swscale_template.c

uclinux 下的vlc播放器源代码

		"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++)
	{
		int b= 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 r= 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;
	}
#endif
}

static inline void RENAME(bgr16ToY)(uint8_t *dst, uint8_t *src, int width)
{
	int i;
	for(i=0; i<width; i++)
	{
		int d= ((uint16_t*)src)[i];
		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++)
	{
		int d0= ((uint32_t*)src1)[i];
		int d1= ((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;
		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= ((uint16_t*)src)[i];
		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++)
	{
		int d0= ((uint32_t*)src1)[i];
		int d1= ((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;
		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=  ((uint32_t*)src)[i]&0xFF;
		int g= (((uint32_t*)src)[i]>>8)&0xFF;
		int b= (((uint32_t*)src)[i]>>16)&0xFF;

		dst[i]= ((RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT);
	}
}

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++)
	{
		const int a= ((uint32_t*)src1)[2*i+0];
		const int e= ((uint32_t*)src1)[2*i+1];
		const int c= ((uint32_t*)src2)[2*i+0];
		const int d= ((uint32_t*)src2)[2*i+1];
		const int l= (a&0xFF00FF) + (e&0xFF00FF) + (c&0xFF00FF) + (d&0xFF00FF);
		const int h= (a&0x00FF00) + (e&0x00FF00) + (c&0x00FF00) + (d&0x00FF00);
 		const int r=  l&0x3FF;
		const int g=  h>>8;
		const int b=  l>>16;

		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 + (33<<(RGB2YUV_SHIFT-1)) )>>RGB2YUV_SHIFT);
	}
}

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,
				  int16_t *filter, int16_t *filterPos, int filterSize)
{
#ifdef HAVE_MMX
	assert(filterSize % 4 == 0 && filterSize>0);
	if(filterSize==4) // allways true for upscaling, sometimes for down too
	{
		int counter= -2*dstW;
		filter-= counter*2;
		filterPos-= counter/2;
		dst-= counter/2;
		asm volatile(
			"pxor %%mm7, %%mm7		\n\t"
			"movq "MANGLE(w02)", %%mm6	\n\t"
			"pushl %%ebp			\n\t" // we use 7 regs here ...
			"movl %%eax, %%ebp		\n\t"
			".balign 16			\n\t"
			"1:				\n\t"
			"movzwl (%2, %%ebp), %%eax	\n\t"
			"movzwl 2(%2, %%ebp), %%ebx	\n\t"
			"movq (%1, %%ebp, 4), %%mm1	\n\t"
			"movq 8(%1, %%ebp, 4), %%mm3	\n\t"
			"movd (%3, %%eax), %%mm0	\n\t"
			"movd (%3, %%ebx), %%mm2	\n\t"
			"punpcklbw %%mm7, %%mm0		\n\t"
			"punpcklbw %%mm7, %%mm2		\n\t"
			"pmaddwd %%mm1, %%mm0		\n\t"
			"pmaddwd %%mm2, %%mm3		\n\t"
			"psrad $8, %%mm0		\n\t"
			"psrad $8, %%mm3		\n\t"
			"packssdw %%mm3, %%mm0		\n\t"
			"pmaddwd %%mm6, %%mm0		\n\t"
			"packssdw %%mm0, %%mm0		\n\t"
			"movd %%mm0, (%4, %%ebp)	\n\t"
			"addl $4, %%ebp			\n\t"
			" jnc 1b			\n\t"

			"popl %%ebp			\n\t"
			: "+a" (counter)
			: "c" (filter), "d" (filterPos), "S" (src), "D" (dst)
			: "%ebx"
		);
	}
	else if(filterSize==8)
	{
		int counter= -2*dstW;
		filter-= counter*4;
		filterPos-= counter/2;
		dst-= counter/2;
		asm volatile(
			"pxor %%mm7, %%mm7		\n\t"
			"movq "MANGLE(w02)", %%mm6	\n\t"
			"pushl %%ebp			\n\t" // we use 7 regs here ...
			"movl %%eax, %%ebp		\n\t"
			".balign 16			\n\t"
			"1:				\n\t"
			"movzwl (%2, %%ebp), %%eax	\n\t"
			"movzwl 2(%2, %%ebp), %%ebx	\n\t"
			"movq (%1, %%ebp, 8), %%mm1	\n\t"
			"movq 16(%1, %%ebp, 8), %%mm3	\n\t"
			"movd (%3, %%eax), %%mm0	\n\t"
			"movd (%3, %%ebx), %%mm2	\n\t"
			"punpcklbw %%mm7, %%mm0		\n\t"
			"punpcklbw %%mm7, %%mm2		\n\t"
			"pmaddwd %%mm1, %%mm0		\n\t"
			"pmaddwd %%mm2, %%mm3		\n\t"

			"movq 8(%1, %%ebp, 8), %%mm1	\n\t"
			"movq 24(%1, %%ebp, 8), %%mm5	\n\t"
			"movd 4(%3, %%eax), %%mm4	\n\t"
			"movd 4(%3, %%ebx), %%mm2	\n\t"
			"punpcklbw %%mm7, %%mm4		\n\t"
			"punpcklbw %%mm7, %%mm2		\n\t"
			"pmaddwd %%mm1, %%mm4		\n\t"
			"pmaddwd %%mm2, %%mm5		\n\t"
			"paddd %%mm4, %%mm0		\n\t"
			"paddd %%mm5, %%mm3		\n\t"
						
			"psrad $8, %%mm0		\n\t"
			"psrad $8, %%mm3		\n\t"
			"packssdw %%mm3, %%mm0		\n\t"
			"pmaddwd %%mm6, %%mm0		\n\t"
			"packssdw %%mm0, %%mm0		\n\t"
			"movd %%mm0, (%4, %%ebp)	\n\t"
			"addl $4, %%ebp			\n\t"
			" jnc 1b			\n\t"

			"popl %%ebp			\n\t"
			: "+a" (counter)
			: "c" (filter), "d" (filterPos), "S" (src), "D" (dst)
			: "%ebx"
		);
	}
	else
	{
		int counter= -2*dstW;
//		filter-= counter*filterSize/2;
		filterPos-= counter/2;
		dst-= counter/2;
		asm volatile(
			"pxor %%mm7, %%mm7		\n\t"
			"movq "MANGLE(w02)", %%mm6	\n\t"
			".balign 16			\n\t"
			"1:				\n\t"
			"movl %2, %%ecx			\n\t"
			"movzwl (%%ecx, %0), %%eax	\n\t"
			"movzwl 2(%%ecx, %0), %%ebx	\n\t"
			"movl %5, %%ecx			\n\t"
			"pxor %%mm4, %%mm4		\n\t"
			"pxor %%mm5, %%mm5		\n\t"
			"2:				\n\t"
			"movq (%1), %%mm1		\n\t"
			"movq (%1, %6), %%mm3		\n\t"
			"movd (%%ecx, %%eax), %%mm0	\n\t"
			"movd (%%ecx, %%ebx), %%mm2	\n\t"
			"punpcklbw %%mm7, %%mm0		\n\t"
			"punpcklbw %%mm7, %%mm2		\n\t"
			"pmaddwd %%mm1, %%mm0		\n\t"
			"pmaddwd %%mm2, %%mm3		\n\t"
			"paddd %%mm3, %%mm5		\n\t"
			"paddd %%mm0, %%mm4		\n\t"
			"addl $8, %1			\n\t"
			"addl $4, %%ecx			\n\t"
			"cmpl %4, %%ecx			\n\t"
			" jb 2b				\n\t"
			"addl %6, %1			\n\t"
			"psrad $8, %%mm4		\n\t"
			"psrad $8, %%mm5		\n\t"
			"packssdw %%mm5, %%mm4		\n\t"
			"pmaddwd %%mm6, %%mm4		\n\t"
			"packssdw %%mm4, %%mm4		\n\t"
			"movl %3, %%eax			\n\t"
			"movd %%mm4, (%%eax, %0)	\n\t"
			"addl $4, %0			\n\t"
			" jnc 1b			\n\t"

			: "+r" (counter), "+r" (filter)
			: "m" (filterPos), "m" (dst), "m"(src+filterSize),
			  "m" (src), "r" (filterSize*2)
			: "%ebx", "%eax", "%ecx"
		);
	}
#else
#ifdef HAVE_ALTIVEC
	hScale_altivec_real(dst, dstW, src, srcW, xInc, filter, filterPos, filterSize);
#else
	int i;
	for(i=0; i<dstW; i++)
	{
		int j;
		int srcPos= filterPos[i];
		int val=0;
//		printf("filterPos: %d\n", filterPos[i]);
		for(j=0; j<filterSize; j++)
		{
//			printf("filter: %d, src: %d\n", filter[i], src[srcPos + j]);
			val += ((int)src[srcPos + j])*filter[filterSize*i + j];
		}
//		filter += hFilterSize;
		dst[i] = MIN(MAX(0, val>>7), (1<<15)-1); // the cubic equation does overflow ...
//		dst[i] = val>>7;
	}
#endif
#endif
}
      // *** horizontal scale Y line to temp buffer
static inline void RENAME(hyscale)(uint16_t *dst, int dstWidth, uint8_t *src, int srcW, int xInc,
				   int flags, int canMMX2BeUsed,