
@@ -2338,11 +2766,12 @@ static void RENAME(swScale)(SwsContext *c, uint8_t* srcParam[], int srcStridePar
 				uint8_t *src2= src[2]+(lastInChrBuf + 1 - (srcSliceY>>1))*srcStride[2];
 				chrBufIndex++;
 				ASSERT(chrBufIndex < 2*vChrBufSize)
-				ASSERT(lastInChrBuf + 1 - (srcSliceY>>1) < (srcSliceH>>1))
+				ASSERT(lastInChrBuf + 1 - (srcSliceY>>1) < ((srcSliceH+1)>>1))
 				ASSERT(lastInChrBuf + 1 - (srcSliceY>>1) >= 0)
 				RENAME(hcscale)(chrPixBuf[ chrBufIndex ], chrDstW, src1, src2, (srcW+1)>>1, chrXInc,
 						flags, canMMX2BeUsed, hChrFilter, hChrFilterPos, hChrFilterSize,
-						funnyUVCode, c->srcFormat, formatConvBuffer);
+						funnyUVCode, c->srcFormat, formatConvBuffer, 
+						c->chrMmx2Filter, c->chrMmx2FilterPos);
 				lastInChrBuf++;
 			}
 			//wrap buf index around to stay inside the ring buffer