rgb2rgb_template.c

video linux conference

		"punpckldq 8%1, %%mm0\n\t"
		"punpckldq 12%1, %%mm3\n\t"
		"movq	%%mm0, %%mm1\n\t"
		"movq	%%mm0, %%mm2\n\t"
		"movq	%%mm3, %%mm4\n\t"
		"movq	%%mm3, %%mm5\n\t"
		"psrlq	$3, %%mm0\n\t"
		"psrlq	$3, %%mm3\n\t"
		"pand	%2, %%mm0\n\t"
		"pand	%2, %%mm3\n\t"
		"psrlq	$6, %%mm1\n\t"
		"psrlq	$6, %%mm4\n\t"
		"pand	%%mm6, %%mm1\n\t"
		"pand	%%mm6, %%mm4\n\t"
		"psrlq	$9, %%mm2\n\t"
		"psrlq	$9, %%mm5\n\t"
		"pand	%%mm7, %%mm2\n\t"
		"pand	%%mm7, %%mm5\n\t"
		"por	%%mm1, %%mm0\n\t"
		"por	%%mm4, %%mm3\n\t"
		"por	%%mm2, %%mm0\n\t"
		"por	%%mm5, %%mm3\n\t"
		"psllq	$16, %%mm3\n\t"
		"por	%%mm3, %%mm0\n\t"
		MOVNTQ"	%%mm0, %0\n\t"
		:"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");
		d += 4;
		s += 16;
	}
#endif
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS:::"memory");
#endif
	while(s < end)
	{
		const int src= *s; s += 4;
		*d++ = ((src&0xFF)>>3) + ((src&0xF800)>>6) + ((src&0xF80000)>>9);
	}
}

static inline void RENAME(rgb32tobgr15)(const uint8_t *src, uint8_t *dst, unsigned src_size)
{
	const uint8_t *s = src;
	const uint8_t *end;
#ifdef HAVE_MMX
	const uint8_t *mm_end;
#endif
	uint16_t *d = (uint16_t *)dst;
	end = s + src_size;
#ifdef HAVE_MMX
	__asm __volatile(PREFETCH"	%0"::"m"(*src):"memory");
	__asm __volatile(
	    "movq	%0, %%mm7\n\t"
	    "movq	%1, %%mm6\n\t"
	    ::"m"(red_15mask),"m"(green_15mask));
	mm_end = end - 15;
	while(s < mm_end)
	{
	    __asm __volatile(
		PREFETCH" 32%1\n\t"
		"movd	%1, %%mm0\n\t"
		"movd	4%1, %%mm3\n\t"
		"punpckldq 8%1, %%mm0\n\t"
		"punpckldq 12%1, %%mm3\n\t"
		"movq	%%mm0, %%mm1\n\t"
		"movq	%%mm0, %%mm2\n\t"
		"movq	%%mm3, %%mm4\n\t"
		"movq	%%mm3, %%mm5\n\t"
		"psllq	$7, %%mm0\n\t"
		"psllq	$7, %%mm3\n\t"
		"pand	%%mm7, %%mm0\n\t"
		"pand	%%mm7, %%mm3\n\t"
		"psrlq	$6, %%mm1\n\t"
		"psrlq	$6, %%mm4\n\t"
		"pand	%%mm6, %%mm1\n\t"
		"pand	%%mm6, %%mm4\n\t"
		"psrlq	$19, %%mm2\n\t"
		"psrlq	$19, %%mm5\n\t"
		"pand	%2, %%mm2\n\t"
		"pand	%2, %%mm5\n\t"
		"por	%%mm1, %%mm0\n\t"
		"por	%%mm4, %%mm3\n\t"
		"por	%%mm2, %%mm0\n\t"
		"por	%%mm5, %%mm3\n\t"
		"psllq	$16, %%mm3\n\t"
		"por	%%mm3, %%mm0\n\t"
		MOVNTQ"	%%mm0, %0\n\t"
		:"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");
		d += 4;
		s += 16;
	}
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS:::"memory");
#endif
	while(s < end)
	{
		const int src= *s; s += 4;
		*d++ = ((src&0xF8)<<7) + ((src&0xF800)>>6) + ((src&0xF80000)>>19);
	}
}

static inline void RENAME(rgb24to16)(const uint8_t *src, uint8_t *dst, unsigned src_size)
{
	const uint8_t *s = src;
	const uint8_t *end;
#ifdef HAVE_MMX
	const uint8_t *mm_end;
#endif
	uint16_t *d = (uint16_t *)dst;
	end = s + src_size;
#ifdef HAVE_MMX
	__asm __volatile(PREFETCH"	%0"::"m"(*src):"memory");
	__asm __volatile(
	    "movq	%0, %%mm7\n\t"
	    "movq	%1, %%mm6\n\t"
	    ::"m"(red_16mask),"m"(green_16mask));
	mm_end = end - 11;
	while(s < mm_end)
	{
	    __asm __volatile(
		PREFETCH" 32%1\n\t"
		"movd	%1, %%mm0\n\t"
		"movd	3%1, %%mm3\n\t"
		"punpckldq 6%1, %%mm0\n\t"
		"punpckldq 9%1, %%mm3\n\t"
		"movq	%%mm0, %%mm1\n\t"
		"movq	%%mm0, %%mm2\n\t"
		"movq	%%mm3, %%mm4\n\t"
		"movq	%%mm3, %%mm5\n\t"
		"psrlq	$3, %%mm0\n\t"
		"psrlq	$3, %%mm3\n\t"
		"pand	%2, %%mm0\n\t"
		"pand	%2, %%mm3\n\t"
		"psrlq	$5, %%mm1\n\t"
		"psrlq	$5, %%mm4\n\t"
		"pand	%%mm6, %%mm1\n\t"
		"pand	%%mm6, %%mm4\n\t"
		"psrlq	$8, %%mm2\n\t"
		"psrlq	$8, %%mm5\n\t"
		"pand	%%mm7, %%mm2\n\t"
		"pand	%%mm7, %%mm5\n\t"
		"por	%%mm1, %%mm0\n\t"
		"por	%%mm4, %%mm3\n\t"
		"por	%%mm2, %%mm0\n\t"
		"por	%%mm5, %%mm3\n\t"
		"psllq	$16, %%mm3\n\t"
		"por	%%mm3, %%mm0\n\t"
		MOVNTQ"	%%mm0, %0\n\t"
		:"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");
		d += 4;
		s += 12;
	}
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS:::"memory");
#endif
	while(s < end)
	{
		const int b= *s++;
		const int g= *s++;
		const int r= *s++;
		*d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);
	}
}

static inline void RENAME(rgb24tobgr16)(const uint8_t *src, uint8_t *dst, unsigned int src_size)
{
	const uint8_t *s = src;
	const uint8_t *end;
#ifdef HAVE_MMX
	const uint8_t *mm_end;
#endif
	uint16_t *d = (uint16_t *)dst;
	end = s + src_size;
#ifdef HAVE_MMX
	__asm __volatile(PREFETCH"	%0"::"m"(*src):"memory");
	__asm __volatile(
	    "movq	%0, %%mm7\n\t"
	    "movq	%1, %%mm6\n\t"
	    ::"m"(red_16mask),"m"(green_16mask));
	mm_end = end - 15;
	while(s < mm_end)
	{
	    __asm __volatile(
		PREFETCH" 32%1\n\t"
		"movd	%1, %%mm0\n\t"
		"movd	3%1, %%mm3\n\t"
		"punpckldq 6%1, %%mm0\n\t"
		"punpckldq 9%1, %%mm3\n\t"
		"movq	%%mm0, %%mm1\n\t"
		"movq	%%mm0, %%mm2\n\t"
		"movq	%%mm3, %%mm4\n\t"
		"movq	%%mm3, %%mm5\n\t"
		"psllq	$8, %%mm0\n\t"
		"psllq	$8, %%mm3\n\t"
		"pand	%%mm7, %%mm0\n\t"
		"pand	%%mm7, %%mm3\n\t"
		"psrlq	$5, %%mm1\n\t"
		"psrlq	$5, %%mm4\n\t"
		"pand	%%mm6, %%mm1\n\t"
		"pand	%%mm6, %%mm4\n\t"
		"psrlq	$19, %%mm2\n\t"
		"psrlq	$19, %%mm5\n\t"
		"pand	%2, %%mm2\n\t"
		"pand	%2, %%mm5\n\t"
		"por	%%mm1, %%mm0\n\t"
		"por	%%mm4, %%mm3\n\t"
		"por	%%mm2, %%mm0\n\t"
		"por	%%mm5, %%mm3\n\t"
		"psllq	$16, %%mm3\n\t"
		"por	%%mm3, %%mm0\n\t"
		MOVNTQ"	%%mm0, %0\n\t"
		:"=m"(*d):"m"(*s),"m"(blue_16mask):"memory");
		d += 4;
		s += 12;
	}
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS:::"memory");
#endif
	while(s < end)
	{
		const int r= *s++;
		const int g= *s++;
		const int b= *s++;
		*d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8);
	}
}

static inline void RENAME(rgb24to15)(const uint8_t *src, uint8_t *dst, unsigned src_size)
{
	const uint8_t *s = src;
	const uint8_t *end;
#ifdef HAVE_MMX
	const uint8_t *mm_end;
#endif
	uint16_t *d = (uint16_t *)dst;
	end = s + src_size;
#ifdef HAVE_MMX
	__asm __volatile(PREFETCH"	%0"::"m"(*src):"memory");
	__asm __volatile(
	    "movq	%0, %%mm7\n\t"
	    "movq	%1, %%mm6\n\t"
	    ::"m"(red_15mask),"m"(green_15mask));
	mm_end = end - 11;
	while(s < mm_end)
	{
	    __asm __volatile(
		PREFETCH" 32%1\n\t"
		"movd	%1, %%mm0\n\t"
		"movd	3%1, %%mm3\n\t"
		"punpckldq 6%1, %%mm0\n\t"
		"punpckldq 9%1, %%mm3\n\t"
		"movq	%%mm0, %%mm1\n\t"
		"movq	%%mm0, %%mm2\n\t"
		"movq	%%mm3, %%mm4\n\t"
		"movq	%%mm3, %%mm5\n\t"
		"psrlq	$3, %%mm0\n\t"
		"psrlq	$3, %%mm3\n\t"
		"pand	%2, %%mm0\n\t"
		"pand	%2, %%mm3\n\t"
		"psrlq	$6, %%mm1\n\t"
		"psrlq	$6, %%mm4\n\t"
		"pand	%%mm6, %%mm1\n\t"
		"pand	%%mm6, %%mm4\n\t"
		"psrlq	$9, %%mm2\n\t"
		"psrlq	$9, %%mm5\n\t"
		"pand	%%mm7, %%mm2\n\t"
		"pand	%%mm7, %%mm5\n\t"
		"por	%%mm1, %%mm0\n\t"
		"por	%%mm4, %%mm3\n\t"
		"por	%%mm2, %%mm0\n\t"
		"por	%%mm5, %%mm3\n\t"
		"psllq	$16, %%mm3\n\t"
		"por	%%mm3, %%mm0\n\t"
		MOVNTQ"	%%mm0, %0\n\t"
		:"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");
		d += 4;
		s += 12;
	}
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS:::"memory");
#endif
	while(s < end)
	{
		const int b= *s++;
		const int g= *s++;
		const int r= *s++;
		*d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
	}
}

static inline void RENAME(rgb24tobgr15)(const uint8_t *src, uint8_t *dst, unsigned src_size)
{
	const uint8_t *s = src;
	const uint8_t *end;
#ifdef HAVE_MMX
	const uint8_t *mm_end;
#endif
	uint16_t *d = (uint16_t *)dst;
	end = s + src_size;
#ifdef HAVE_MMX
	__asm __volatile(PREFETCH"	%0"::"m"(*src):"memory");
	__asm __volatile(
	    "movq	%0, %%mm7\n\t"
	    "movq	%1, %%mm6\n\t"
	    ::"m"(red_15mask),"m"(green_15mask));
	mm_end = end - 15;
	while(s < mm_end)
	{
	    __asm __volatile(
		PREFETCH" 32%1\n\t"
		"movd	%1, %%mm0\n\t"
		"movd	3%1, %%mm3\n\t"
		"punpckldq 6%1, %%mm0\n\t"
		"punpckldq 9%1, %%mm3\n\t"
		"movq	%%mm0, %%mm1\n\t"
		"movq	%%mm0, %%mm2\n\t"
		"movq	%%mm3, %%mm4\n\t"
		"movq	%%mm3, %%mm5\n\t"
		"psllq	$7, %%mm0\n\t"
		"psllq	$7, %%mm3\n\t"
		"pand	%%mm7, %%mm0\n\t"
		"pand	%%mm7, %%mm3\n\t"
		"psrlq	$6, %%mm1\n\t"
		"psrlq	$6, %%mm4\n\t"
		"pand	%%mm6, %%mm1\n\t"
		"pand	%%mm6, %%mm4\n\t"
		"psrlq	$19, %%mm2\n\t"
		"psrlq	$19, %%mm5\n\t"
		"pand	%2, %%mm2\n\t"
		"pand	%2, %%mm5\n\t"
		"por	%%mm1, %%mm0\n\t"
		"por	%%mm4, %%mm3\n\t"
		"por	%%mm2, %%mm0\n\t"
		"por	%%mm5, %%mm3\n\t"
		"psllq	$16, %%mm3\n\t"
		"por	%%mm3, %%mm0\n\t"
		MOVNTQ"	%%mm0, %0\n\t"
		:"=m"(*d):"m"(*s),"m"(blue_15mask):"memory");
		d += 4;
		s += 12;
	}
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS:::"memory");
#endif
	while(s < end)
	{
		const int r= *s++;
		const int g= *s++;
		const int b= *s++;
		*d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
	}
}

/*
  I use here less accurate approximation by simply
 left-shifting the input
  value and filling the low order bits with
 zeroes. This method improves png's
  compression but this scheme cannot reproduce white exactly, since it does not
  generate an all-ones maximum value; the net effect is to darken the
  image slightly.

  The better method should be "left bit replication":

   4 3 2 1 0
   ---------
   1 1 0 1 1

   7 6 5 4 3  2 1 0
   ----------------
   1 1 0 1 1  1 1 0
   |=======|  |===|
       |      Leftmost Bits Repeated to Fill Open Bits
       |
   Original Bits
*/
static inline void RENAME(rgb15to24)(const uint8_t *src, uint8_t *dst, unsigned src_size)
{
	const uint16_t *end;
#ifdef HAVE_MMX
	const uint16_t *mm_end;
#endif
	uint8_t *d = (uint8_t *)dst;
	const uint16_t *s = (uint16_t *)src;
	end = s + src_size/2;
#ifdef HAVE_MMX
	__asm __volatile(PREFETCH"	%0"::"m"(*s):"memory");
	mm_end = end - 7;
	while(s < mm_end)
	{
	    __asm __volatile(
		PREFETCH" 32%1\n\t"
		"movq	%1, %%mm0\n\t"
		"movq	%1, %%mm1\n\t"
		"movq	%1, %%mm2\n\t"
		"pand	%2, %%mm0\n\t"
		"pand	%3, %%mm1\n\t"
		"pand	%4, %%mm2\n\t"
		"psllq	$3, %%mm0\n\t"
		"psrlq	$2, %%mm1\n\t"
		"psrlq	$7, %%mm2\n\t"
		"movq	%%mm0, %%mm3\n\t"
		"movq	%%mm1, %%mm4\n\t"
		"movq	%%mm2, %%mm5\n\t"
		"punpcklwd %5, %%mm0\n\t"
		"punpcklwd %5, %%mm1\n\t"
		"punpcklwd %5, %%mm2\n\t"
		"punpckhwd %5, %%mm3\n\t"
		"punpckhwd %5, %%mm4\n\t"
		"punpckhwd %5, %%mm5\n\t"
		"psllq	$8, %%mm1\n\t"
		"psllq	$16, %%mm2\n\t"
		"por	%%mm1, %%mm0\n\t"
		"por	%%mm2, %%mm0\n\t"
		"psllq	$8, %%mm4\n\t"
		"psllq	$16, %%mm5\n\t"
		"por	%%mm4, %%mm3\n\t"
		"por	%%mm5, %%mm3\n\t"

		"movq	%%mm0, %%mm6\n\t"
		"movq	%%mm3, %%mm7\n\t"
		
		"movq	8%1, %%mm0\n\t"
		"movq	8%1, %%mm1\n\t"
		"movq	8%1, %%mm2\n\t"
		"pand	%2, %%mm0\n\t"
		"pand	%3, %%mm1\n\t"
		"pand	%4, %%mm2\n\t"
		"psllq	$3, %%mm0\n\t"
		"psrlq	$2, %%mm1\n\t"
		"psrlq	$7, %%mm2\n\t"
		"movq	%%mm0, %%mm3\n\t"
		"movq	%%mm1, %%mm4\n\t"
		"movq	%%mm2, %%mm5\n\t"
		"punpcklwd %5, %%mm0\n\t"
		"punpcklwd %5, %%mm1\n\t"
		"punpcklwd %5, %%mm2\n\t"
		"punpckhwd %5, %%mm3\n\t"
		"punpckhwd %5, %%mm4\n\t"
		"punpckhwd %5, %%mm5\n\t"
		"psllq	$8, %%mm1\n\t"
		"psllq	$16, %%mm2\n\t"
		"por	%%mm1, %%mm0\n\t"
		"por	%%mm2, %%mm0\n\t"
		"psllq	$8, %%mm4\n\t"
		"psllq	$16, %%mm5\n\t"
		"por	%%mm4, %%mm3\n\t"
		"por	%%mm5, %%mm3\n\t"

		:"=m"(*d)
		:"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r), "m"(mmx_null)
		:"memory");
	    /* Borrowed 32 to 24 */
	    __asm __volatile(
		"movq	%%mm0, %%mm4\n\t"
		"movq	%%mm3, %%mm5\n\t"
		"movq	%%mm6, %%mm0\n\t"
		"movq	%%mm7, %%mm1\n\t"
		
		"movq	%%mm4, %%mm6\n\t"
		"movq	%%mm5, %%mm7\n\t"
		"movq	%%mm0, %%mm2\n\t"
		"movq	%%mm1, %%mm3\n\t"

		"psrlq	$8, %%mm2\n\t"
		"psrlq	$8, %%mm3\n\t"
		"psrlq	$8, %%mm6\n\t"
		"psrlq	$8, %%mm7\n\t"
		"pand	%2, %%mm0\n\t"
		"pand	%2, %%mm1\n\t"
		"pand	%2, %%mm4\n\t"
		"pand	%2, %%mm5\n\t"
		"pand	%3, %%mm2\n\t"
		"pand	%3, %%mm3\n\t"
		"pand	%3, %%mm6\n\t"
		"pand	%3, %%mm7\n\t"
		"por	%%mm2, %%mm0\n\t"
		"por	%%mm3, %%mm1\n\t"
		"por	%%mm6, %%mm4\n\t"
		"por	%%mm7, %%mm5\n\t"

		"movq	%%mm1, %%mm2\n\t"
		"movq	%%mm4, %%mm3\n\t"
		"psllq	$48, %%mm2\n\t"
		"psllq	$32, %%mm3\n\t"
		"pand	%4, %%mm2\n\t"
		"pand	%5, %%mm3\n\t"
		"por	%%mm2, %%mm0\n\t"
		"psrlq	$16, %%mm1\n\t"
		"psrlq	$32, %%mm4\n\t"
		"psllq	$16, %%mm5\n\t"
		"por	%%mm3, %%mm1\n\t"
		"pand	%6, %%mm5\n\t"
		"por	%%mm5, %%mm4\n\t"

		MOVNTQ"	%%mm0, %0\n\t"
		MOVNTQ"	%%mm1, 8%0\n\t"
		MOVNTQ"	%%mm4, 16%0"

		:"=m"(*d)
		:"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh)
		:"memory");
		d += 24;
		s += 8;
	}
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS:::"memory");
#endif
	while(s < end)
	{
		register uint16_t bgr;
		bgr = *s++;
		*d++ = (bgr&0x1F)<<3;