video_out_rgb.c

最新osg包

			::: "memory");

		while(n--)
		{
			__asm__ __volatile__(

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"punpcklbw %%mm1, %%mm2\n\t" /* mm1 = [g3 b3 g2 b2 g1 b1 g0 b0] */
				"punpcklbw %%mm3, %%mm0\n\t" /* mm3 = [0 r3 0 r2 0 r1 0 r0] */

				"movq %%mm2, %%mm4\n\t" /* mm4 = [g3 b3 g2 b2 g1 b1 g0 b0] */
				"punpcklwd %%mm0, %%mm4\n\t" /* mm4 = [0 r1 g1 b1 0 r0 g0 b0] */
				"movd %%mm4, (%3)\n\t"
				"punpckhdq %%mm4, %%mm4\n\t" /* mm4 = [0 r1 g1 b1 0 r1 g1 b1] */
				"movd %%mm4, 3(%3)\n\t"

				"punpckhwd %%mm0, %%mm2\n\t" /* mm2 = [0 r3 g3 b3 0 r2 g2 b2] */
				"movd %%mm2, 6(%3)\n\t"
				"punpckhdq %%mm2, %%mm2\n\t" /* mm2 = [0 r3 g3 b3 0 r3 g3 b3] */
				"movd %%mm2, 9(%3)\n\t"

				"addl $4, %0\n\t"
				"addl $4, %1\n\t"
				"addl $4, %2\n\t"
				"addl $12, %3\n\t"

				: "=&r" (r_data), "=&r" (g_data), "=&r" (b_data), "=&r" (buffer)
				: "0" (r_data), "1" (g_data), "2" (b_data), "3" (buffer)
				: "memory");

		}

		__asm__ __volatile__("emms\n\t");

	}
#endif

	while(rest--)
	{
		*(buffer)     = *(b_data++);
		*(buffer + 1) = *(g_data++);
		*(buffer + 2) = *(r_data++);

		buffer += 3;
	}

}


static void
__pack_rgb16(rgb_planar_t* data, void* dest,
		uint32_t pixels, uint32_t accel)
{
	uint16_t* buffer = (uint16_t*) dest;
	uint8_t* r_data = data->r;
	uint8_t* g_data = data->g;
	uint8_t* b_data = data->b;
	uint32_t rest = pixels;

#ifdef COMPILE_ASSEMBLY

	if((accel & MM_MMXEXT) == MM_MMXEXT)
	{
		uint32_t n = pixels >> 2; /* (width * height) / 4 */
		rest = pixels & 3; /* pixels - (n * 4) */


		__asm__ __volatile__(
			"pxor %%mm3, %%mm3\n\t"
			::: "memory");

		while(n--)
		{
			__asm__ __volatile__(

				"prefetchw 320(%3)\n\t"

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"punpcklbw %%mm3, %%mm0\n\t" /* mm0 = [0 r3 0 r2 0 r1 0 r0] */
				"punpcklbw %%mm3, %%mm1\n\t" /* mm1 = [0 g3 0 g2 0 g1 0 g0] */
				"punpcklbw %%mm3, %%mm2\n\t" /* mm2 = [0 b3 0 b2 0 b1 0 b0] */

				"psrlw $3, %%mm0\n\t" /* word = [00000000 000rrrrr] */
				"psrlw $2, %%mm1\n\t" /* word = [00000000 00gggggg] */
				"psrlw $3, %%mm2\n\t" /* word = [00000000 000bbbbb] */

				"psllw $11, %%mm0\n\t" /* word = [rrrrr000 00000000] */
				"psllw $5,  %%mm1\n\t" /* word = [00000ggg ggg00000] */
				"por %%mm2, %%mm1\n\t" /* word = [00000ggg gggbbbbb] */
				"por %%mm1, %%mm0\n\t" /* word = [rrrrrggg gggbbbbb] */

				"movntq %%mm0, (%3)\n\t"

				"addl $4, %0\n\t"
				"addl $4, %1\n\t"
				"addl $4, %2\n\t"
				"addl $8, %3\n\t"

				: "=&r" (r_data), "=&r" (g_data), "=&r" (b_data), "=&r" (buffer)
				: "0" (r_data), "1" (g_data), "2" (b_data), "3" (buffer)
				: "memory");
			}

		__asm__ __volatile__("sfence; emms\n\t");

	}
	else if((accel & MM_MMX) == MM_MMX)
	{
		uint32_t n = pixels >> 2; /* (width * height) / 4 */
		rest = pixels & 3; /* pixels - (n * 4) */


		__asm__ __volatile__(
			"pxor %%mm3, %%mm3\n\t"
			::: "memory");

		while(n--)
		{
			__asm__ __volatile__(

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"punpcklbw %%mm3, %%mm0\n\t" /* mm0 = [0 r3 0 r2 0 r1 0 r0] */
				"punpcklbw %%mm3, %%mm1\n\t" /* mm1 = [0 g3 0 g2 0 g1 0 g0] */
				"punpcklbw %%mm3, %%mm2\n\t" /* mm2 = [0 b3 0 b2 0 b1 0 b0] */

				"psrlw $3, %%mm0\n\t" /* word = [00000000 000rrrrr] */
				"psrlw $2, %%mm1\n\t" /* word = [00000000 00gggggg] */
				"psrlw $3, %%mm2\n\t" /* word = [00000000 000bbbbb] */

				"psllw $11, %%mm0\n\t" /* word = [rrrrr000 00000000] */
				"psllw $5,  %%mm1\n\t" /* word = [00000ggg ggg00000] */
				"por %%mm2, %%mm1\n\t" /* word = [00000ggg gggbbbbb] */
				"por %%mm1, %%mm0\n\t" /* word = [rrrrrggg gggbbbbb] */

				"movq %%mm0, (%3)\n\t"

				"addl $4, %0\n\t"
				"addl $4, %1\n\t"
				"addl $4, %2\n\t"
				"addl $8, %3\n\t"

				: "=&r" (r_data), "=&r" (g_data), "=&r" (b_data), "=&r" (buffer)
				: "0" (r_data), "1" (g_data), "2" (b_data), "3" (buffer)
				: "memory");
			}

		__asm__ __volatile__("emms\n\t");

	}
#endif

	while(rest--)
	{
		uint16_t r16, g16, b16;

		r16 = *(r_data++) >> 3;
		g16 = *(g_data++) >> 2;
		b16 = *(b_data++) >> 3;

		*(buffer++) = ((r16 << 11) | (g16 << 5) | b16);
	}

}


static void
__pack_bgra(rgb_planar_t* data, void* dest,
		 uint32_t pixels, uint32_t accel)
{
	static const uint32_t alpha[] =
	{
		0xffffffff,
		0xffffffff
	};
	uint8_t* buffer = (uint8_t*) dest;
	uint8_t* r_data = data->r;
	uint8_t* g_data = data->g;
	uint8_t* b_data = data->b;
	uint32_t rest = pixels;


#ifdef COMPILE_ASSEMBLY
	if((accel & MM_MMXEXT) == MM_MMXEXT)
	{
		uint32_t n = pixels >> 2; /* (width * height) / 4 */
		rest = pixels & 3; /* pixels - (n * 4) */


		while(n--)
		{
			__asm__ __volatile__(

				"prefetchw 320(%3)\n\t"

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"por (%4), %%mm3\n\t" /* mm3 = [ff ff ff ff ff ff ff ff] */
				"punpcklbw %%mm0, %%mm3\n\t" /* mm3 = [r3 ff r2 ff r1 ff r0 ff] */
				"punpcklbw %%mm2, %%mm1\n\t" /* mm1 = [b3 g3 b2 g2 b1 g1 b0 g0] */

				"movq %%mm3, %%mm4\n\t" /* mm4 = [r3 ff r2 ff r1 ff r0 ff] */
				"punpcklwd %%mm1, %%mm4\n\t" /* mm4 = [b1 g1 r1 ff b1 g1 r1 ff] */
				"movntq %%mm4, (%3)\n\t"

				"punpckhwd %%mm1, %%mm3\n\t" /* mm3 = [b3 g3 r3 ff b2 g2 r2 ff] */
				"movntq %%mm0, 8(%3)\n\t"

				"addl $4, %0\n\t"
				"addl $4, %1\n\t"
				"addl $4, %2\n\t"
				"addl $16, %3\n\t"

				: "=&r" (r_data), "=&r" (g_data), "=&r" (b_data), "=&r" (buffer)
				: "r" (alpha), "0" (r_data), "1" (g_data), "2" (b_data), "3" (buffer)
				: "memory");
		}

		__asm__ __volatile__("sfence; emms\n\t");

	}
	else if((accel & MM_MMX) == MM_MMX)
	{
		uint32_t n = pixels >> 2; /* (width * height) / 4 */
		rest = pixels & 3; /* pixels - (n * 4) */


		while(n--)
		{
			__asm__ __volatile__(

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"por (%4), %%mm3\n\t" /* mm3 = [ff ff ff ff ff ff ff ff] */
				"punpcklbw %%mm0, %%mm3\n\t" /* mm3 = [r3 ff r2 ff r1 ff r0 ff] */
				"punpcklbw %%mm2, %%mm1\n\t" /* mm1 = [b3 g3 b2 g2 b1 g1 b0 g0] */

				"movq %%mm3, %%mm4\n\t" /* mm4 = [r3 ff r2 ff r1 ff r0 ff] */
				"punpcklwd %%mm1, %%mm4\n\t" /* mm4 = [b1 g1 r1 ff b1 g1 r1 ff] */
				"movq %%mm4, (%3)\n\t"

				"punpckhwd %%mm1, %%mm3\n\t" /* mm3 = [b3 g3 r3 ff b2 g2 r2 ff] */
				"movq %%mm0, 8(%3)\n\t"

				"addl $4, %0\n\t"
				"addl $4, %1\n\t"
				"addl $4, %2\n\t"
				"addl $16, %3\n\t"

				: "=&r" (r_data), "=&r" (g_data), "=&r" (b_data), "=&r" (buffer)
				: "r" (alpha), "0" (r_data), "1" (g_data), "2" (b_data), "3" (buffer)
				: "memory");
		}

		__asm__ __volatile__("emms\n\t");

	}

#endif
	while(rest--)
	{
		*(buffer)     = 0xff;
		*(buffer + 1) = *(r_data++);
		*(buffer + 2) = *(g_data++);
		*(buffer + 3) = *(b_data++);

		buffer += 4;
	}

}


static void
__pack_bgra5551(rgb_planar_t* data, void* dest,
		 uint32_t pixels, uint32_t accel)
{
	static const uint16_t alpha[] =
	{
		1, 1, 1, 1
	};
	uint16_t* buffer = (uint16_t*) dest;
	uint8_t* r_data = data->r;
	uint8_t* g_data = data->g;
	uint8_t* b_data = data->b;
	uint32_t rest = pixels;


#ifdef COMPILE_ASSEMBLY
	if((accel & MM_MMXEXT) == MM_MMXEXT)
	{
		uint32_t n = pixels >> 2; /* (width * height) / 4 */
		rest = pixels & 3; /* pixels - (n * 4) */


		__asm__ __volatile__(
			"pxor %%mm3, %%mm3\n\t"
			::: "memory");

		while(n--)
		{
			__asm__ __volatile__(

				"prefetchw 320(%3)\n\t"

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"punpcklbw %%mm3, %%mm0\n\t" /* mm0 = [0 r3 0 r2 0 r1 0 r0] */
				"punpcklbw %%mm3, %%mm1\n\t" /* mm1 = [0 g3 0 g2 0 g1 0 g0] */
				"punpcklbw %%mm3, %%mm2\n\t" /* mm2 = [0 b3 0 b2 0 b1 0 b0] */

				"psrlw $2, %%mm0\n\t" /* word = [00000000 00rrrrrr] */
				"psrlw $3, %%mm1\n\t" /* word = [00000000 000ggggg] */
				"psrlw $3, %%mm2\n\t" /* word = [00000000 000bbbbb] */

				"psllw $11, %%mm2\n\t" /* word = [bbbbb000 00000000] */
				"psllw $6, %%mm1\n\t"  /* word = [00000ggg gg000000] */
				"por (%4), %%mm0\n\t"  /* word = [00000000 00rrrrra] */
				"por %%mm2, %%mm1\n\t" /* word = [bbbbbggg gg000000] */
				"por %%mm1, %%mm0\n\t" /* word = [bbbbbggg ggrrrrra] */

				"movntq %%mm0, (%3)\n\t"

				"addl $4, %0\n\t"
				"addl $4, %1\n\t"
				"addl $4, %2\n\t"
				"addl $8, %3\n\t"

				: "=&r" (r_data), "=&r" (g_data), "=&r" (b_data), "=&r" (buffer)
				: "r" (alpha), "0" (r_data), "1" (g_data), "2" (b_data), "3" (buffer)
				: "memory");
		}

		__asm__ __volatile__("sfence; emms\n\t");

	}
	else if((accel & MM_MMX) == MM_MMX)
	{
		uint32_t n = pixels >> 2; /* (width * height) / 4 */
		rest = pixels & 3; /* pixels - (n * 4) */


		__asm__ __volatile__(
			"pxor %%mm3, %%mm3\n\t"
			::: "memory");

		while(n--)
		{
			__asm__ __volatile__(

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"punpcklbw %%mm3, %%mm0\n\t" /* mm0 = [0 r3 0 r2 0 r1 0 r0] */
				"punpcklbw %%mm3, %%mm1\n\t" /* mm1 = [0 g3 0 g2 0 g1 0 g0] */
				"punpcklbw %%mm3, %%mm2\n\t" /* mm2 = [0 b3 0 b2 0 b1 0 b0] */

				"psrlw $2, %%mm0\n\t" /* word = [00000000 00rrrrrr] */
				"psrlw $3, %%mm1\n\t" /* word = [00000000 000ggggg] */
				"psrlw $3, %%mm2\n\t" /* word = [00000000 000bbbbb] */

				"psllw $11, %%mm2\n\t" /* word = [bbbbb000 00000000] */
				"psllw $6, %%mm1\n\t"  /* word = [00000ggg gg000000] */
				"por (%4), %%mm0\n\t"  /* word = [00000000 00rrrrra] */
				"por %%mm2, %%mm1\n\t" /* word = [bbbbbggg gg000000] */
				"por %%mm1, %%mm0\n\t" /* word = [bbbbbggg ggrrrrra] */

				"movq %%mm0, (%3)\n\t"

				"addl $4, %0\n\t"
				"addl $4, %1\n\t"
				"addl $4, %2\n\t"
				"addl $8, %3\n\t"

				: "=&r" (r_data), "=&r" (g_data), "=&r" (b_data), "=&r" (buffer)
				: "r" (alpha), "0" (r_data), "1" (g_data), "2" (b_data), "3" (buffer)
				: "memory");
		}

		__asm__ __volatile__("emms\n\t");

	}
#endif

	while(rest--)
	{
		uint16_t r5, g5, b5;

		r5 = *(r_data++) >> 2;
		g5 = *(g_data++) >> 3;
		b5 = *(b_data++) >> 3;

		*(buffer++) = ((b5 << 11) | (g5 << 6) | r5 | 1);
	}

}


static void
__pack_bgr32(rgb_planar_t* data, void* dest,
		 uint32_t pixels, uint32_t accel)
{
	uint8_t* buffer = (uint8_t*) dest;
	uint8_t* r_data = data->r;
	uint8_t* g_data = data->g;
	uint8_t* b_data = data->b;
	uint32_t rest = pixels;

#ifdef COMPILE_ASSEMBLY

	if((accel & MM_MMXEXT) == MM_MMXEXT)
	{
		uint32_t n = pixels >> 2; /* (width * height) / 4 */
		rest = pixels & 3; /* pixels - (n * 4) */


		__asm__ __volatile__(
			"pxor %%mm3, %%mm3\n\t"
			::: "memory");

		while(n--)
		{
			__asm__ __volatile__(

				"prefetchw 320(%0)\n\t"

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"punpcklbw %%mm1, %%mm0\n\t" /* mm0 = [g3 r3 g2 r2 g1 r1 g0 r0] */
				"punpcklbw %%mm3, %%mm2\n\t" /* mm2 = [b3 0 b2 0 b1 0 b0 0] */

				"movq %%mm0, %%mm4\n\t" /* mm4 = [g3 r3 g2 r2 g1 r1 g0 r0] */
				"punpcklwd %%mm2, %%mm4\n\t" /* mm4 = [0 b1 g1 r1 0 b0 g0 r0] */
				"movntq %%mm4, (%3)\n\t"

				"punpckhwd %%mm2, %%mm0\n\t" /* mm0 = [0 b3 g3 r3 0 b2 g2 r2] */
				"movntq %%mm0, 8(%3)\n\t"

				"addl $4, %0\n\t"
				"addl $4, %1\n\t"
				"addl $4, %2\n\t"
				"addl $16, %3\n\t"

				: "=&r" (r_data), "=&r" (g_data), "=&r" (b_data), "=&r" (buffer)
				: "0" (r_data), "1" (g_data), "2" (b_data), "3" (buffer)
				: "memory");
		}

		__asm__ __volatile__("sfence; emms\n\t");

	}
	else if((accel & MM_MMX) == MM_MMX)
	{
		uint32_t n = pixels >> 2; /* (width * height) / 4 */
		rest = pixels & 3; /* pixels - (n * 4) */


		__asm__ __volatile__(
			"pxor %%mm3, %%mm3\n\t"
			::: "memory");

		while(n--)
		{
			__asm__ __volatile__(

				"movd (%0), %%mm0\n\t" /* mm0 = [0 0 0 0 r3 r2 r1 r0] */
				"movd (%1), %%mm1\n\t" /* mm1 = [0 0 0 0 g3 g2 g1 g0] */
				"movd (%2), %%mm2\n\t" /* mm2 = [0 0 0 0 b3 b2 b1 b0] */

				"punpcklbw %%mm1, %%mm0\n\t" /* mm0 = [g3 r3 g2 r2 g1 r1 g0 r0] */
				"punpcklbw %%mm3, %%mm2\n\t" /* mm2 = [b3 0 b2 0 b1 0 b0 0] */

				"movq %%mm0, %%mm4\n\t" /* mm4 = [g3 r3 g2 r2 g1 r1 g0 r0] */
				"punpcklwd %%mm2, %%mm4\n\t" /* mm4 = [0 b1 g1 r1 0 b0 g0 r0] */
				"movq %%mm4, (%3)\n\t"