runtime cpu detection

git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@3127 b3059339-0415-0410-9bf9-f77b7e298cf2

2 files changed, 87 insertions, 1824 deletions

diff --git a/postproc/swscale.c b/postproc/swscale.c
index 56e53f8bf6..7afb61b2b4 100644
--- a/postproc/swscale.c
+++ b/postproc/swscale.c
@@ -10,6 +10,7 @@
 #include "../config.h"
 #include "swscale.h"
 #include "../mmx_defs.h"
+#include "../cpudetect.h"
 #undef MOVNTQ
 #undef PAVGB
 
@@ -37,26 +38,18 @@ more intelligent missalignment avoidance for the horizontal scaler
 bicubic scaler
 dither in C
 change the distance of the u & v buffer
+how to differenciate between x86 an C at runtime ?! (using C for now)
 */
 
 #define ABS(a) ((a) > 0 ? (a) : (-(a)))
 #define MIN(a,b) ((a) > (b) ? (b) : (a))
 #define MAX(a,b) ((a) < (b) ? (b) : (a))
 
-#ifdef HAVE_MMX2
-#define PAVGB(a,b) "pavgb " #a ", " #b " \n\t"
-#elif defined (HAVE_3DNOW)
-#define PAVGB(a,b) "pavgusb " #a ", " #b " \n\t"
-#endif
-
-#ifdef HAVE_MMX2
-#define MOVNTQ(a,b) "movntq " #a ", " #b " \n\t"
-#else
-#define MOVNTQ(a,b) "movq " #a ", " #b " \n\t"
+#ifdef ARCH_X86
+#define CAN_COMPILE_X86_ASM
 #endif
 
-
-#ifdef HAVE_MMX
+#ifdef CAN_COMPILE_X86_ASM
 static uint64_t __attribute__((aligned(8))) yCoeff=    0x2568256825682568LL;
 static uint64_t __attribute__((aligned(8))) vrCoeff=   0x3343334333433343LL;
 static uint64_t __attribute__((aligned(8))) ubCoeff=   0x40cf40cf40cf40cfLL;
@@ -98,11 +91,9 @@ static uint64_t __attribute__((aligned(8))) M24C=   0x0000FF0000FF0000LL;
 static uint64_t __attribute__((aligned(8))) temp0;
 static uint64_t __attribute__((aligned(8))) asm_yalpha1;
 static uint64_t __attribute__((aligned(8))) asm_uvalpha1;
-#endif
 
 // temporary storage for 4 yuv lines:
 // 16bit for now (mmx likes it more compact)
-#ifdef HAVE_MMX
 static uint16_t __attribute__((aligned(8))) pix_buf_y[4][2048];
 static uint16_t __attribute__((aligned(8))) pix_buf_uv[2][2048*2];
 #else
@@ -127,483 +118,14 @@ static    int yuvtab_0c92[256];
 static    int yuvtab_1a1e[256];
 static    int yuvtab_40cf[256];
 
-#ifdef HAVE_MMX2
+#ifdef CAN_COMPILE_X86_ASM
 static uint8_t funnyYCode[10000];
 static uint8_t funnyUVCode[10000];
 #endif
 
 static int canMMX2BeUsed=0;
 
-#define FULL_YSCALEYUV2RGB \
-		"pxor %%mm7, %%mm7		\n\t"\
-		"movd %6, %%mm6			\n\t" /*yalpha1*/\
-		"punpcklwd %%mm6, %%mm6		\n\t"\
-		"punpcklwd %%mm6, %%mm6		\n\t"\
-		"movd %7, %%mm5			\n\t" /*uvalpha1*/\
-		"punpcklwd %%mm5, %%mm5		\n\t"\
-		"punpcklwd %%mm5, %%mm5		\n\t"\
-		"xorl %%eax, %%eax		\n\t"\
-		".balign 16			\n\t"\
-		"1:				\n\t"\
-		"movq (%0, %%eax, 2), %%mm0	\n\t" /*buf0[eax]*/\
-		"movq (%1, %%eax, 2), %%mm1	\n\t" /*buf1[eax]*/\
-		"movq (%2, %%eax,2), %%mm2	\n\t" /* uvbuf0[eax]*/\
-		"movq (%3, %%eax,2), %%mm3	\n\t" /* uvbuf1[eax]*/\
-		"psubw %%mm1, %%mm0		\n\t" /* buf0[eax] - buf1[eax]*/\
-		"psubw %%mm3, %%mm2		\n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
-		"pmulhw %%mm6, %%mm0		\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
-		"pmulhw %%mm5, %%mm2		\n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
-		"psraw $4, %%mm1		\n\t" /* buf0[eax] - buf1[eax] >>4*/\
-		"movq 4096(%2, %%eax,2), %%mm4	\n\t" /* uvbuf0[eax+2048]*/\
-		"psraw $4, %%mm3		\n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
-		"paddw %%mm0, %%mm1		\n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
-		"movq 4096(%3, %%eax,2), %%mm0	\n\t" /* uvbuf1[eax+2048]*/\
-		"paddw %%mm2, %%mm3		\n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
-		"psubw %%mm0, %%mm4		\n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
-		"psubw w80, %%mm1		\n\t" /* 8(Y-16)*/\
-		"psubw w400, %%mm3		\n\t" /* 8(U-128)*/\
-		"pmulhw yCoeff, %%mm1		\n\t"\
-\
-\
-		"pmulhw %%mm5, %%mm4		\n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
-		"movq %%mm3, %%mm2		\n\t" /* (U-128)8*/\
-		"pmulhw ubCoeff, %%mm3		\n\t"\
-		"psraw $4, %%mm0		\n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
-		"pmulhw ugCoeff, %%mm2		\n\t"\
-		"paddw %%mm4, %%mm0		\n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
-		"psubw w400, %%mm0		\n\t" /* (V-128)8*/\
-\
-\
-		"movq %%mm0, %%mm4		\n\t" /* (V-128)8*/\
-		"pmulhw vrCoeff, %%mm0		\n\t"\
-		"pmulhw vgCoeff, %%mm4		\n\t"\
-		"paddw %%mm1, %%mm3		\n\t" /* B*/\
-		"paddw %%mm1, %%mm0		\n\t" /* R*/\
-		"packuswb %%mm3, %%mm3		\n\t"\
-\
-		"packuswb %%mm0, %%mm0		\n\t"\
-		"paddw %%mm4, %%mm2		\n\t"\
-		"paddw %%mm2, %%mm1		\n\t" /* G*/\
-\
-		"packuswb %%mm1, %%mm1		\n\t"
-
-#define YSCALEYUV2RGB \
-		"movd %6, %%mm6			\n\t" /*yalpha1*/\
-		"punpcklwd %%mm6, %%mm6		\n\t"\
-		"punpcklwd %%mm6, %%mm6		\n\t"\
-		"movq %%mm6, asm_yalpha1	\n\t"\
-		"movd %7, %%mm5			\n\t" /*uvalpha1*/\
-		"punpcklwd %%mm5, %%mm5		\n\t"\
-		"punpcklwd %%mm5, %%mm5		\n\t"\
-		"movq %%mm5, asm_uvalpha1	\n\t"\
-		"xorl %%eax, %%eax		\n\t"\
-		".balign 16			\n\t"\
-		"1:				\n\t"\
-		"movq (%2, %%eax), %%mm2	\n\t" /* uvbuf0[eax]*/\
-		"movq (%3, %%eax), %%mm3	\n\t" /* uvbuf1[eax]*/\
-		"movq 4096(%2, %%eax), %%mm5	\n\t" /* uvbuf0[eax+2048]*/\
-		"movq 4096(%3, %%eax), %%mm4	\n\t" /* uvbuf1[eax+2048]*/\
-		"psubw %%mm3, %%mm2		\n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
-		"psubw %%mm4, %%mm5		\n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
-		"movq asm_uvalpha1, %%mm0	\n\t"\
-		"pmulhw %%mm0, %%mm2		\n\t" /* (uvbuf0[eax] - uvbuf1[eax])uvalpha1>>16*/\
-		"pmulhw %%mm0, %%mm5		\n\t" /* (uvbuf0[eax+2048] - uvbuf1[eax+2048])uvalpha1>>16*/\
-		"psraw $4, %%mm3		\n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
-		"psraw $4, %%mm4		\n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
-		"paddw %%mm2, %%mm3		\n\t" /* uvbuf0[eax]uvalpha1 - uvbuf1[eax](1-uvalpha1)*/\
-		"paddw %%mm5, %%mm4		\n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
-		"psubw w400, %%mm3		\n\t" /* (U-128)8*/\
-		"psubw w400, %%mm4		\n\t" /* (V-128)8*/\
-		"movq %%mm3, %%mm2		\n\t" /* (U-128)8*/\
-		"movq %%mm4, %%mm5		\n\t" /* (V-128)8*/\
-		"pmulhw ugCoeff, %%mm3		\n\t"\
-		"pmulhw vgCoeff, %%mm4		\n\t"\
-	/* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
-		"movq (%0, %%eax, 2), %%mm0	\n\t" /*buf0[eax]*/\
-		"movq (%1, %%eax, 2), %%mm1	\n\t" /*buf1[eax]*/\
-		"movq 8(%0, %%eax, 2), %%mm6	\n\t" /*buf0[eax]*/\
-		"movq 8(%1, %%eax, 2), %%mm7	\n\t" /*buf1[eax]*/\
-		"psubw %%mm1, %%mm0		\n\t" /* buf0[eax] - buf1[eax]*/\
-		"psubw %%mm7, %%mm6		\n\t" /* buf0[eax] - buf1[eax]*/\
-		"pmulhw asm_yalpha1, %%mm0	\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
-		"pmulhw asm_yalpha1, %%mm6	\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
-		"psraw $4, %%mm1		\n\t" /* buf0[eax] - buf1[eax] >>4*/\
-		"psraw $4, %%mm7		\n\t" /* buf0[eax] - buf1[eax] >>4*/\
-		"paddw %%mm0, %%mm1		\n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
-		"paddw %%mm6, %%mm7		\n\t" /* buf0[eax]yalpha1 + buf1[eax](1-yalpha1) >>16*/\
-		"pmulhw ubCoeff, %%mm2		\n\t"\
-		"pmulhw vrCoeff, %%mm5		\n\t"\
-		"psubw w80, %%mm1		\n\t" /* 8(Y-16)*/\
-		"psubw w80, %%mm7		\n\t" /* 8(Y-16)*/\
-		"pmulhw yCoeff, %%mm1		\n\t"\
-		"pmulhw yCoeff, %%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"\
-		"movq %%mm5, %%mm6		\n\t"\
-		"movq %%mm4, %%mm3		\n\t"\
-		"punpcklwd %%mm2, %%mm2		\n\t"\
-		"punpcklwd %%mm5, %%mm5		\n\t"\
-		"punpcklwd %%mm4, %%mm4		\n\t"\
-		"paddw %%mm1, %%mm2		\n\t"\
-		"paddw %%mm1, %%mm5		\n\t"\
-		"paddw %%mm1, %%mm4		\n\t"\
-		"punpckhwd %%mm0, %%mm0		\n\t"\
-		"punpckhwd %%mm6, %%mm6		\n\t"\
-		"punpckhwd %%mm3, %%mm3		\n\t"\
-		"paddw %%mm7, %%mm0		\n\t"\
-		"paddw %%mm7, %%mm6		\n\t"\
-		"paddw %%mm7, %%mm3		\n\t"\
-		/* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
-		"packuswb %%mm0, %%mm2		\n\t"\
-		"packuswb %%mm6, %%mm5		\n\t"\
-		"packuswb %%mm3, %%mm4		\n\t"\
-		"pxor %%mm7, %%mm7		\n\t"
-
-#define YSCALEYUV2RGB1 \
-		"xorl %%eax, %%eax		\n\t"\
-		".balign 16			\n\t"\
-		"1:				\n\t"\
-		"movq (%2, %%eax), %%mm3	\n\t" /* uvbuf0[eax]*/\
-		"movq 4096(%2, %%eax), %%mm4	\n\t" /* uvbuf0[eax+2048]*/\
-		"psraw $4, %%mm3		\n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
-		"psraw $4, %%mm4		\n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
-		"psubw w400, %%mm3		\n\t" /* (U-128)8*/\
-		"psubw w400, %%mm4		\n\t" /* (V-128)8*/\
-		"movq %%mm3, %%mm2		\n\t" /* (U-128)8*/\
-		"movq %%mm4, %%mm5		\n\t" /* (V-128)8*/\
-		"pmulhw ugCoeff, %%mm3		\n\t"\
-		"pmulhw vgCoeff, %%mm4		\n\t"\
-	/* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
-		"movq (%0, %%eax, 2), %%mm1	\n\t" /*buf0[eax]*/\
-		"movq 8(%0, %%eax, 2), %%mm7	\n\t" /*buf0[eax]*/\
-		"psraw $4, %%mm1		\n\t" /* buf0[eax] - buf1[eax] >>4*/\
-		"psraw $4, %%mm7		\n\t" /* buf0[eax] - buf1[eax] >>4*/\
-		"pmulhw ubCoeff, %%mm2		\n\t"\
-		"pmulhw vrCoeff, %%mm5		\n\t"\
-		"psubw w80, %%mm1		\n\t" /* 8(Y-16)*/\
-		"psubw w80, %%mm7		\n\t" /* 8(Y-16)*/\
-		"pmulhw yCoeff, %%mm1		\n\t"\
-		"pmulhw yCoeff, %%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"\
-		"movq %%mm5, %%mm6		\n\t"\
-		"movq %%mm4, %%mm3		\n\t"\
-		"punpcklwd %%mm2, %%mm2		\n\t"\
-		"punpcklwd %%mm5, %%mm5		\n\t"\
-		"punpcklwd %%mm4, %%mm4		\n\t"\
-		"paddw %%mm1, %%mm2		\n\t"\
-		"paddw %%mm1, %%mm5		\n\t"\
-		"paddw %%mm1, %%mm4		\n\t"\
-		"punpckhwd %%mm0, %%mm0		\n\t"\
-		"punpckhwd %%mm6, %%mm6		\n\t"\
-		"punpckhwd %%mm3, %%mm3		\n\t"\
-		"paddw %%mm7, %%mm0		\n\t"\
-		"paddw %%mm7, %%mm6		\n\t"\
-		"paddw %%mm7, %%mm3		\n\t"\
-		/* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
-		"packuswb %%mm0, %%mm2		\n\t"\
-		"packuswb %%mm6, %%mm5		\n\t"\
-		"packuswb %%mm3, %%mm4		\n\t"\
-		"pxor %%mm7, %%mm7		\n\t"
-
-// do vertical chrominance interpolation
-#define YSCALEYUV2RGB1b \
-		"xorl %%eax, %%eax		\n\t"\
-		".balign 16			\n\t"\
-		"1:				\n\t"\
-		"movq (%2, %%eax), %%mm2	\n\t" /* uvbuf0[eax]*/\
-		"movq (%3, %%eax), %%mm3	\n\t" /* uvbuf1[eax]*/\
-		"movq 4096(%2, %%eax), %%mm5	\n\t" /* uvbuf0[eax+2048]*/\
-		"movq 4096(%3, %%eax), %%mm4	\n\t" /* uvbuf1[eax+2048]*/\
-		"paddw %%mm2, %%mm3		\n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
-		"paddw %%mm5, %%mm4		\n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
-		"psrlw $5, %%mm3		\n\t"\
-		"psrlw $5, %%mm4		\n\t"\
-		"psubw w400, %%mm3		\n\t" /* (U-128)8*/\
-		"psubw w400, %%mm4		\n\t" /* (V-128)8*/\
-		"movq %%mm3, %%mm2		\n\t" /* (U-128)8*/\
-		"movq %%mm4, %%mm5		\n\t" /* (V-128)8*/\
-		"pmulhw ugCoeff, %%mm3		\n\t"\
-		"pmulhw vgCoeff, %%mm4		\n\t"\
-	/* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
-		"movq (%0, %%eax, 2), %%mm1	\n\t" /*buf0[eax]*/\
-		"movq 8(%0, %%eax, 2), %%mm7	\n\t" /*buf0[eax]*/\
-		"psraw $4, %%mm1		\n\t" /* buf0[eax] - buf1[eax] >>4*/\
-		"psraw $4, %%mm7		\n\t" /* buf0[eax] - buf1[eax] >>4*/\
-		"pmulhw ubCoeff, %%mm2		\n\t"\
-		"pmulhw vrCoeff, %%mm5		\n\t"\
-		"psubw w80, %%mm1		\n\t" /* 8(Y-16)*/\
-		"psubw w80, %%mm7		\n\t" /* 8(Y-16)*/\
-		"pmulhw yCoeff, %%mm1		\n\t"\
-		"pmulhw yCoeff, %%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"\
-		"movq %%mm5, %%mm6		\n\t"\
-		"movq %%mm4, %%mm3		\n\t"\
-		"punpcklwd %%mm2, %%mm2		\n\t"\
-		"punpcklwd %%mm5, %%mm5		\n\t"\
-		"punpcklwd %%mm4, %%mm4		\n\t"\
-		"paddw %%mm1, %%mm2		\n\t"\
-		"paddw %%mm1, %%mm5		\n\t"\
-		"paddw %%mm1, %%mm4		\n\t"\
-		"punpckhwd %%mm0, %%mm0		\n\t"\
-		"punpckhwd %%mm6, %%mm6		\n\t"\
-		"punpckhwd %%mm3, %%mm3		\n\t"\
-		"paddw %%mm7, %%mm0		\n\t"\
-		"paddw %%mm7, %%mm6		\n\t"\
-		"paddw %%mm7, %%mm3		\n\t"\
-		/* mm0=B1, mm2=B2, mm3=G2, mm4=G1, mm5=R1, mm6=R2 */\
-		"packuswb %%mm0, %%mm2		\n\t"\
-		"packuswb %%mm6, %%mm5		\n\t"\
-		"packuswb %%mm3, %%mm4		\n\t"\
-		"pxor %%mm7, %%mm7		\n\t"
-
-#define WRITEBGR32 \
-		/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
-			"movq %%mm2, %%mm1		\n\t" /* B */\
-			"movq %%mm5, %%mm6		\n\t" /* R */\
-			"punpcklbw %%mm4, %%mm2		\n\t" /* GBGBGBGB 0 */\
-			"punpcklbw %%mm7, %%mm5		\n\t" /* 0R0R0R0R 0 */\
-			"punpckhbw %%mm4, %%mm1		\n\t" /* GBGBGBGB 2 */\
-			"punpckhbw %%mm7, %%mm6		\n\t" /* 0R0R0R0R 2 */\
-			"movq %%mm2, %%mm0		\n\t" /* GBGBGBGB 0 */\
-			"movq %%mm1, %%mm3		\n\t" /* GBGBGBGB 2 */\
-			"punpcklwd %%mm5, %%mm0		\n\t" /* 0RGB0RGB 0 */\
-			"punpckhwd %%mm5, %%mm2		\n\t" /* 0RGB0RGB 1 */\
-			"punpcklwd %%mm6, %%mm1		\n\t" /* 0RGB0RGB 2 */\
-			"punpckhwd %%mm6, %%mm3		\n\t" /* 0RGB0RGB 3 */\
-\
-			MOVNTQ(%%mm0, (%4, %%eax, 4))\
-			MOVNTQ(%%mm2, 8(%4, %%eax, 4))\
-			MOVNTQ(%%mm1, 16(%4, %%eax, 4))\
-			MOVNTQ(%%mm3, 24(%4, %%eax, 4))\
-\
-			"addl $8, %%eax			\n\t"\
-			"cmpl %5, %%eax			\n\t"\
-			" jb 1b				\n\t"
-
-#define WRITEBGR16 \
-			"pand bF8, %%mm2		\n\t" /* B */\
-			"pand bFC, %%mm4		\n\t" /* G */\
-			"pand bF8, %%mm5		\n\t" /* R */\
-			"psrlq $3, %%mm2		\n\t"\
-\
-			"movq %%mm2, %%mm1		\n\t"\
-			"movq %%mm4, %%mm3		\n\t"\
-\
-			"punpcklbw %%mm7, %%mm3		\n\t"\
-			"punpcklbw %%mm5, %%mm2		\n\t"\
-			"punpckhbw %%mm7, %%mm4		\n\t"\
-			"punpckhbw %%mm5, %%mm1		\n\t"\
-\
-			"psllq $3, %%mm3		\n\t"\
-			"psllq $3, %%mm4		\n\t"\
-\
-			"por %%mm3, %%mm2		\n\t"\
-			"por %%mm4, %%mm1		\n\t"\
-\
-			MOVNTQ(%%mm2, (%4, %%eax, 2))\
-			MOVNTQ(%%mm1, 8(%4, %%eax, 2))\
-\
-			"addl $8, %%eax			\n\t"\
-			"cmpl %5, %%eax			\n\t"\
-			" jb 1b				\n\t"
-
-#define WRITEBGR15 \
-			"pand bF8, %%mm2		\n\t" /* B */\
-			"pand bF8, %%mm4		\n\t" /* G */\
-			"pand bF8, %%mm5		\n\t" /* R */\
-			"psrlq $3, %%mm2		\n\t"\
-			"psrlq $1, %%mm5		\n\t"\
-\
-			"movq %%mm2, %%mm1		\n\t"\
-			"movq %%mm4, %%mm3		\n\t"\
-\
-			"punpcklbw %%mm7, %%mm3		\n\t"\
-			"punpcklbw %%mm5, %%mm2		\n\t"\
-			"punpckhbw %%mm7, %%mm4		\n\t"\
-			"punpckhbw %%mm5, %%mm1		\n\t"\
-\
-			"psllq $2, %%mm3		\n\t"\
-			"psllq $2, %%mm4		\n\t"\
-\
-			"por %%mm3, %%mm2		\n\t"\
-			"por %%mm4, %%mm1		\n\t"\
-\
-			MOVNTQ(%%mm2, (%4, %%eax, 2))\
-			MOVNTQ(%%mm1, 8(%4, %%eax, 2))\
-\
-			"addl $8, %%eax			\n\t"\
-			"cmpl %5, %%eax			\n\t"\
-			" jb 1b				\n\t"
-
-#define WRITEBGR24OLD \
-		/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
-			"movq %%mm2, %%mm1		\n\t" /* B */\
-			"movq %%mm5, %%mm6		\n\t" /* R */\
-			"punpcklbw %%mm4, %%mm2		\n\t" /* GBGBGBGB 0 */\
-			"punpcklbw %%mm7, %%mm5		\n\t" /* 0R0R0R0R 0 */\
-			"punpckhbw %%mm4, %%mm1		\n\t" /* GBGBGBGB 2 */\
-			"punpckhbw %%mm7, %%mm6		\n\t" /* 0R0R0R0R 2 */\
-			"movq %%mm2, %%mm0		\n\t" /* GBGBGBGB 0 */\
-			"movq %%mm1, %%mm3		\n\t" /* GBGBGBGB 2 */\
-			"punpcklwd %%mm5, %%mm0		\n\t" /* 0RGB0RGB 0 */\
-			"punpckhwd %%mm5, %%mm2		\n\t" /* 0RGB0RGB 1 */\
-			"punpcklwd %%mm6, %%mm1		\n\t" /* 0RGB0RGB 2 */\
-			"punpckhwd %%mm6, %%mm3		\n\t" /* 0RGB0RGB 3 */\
-\
-			"movq %%mm0, %%mm4		\n\t" /* 0RGB0RGB 0 */\
-			"psrlq $8, %%mm0		\n\t" /* 00RGB0RG 0 */\
-			"pand bm00000111, %%mm4		\n\t" /* 00000RGB 0 */\
-			"pand bm11111000, %%mm0		\n\t" /* 00RGB000 0.5 */\
-			"por %%mm4, %%mm0		\n\t" /* 00RGBRGB 0 */\
-			"movq %%mm2, %%mm4		\n\t" /* 0RGB0RGB 1 */\
-			"psllq $48, %%mm2		\n\t" /* GB000000 1 */\
-			"por %%mm2, %%mm0		\n\t" /* GBRGBRGB 0 */\
-\
-			"movq %%mm4, %%mm2		\n\t" /* 0RGB0RGB 1 */\
-			"psrld $16, %%mm4		\n\t" /* 000R000R 1 */\
-			"psrlq $24, %%mm2		\n\t" /* 0000RGB0 1.5 */\
-			"por %%mm4, %%mm2		\n\t" /* 000RRGBR 1 */\
-			"pand bm00001111, %%mm2		\n\t" /* 0000RGBR 1 */\
-			"movq %%mm1, %%mm4		\n\t" /* 0RGB0RGB 2 */\
-			"psrlq $8, %%mm1		\n\t" /* 00RGB0RG 2 */\
-			"pand bm00000111, %%mm4		\n\t" /* 00000RGB 2 */\
-			"pand bm11111000, %%mm1		\n\t" /* 00RGB000 2.5 */\
-			"por %%mm4, %%mm1		\n\t" /* 00RGBRGB 2 */\
-			"movq %%mm1, %%mm4		\n\t" /* 00RGBRGB 2 */\
-			"psllq $32, %%mm1		\n\t" /* BRGB0000 2 */\
-			"por %%mm1, %%mm2		\n\t" /* BRGBRGBR 1 */\
-\
-			"psrlq $32, %%mm4		\n\t" /* 000000RG 2.5 */\
-			"movq %%mm3, %%mm5		\n\t" /* 0RGB0RGB 3 */\
-			"psrlq $8, %%mm3		\n\t" /* 00RGB0RG 3 */\
-			"pand bm00000111, %%mm5		\n\t" /* 00000RGB 3 */\
-			"pand bm11111000, %%mm3		\n\t" /* 00RGB000 3.5 */\
-			"por %%mm5, %%mm3		\n\t" /* 00RGBRGB 3 */\
-			"psllq $16, %%mm3		\n\t" /* RGBRGB00 3 */\
-			"por %%mm4, %%mm3		\n\t" /* RGBRGBRG 2.5 */\
-\
-			MOVNTQ(%%mm0, (%%ebx))\
-			MOVNTQ(%%mm2, 8(%%ebx))\
-			MOVNTQ(%%mm3, 16(%%ebx))\
-			"addl $24, %%ebx		\n\t"\
-\
-			"addl $8, %%eax			\n\t"\
-			"cmpl %5, %%eax			\n\t"\
-			" jb 1b				\n\t"
-
-#define WRITEBGR24MMX \
-		/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
-			"movq %%mm2, %%mm1		\n\t" /* B */\
-			"movq %%mm5, %%mm6		\n\t" /* R */\
-			"punpcklbw %%mm4, %%mm2		\n\t" /* GBGBGBGB 0 */\
-			"punpcklbw %%mm7, %%mm5		\n\t" /* 0R0R0R0R 0 */\
-			"punpckhbw %%mm4, %%mm1		\n\t" /* GBGBGBGB 2 */\
-			"punpckhbw %%mm7, %%mm6		\n\t" /* 0R0R0R0R 2 */\
-			"movq %%mm2, %%mm0		\n\t" /* GBGBGBGB 0 */\
-			"movq %%mm1, %%mm3		\n\t" /* GBGBGBGB 2 */\
-			"punpcklwd %%mm5, %%mm0		\n\t" /* 0RGB0RGB 0 */\
-			"punpckhwd %%mm5, %%mm2		\n\t" /* 0RGB0RGB 1 */\
-			"punpcklwd %%mm6, %%mm1		\n\t" /* 0RGB0RGB 2 */\
-			"punpckhwd %%mm6, %%mm3		\n\t" /* 0RGB0RGB 3 */\
-\
-			"movq %%mm0, %%mm4		\n\t" /* 0RGB0RGB 0 */\
-			"movq %%mm2, %%mm6		\n\t" /* 0RGB0RGB 1 */\
-			"movq %%mm1, %%mm5		\n\t" /* 0RGB0RGB 2 */\
-			"movq %%mm3, %%mm7		\n\t" /* 0RGB0RGB 3 */\
-\
-			"psllq $40, %%mm0		\n\t" /* RGB00000 0 */\
-			"psllq $40, %%mm2		\n\t" /* RGB00000 1 */\
-			"psllq $40, %%mm1		\n\t" /* RGB00000 2 */\
-			"psllq $40, %%mm3		\n\t" /* RGB00000 3 */\
-\
-			"punpckhdq %%mm4, %%mm0		\n\t" /* 0RGBRGB0 0 */\
-			"punpckhdq %%mm6, %%mm2		\n\t" /* 0RGBRGB0 1 */\
-			"punpckhdq %%mm5, %%mm1		\n\t" /* 0RGBRGB0 2 */\
-			"punpckhdq %%mm7, %%mm3		\n\t" /* 0RGBRGB0 3 */\
-\
-			"psrlq $8, %%mm0		\n\t" /* 00RGBRGB 0 */\
-			"movq %%mm2, %%mm6		\n\t" /* 0RGBRGB0 1 */\
-			"psllq $40, %%mm2		\n\t" /* GB000000 1 */\
-			"por %%mm2, %%mm0		\n\t" /* GBRGBRGB 0 */\
-			MOVNTQ(%%mm0, (%%ebx))\
-\
-			"psrlq $24, %%mm6		\n\t" /* 0000RGBR 1 */\
-			"movq %%mm1, %%mm5		\n\t" /* 0RGBRGB0 2 */\
-			"psllq $24, %%mm1		\n\t" /* BRGB0000 2 */\
-			"por %%mm1, %%mm6		\n\t" /* BRGBRGBR 1 */\
-			MOVNTQ(%%mm6, 8(%%ebx))\
-\
-			"psrlq $40, %%mm5		\n\t" /* 000000RG 2 */\
-			"psllq $8, %%mm3		\n\t" /* RGBRGB00 3 */\
-			"por %%mm3, %%mm5		\n\t" /* RGBRGBRG 2 */\
-			MOVNTQ(%%mm5, 16(%%ebx))\
-\
-			"addl $24, %%ebx		\n\t"\
-\
-			"addl $8, %%eax			\n\t"\
-			"cmpl %5, %%eax			\n\t"\
-			" jb 1b				\n\t"
-
-#define WRITEBGR24MMX2 \
-		/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */\
-			"movq M24A, %%mm0		\n\t"\
-			"movq M24C, %%mm7		\n\t"\
-			"pshufw $0x50, %%mm2, %%mm1	\n\t" /* B3 B2 B3 B2  B1 B0 B1 B0 */\
-			"pshufw $0x50, %%mm4, %%mm3	\n\t" /* G3 G2 G3 G2  G1 G0 G1 G0 */\
-			"pshufw $0x00, %%mm5, %%mm6	\n\t" /* R1 R0 R1 R0  R1 R0 R1 R0 */\
-\
-			"pand %%mm0, %%mm1		\n\t" /*    B2        B1       B0 */\
-			"pand %%mm0, %%mm3		\n\t" /*    G2        G1       G0 */\
-			"pand %%mm7, %%mm6		\n\t" /*       R1        R0       */\
-\
-			"psllq $8, %%mm3		\n\t" /* G2        G1       G0    */\
-			"por %%mm1, %%mm6		\n\t"\
-			"por %%mm3, %%mm6		\n\t"\
-			MOVNTQ(%%mm6, (%%ebx))\
-\
-			"psrlq $8, %%mm4		\n\t" /* 00 G7 G6 G5  G4 G3 G2 G1 */\
-			"pshufw $0xA5, %%mm2, %%mm1	\n\t" /* B5 B4 B5 B4  B3 B2 B3 B2 */\
-			"pshufw $0x55, %%mm4, %%mm3	\n\t" /* G4 G3 G4 G3  G4 G3 G4 G3 */\
-			"pshufw $0xA5, %%mm5, %%mm6	\n\t" /* R5 R4 R5 R4  R3 R2 R3 R2 */\
-\
-			"pand M24B, %%mm1		\n\t" /* B5       B4        B3    */\
-			"pand %%mm7, %%mm3		\n\t" /*       G4        G3       */\
-			"pand %%mm0, %%mm6		\n\t" /*    R4        R3       R2 */\
-\
-			"por %%mm1, %%mm3		\n\t" /* B5    G4 B4     G3 B3    */\
-			"por %%mm3, %%mm6		\n\t"\
-			MOVNTQ(%%mm6, 8(%%ebx))\
-\
-			"pshufw $0xFF, %%mm2, %%mm1	\n\t" /* B7 B6 B7 B6  B7 B6 B6 B7 */\
-			"pshufw $0xFA, %%mm4, %%mm3	\n\t" /* 00 G7 00 G7  G6 G5 G6 G5 */\
-			"pshufw $0xFA, %%mm5, %%mm6	\n\t" /* R7 R6 R7 R6  R5 R4 R5 R4 */\
-\
-			"pand %%mm7, %%mm1		\n\t" /*       B7        B6       */\
-			"pand %%mm0, %%mm3		\n\t" /*    G7        G6       G5 */\
-			"pand M24B, %%mm6		\n\t" /* R7       R6        R5    */\
-\
-			"por %%mm1, %%mm3		\n\t"\
-			"por %%mm3, %%mm6		\n\t"\
-			MOVNTQ(%%mm6, 16(%%ebx))\
-\
-			"addl $24, %%ebx		\n\t"\
-\
-			"addl $8, %%eax			\n\t"\
-			"cmpl %5, %%eax			\n\t"\
-			" jb 1b				\n\t"
-
-#ifdef HAVE_MMX2
-#define WRITEBGR24 WRITEBGR24MMX2
-#else
-#define WRITEBGR24 WRITEBGR24MMX
-#endif
-
-#ifdef HAVE_MMX
+#ifdef CAN_COMPILE_X86_ASM
 void in_asm_used_var_warning_killer()
 {
  int i= yCoeff+vrCoeff+ubCoeff+vgCoeff+ugCoeff+bF8+bFC+w400+w80+w10+
@@ -613,888 +135,57 @@ void in_asm_used_var_warning_killer()
 }
 #endif
 
-static inline void yuv2yuv(uint16_t *buf0, uint16_t *buf1, uint16_t *uvbuf0, uint16_t *uvbuf1,
-			   uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstw, int yalpha, int uvalpha)
-{
-	int yalpha1=yalpha^4095;
-	int uvalpha1=uvalpha^4095;
-	int i;
-
-	asm volatile ("\n\t"::: "memory");
-
-	for(i=0;i<dstw;i++)
-	{
-		((uint8_t*)dest)[i] = (buf0[i]*yalpha1+buf1[i]*yalpha)>>19;
-	}
-
-	if(uvalpha != -1)
-	{
-		for(i=0; i<(dstw>>1); i++)
-		{
-			((uint8_t*)uDest)[i] = (uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19;
-			((uint8_t*)vDest)[i] = (uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19;
-		}
-	}
-}
-
-/**
- * vertical scale YV12 to RGB
- */
-static inline void yuv2rgbX(uint16_t *buf0, uint16_t *buf1, uint16_t *uvbuf0, uint16_t *uvbuf1,
-			    uint8_t *dest, int dstw, int yalpha, int uvalpha, int dstbpp)
-{
-	int yalpha1=yalpha^4095;
-	int uvalpha1=uvalpha^4095;
-
-	if(fullUVIpol)
-	{
-
-#ifdef HAVE_MMX
-		if(dstbpp == 32)
-		{
-			asm volatile(
-
-
-FULL_YSCALEYUV2RGB
-			"punpcklbw %%mm1, %%mm3		\n\t" // BGBGBGBG
-			"punpcklbw %%mm7, %%mm0		\n\t" // R0R0R0R0
-
-			"movq %%mm3, %%mm1		\n\t"
-			"punpcklwd %%mm0, %%mm3		\n\t" // BGR0BGR0
-			"punpckhwd %%mm0, %%mm1		\n\t" // BGR0BGR0
-
-			MOVNTQ(%%mm3, (%4, %%eax, 4))
-			MOVNTQ(%%mm1, 8(%4, %%eax, 4))
-
-			"addl $4, %%eax			\n\t"
-			"cmpl %5, %%eax			\n\t"
-			" jb 1b				\n\t"
-
-
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-		else if(dstbpp==24)
-		{
-			asm volatile(
-
-FULL_YSCALEYUV2RGB
-
-								// lsb ... msb
-			"punpcklbw %%mm1, %%mm3		\n\t" // BGBGBGBG
-			"punpcklbw %%mm7, %%mm0		\n\t" // R0R0R0R0
-
-			"movq %%mm3, %%mm1		\n\t"
-			"punpcklwd %%mm0, %%mm3		\n\t" // BGR0BGR0
-			"punpckhwd %%mm0, %%mm1		\n\t" // BGR0BGR0
-
-			"movq %%mm3, %%mm2		\n\t" // BGR0BGR0
-			"psrlq $8, %%mm3		\n\t" // GR0BGR00
-			"pand bm00000111, %%mm2		\n\t" // BGR00000
-			"pand bm11111000, %%mm3		\n\t" // 000BGR00
-			"por %%mm2, %%mm3		\n\t" // BGRBGR00
-			"movq %%mm1, %%mm2		\n\t"
-			"psllq $48, %%mm1		\n\t" // 000000BG
-			"por %%mm1, %%mm3		\n\t" // BGRBGRBG
-
-			"movq %%mm2, %%mm1		\n\t" // BGR0BGR0
-			"psrld $16, %%mm2		\n\t" // R000R000
-			"psrlq $24, %%mm1		\n\t" // 0BGR0000
-			"por %%mm2, %%mm1		\n\t" // RBGRR000
-
-			"movl %4, %%ebx			\n\t"
-			"addl %%eax, %%ebx		\n\t"
-
-#ifdef HAVE_MMX2
-			//FIXME Alignment
-			"movntq %%mm3, (%%ebx, %%eax, 2)\n\t"
-			"movntq %%mm1, 8(%%ebx, %%eax, 2)\n\t"
-#else
-			"movd %%mm3, (%%ebx, %%eax, 2)	\n\t"
-			"psrlq $32, %%mm3		\n\t"
-			"movd %%mm3, 4(%%ebx, %%eax, 2)	\n\t"
-			"movd %%mm1, 8(%%ebx, %%eax, 2)	\n\t"
-#endif
-			"addl $4, %%eax			\n\t"
-			"cmpl %5, %%eax			\n\t"
-			" jb 1b				\n\t"
-
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "m" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax", "%ebx"
-			);
-		}
-		else if(dstbpp==15)
-		{
-			asm volatile(
-
-FULL_YSCALEYUV2RGB
-#ifdef DITHER1XBPP
-			"paddusb g5Dither, %%mm1	\n\t"
-			"paddusb r5Dither, %%mm0	\n\t"
-			"paddusb b5Dither, %%mm3	\n\t"
-#endif
-			"punpcklbw %%mm7, %%mm1		\n\t" // 0G0G0G0G
-			"punpcklbw %%mm7, %%mm3		\n\t" // 0B0B0B0B
-			"punpcklbw %%mm7, %%mm0		\n\t" // 0R0R0R0R
-
-			"psrlw $3, %%mm3		\n\t"
-			"psllw $2, %%mm1		\n\t"
-			"psllw $7, %%mm0		\n\t"
-			"pand g15Mask, %%mm1		\n\t"
-			"pand r15Mask, %%mm0		\n\t"
-
-			"por %%mm3, %%mm1		\n\t"
-			"por %%mm1, %%mm0		\n\t"
-
-			MOVNTQ(%%mm0, (%4, %%eax, 2))
-
-			"addl $4, %%eax			\n\t"
-			"cmpl %5, %%eax			\n\t"
-			" jb 1b				\n\t"
-
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-		else if(dstbpp==16)
-		{
-			asm volatile(
-
-FULL_YSCALEYUV2RGB
-#ifdef DITHER1XBPP
-			"paddusb g6Dither, %%mm1	\n\t"
-			"paddusb r5Dither, %%mm0	\n\t"
-			"paddusb b5Dither, %%mm3	\n\t"
-#endif
-			"punpcklbw %%mm7, %%mm1		\n\t" // 0G0G0G0G
-			"punpcklbw %%mm7, %%mm3		\n\t" // 0B0B0B0B
-			"punpcklbw %%mm7, %%mm0		\n\t" // 0R0R0R0R
-
-			"psrlw $3, %%mm3		\n\t"
-			"psllw $3, %%mm1		\n\t"
-			"psllw $8, %%mm0		\n\t"
-			"pand g16Mask, %%mm1		\n\t"
-			"pand r16Mask, %%mm0		\n\t"
-
-			"por %%mm3, %%mm1		\n\t"
-			"por %%mm1, %%mm0		\n\t"
-
-			MOVNTQ(%%mm0, (%4, %%eax, 2))
-
-			"addl $4, %%eax			\n\t"
-			"cmpl %5, %%eax			\n\t"
-			" jb 1b				\n\t"
-
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-#else
-		asm volatile ("\n\t"::: "memory");
-
-		if(dstbpp==32 || dstbpp==24)
-		{
-			int i;
-			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)];
-				int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19);
-				int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19);
-				dest[0]=clip_table[((Y + yuvtab_40cf[U]) >>13)];
-				dest[1]=clip_table[((Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13)];
-				dest[2]=clip_table[((Y + yuvtab_3343[V]) >>13)];
-				dest+=dstbpp>>3;
-			}
-		}
-		else if(dstbpp==16)
-		{
-			int i;
-			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)];
-				int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19);
-				int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19);
-
-				((uint16_t*)dest)[i] =
-					clip_table16b[(Y + yuvtab_40cf[U]) >>13] |
-					clip_table16g[(Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13] |
-					clip_table16r[(Y + yuvtab_3343[V]) >>13];
-			}
-		}
-		else if(dstbpp==15)
-		{
-			int i;
-			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)];
-				int U=((uvbuf0[i]*uvalpha1+uvbuf1[i]*uvalpha)>>19);
-				int V=((uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19);
-
-				((uint16_t*)dest)[i] =
-					clip_table15b[(Y + yuvtab_40cf[U]) >>13] |
-					clip_table15g[(Y + yuvtab_1a1e[V] + yuvtab_0c92[U]) >>13] |
-					clip_table15r[(Y + yuvtab_3343[V]) >>13];
-			}
-		}
-#endif
-	}//FULL_UV_IPOL
-	else
-	{
-#ifdef HAVE_MMX
-		if(dstbpp == 32)
-		{
-			asm volatile(
-				YSCALEYUV2RGB
-				WRITEBGR32
-
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-		else if(dstbpp==24)
-		{
-			asm volatile(
-				"movl %4, %%ebx			\n\t"
-				YSCALEYUV2RGB
-				WRITEBGR24
-
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "m" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax", "%ebx"
-			);
-		}
-		else if(dstbpp==15)
-		{
-			asm volatile(
-				YSCALEYUV2RGB
-		/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
-#ifdef DITHER1XBPP
-				"paddusb b5Dither, %%mm2	\n\t"
-				"paddusb g5Dither, %%mm4	\n\t"
-				"paddusb r5Dither, %%mm5	\n\t"
-#endif
-
-				WRITEBGR15
-
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-		else if(dstbpp==16)
-		{
-			asm volatile(
-				YSCALEYUV2RGB
-		/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
-#ifdef DITHER1XBPP
-				"paddusb b5Dither, %%mm2	\n\t"
-				"paddusb g6Dither, %%mm4	\n\t"
-				"paddusb r5Dither, %%mm5	\n\t"
-#endif
-
-				WRITEBGR16
-
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-#else
-		asm volatile ("\n\t"::: "memory");
-
-		if(dstbpp==32)
-		{
-			int i;
-			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)];
-				int Y2=yuvtab_2568[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19)];
-				int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
-				int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
-
-				int Cb= yuvtab_40cf[U];
-				int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
-				int Cr= yuvtab_3343[V];
-
-				dest[4*i+0]=clip_table[((Y1 + Cb) >>13)];
-				dest[4*i+1]=clip_table[((Y1 + Cg) >>13)];
-				dest[4*i+2]=clip_table[((Y1 + Cr) >>13)];
-
-				dest[4*i+4]=clip_table[((Y2 + Cb) >>13)];
-				dest[4*i+5]=clip_table[((Y2 + Cg) >>13)];
-				dest[4*i+6]=clip_table[((Y2 + Cr) >>13)];
-			}
-		}
-		if(dstbpp==24)
-		{
-			int i;
-			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)];
-				int Y2=yuvtab_2568[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19)];
-				int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
-				int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
-
-				int Cb= yuvtab_40cf[U];
-				int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
-				int Cr= yuvtab_3343[V];
-
-				dest[0]=clip_table[((Y1 + Cb) >>13)];
-				dest[1]=clip_table[((Y1 + Cg) >>13)];
-				dest[2]=clip_table[((Y1 + Cr) >>13)];
-
-				dest[3]=clip_table[((Y2 + Cb) >>13)];
-				dest[4]=clip_table[((Y2 + Cg) >>13)];
-				dest[5]=clip_table[((Y2 + Cr) >>13)];
-				dest+=6;
-			}
-		}
-		else if(dstbpp==16)
-		{
-			int i;
-			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)];
-				int Y2=yuvtab_2568[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19)];
-				int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
-				int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
-
-				int Cb= yuvtab_40cf[U];
-				int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
-				int Cr= yuvtab_3343[V];
-
-				((uint16_t*)dest)[i] =
-					clip_table16b[(Y1 + Cb) >>13] |
-					clip_table16g[(Y1 + Cg) >>13] |
-					clip_table16r[(Y1 + Cr) >>13];
-
-				((uint16_t*)dest)[i+1] =
-					clip_table16b[(Y2 + Cb) >>13] |
-					clip_table16g[(Y2 + Cg) >>13] |
-					clip_table16r[(Y2 + Cr) >>13];
-			}
-		}
-		else if(dstbpp==15)
-		{
-			int i;
-			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)];
-				int Y2=yuvtab_2568[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19)];
-				int U=((uvbuf0[i>>1]*uvalpha1+uvbuf1[i>>1]*uvalpha)>>19);
-				int V=((uvbuf0[(i>>1)+2048]*uvalpha1+uvbuf1[(i>>1)+2048]*uvalpha)>>19);
-
-				int Cb= yuvtab_40cf[U];
-				int Cg= yuvtab_1a1e[V] + yuvtab_0c92[U];
-				int Cr= yuvtab_3343[V];
-
-				((uint16_t*)dest)[i] =
-					clip_table15b[(Y1 + Cb) >>13] |
-					clip_table15g[(Y1 + Cg) >>13] |
-					clip_table15r[(Y1 + Cr) >>13];
-
-				((uint16_t*)dest)[i+1] =
-					clip_table15b[(Y2 + Cb) >>13] |
-					clip_table15g[(Y2 + Cg) >>13] |
-					clip_table15r[(Y2 + Cr) >>13];
-			}
-		}
-#endif
-	} //!FULL_UV_IPOL
-}
-
-/**
- * YV12 to RGB without scaling or interpolating
- */
-static inline void yuv2rgb1(uint16_t *buf0, uint16_t *buf1, uint16_t *uvbuf0, uint16_t *uvbuf1,
-			    uint8_t *dest, int dstw, int yalpha, int uvalpha, int dstbpp)
-{
-	int uvalpha1=uvalpha^4095;
-#ifdef HAVE_MMX
-	int yalpha1=yalpha^4095;
-#endif
-
-	if(fullUVIpol || allwaysIpol)
-	{
-		yuv2rgbX(buf0, buf1, uvbuf0, uvbuf1, dest, dstw, yalpha, uvalpha, dstbpp);
-		return;
-	}
-	if( yalpha > 2048 ) buf0 = buf1;
-
-#ifdef HAVE_MMX
-	if( uvalpha < 2048 ) // note this is not correct (shifts chrominance by 0.5 pixels) but its a bit faster
-	{
-		if(dstbpp == 32)
-		{
-			asm volatile(
-				YSCALEYUV2RGB1
-				WRITEBGR32
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-		else if(dstbpp==24)
-		{
-			asm volatile(
-				"movl %4, %%ebx			\n\t"
-				YSCALEYUV2RGB1
-				WRITEBGR24
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "m" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax", "%ebx"
-			);
-		}
-		else if(dstbpp==15)
-		{
-			asm volatile(
-				YSCALEYUV2RGB1
-		/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
-#ifdef DITHER1XBPP
-				"paddusb b5Dither, %%mm2	\n\t"
-				"paddusb g5Dither, %%mm4	\n\t"
-				"paddusb r5Dither, %%mm5	\n\t"
-#endif
-				WRITEBGR15
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-		else if(dstbpp==16)
-		{
-			asm volatile(
-				YSCALEYUV2RGB1
-		/* mm2=B, %%mm4=G, %%mm5=R, %%mm7=0 */
-#ifdef DITHER1XBPP
-				"paddusb b5Dither, %%mm2	\n\t"
-				"paddusb g6Dither, %%mm4	\n\t"
-				"paddusb r5Dither, %%mm5	\n\t"
-#endif
-
-				WRITEBGR16
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-	}
-	else
-	{
-		if(dstbpp == 32)
-		{
-			asm volatile(
-				YSCALEYUV2RGB1b
-				WRITEBGR32
-			:: "r" (buf0), "r" (buf1), "r" (uvbuf0), "r" (uvbuf1), "r" (dest), "m" (dstw),
-			"m" (yalpha1), "m" (uvalpha1)
-			: "%eax"
-			);
-		}
-		else if(dstbpp==24)
-		{
-			asm volatile(
-				"movl %4, %%ebx			\n\t"