From dabcdbc493b02d3b477cb056beaf0fb7a1879c1e Mon Sep 17 00:00:00 2001
From: Michael Niedermayer <michaelni@gmx.at>
Date: Mon, 5 Nov 2001 18:50:58 +0000
Subject: [PATCH] yuy2toyv12 fixed and speedup

Originally committed as revision 2725 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc
---
 postproc/rgb2rgb.c          | 171 ++++++++++++++++++++++++++++----------------
 postproc/rgb2rgb.h          |   3 +-
 postproc/rgb2rgb_template.c | 171 ++++++++++++++++++++++++++++----------------
 3 files changed, 220 insertions(+), 125 deletions(-)

diff --git a/postproc/rgb2rgb.c b/postproc/rgb2rgb.c
index 5d538ec..566a785 100644
--- a/postproc/rgb2rgb.c
+++ b/postproc/rgb2rgb.c
@@ -291,7 +291,8 @@ void palette8torgb15(const uint8_t *src, uint8_t *dst, unsigned num_pixels, cons
 }
 /**
  *
- * width must be a multiple of 16 for the MMX version
+ * height should be a multiple of 2 and width should be a multiple of 16 (if this is a
+ * problem for anyone then tell me, and ill fix it)
  */
 void yv12toyuy2(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
 	int width, int height, int lumStride, int chromStride, int dstStride)
@@ -359,70 +360,116 @@ asm(    EMMS" \n\t"
 #endif
 }
 
-void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, unsigned num_pixels)
+/**
+ *
+ * height should be a multiple of 2 and width should be a multiple of 16 (if this is a
+ * problem for anyone then tell me, and ill fix it)
+ */
+void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
+	int width, int height, int lumStride, int chromStride, int srcStride)
 {
+	int y;
+	const int chromWidth= width>>1;
+	for(y=0; y<height; y+=2)
+	{
 #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"
-	);
+		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"
+
+			"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 16(%0, %%eax, 4), %%mm2	\n\t" // YUYV YUYV(8)
+			"movq 24(%0, %%eax, 4), %%mm3	\n\t" // YUYV YUYV(12)
+			"pand %%mm7, %%mm0		\n\t" // Y0Y0 Y0Y0(0)
+			"pand %%mm7, %%mm1		\n\t" // Y0Y0 Y0Y0(4)
+			"pand %%mm7, %%mm2		\n\t" // Y0Y0 Y0Y0(8)
+			"pand %%mm7, %%mm3		\n\t" // Y0Y0 Y0Y0(12)
+			"packuswb %%mm1, %%mm0		\n\t" // YYYY YYYY(0)
+			"packuswb %%mm3, %%mm2		\n\t" // YYYY YYYY(8)
+
+			MOVNTQ" %%mm0, (%1, %%eax, 2)	\n\t"
+			MOVNTQ" %%mm2, 8(%1, %%eax, 2)	\n\t"
+
+			"addl $8, %%eax			\n\t"
+			"cmpl %5, %%eax			\n\t"
+			" jb 1b				\n\t"
+
+			::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "r" (chromWidth), "m"(width)
+			: "memory", "%eax"
+		);
 #else
-	int i;
-	num_pixels>>=1;
-	for(i=0; i<num_pixels; i++)
-	{
-		 ydst[2*i+0] 	= src[4*i+0];
-		 udst[i] 	= src[4*i+1];
-		 ydst[2*i+1] 	= src[4*i+2];
-		 vdst[i] 	= src[4*i+3];
+		int i;
+		for(i=0; i<chromWidth; i++)
+		{
+			ydst[2*i+0] 	= src[4*i+0];
+			udst[i] 	= src[4*i+1];
+			ydst[2*i+1] 	= src[4*i+2];
+			vdst[i] 	= src[4*i+3];
+		}
+		ydst += lumStride;
+		src  += srcStride;
+
+		for(i=0; i<chromWidth; i++)
+		{
+			ydst[2*i+0] 	= src[4*i+0];
+			ydst[2*i+1] 	= src[4*i+2];
+		}
+#endif
+		udst += chromStride;
+		vdst += chromStride;
+		ydst += lumStride;
+		src  += srcStride;
 	}
+#ifdef HAVE_MMX
+asm(    EMMS" \n\t"
+        SFENCE" \n\t"
+        :::"memory");
 #endif
 }
diff --git a/postproc/rgb2rgb.h b/postproc/rgb2rgb.h
index db29be3..dc900a9 100644
--- a/postproc/rgb2rgb.h
+++ b/postproc/rgb2rgb.h
@@ -23,6 +23,7 @@ extern void palette8torgb24(const uint8_t *src, uint8_t *dst, unsigned num_pixel
 
 extern void yv12toyuy2(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
 	int width, int height, int lumStride, int chromStride, int dstStride);
-extern void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, unsigned num_pixels);
+extern void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
+	int width, int height, int lumStride, int chromStride, int srcStride);
 
 #endif
diff --git a/postproc/rgb2rgb_template.c b/postproc/rgb2rgb_template.c
index 5d538ec..566a785 100644
--- a/postproc/rgb2rgb_template.c
+++ b/postproc/rgb2rgb_template.c
@@ -291,7 +291,8 @@ void palette8torgb15(const uint8_t *src, uint8_t *dst, unsigned num_pixels, cons
 }
 /**
  *
- * width must be a multiple of 16 for the MMX version
+ * height should be a multiple of 2 and width should be a multiple of 16 (if this is a
+ * problem for anyone then tell me, and ill fix it)
  */
 void yv12toyuy2(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst,
 	int width, int height, int lumStride, int chromStride, int dstStride)
@@ -359,70 +360,116 @@ asm(    EMMS" \n\t"
 #endif
 }
 
-void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, unsigned num_pixels)
+/**
+ *
+ * height should be a multiple of 2 and width should be a multiple of 16 (if this is a
+ * problem for anyone then tell me, and ill fix it)
+ */
+void yuy2toyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
+	int width, int height, int lumStride, int chromStride, int srcStride)
 {
+	int y;
+	const int chromWidth= width>>1;
+	for(y=0; y<height; y+=2)
+	{
 #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"
-	);
+		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"
+
+			"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 16(%0, %%eax, 4), %%mm2	\n\t" // YUYV YUYV(8)
+			"movq 24(%0, %%eax, 4), %%mm3	\n\t" // YUYV YUYV(12)
+			"pand %%mm7, %%mm0		\n\t" // Y0Y0 Y0Y0(0)
+			"pand %%mm7, %%mm1		\n\t" // Y0Y0 Y0Y0(4)
+			"pand %%mm7, %%mm2		\n\t" // Y0Y0 Y0Y0(8)
+			"pand %%mm7, %%mm3		\n\t" // Y0Y0 Y0Y0(12)
+			"packuswb %%mm1, %%mm0		\n\t" // YYYY YYYY(0)
+			"packuswb %%mm3, %%mm2		\n\t" // YYYY YYYY(8)
+
+			MOVNTQ" %%mm0, (%1, %%eax, 2)	\n\t"
+			MOVNTQ" %%mm2, 8(%1, %%eax, 2)	\n\t"
+
+			"addl $8, %%eax			\n\t"
+			"cmpl %5, %%eax			\n\t"
+			" jb 1b				\n\t"
+
+			::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "r" (chromWidth), "m"(width)
+			: "memory", "%eax"
+		);
 #else
-	int i;
-	num_pixels>>=1;
-	for(i=0; i<num_pixels; i++)
-	{
-		 ydst[2*i+0] 	= src[4*i+0];
-		 udst[i] 	= src[4*i+1];
-		 ydst[2*i+1] 	= src[4*i+2];
-		 vdst[i] 	= src[4*i+3];
+		int i;
+		for(i=0; i<chromWidth; i++)
+		{
+			ydst[2*i+0] 	= src[4*i+0];
+			udst[i] 	= src[4*i+1];
+			ydst[2*i+1] 	= src[4*i+2];
+			vdst[i] 	= src[4*i+3];
+		}
+		ydst += lumStride;
+		src  += srcStride;
+
+		for(i=0; i<chromWidth; i++)
+		{
+			ydst[2*i+0] 	= src[4*i+0];
+			ydst[2*i+1] 	= src[4*i+2];
+		}
+#endif
+		udst += chromStride;
+		vdst += chromStride;
+		ydst += lumStride;
+		src  += srcStride;
 	}
+#ifdef HAVE_MMX
+asm(    EMMS" \n\t"
+        SFENCE" \n\t"
+        :::"memory");
 #endif
 }
-- 
2.7.4