swscale_template.c

uclinux 下的vlc播放器源代码

/*
    Copyright (C) 2001-2003 Michael Niedermayer <[e-mail]>

    This program 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.

    This program 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 this program; if not, write to the Free Software
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
*/

#undef MOVNTQ
#undef PAVGB
#undef PREFETCH
#undef PREFETCHW
#undef EMMS
#undef SFENCE

#ifdef HAVE_3DNOW
/* On K6 femms is faster of emms. On K7 femms is directly mapped on emms. */
#define EMMS     "femms"
#else
#define EMMS     "emms"
#endif

#ifdef HAVE_3DNOW
#define PREFETCH  "prefetch"
#define PREFETCHW "prefetchw"
#elif defined ( HAVE_MMX2 )
#define PREFETCH "prefetchnta"
#define PREFETCHW "prefetcht0"
#else
#define PREFETCH "/nop"
#define PREFETCHW "/nop"
#endif

#ifdef HAVE_MMX2
#define SFENCE "sfence"
#else
#define SFENCE "/nop"
#endif

#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"
#endif

#ifdef HAVE_ALTIVEC
#include "swscale_altivec_template.c"
#endif

#define YSCALEYUV2YV12X(x, offset) \
			"xorl %%eax, %%eax		\n\t"\
			"movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\
			"movq %%mm3, %%mm4		\n\t"\
			"leal " offset "(%0), %%edx	\n\t"\
			"movl (%%edx), %%esi		\n\t"\
			".balign 16			\n\t" /* FIXME Unroll? */\
			"1:				\n\t"\
			"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"\
			" jnz 1b			\n\t"\
			"psraw $3, %%mm3		\n\t"\
			"psraw $3, %%mm4		\n\t"\
			"packuswb %%mm4, %%mm3		\n\t"\
			MOVNTQ(%%mm3, (%1, %%eax))\
			"addl $8, %%eax			\n\t"\
			"cmpl %2, %%eax			\n\t"\
			"movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\
			"movq %%mm3, %%mm4		\n\t"\
			"leal " offset "(%0), %%edx	\n\t"\
			"movl (%%edx), %%esi		\n\t"\
			"jb 1b				\n\t"

#define YSCALEYUV2YV121 \
			"movl %2, %%eax			\n\t"\
			".balign 16			\n\t" /* FIXME Unroll? */\
			"1:				\n\t"\
			"movq (%0, %%eax, 2), %%mm0	\n\t"\
			"movq 8(%0, %%eax, 2), %%mm1	\n\t"\
			"psraw $7, %%mm0		\n\t"\
			"psraw $7, %%mm1		\n\t"\
			"packuswb %%mm1, %%mm0		\n\t"\
			MOVNTQ(%%mm0, (%1, %%eax))\
			"addl $8, %%eax			\n\t"\
			"jnc 1b				\n\t"

/*
			:: "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"
*/
#define YSCALEYUV2PACKEDX \
		"xorl %%eax, %%eax		\n\t"\
		".balign 16			\n\t"\
		"nop				\n\t"\
		"1:				\n\t"\
		"leal "CHR_MMX_FILTER_OFFSET"(%0), %%edx	\n\t"\
		"movl (%%edx), %%esi		\n\t"\
		"movq "VROUNDER_OFFSET"(%0), %%mm3\n\t"\
		"movq %%mm3, %%mm4		\n\t"\
		".balign 16			\n\t"\
		"2:				\n\t"\
		"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"\
		"testl %%esi, %%esi		\n\t"\
		" jnz 2b			\n\t"\
\
		"leal "LUM_MMX_FILTER_OFFSET"(%0), %%edx	\n\t"\
		"movl (%%edx), %%esi		\n\t"\
		"movq "VROUNDER_OFFSET"(%0), %%mm1\n\t"\
		"movq %%mm1, %%mm7		\n\t"\
		".balign 16			\n\t"\
		"2:				\n\t"\
		"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"\
		"testl %%esi, %%esi		\n\t"\
		" jnz 2b			\n\t"\


#define YSCALEYUV2RGBX \
		YSCALEYUV2PACKEDX\
		"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 "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 "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"\
		"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"
#if 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 "MANGLE(w80)", %%mm1	\n\t" /* 8(Y-16)*/\
		"psubw "MANGLE(w400)", %%mm3	\n\t" /* 8(U-128)*/\
		"pmulhw "MANGLE(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 "MANGLE(ubCoeff)", %%mm3\n\t"\
		"psraw $4, %%mm0		\n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
		"pmulhw "MANGLE(ugCoeff)", %%mm2\n\t"\
		"paddw %%mm4, %%mm0		\n\t" /* uvbuf0[eax+2048]uvalpha1 - uvbuf1[eax+2048](1-uvalpha1)*/\
		"psubw "MANGLE(w400)", %%mm0	\n\t" /* (V-128)8*/\
\
\
		"movq %%mm0, %%mm4		\n\t" /* (V-128)8*/\
		"pmulhw "MANGLE(vrCoeff)", %%mm0\n\t"\
		"pmulhw "MANGLE(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"
#endif

#define YSCALEYUV2PACKED(index, c) \
		"movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t"\
		"movq "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm1\n\t"\
		"psraw $3, %%mm0		\n\t"\
		"psraw $3, %%mm1		\n\t"\
		"movq %%mm0, "CHR_MMX_FILTER_OFFSET"+8("#c")\n\t"\
		"movq %%mm1, "LUM_MMX_FILTER_OFFSET"+8("#c")\n\t"\
		"xorl "#index", "#index"		\n\t"\
		".balign 16			\n\t"\
		"1:				\n\t"\
		"movq (%2, "#index"), %%mm2	\n\t" /* uvbuf0[eax]*/\
		"movq (%3, "#index"), %%mm3	\n\t" /* uvbuf1[eax]*/\
		"movq 4096(%2, "#index"), %%mm5	\n\t" /* uvbuf0[eax+2048]*/\
		"movq 4096(%3, "#index"), %%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 "CHR_MMX_FILTER_OFFSET"+8("#c"), %%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 $7, %%mm3		\n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
		"psraw $7, %%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)*/\
		"movq (%0, "#index", 2), %%mm0	\n\t" /*buf0[eax]*/\
		"movq (%1, "#index", 2), %%mm1	\n\t" /*buf1[eax]*/\
		"movq 8(%0, "#index", 2), %%mm6	\n\t" /*buf0[eax]*/\
		"movq 8(%1, "#index", 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 "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
		"pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm6\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
		"psraw $7, %%mm1		\n\t" /* buf0[eax] - buf1[eax] >>4*/\
		"psraw $7, %%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*/\
                
#define YSCALEYUV2RGB(index, c) \
		"xorl "#index", "#index"	\n\t"\
		".balign 16			\n\t"\
		"1:				\n\t"\
		"movq (%2, "#index"), %%mm2	\n\t" /* uvbuf0[eax]*/\
		"movq (%3, "#index"), %%mm3	\n\t" /* uvbuf1[eax]*/\
		"movq 4096(%2, "#index"), %%mm5\n\t" /* uvbuf0[eax+2048]*/\
		"movq 4096(%3, "#index"), %%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 "CHR_MMX_FILTER_OFFSET"+8("#c"), %%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 "U_OFFSET"("#c"), %%mm3	\n\t" /* (U-128)8*/\
		"psubw "V_OFFSET"("#c"), %%mm4	\n\t" /* (V-128)8*/\
		"movq %%mm3, %%mm2		\n\t" /* (U-128)8*/\
		"movq %%mm4, %%mm5		\n\t" /* (V-128)8*/\
		"pmulhw "UG_COEFF"("#c"), %%mm3\n\t"\
		"pmulhw "VG_COEFF"("#c"), %%mm4\n\t"\
	/* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
		"movq (%0, "#index", 2), %%mm0	\n\t" /*buf0[eax]*/\
		"movq (%1, "#index", 2), %%mm1	\n\t" /*buf1[eax]*/\
		"movq 8(%0, "#index", 2), %%mm6\n\t" /*buf0[eax]*/\
		"movq 8(%1, "#index", 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 "LUM_MMX_FILTER_OFFSET"+8("#c"), %%mm0\n\t" /* (buf0[eax] - buf1[eax])yalpha1>>16*/\
		"pmulhw "LUM_MMX_FILTER_OFFSET"+8("#c"), %%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 "UB_COEFF"("#c"), %%mm2\n\t"\
		"pmulhw "VR_COEFF"("#c"), %%mm5\n\t"\
		"psubw "Y_OFFSET"("#c"), %%mm1	\n\t" /* 8(Y-16)*/\
		"psubw "Y_OFFSET"("#c"), %%mm7	\n\t" /* 8(Y-16)*/\
		"pmulhw "Y_COEFF"("#c"), %%mm1	\n\t"\
		"pmulhw "Y_COEFF"("#c"), %%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 YSCALEYUV2PACKED1(index, c) \
		"xorl "#index", "#index"		\n\t"\
		".balign 16			\n\t"\
		"1:				\n\t"\
		"movq (%2, "#index"), %%mm3	\n\t" /* uvbuf0[eax]*/\
		"movq 4096(%2, "#index"), %%mm4	\n\t" /* uvbuf0[eax+2048]*/\
		"psraw $7, %%mm3		\n\t" \
		"psraw $7, %%mm4		\n\t" \
		"movq (%0, "#index", 2), %%mm1	\n\t" /*buf0[eax]*/\
		"movq 8(%0, "#index", 2), %%mm7	\n\t" /*buf0[eax]*/\
		"psraw $7, %%mm1		\n\t" \
		"psraw $7, %%mm7		\n\t" \
                
#define YSCALEYUV2RGB1(index, c) \
		"xorl "#index", "#index"	\n\t"\
		".balign 16			\n\t"\
		"1:				\n\t"\
		"movq (%2, "#index"), %%mm3	\n\t" /* uvbuf0[eax]*/\
		"movq 4096(%2, "#index"), %%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 "U_OFFSET"("#c"), %%mm3	\n\t" /* (U-128)8*/\
		"psubw "V_OFFSET"("#c"), %%mm4	\n\t" /* (V-128)8*/\
		"movq %%mm3, %%mm2		\n\t" /* (U-128)8*/\
		"movq %%mm4, %%mm5		\n\t" /* (V-128)8*/\
		"pmulhw "UG_COEFF"("#c"), %%mm3\n\t"\
		"pmulhw "VG_COEFF"("#c"), %%mm4\n\t"\
	/* mm2=(U-128)8, mm3=ug, mm4=vg mm5=(V-128)8 */\
		"movq (%0, "#index", 2), %%mm1	\n\t" /*buf0[eax]*/\
		"movq 8(%0, "#index", 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 "UB_COEFF"("#c"), %%mm2\n\t"\
		"pmulhw "VR_COEFF"("#c"), %%mm5\n\t"\
		"psubw "Y_OFFSET"("#c"), %%mm1	\n\t" /* 8(Y-16)*/\
		"psubw "Y_OFFSET"("#c"), %%mm7	\n\t" /* 8(Y-16)*/\
		"pmulhw "Y_COEFF"("#c"), %%mm1	\n\t"\
		"pmulhw "Y_COEFF"("#c"), %%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 YSCALEYUV2PACKED1b(index, c) \
		"xorl "#index", "#index"		\n\t"\
		".balign 16			\n\t"\
		"1:				\n\t"\
		"movq (%2, "#index"), %%mm2	\n\t" /* uvbuf0[eax]*/\
		"movq (%3, "#index"), %%mm3	\n\t" /* uvbuf1[eax]*/\
		"movq 4096(%2, "#index"), %%mm5	\n\t" /* uvbuf0[eax+2048]*/\
		"movq 4096(%3, "#index"), %%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 $8, %%mm3		\n\t" \
		"psrlw $8, %%mm4		\n\t" \
		"movq (%0, "#index", 2), %%mm1	\n\t" /*buf0[eax]*/\
		"movq 8(%0, "#index", 2), %%mm7	\n\t" /*buf0[eax]*/\
		"psraw $7, %%mm1		\n\t" \
		"psraw $7, %%mm7		\n\t" 
                
// do vertical chrominance interpolation
#define YSCALEYUV2RGB1b(index, c) \
		"xorl "#index", "#index"		\n\t"\
		".balign 16			\n\t"\
		"1:				\n\t"\
		"movq (%2, "#index"), %%mm2	\n\t" /* uvbuf0[eax]*/\
		"movq (%3, "#index"), %%mm3	\n\t" /* uvbuf1[eax]*/\
		"movq 4096(%2, "#index"), %%mm5	\n\t" /* uvbuf0[eax+2048]*/\
		"movq 4096(%3, "#index"), %%mm4	\n\t" /* uvbuf1[eax+2048]*/\