yuy2toyv12 in MMX

Originally committed as revision 2705 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc

postproc/rgb2rgb.c

@@ -317,6 +317,60 @@ void yv12toyuy2(uint8_t *ysrc, uint8_t *usrc, uint8_t *vsrc, uint8_t *dst, int n
 
 void yuy2toyv12(uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, int num_pixels)
 {
+#ifdef HAVE_MMX
+	asm volatile(
+		"xorl %%eax, %%eax		\n\t"
+		"pcmpeqw %%mm7, %%mm7		\n\t"
+		"psrlw $8, %%mm7		\n\t" // FF,00,FF,00...
+		"1:				\n\t"
+		PREFETCH" 64(%0, %%eax, 4)	\n\t"
+		"movq (%0, %%eax, 4), %%mm0	\n\t" // YUYV YUYV(0)
+		"movq 8(%0, %%eax, 4), %%mm1	\n\t" // YUYV YUYV(4)
+		"movq %%mm0, %%mm2		\n\t" // YUYV YUYV(0)
+		"movq %%mm1, %%mm3		\n\t" // YUYV YUYV(4)
+		"psrlw $8, %%mm0		\n\t" // U0V0 U0V0(0)
+		"psrlw $8, %%mm1		\n\t" // U0V0 U0V0(4)
+		"pand %%mm7, %%mm2		\n\t" // Y0Y0 Y0Y0(0)
+		"pand %%mm7, %%mm3		\n\t" // Y0Y0 Y0Y0(4)
+		"packuswb %%mm1, %%mm0		\n\t" // UVUV UVUV(0)
+		"packuswb %%mm3, %%mm2		\n\t" // YYYY YYYY(0)
+
+		MOVNTQ" %%mm2, (%1, %%eax, 2)	\n\t"
+
+		"movq 16(%0, %%eax, 4), %%mm1	\n\t" // YUYV YUYV(8)
+		"movq 24(%0, %%eax, 4), %%mm2	\n\t" // YUYV YUYV(12)
+		"movq %%mm1, %%mm3		\n\t" // YUYV YUYV(8)
+		"movq %%mm2, %%mm4		\n\t" // YUYV YUYV(12)
+		"psrlw $8, %%mm1		\n\t" // U0V0 U0V0(8)
+		"psrlw $8, %%mm2		\n\t" // U0V0 U0V0(12)
+		"pand %%mm7, %%mm3		\n\t" // Y0Y0 Y0Y0(8)
+		"pand %%mm7, %%mm4		\n\t" // Y0Y0 Y0Y0(12)
+		"packuswb %%mm2, %%mm1		\n\t" // UVUV UVUV(8)
+		"packuswb %%mm4, %%mm3		\n\t" // YYYY YYYY(8)
+
+		MOVNTQ" %%mm3, 8(%1, %%eax, 2)	\n\t"
+
+		"movq %%mm0, %%mm2		\n\t" // UVUV UVUV(0)
+		"movq %%mm1, %%mm3		\n\t" // UVUV UVUV(8)
+		"psrlw $8, %%mm0		\n\t" // V0V0 V0V0(0)
+		"psrlw $8, %%mm1		\n\t" // V0V0 V0V0(8)
+		"pand %%mm7, %%mm2		\n\t" // U0U0 U0U0(0)
+		"pand %%mm7, %%mm3		\n\t" // U0U0 U0U0(8)
+		"packuswb %%mm1, %%mm0		\n\t" // VVVV VVVV(0)
+		"packuswb %%mm3, %%mm2		\n\t" // UUUU UUUU(0)
+
+		MOVNTQ" %%mm0, (%3, %%eax)	\n\t"
+		MOVNTQ" %%mm2, (%2, %%eax)	\n\t"
+
+		"addl $8, %%eax			\n\t"
+		"cmpl %4, %%eax			\n\t"
+		" jb 1b				\n\t"
+		EMMS" \n\t"
+		SFENCE
+		::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "r" (num_pixels>>1)
+		: "memory", "%eax"
+	);
+#else
 	int i;
 	num_pixels>>=1;
 	for(i=0; i<num_pixels; i++)
@@ -326,4 +380,5 @@ void yuy2toyv12(uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, int n
 		 ydst[2*i+1] 	= src[4*i+2];
 		 vdst[i] 	= src[4*i+3];
 	}
+#endif
 }

postproc/rgb2rgb_template.c

@@ -317,6 +317,60 @@ void yv12toyuy2(uint8_t *ysrc, uint8_t *usrc, uint8_t *vsrc, uint8_t *dst, int n
 
 void yuy2toyv12(uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, int num_pixels)
 {
+#ifdef HAVE_MMX
+	asm volatile(
+		"xorl %%eax, %%eax		\n\t"
+		"pcmpeqw %%mm7, %%mm7		\n\t"
+		"psrlw $8, %%mm7		\n\t" // FF,00,FF,00...
+		"1:				\n\t"
+		PREFETCH" 64(%0, %%eax, 4)	\n\t"
+		"movq (%0, %%eax, 4), %%mm0	\n\t" // YUYV YUYV(0)
+		"movq 8(%0, %%eax, 4), %%mm1	\n\t" // YUYV YUYV(4)
+		"movq %%mm0, %%mm2		\n\t" // YUYV YUYV(0)
+		"movq %%mm1, %%mm3		\n\t" // YUYV YUYV(4)
+		"psrlw $8, %%mm0		\n\t" // U0V0 U0V0(0)
+		"psrlw $8, %%mm1		\n\t" // U0V0 U0V0(4)
+		"pand %%mm7, %%mm2		\n\t" // Y0Y0 Y0Y0(0)
+		"pand %%mm7, %%mm3		\n\t" // Y0Y0 Y0Y0(4)
+		"packuswb %%mm1, %%mm0		\n\t" // UVUV UVUV(0)
+		"packuswb %%mm3, %%mm2		\n\t" // YYYY YYYY(0)
+
+		MOVNTQ" %%mm2, (%1, %%eax, 2)	\n\t"
+
+		"movq 16(%0, %%eax, 4), %%mm1	\n\t" // YUYV YUYV(8)
+		"movq 24(%0, %%eax, 4), %%mm2	\n\t" // YUYV YUYV(12)
+		"movq %%mm1, %%mm3		\n\t" // YUYV YUYV(8)
+		"movq %%mm2, %%mm4		\n\t" // YUYV YUYV(12)
+		"psrlw $8, %%mm1		\n\t" // U0V0 U0V0(8)
+		"psrlw $8, %%mm2		\n\t" // U0V0 U0V0(12)
+		"pand %%mm7, %%mm3		\n\t" // Y0Y0 Y0Y0(8)
+		"pand %%mm7, %%mm4		\n\t" // Y0Y0 Y0Y0(12)
+		"packuswb %%mm2, %%mm1		\n\t" // UVUV UVUV(8)
+		"packuswb %%mm4, %%mm3		\n\t" // YYYY YYYY(8)
+
+		MOVNTQ" %%mm3, 8(%1, %%eax, 2)	\n\t"
+
+		"movq %%mm0, %%mm2		\n\t" // UVUV UVUV(0)
+		"movq %%mm1, %%mm3		\n\t" // UVUV UVUV(8)
+		"psrlw $8, %%mm0		\n\t" // V0V0 V0V0(0)
+		"psrlw $8, %%mm1		\n\t" // V0V0 V0V0(8)
+		"pand %%mm7, %%mm2		\n\t" // U0U0 U0U0(0)
+		"pand %%mm7, %%mm3		\n\t" // U0U0 U0U0(8)
+		"packuswb %%mm1, %%mm0		\n\t" // VVVV VVVV(0)
+		"packuswb %%mm3, %%mm2		\n\t" // UUUU UUUU(0)
+
+		MOVNTQ" %%mm0, (%3, %%eax)	\n\t"
+		MOVNTQ" %%mm2, (%2, %%eax)	\n\t"
+
+		"addl $8, %%eax			\n\t"
+		"cmpl %4, %%eax			\n\t"
+		" jb 1b				\n\t"
+		EMMS" \n\t"
+		SFENCE
+		::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "r" (num_pixels>>1)
+		: "memory", "%eax"
+	);
+#else
 	int i;
 	num_pixels>>=1;
 	for(i=0; i<num_pixels; i++)
@@ -326,4 +380,5 @@ void yuy2toyv12(uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, int n
 		 ydst[2*i+1] 	= src[4*i+2];
 		 vdst[i] 	= src[4*i+3];
 	}
+#endif
 }