From 97c50cb77949856573d7f5f5a3c28cb73e61e011 Mon Sep 17 00:00:00 2001
From: Mike Melanson <mike@multimedia.cx>
Date: Tue, 19 Apr 2005 05:16:45 +0000
Subject: sync to FFmpeg build 4752

CVS patchset: 7463
CVS date: 2005/04/19 05:16:45
---
 src/libffmpeg/libavcodec/libpostproc/mangle.h      |    9 +
 src/libffmpeg/libavcodec/libpostproc/postprocess.c |  266 +++-
 .../libpostproc/postprocess_altivec_template.c     |  721 ++++++++--
 .../libavcodec/libpostproc/postprocess_internal.h  |   10 +
 .../libavcodec/libpostproc/postprocess_template.c  | 1477 ++++++++++++++------
 5 files changed, 1853 insertions(+), 630 deletions(-)

(limited to 'src/libffmpeg/libavcodec/libpostproc')

diff --git a/src/libffmpeg/libavcodec/libpostproc/mangle.h b/src/libffmpeg/libavcodec/libpostproc/mangle.h
index f3894cc33..aa09cd6bf 100644
--- a/src/libffmpeg/libavcodec/libpostproc/mangle.h
+++ b/src/libffmpeg/libavcodec/libpostproc/mangle.h
@@ -8,12 +8,21 @@
 #define __MANGLE_H
 
 /* Feel free to add more to the list, eg. a.out IMO */
+/* Use rip-relative addressing if compiling PIC code on x86-64. */
 #if defined(__CYGWIN__) || defined(__MINGW32__) || defined(__OS2__) || \
    (defined(__OpenBSD__) && !defined(__ELF__))
+#if defined(ARCH_X86_64) && defined(PIC)
+#define MANGLE(a) "_" #a"(%%rip)"
+#else
 #define MANGLE(a) "_" #a
+#endif
+#else
+#if defined(ARCH_X86_64) && defined(PIC)
+#define MANGLE(a) #a"(%%rip)"
 #else
 #define MANGLE(a) #a
 #endif
+#endif
 
 #endif /* !__MANGLE_H */
 
diff --git a/src/libffmpeg/libavcodec/libpostproc/postprocess.c b/src/libffmpeg/libavcodec/libpostproc/postprocess.c
index a03ff133d..e7ca0191d 100644
--- a/src/libffmpeg/libavcodec/libpostproc/postprocess.c
+++ b/src/libffmpeg/libavcodec/libpostproc/postprocess.c
@@ -29,10 +29,11 @@ isVertDC		Ec	Ec			Ec
 isVertMinMaxOk		Ec	Ec			Ec
 doVertLowPass		E		e	e	Ec
 doVertDefFilter		Ec	Ec	e	e	Ec
-isHorizDC		Ec	Ec
-isHorizMinMaxOk		a	E
-doHorizLowPass		E		e	e
-doHorizDefFilter	Ec	Ec	e	e
+isHorizDC		Ec	Ec			Ec
+isHorizMinMaxOk		a	E			Ec
+doHorizLowPass		E		e	e	Ec
+doHorizDefFilter	Ec	Ec	e	e	Ec
+do_a_deblock		Ec	E	Ec	E
 deRing			E		e	e*	Ecp
 Vertical RKAlgo1	E		a	a
 Horizontal RKAlgo1			a	a
@@ -42,7 +43,7 @@ LinIpolDeinterlace	e		E	E*
 CubicIpolDeinterlace	a		e	e*
 LinBlendDeinterlace	e		E	E*
 MedianDeinterlace#	E	Ec	Ec
-TempDeNoiser#		E		e	e
+TempDeNoiser#		E		e	e	Ec
 
 * i dont have a 3dnow CPU -> its untested, but noone said it doesnt work so it seems to work
 # more or less selfinvented filters so the exactness isnt too meaningfull
@@ -91,6 +92,10 @@ try to unroll inner for(x=0 ... loop to avoid these damn if(x ... checks
 
 #include "mangle.h" //FIXME should be supressed
 
+#ifdef HAVE_ALTIVEC_H
+#include <altivec.h>
+#endif
+
 #ifndef HAVE_MEMALIGN
 #define memalign(a,b) malloc(b)
 #endif
@@ -108,12 +113,15 @@ try to unroll inner for(x=0 ... loop to avoid these damn if(x ... checks
 
 #if defined(__GNUC__) && (__GNUC__ > 3 || __GNUC__ == 3 && __GNUC_MINOR__ > 0)
 #    define attribute_used __attribute__((used))
+#    define always_inline __attribute__((always_inline)) inline
 #else
 #    define attribute_used
+#    define always_inline inline
 #endif
 
-#ifdef ARCH_X86
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
 static uint64_t __attribute__((aligned(8))) attribute_used w05=		0x0005000500050005LL;
+static uint64_t __attribute__((aligned(8))) attribute_used w04=		0x0004000400040004LL;
 static uint64_t __attribute__((aligned(8))) attribute_used w20=		0x0020002000200020LL;
 static uint64_t __attribute__((aligned(8))) attribute_used b00= 		0x0000000000000000LL;
 static uint64_t __attribute__((aligned(8))) attribute_used b01= 		0x0101010101010101LL;
@@ -122,7 +130,6 @@ static uint64_t __attribute__((aligned(8))) attribute_used b08= 		0x080808080808
 static uint64_t __attribute__((aligned(8))) attribute_used b80= 		0x8080808080808080LL;
 #endif
 
-
 static uint8_t clip_table[3*256];
 static uint8_t * const clip_tab= clip_table + 256;
 
@@ -139,6 +146,8 @@ static struct PPFilter filters[]=
 	{"vr", "rkvdeblock", 		1, 2, 4, V_RK1_FILTER},*/
 	{"h1", "x1hdeblock", 		1, 1, 3, H_X1_FILTER},
 	{"v1", "x1vdeblock", 		1, 2, 4, V_X1_FILTER},
+	{"ha", "ahdeblock", 		1, 1, 3, H_A_DEBLOCK},
+	{"va", "avdeblock", 		1, 2, 4, V_A_DEBLOCK},
 	{"dr", "dering", 		1, 5, 6, DERING},
 	{"al", "autolevels", 		0, 1, 2, LEVEL_FIX},
 	{"lb", "linblenddeint", 	1, 1, 4, LINEAR_BLEND_DEINT_FILTER},
@@ -154,15 +163,16 @@ static struct PPFilter filters[]=
 
 static char *replaceTable[]=
 {
-	"default", 	"hdeblock:a,vdeblock:a,dering:a,autolevels,tmpnoise:a:150:200:400",
-	"de", 		"hdeblock:a,vdeblock:a,dering:a,autolevels,tmpnoise:a:150:200:400",
-	"fast", 	"x1hdeblock:a,x1vdeblock:a,dering:a,autolevels,tmpnoise:a:150:200:400",
-	"fa", 		"x1hdeblock:a,x1vdeblock:a,dering:a,autolevels,tmpnoise:a:150:200:400",
+	"default", 	"hdeblock:a,vdeblock:a,dering:a",
+	"de", 		"hdeblock:a,vdeblock:a,dering:a",
+	"fast", 	"x1hdeblock:a,x1vdeblock:a,dering:a",
+	"fa", 		"x1hdeblock:a,x1vdeblock:a,dering:a",
+	"ac", 		"ha:a:128:7,va:a,dering:a",
 	NULL //End Marker
 };
 
 
-#ifdef ARCH_X86
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
 static inline void prefetchnta(void *p)
 {
 	asm volatile(	"prefetchnta (%0)\n\t"
@@ -372,32 +382,32 @@ static inline void doHorizDefFilter_C(uint8_t dst[], int stride, PPContext *c)
  */
 static inline void doHorizLowPass_C(uint8_t dst[], int stride, PPContext *c)
 {
-
 	int y;
 	for(y=0; y<BLOCK_SIZE; y++)
 	{
 		const int first= ABS(dst[-1] - dst[0]) < c->QP ? dst[-1] : dst[0];
 		const int last= ABS(dst[8] - dst[7]) < c->QP ? dst[8] : dst[7];
 
-		int sums[9];
-		sums[0] = first + dst[0];
-		sums[1] = dst[0] + dst[1];
-		sums[2] = dst[1] + dst[2];
-		sums[3] = dst[2] + dst[3];
-		sums[4] = dst[3] + dst[4];
-		sums[5] = dst[4] + dst[5];
-		sums[6] = dst[5] + dst[6];
-		sums[7] = dst[6] + dst[7];
-		sums[8] = dst[7] + last;
-
-		dst[0]= ((sums[0]<<2) + ((first + sums[2])<<1) + sums[4] + 8)>>4;
-		dst[1]= ((dst[1]<<2) + ((first + sums[0] + sums[3])<<1) + sums[5] + 8)>>4;
-		dst[2]= ((dst[2]<<2) + ((first + sums[1] + sums[4])<<1) + sums[6] + 8)>>4;
-		dst[3]= ((dst[3]<<2) + ((sums[2] + sums[5])<<1) + sums[0] + sums[7] + 8)>>4;
-		dst[4]= ((dst[4]<<2) + ((sums[3] + sums[6])<<1) + sums[1] + sums[8] + 8)>>4;
-		dst[5]= ((dst[5]<<2) + ((last + sums[7] + sums[4])<<1) + sums[2] + 8)>>4;
-		dst[6]= (((last + dst[6])<<2) + ((dst[7] + sums[5])<<1) + sums[3] + 8)>>4;
-		dst[7]= ((sums[8]<<2) + ((last + sums[6])<<1) + sums[4] + 8)>>4;
+		int sums[10];
+		sums[0] = 4*first + dst[0] + dst[1] + dst[2] + 4;
+		sums[1] = sums[0] - first  + dst[3];
+		sums[2] = sums[1] - first  + dst[4];
+		sums[3] = sums[2] - first  + dst[5];
+		sums[4] = sums[3] - first  + dst[6];
+		sums[5] = sums[4] - dst[0] + dst[7];
+		sums[6] = sums[5] - dst[1] + last;
+		sums[7] = sums[6] - dst[2] + last;
+		sums[8] = sums[7] - dst[3] + last;
+		sums[9] = sums[8] - dst[4] + last;
+
+		dst[0]= (sums[0] + sums[2] + 2*dst[0])>>4;
+		dst[1]= (sums[1] + sums[3] + 2*dst[1])>>4;
+		dst[2]= (sums[2] + sums[4] + 2*dst[2])>>4;
+		dst[3]= (sums[3] + sums[5] + 2*dst[3])>>4;
+		dst[4]= (sums[4] + sums[6] + 2*dst[4])>>4;
+		dst[5]= (sums[5] + sums[7] + 2*dst[5])>>4;
+		dst[6]= (sums[6] + sums[8] + 2*dst[6])>>4;
+		dst[7]= (sums[7] + sums[9] + 2*dst[7])>>4;
 
 		dst+= stride;
 	}
@@ -469,6 +479,111 @@ static inline void horizX1Filter(uint8_t *src, int stride, int QP)
 	}
 }
 
+/**
+ * accurate deblock filter
+ */
+static always_inline void do_a_deblock_C(uint8_t *src, int step, int stride, PPContext *c){
+	int y;
+	const int QP= c->QP;
+	const int dcOffset= ((c->nonBQP*c->ppMode.baseDcDiff)>>8) + 1;
+	const int dcThreshold= dcOffset*2 + 1;
+//START_TIMER
+	src+= step*4; // src points to begin of the 8x8 Block
+	for(y=0; y<8; y++){
+		int numEq= 0;
+
+		if(((unsigned)(src[-1*step] - src[0*step] + dcOffset)) < dcThreshold) numEq++;
+		if(((unsigned)(src[ 0*step] - src[1*step] + dcOffset)) < dcThreshold) numEq++;
+		if(((unsigned)(src[ 1*step] - src[2*step] + dcOffset)) < dcThreshold) numEq++;
+		if(((unsigned)(src[ 2*step] - src[3*step] + dcOffset)) < dcThreshold) numEq++;
+		if(((unsigned)(src[ 3*step] - src[4*step] + dcOffset)) < dcThreshold) numEq++;
+		if(((unsigned)(src[ 4*step] - src[5*step] + dcOffset)) < dcThreshold) numEq++;
+		if(((unsigned)(src[ 5*step] - src[6*step] + dcOffset)) < dcThreshold) numEq++;
+		if(((unsigned)(src[ 6*step] - src[7*step] + dcOffset)) < dcThreshold) numEq++;
+		if(((unsigned)(src[ 7*step] - src[8*step] + dcOffset)) < dcThreshold) numEq++;
+		if(numEq > c->ppMode.flatnessThreshold){
+			int min, max, x;
+			
+			if(src[0] > src[step]){
+			    max= src[0];
+			    min= src[step];
+			}else{
+			    max= src[step];
+			    min= src[0];
+			}
+			for(x=2; x<8; x+=2){
+				if(src[x*step] > src[(x+1)*step]){
+					if(src[x    *step] > max) max= src[ x   *step];
+					if(src[(x+1)*step] < min) min= src[(x+1)*step];
+				}else{
+					if(src[(x+1)*step] > max) max= src[(x+1)*step];
+					if(src[ x   *step] < min) min= src[ x   *step];
+				}
+			}
+			if(max-min < 2*QP){
+				const int first= ABS(src[-1*step] - src[0]) < QP ? src[-1*step] : src[0];
+				const int last= ABS(src[8*step] - src[7*step]) < QP ? src[8*step] : src[7*step];
+				
+				int sums[10];
+				sums[0] = 4*first + src[0*step] + src[1*step] + src[2*step] + 4;
+				sums[1] = sums[0] - first       + src[3*step];
+				sums[2] = sums[1] - first       + src[4*step];
+				sums[3] = sums[2] - first       + src[5*step];
+				sums[4] = sums[3] - first       + src[6*step];
+				sums[5] = sums[4] - src[0*step] + src[7*step];
+				sums[6] = sums[5] - src[1*step] + last;
+				sums[7] = sums[6] - src[2*step] + last;
+				sums[8] = sums[7] - src[3*step] + last;
+				sums[9] = sums[8] - src[4*step] + last;
+
+				src[0*step]= (sums[0] + sums[2] + 2*src[0*step])>>4;
+				src[1*step]= (sums[1] + sums[3] + 2*src[1*step])>>4;
+				src[2*step]= (sums[2] + sums[4] + 2*src[2*step])>>4;
+				src[3*step]= (sums[3] + sums[5] + 2*src[3*step])>>4;
+				src[4*step]= (sums[4] + sums[6] + 2*src[4*step])>>4;
+				src[5*step]= (sums[5] + sums[7] + 2*src[5*step])>>4;
+				src[6*step]= (sums[6] + sums[8] + 2*src[6*step])>>4;
+				src[7*step]= (sums[7] + sums[9] + 2*src[7*step])>>4;
+			}
+		}else{
+			const int middleEnergy= 5*(src[4*step] - src[3*step]) + 2*(src[2*step] - src[5*step]);
+
+			if(ABS(middleEnergy) < 8*QP)
+			{
+				const int q=(src[3*step] - src[4*step])/2;
+				const int leftEnergy=  5*(src[2*step] - src[1*step]) + 2*(src[0*step] - src[3*step]);
+				const int rightEnergy= 5*(src[6*step] - src[5*step]) + 2*(src[4*step] - src[7*step]);
+
+				int d= ABS(middleEnergy) - MIN( ABS(leftEnergy), ABS(rightEnergy) );
+				d= MAX(d, 0);
+	
+				d= (5*d + 32) >> 6;
+				d*= SIGN(-middleEnergy);
+	
+				if(q>0)
+				{
+					d= d<0 ? 0 : d;
+					d= d>q ? q : d;
+				}
+				else
+				{
+					d= d>0 ? 0 : d;
+					d= d<q ? q : d;
+				}
+	
+				src[3*step]-= d;
+				src[4*step]+= d;
+			}
+		}
+
+		src += stride;
+	}
+/*if(step==16){
+    STOP_TIMER("step16")
+}else{
+    STOP_TIMER("stepX")
+}*/
+}
 
 //Note: we have C, MMX, MMX2, 3DNOW version there is no 3DNOW+MMX2 one
 //Plain C versions
@@ -479,15 +594,10 @@ static inline void horizX1Filter(uint8_t *src, int stride, int QP)
 #ifdef ARCH_POWERPC
 #ifdef HAVE_ALTIVEC
 #define COMPILE_ALTIVEC
-#ifndef CONFIG_DARWIN
-#warning "################################################################################"
-#warning  "WARNING: No gcc available as of today (2004-05-25) seems to be able to compile properly some of the code under non-Darwin PPC OSes. Some functions result in wrong results, while others simply won't compile (gcc explodes after allocating 1GiB+)."
-#warning "################################################################################"
-#endif //CONFIG_DARWIN
 #endif //HAVE_ALTIVEC
 #endif //ARCH_POWERPC
 
-#ifdef ARCH_X86
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
 
 #if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
 #define COMPILE_MMX
@@ -506,13 +616,11 @@ static inline void horizX1Filter(uint8_t *src, int stride, int QP)
 #undef HAVE_MMX2
 #undef HAVE_3DNOW
 #undef HAVE_ALTIVEC
-#undef ARCH_X86
 
 #ifdef COMPILE_C
 #undef HAVE_MMX
 #undef HAVE_MMX2
 #undef HAVE_3DNOW
-#undef ARCH_X86
 #define RENAME(a) a ## _C
 #include "postprocess_template.c"
 #endif
@@ -533,7 +641,6 @@ static inline void horizX1Filter(uint8_t *src, int stride, int QP)
 #define HAVE_MMX
 #undef HAVE_MMX2
 #undef HAVE_3DNOW
-#define ARCH_X86
 #define RENAME(a) a ## _MMX
 #include "postprocess_template.c"
 #endif
@@ -544,7 +651,6 @@ static inline void horizX1Filter(uint8_t *src, int stride, int QP)
 #define HAVE_MMX
 #define HAVE_MMX2
 #undef HAVE_3DNOW
-#define ARCH_X86
 #define RENAME(a) a ## _MMX2
 #include "postprocess_template.c"
 #endif
@@ -555,7 +661,6 @@ static inline void horizX1Filter(uint8_t *src, int stride, int QP)
 #define HAVE_MMX
 #undef HAVE_MMX2
 #define HAVE_3DNOW
-#define ARCH_X86
 #define RENAME(a) a ## _3DNow
 #include "postprocess_template.c"
 #endif
@@ -573,7 +678,7 @@ static inline void postProcess(uint8_t src[], int srcStride, uint8_t dst[], int
 	// difference wouldnt be messureable here but its much better because
 	// someone might exchange the cpu whithout restarting mplayer ;)
 #ifdef RUNTIME_CPUDETECT
-#ifdef ARCH_X86
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
 	// ordered per speed fasterst first
 	if(c->cpuCaps & PP_CPU_CAPS_MMX2)
 		postProcess_MMX2(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c);
@@ -586,7 +691,7 @@ static inline void postProcess(uint8_t src[], int srcStride, uint8_t dst[], int
 #else
 #ifdef ARCH_POWERPC
 #ifdef HAVE_ALTIVEC
-        else if(c->cpuCaps & PP_CPU_CAPS_ALTIVEC)
+        if(c->cpuCaps & PP_CPU_CAPS_ALTIVEC)
 		postProcess_altivec(src, srcStride, dst, dstStride, width, height, QPs, QPStride, isColor, c);
         else
 #endif
@@ -614,24 +719,21 @@ static inline void postProcess(uint8_t src[], int srcStride, uint8_t dst[], int
 /* -pp Command line Help
 */
 char *pp_help=
-"<filterName>[:<option>[:<option>...]][[,|/][-]<filterName>[:<option>...]]...\n"
-"long form example:\n"
-"vdeblock:autoq/hdeblock:autoq/linblenddeint	default,-vdeblock\n"
-"short form example:\n"
-"vb:a/hb:a/lb					de,-vb\n"
-"more examples:\n"
-"tn:64:128:256\n"
+"Available postprocessing filters:\n"
 "Filters			Options\n"
 "short	long name	short	long option	Description\n"
 "*	*		a	autoq		CPU power dependent enabler\n"
 "			c	chrom		chrominance filtering enabled\n"
 "			y	nochrom		chrominance filtering disabled\n"
+"			n	noluma		luma filtering disabled\n"
 "hb	hdeblock	(2 threshold)		horizontal deblocking filter\n"
 "	1. difference factor: default=32, higher -> more deblocking\n"
 "	2. flatness threshold: default=39, lower -> more deblocking\n"
 "			the h & v deblocking filters share these\n"
 "			so you can't set different thresholds for h / v\n"
 "vb	vdeblock	(2 threshold)		vertical deblocking filter\n"
+"ha	hadeblock	(2 threshold)		horizontal deblocking filter\n"
+"va	vadeblock	(2 threshold)		vertical deblocking filter\n"
 "h1	x1hdeblock				experimental h deblock filter 1\n"
 "v1	x1vdeblock				experimental v deblock filter 1\n"
 "dr	dering					deringing filter\n"
@@ -642,11 +744,20 @@ char *pp_help=
 "ci	cubicipoldeint				cubic interpolating deinterlacer\n"
 "md	mediandeint				median deinterlacer\n"
 "fd	ffmpegdeint				ffmpeg deinterlacer\n"
-"de	default					hb:a,vb:a,dr:a,al\n"
-"fa	fast					h1:a,v1:a,dr:a,al\n"
+"l5	lowpass5				FIR lowpass deinterlacer\n"
+"de	default					hb:a,vb:a,dr:a\n"
+"fa	fast					h1:a,v1:a,dr:a\n"
 "tn	tmpnoise	(3 threshold)		temporal noise reducer\n"
 "			1. <= 2. <= 3.		larger -> stronger filtering\n"
 "fq	forceQuant	<quantizer>		force quantizer\n"
+"Usage:\n"
+"<filterName>[:<option>[:<option>...]][[,|/][-]<filterName>[:<option>...]]...\n"
+"long form example:\n"
+"vdeblock:autoq/hdeblock:autoq/linblenddeint	default,-vdeblock\n"
+"short form example:\n"
+"vb:a/hb:a/lb					de,-vb\n"
+"more examples:\n"
+"tn:64:128:256\n"
 ;
 
 pp_mode_t *pp_get_mode_by_name_and_quality(char *name, int quality)
@@ -680,6 +791,7 @@ pp_mode_t *pp_get_mode_by_name_and_quality(char *name, int quality)
 		char *filterName;
 		int q= 1000000; //PP_QUALITY_MAX;
 		int chrom=-1;
+		int luma=-1;
 		char *option;
 		char *options[OPTIONS_ARRAY_SIZE];
 		int i;
@@ -707,6 +819,7 @@ pp_mode_t *pp_get_mode_by_name_and_quality(char *name, int quality)
 			if(!strcmp("autoq", option) || !strcmp("a", option)) q= quality;
 			else if(!strcmp("nochrom", option) || !strcmp("y", option)) chrom=0;
 			else if(!strcmp("chrom", option) || !strcmp("c", option)) chrom=1;
+			else if(!strcmp("noluma", option) || !strcmp("n", option)) luma=0;
 			else
 			{
 				options[numOfUnknownOptions] = option;
@@ -753,7 +866,7 @@ pp_mode_t *pp_get_mode_by_name_and_quality(char *name, int quality)
 				filterNameOk=1;
 				if(!enable) break; // user wants to disable it
 
-				if(q >= filters[i].minLumQuality)
+				if(q >= filters[i].minLumQuality && luma)
 					ppMode->lumMode|= filters[i].mask;
 				if(chrom==1 || (chrom==-1 && filters[i].chromDefault))
 					if(q >= filters[i].minChromQuality)
@@ -793,7 +906,8 @@ pp_mode_t *pp_get_mode_by_name_and_quality(char *name, int quality)
 						}
 					}
 				}
-				else if(filters[i].mask == V_DEBLOCK || filters[i].mask == H_DEBLOCK)
+				else if(filters[i].mask == V_DEBLOCK   || filters[i].mask == H_DEBLOCK 
+				     || filters[i].mask == V_A_DEBLOCK || filters[i].mask == H_A_DEBLOCK)
 				{
 					int o;
 
@@ -940,18 +1054,20 @@ void  pp_postprocess(uint8_t * src[3], int srcStride[3],
 	int mbHeight= (height+15)>>4;
 	PPMode *mode = (PPMode*)vm;
 	PPContext *c = (PPContext*)vc;
-        int minStride= MAX(srcStride[0], dstStride[0]);
+	int minStride= MAX(ABS(srcStride[0]), ABS(dstStride[0]));
+	int absQPStride = ABS(QPStride);
 
-	if(c->stride < minStride || c->qpStride < QPStride)
+	// c->stride and c->QPStride are always positive
+	if(c->stride < minStride || c->qpStride < absQPStride)
 		reallocBuffers(c, width, height, 
 				MAX(minStride, c->stride), 
-				MAX(c->qpStride, QPStride));
+				MAX(c->qpStride, absQPStride));
 
 	if(QP_store==NULL || (mode->lumMode & FORCE_QUANT)) 
 	{
 		int i;
 		QP_store= c->forcedQPTable;
-		QPStride= 0;
+		absQPStride = QPStride = 0;
 		if(mode->lumMode & FORCE_QUANT)
 			for(i=0; i<mbWidth; i++) QP_store[i]= mode->forcedQuant;
 		else
@@ -961,7 +1077,7 @@ void  pp_postprocess(uint8_t * src[3], int srcStride[3],
 
 	if(pict_type & PP_PICT_TYPE_QP2){
 		int i;
-		const int count= mbHeight * QPStride;
+		const int count= mbHeight * absQPStride;
 		for(i=0; i<(count>>2); i++){
 			((uint32_t*)c->stdQPTable)[i] = (((uint32_t*)QP_store)[i]>>1) & 0x7F7F7F7F;
 		}
@@ -969,6 +1085,7 @@ void  pp_postprocess(uint8_t * src[3], int srcStride[3],
 			c->stdQPTable[i] = QP_store[i]>>1;
 		}
                 QP_store= c->stdQPTable;
+		QPStride= absQPStride;		
 	}
 
 if(0){
@@ -984,13 +1101,22 @@ for(y=0; y<mbHeight; y++){
 
 	if((pict_type&7)!=3)
 	{
-		int i;
-		const int count= mbHeight * QPStride;
-		for(i=0; i<(count>>2); i++){
-			((uint32_t*)c->nonBQPTable)[i] = ((uint32_t*)QP_store)[i] & 0x3F3F3F3F;
-		}
-		for(i<<=2; i<count; i++){
-			c->nonBQPTable[i] = QP_store[i] & 0x3F;
+		if (QPStride >= 0) {
+			int i;
+			const int count= mbHeight * QPStride;
+			for(i=0; i<(count>>2); i++){
+				((uint32_t*)c->nonBQPTable)[i] = ((uint32_t*)QP_store)[i] & 0x3F3F3F3F;
+			}
+			for(i<<=2; i<count; i++){
+				c->nonBQPTable[i] = QP_store[i] & 0x3F;
+			}
+		} else {
+			int i,j;
+			for(i=0; i<mbHeight; i++) {
+		    		for(j=0; j<absQPStride; j++) {
+					c->nonBQPTable[i*absQPStride+j] = QP_store[i*QPStride+j] & 0x3F;
+				}
+			}
 		}
 	}
 
@@ -1014,8 +1140,8 @@ for(y=0; y<mbHeight; y++){
 	}
 	else if(srcStride[1] == dstStride[1] && srcStride[2] == dstStride[2])
 	{
-		memcpy(dst[1], src[1], srcStride[1]*height);
-		memcpy(dst[2], src[2], srcStride[2]*height);
+		linecpy(dst[1], src[1], height, srcStride[1]);
+		linecpy(dst[2], src[2], height, srcStride[2]);
 	}
 	else
 	{
diff --git a/src/libffmpeg/libavcodec/libpostproc/postprocess_altivec_template.c b/src/libffmpeg/libavcodec/libpostproc/postprocess_altivec_template.c
index 0c84873cc..1c59b9465 100644
--- a/src/libffmpeg/libavcodec/libpostproc/postprocess_altivec_template.c
+++ b/src/libffmpeg/libavcodec/libpostproc/postprocess_altivec_template.c
@@ -25,6 +25,39 @@
 #define AVV(x...) {x}
 #endif
 
+#define ALTIVEC_TRANSPOSE_8x8_SHORT(src_a,src_b,src_c,src_d,src_e,src_f,src_g,src_h) \
+  do {									\
+    __typeof__(src_a) tempA1, tempB1, tempC1, tempD1;			\
+    __typeof__(src_a) tempE1, tempF1, tempG1, tempH1;			\
+    __typeof__(src_a) tempA2, tempB2, tempC2, tempD2;			\
+    __typeof__(src_a) tempE2, tempF2, tempG2, tempH2;			\
+    tempA1 = vec_mergeh (src_a, src_e);					\
+    tempB1 = vec_mergel (src_a, src_e);					\
+    tempC1 = vec_mergeh (src_b, src_f);					\
+    tempD1 = vec_mergel (src_b, src_f);					\
+    tempE1 = vec_mergeh (src_c, src_g);					\
+    tempF1 = vec_mergel (src_c, src_g);					\
+    tempG1 = vec_mergeh (src_d, src_h);					\
+    tempH1 = vec_mergel (src_d, src_h);					\
+    tempA2 = vec_mergeh (tempA1, tempE1);				\
+    tempB2 = vec_mergel (tempA1, tempE1);				\
+    tempC2 = vec_mergeh (tempB1, tempF1);				\
+    tempD2 = vec_mergel (tempB1, tempF1);				\
+    tempE2 = vec_mergeh (tempC1, tempG1);				\
+    tempF2 = vec_mergel (tempC1, tempG1);				\
+    tempG2 = vec_mergeh (tempD1, tempH1);				\
+    tempH2 = vec_mergel (tempD1, tempH1);				\
+    src_a = vec_mergeh (tempA2, tempE2);				\
+    src_b = vec_mergel (tempA2, tempE2);				\
+    src_c = vec_mergeh (tempB2, tempF2);				\
+    src_d = vec_mergel (tempB2, tempF2);				\
+    src_e = vec_mergeh (tempC2, tempG2);				\
+    src_f = vec_mergel (tempC2, tempG2);				\
+    src_g = vec_mergeh (tempD2, tempH2);				\
+    src_h = vec_mergel (tempD2, tempH2);				\
+  } while (0)
+
+
 static inline int vertClassify_altivec(uint8_t src[], int stride, PPContext *c) {
   /*
     this code makes no assumption on src or stride.
@@ -40,7 +73,9 @@ static inline int vertClassify_altivec(uint8_t src[], int stride, PPContext *c)
   vector signed short v2QP;
   vector unsigned short v4QP;
   vector unsigned short v_dcThreshold;
-  int two_vectors = ((((unsigned long)src2 % 16) > 8) || (stride % 16)) ? 1 : 0;
+  const int properStride = (stride % 16);
+  const int srcAlign = ((unsigned long)src2 % 16);
+  const int two_vectors = ((srcAlign > 8) || properStride) ? 1 : 0;
   const vector signed int zero = vec_splat_s32(0);
   const vector signed short mask = vec_splat_s16(1);
   vector signed int v_numEq = vec_splat_s32(0);
@@ -57,6 +92,8 @@ static inline int vertClassify_altivec(uint8_t src[], int stride, PPContext *c)
 
   src2 += stride * 4;
 
+  vector signed short v_srcAss0, v_srcAss1, v_srcAss2, v_srcAss3, v_srcAss4, v_srcAss5, v_srcAss6, v_srcAss7;
+
 #define LOAD_LINE(i)							\
   register int j##i = i * stride;					\
   vector unsigned char perm##i = vec_lvsl(j##i, src2);			\
@@ -66,19 +103,41 @@ static inline int vertClassify_altivec(uint8_t src[], int stride, PPContext *c)
     v_srcA2##i = vec_ld(j##i + 16, src2);				\
   const vector unsigned char v_srcA##i =				\
     vec_perm(v_srcA1##i, v_srcA2##i, perm##i);				\
-  vector signed short v_srcAss##i =					\
+  v_srcAss##i =                                                         \
     (vector signed short)vec_mergeh((vector signed char)zero,		\
 				    (vector signed char)v_srcA##i)
 
-  LOAD_LINE(0);
-  LOAD_LINE(1);
-  LOAD_LINE(2);
-  LOAD_LINE(3);
-  LOAD_LINE(4);
-  LOAD_LINE(5);
-  LOAD_LINE(6);
-  LOAD_LINE(7);
+#define LOAD_LINE_ALIGNED(i)                                            \
+  register int j##i = i * stride;                                       \
+  const vector unsigned char v_srcA##i = vec_ld(j##i, src2);            \
+  v_srcAss##i =                                                         \
+    (vector signed short)vec_mergeh((vector signed char)zero,		\
+				    (vector signed char)v_srcA##i)
+
+    // special casing the aligned case is worthwhile, as all call from
+    // the (transposed) horizontable deblocks will be aligned, i naddition
+    // to the naturraly aligned vertical deblocks.
+    if (properStride && srcAlign) {
+      LOAD_LINE_ALIGNED(0);
+      LOAD_LINE_ALIGNED(1);
+      LOAD_LINE_ALIGNED(2);
+      LOAD_LINE_ALIGNED(3);
+      LOAD_LINE_ALIGNED(4);
+      LOAD_LINE_ALIGNED(5);
+      LOAD_LINE_ALIGNED(6);
+      LOAD_LINE_ALIGNED(7);
+    } else {
+      LOAD_LINE(0);
+      LOAD_LINE(1);
+      LOAD_LINE(2);
+      LOAD_LINE(3);
+      LOAD_LINE(4);
+      LOAD_LINE(5);
+      LOAD_LINE(6);
+      LOAD_LINE(7);
+    }
 #undef LOAD_LINE
+#undef LOAD_LINE_ALIGNED
 
 #define ITER(i, j)							\
   const vector signed short v_diff##i =					\
@@ -133,7 +192,6 @@ static inline int vertClassify_altivec(uint8_t src[], int stride, PPContext *c)
   else return 2; 
 }
 
-
 static inline void doVertLowPass_altivec(uint8_t *src, int stride, PPContext *c) {
   /*
     this code makes no assumption on src or stride.
@@ -145,112 +203,130 @@ static inline void doVertLowPass_altivec(uint8_t *src, int stride, PPContext *c)
   */
   uint8_t *src2 = src;
   const vector signed int zero = vec_splat_s32(0);
+  const int properStride = (stride % 16);
+  const int srcAlign = ((unsigned long)src2 % 16);
   short __attribute__ ((aligned(16))) qp[8];
   qp[0] = c->QP;
   vector signed short vqp = vec_ld(0, qp);
   vqp = vec_splat(vqp, 0);
 	
+  src2 += stride*3;
+
+  vector signed short vb0, vb1, vb2, vb3, vb4, vb5, vb6, vb7, vb8, vb9;
+  vector unsigned char vbA0, vbA1, vbA2, vbA3, vbA4, vbA5, vbA6, vbA7, vbA8, vbA9;
+  vector unsigned char vbB0, vbB1, vbB2, vbB3, vbB4, vbB5, vbB6, vbB7, vbB8, vbB9;
+  vector unsigned char vbT0, vbT1, vbT2, vbT3, vbT4, vbT5, vbT6, vbT7, vbT8, vbT9;
+	
 #define LOAD_LINE(i)                                                    \
   const vector unsigned char perml##i =					\
     vec_lvsl(i * stride, src2);						\
-  const vector unsigned char vbA##i =					\
-    vec_ld(i * stride, src2);						\
-  const vector unsigned char vbB##i =					\
-    vec_ld(i * stride + 16, src2);					\
-  const vector unsigned char vbT##i =					\
-    vec_perm(vbA##i, vbB##i, perml##i);					\
-  const vector signed short vb##i =					\
+  vbA##i = vec_ld(i * stride, src2);                                    \
+  vbB##i = vec_ld(i * stride + 16, src2);                               \
+  vbT##i = vec_perm(vbA##i, vbB##i, perml##i);                          \
+  vb##i =                                                               \
     (vector signed short)vec_mergeh((vector unsigned char)zero,		\
 				    (vector unsigned char)vbT##i)
-	
-  src2 += stride*3;
 
-  LOAD_LINE(0);
-  LOAD_LINE(1);
-  LOAD_LINE(2);
-  LOAD_LINE(3);
-  LOAD_LINE(4);
-  LOAD_LINE(5);
-  LOAD_LINE(6);
-  LOAD_LINE(7);
-  LOAD_LINE(8);
-  LOAD_LINE(9);
+#define LOAD_LINE_ALIGNED(i)                                            \
+  register int j##i = i * stride;                                       \
+  vbT##i = vec_ld(j##i, src2);                                          \
+  vb##i =                                                               \
+    (vector signed short)vec_mergeh((vector signed char)zero,		\
+				    (vector signed char)vbT##i)
+
+    // special casing the aligned case is worthwhile, as all call from
+    // the (transposed) horizontable deblocks will be aligned, in addition
+    // to the naturraly aligned vertical deblocks.
+    if (properStride && srcAlign) {
+      LOAD_LINE_ALIGNED(0);
+      LOAD_LINE_ALIGNED(1);
+      LOAD_LINE_ALIGNED(2);
+      LOAD_LINE_ALIGNED(3);
+      LOAD_LINE_ALIGNED(4);
+      LOAD_LINE_ALIGNED(5);
+      LOAD_LINE_ALIGNED(6);
+      LOAD_LINE_ALIGNED(7);
+      LOAD_LINE_ALIGNED(8);
+      LOAD_LINE_ALIGNED(9);
+    } else {
+      LOAD_LINE(0);
+      LOAD_LINE(1);
+      LOAD_LINE(2);
+      LOAD_LINE(3);
+      LOAD_LINE(4);
+      LOAD_LINE(5);
+      LOAD_LINE(6);
+      LOAD_LINE(7);
+      LOAD_LINE(8);
+      LOAD_LINE(9);
+    }
 #undef LOAD_LINE
+#undef LOAD_LINE_ALIGNED
 
   const vector unsigned short v_1 = vec_splat_u16(1);
   const vector unsigned short v_2 = vec_splat_u16(2);
   const vector unsigned short v_4 = vec_splat_u16(4);
-  const vector signed short v_8 = vec_splat_s16(8);
-
-  const vector signed short v_first = vec_sel(vb1, vb0,
-                                              vec_cmplt(vec_abs(vec_sub(vb0, vb1)),
-                                                        vqp));
-  const vector signed short v_last = vec_sel(vb8, vb9,
-                                             vec_cmplt(vec_abs(vec_sub(vb8, vb9)),
-                                                       vqp));
-
-  const vector signed short v_sums0 = vec_add(v_first, vb1);
-  const vector signed short v_sums1 = vec_add(vb1, vb2);
-  const vector signed short v_sums2 = vec_add(vb2, vb3);
-  const vector signed short v_sums3 = vec_add(vb3, vb4);
-  const vector signed short v_sums4 = vec_add(vb4, vb5);
-  const vector signed short v_sums5 = vec_add(vb5, vb6);
-  const vector signed short v_sums6 = vec_add(vb6, vb7);
-  const vector signed short v_sums7 = vec_add(vb7, vb8);
-  const vector signed short v_sums8 = vec_add(vb8, v_last);
-
-  const vector signed short vr1 = vec_sra(vec_add(vec_add(vec_sl(v_sums0, v_2),
-                                                          vec_sl(vec_add(v_first, v_sums2), v_1)),
-                                                  vec_add(v_sums4, v_8)),
-                                          v_4);
-  const vector signed short vr2 = vec_sra(vec_add(vec_add(vec_sl(vb2, v_2),
-                                                          v_sums5),
-                                                  vec_add(v_8,
-                                                          vec_sl(vec_add(v_first,
-                                                                         vec_add(v_sums0, v_sums3)),
-                                                                 v_1))),
-                                          v_4);
-  const vector signed short vr3 = vec_sra(vec_add(vec_add(vec_sl(vb3, v_2),
-                                                          v_sums6),
-                                                  vec_add(v_8,
-                                                          vec_sl(vec_add(v_first,
-                                                                         vec_add(v_sums1, v_sums4)),
-                                                                 v_1))),
-                                          v_4);
-  const vector signed short vr4 = vec_sra(vec_add(vec_add(vec_sl(vb4, v_2),
-                                                          v_sums7),
-                                                  vec_add(v_8,
-                                                          vec_add(v_sums0,
-                                                                  vec_sl(vec_add(v_sums2, v_sums5),
-                                                                         v_1)))),
-                                          v_4);
-  const vector signed short vr5 = vec_sra(vec_add(vec_add(vec_sl(vb5, v_2),
-                                                          v_sums8),
-                                                  vec_add(v_8,
-                                                          vec_add(v_sums1,
-                                                                  vec_sl(vec_add(v_sums3, v_sums6),
-                                                                         v_1)))),
-                                          v_4);
-  const vector signed short vr6 = vec_sra(vec_add(vec_add(vec_sl(vb6, v_2),
-                                                          v_sums2),
-                                                  vec_add(v_8,
-                                                          vec_sl(vec_add(v_last,
-                                                                         vec_add(v_sums7, v_sums4)),
-                                                                 v_1))),
-                                          v_4);
-  const vector signed short vr7 = vec_sra(vec_add(vec_add(vec_sl(vec_add(v_last, vb7), v_2),
-                                                          vec_sl(vec_add(vb8, v_sums5), v_1)),
-                                                  vec_add(v_8, v_sums3)),
-                                          v_4);
-  const vector signed short vr8 = vec_sra(vec_add(vec_add(vec_sl(v_sums8, v_2),
-                                                          vec_sl(vec_add(v_last, v_sums6), v_1)),
-                                                  vec_add(v_sums4, v_8)),
-                                          v_4);
-
-  const vector unsigned char neg1 = (vector unsigned char)AVV(-1, -1, -1, -1, -1, -1, -1, -1,
-							      -1, -1, -1, -1, -1, -1, -1, -1);
-  const vector unsigned char permHH = (vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
-								0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
+
+  const vector signed short v_diff01 = vec_sub(vb0, vb1);
+  const vector unsigned short v_cmp01 =
+    (const vector unsigned short) vec_cmplt(vec_abs(v_diff01), vqp);
+  const vector signed short v_first = vec_sel(vb1, vb0, v_cmp01);
+  const vector signed short v_diff89 = vec_sub(vb8, vb9);
+  const vector unsigned short v_cmp89 =
+    (const vector unsigned short) vec_cmplt(vec_abs(v_diff89), vqp);
+  const vector signed short v_last = vec_sel(vb8, vb9, v_cmp89);
+  
+  const vector signed short temp01 = vec_mladd(v_first, (vector signed short)v_4, vb1);
+  const vector signed short temp02 = vec_add(vb2, vb3);
+  const vector signed short temp03 = vec_add(temp01, (vector signed short)v_4);
+  const vector signed short v_sumsB0 = vec_add(temp02, temp03);
+
+  const vector signed short temp11 = vec_sub(v_sumsB0, v_first);
+  const vector signed short v_sumsB1 = vec_add(temp11, vb4);
+
+  const vector signed short temp21 = vec_sub(v_sumsB1, v_first);
+  const vector signed short v_sumsB2 = vec_add(temp21, vb5);
+
+  const vector signed short temp31 = vec_sub(v_sumsB2, v_first);
+  const vector signed short v_sumsB3 = vec_add(temp31, vb6);
+
+  const vector signed short temp41 = vec_sub(v_sumsB3, v_first);
+  const vector signed short v_sumsB4 = vec_add(temp41, vb7);
+
+  const vector signed short temp51 = vec_sub(v_sumsB4, vb1);
+  const vector signed short v_sumsB5 = vec_add(temp51, vb8);
+
+  const vector signed short temp61 = vec_sub(v_sumsB5, vb2);
+  const vector signed short v_sumsB6 = vec_add(temp61, v_last);
+
+  const vector signed short temp71 = vec_sub(v_sumsB6, vb3);
+  const vector signed short v_sumsB7 = vec_add(temp71, v_last);
+
+  const vector signed short temp81 = vec_sub(v_sumsB7, vb4);
+  const vector signed short v_sumsB8 = vec_add(temp81, v_last);
+
+  const vector signed short temp91 = vec_sub(v_sumsB8, vb5);
+  const vector signed short v_sumsB9 = vec_add(temp91, v_last);
+
+#define COMPUTE_VR(i, j, k)						\
+  const vector signed short temps1##i =					\
+    vec_add(v_sumsB##i, v_sumsB##k);					\
+  const vector signed short temps2##i =					\
+    vec_mladd(vb##j, (vector signed short)v_2, temps1##i);		\
+  const vector signed short  vr##j = vec_sra(temps2##i, v_4)
+
+  COMPUTE_VR(0, 1, 2);
+  COMPUTE_VR(1, 2, 3);
+  COMPUTE_VR(2, 3, 4);
+  COMPUTE_VR(3, 4, 5);
+  COMPUTE_VR(4, 5, 6);
+  COMPUTE_VR(5, 6, 7);
+  COMPUTE_VR(6, 7, 8);
+  COMPUTE_VR(7, 8, 9);
+
+  const vector signed char neg1 = vec_splat_s8(-1);
+  const vector unsigned char permHH = (const vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+								      0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
 
 #define PACK_AND_STORE(i)					\
   const vector unsigned char perms##i =				\
@@ -260,7 +336,7 @@ static inline void doVertLowPass_altivec(uint8_t *src, int stride, PPContext *c)
   const vector unsigned char vg##i =				\
     vec_perm(vf##i, vbT##i, permHH);				\
   const vector unsigned char mask##i =				\
-    vec_perm((vector unsigned char)zero, neg1, perms##i);	\
+    vec_perm((vector unsigned char)zero, (vector unsigned char)neg1, perms##i);	\
   const vector unsigned char vg2##i =				\
     vec_perm(vg##i, vg##i, perms##i);				\
   const vector unsigned char svA##i =				\
@@ -270,16 +346,37 @@ static inline void doVertLowPass_altivec(uint8_t *src, int stride, PPContext *c)
   vec_st(svA##i, i * stride, src2);				\
   vec_st(svB##i, i * stride + 16, src2)
 
-  PACK_AND_STORE(1);
-  PACK_AND_STORE(2);
-  PACK_AND_STORE(3);
-  PACK_AND_STORE(4);
-  PACK_AND_STORE(5);
-  PACK_AND_STORE(6);
-  PACK_AND_STORE(7);
-  PACK_AND_STORE(8);
+#define PACK_AND_STORE_ALIGNED(i)				\
+  const vector unsigned char vf##i =				\
+    vec_packsu(vr##i, (vector signed short)zero);		\
+  const vector unsigned char vg##i =				\
+    vec_perm(vf##i, vbT##i, permHH);				\
+  vec_st(vg##i, i * stride, src2)
 
+  // special casing the aligned case is worthwhile, as all call from
+  // the (transposed) horizontable deblocks will be aligned, in addition
+  // to the naturraly aligned vertical deblocks.
+  if (properStride && srcAlign) {
+    PACK_AND_STORE_ALIGNED(1);
+    PACK_AND_STORE_ALIGNED(2);
+    PACK_AND_STORE_ALIGNED(3);
+    PACK_AND_STORE_ALIGNED(4);
+    PACK_AND_STORE_ALIGNED(5);
+    PACK_AND_STORE_ALIGNED(6);
+    PACK_AND_STORE_ALIGNED(7);
+    PACK_AND_STORE_ALIGNED(8);
+  } else {
+    PACK_AND_STORE(1);
+    PACK_AND_STORE(2);
+    PACK_AND_STORE(3);
+    PACK_AND_STORE(4);
+    PACK_AND_STORE(5);
+    PACK_AND_STORE(6);
+    PACK_AND_STORE(7);
+    PACK_AND_STORE(8);
+  }
 #undef PACK_AND_STORE
+#undef PACK_AND_STORE_ALIGNED
 }
 
 
@@ -383,12 +480,10 @@ static inline void doVertDefFilter_altivec(uint8_t src[], int stride, PPContext
   /* finally, stores */
   const vector unsigned char st4 = vec_packsu(vb4minusd, (vector signed short)zero);
   const vector unsigned char st5 = vec_packsu(vb5plusd, (vector signed short)zero);
-	
-  const vector unsigned char neg1 = (vector unsigned char)AVV(-1, -1, -1, -1, -1, -1, -1, -1,
-							      -1, -1, -1, -1, -1, -1, -1, -1);
-	
-  const vector unsigned char permHH = (vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
-								0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
+  
+  const vector signed char neg1 = vec_splat_s8(-1);
+  const vector unsigned char permHH = (const vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+								      0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
 	
 #define STORE(i)						\
   const vector unsigned char perms##i =				\
@@ -396,7 +491,7 @@ static inline void doVertDefFilter_altivec(uint8_t src[], int stride, PPContext
   const vector unsigned char vg##i =				\
     vec_perm(st##i, vbT##i, permHH);				\
   const vector unsigned char mask##i =				\
-    vec_perm((vector unsigned char)zero, neg1, perms##i);	\
+    vec_perm((vector unsigned char)zero, (vector unsigned char)neg1, perms##i);	\
   const vector unsigned char vg2##i =				\
     vec_perm(vg##i, vg##i, perms##i);				\
   const vector unsigned char svA##i =				\
@@ -680,7 +775,7 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) {
   ITER(6, 7, 8);
   ITER(7, 8, 9);
 
-  const vector signed char neg1 = vec_splat_s8( -1 );
+  const vector signed char neg1 = vec_splat_s8(-1);
 	
 #define STORE_LINE(i)					\
   const vector unsigned char permST##i =		\
@@ -708,6 +803,394 @@ static inline void dering_altivec(uint8_t src[], int stride, PPContext *c) {
 #undef F2
 }
 
-#define horizClassify_altivec(a...) horizClassify_C(a)
 #define doHorizLowPass_altivec(a...) doHorizLowPass_C(a)
 #define doHorizDefFilter_altivec(a...) doHorizDefFilter_C(a)
+#define do_a_deblock_altivec(a...) do_a_deblock_C(a)
+
+static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
+				    uint8_t *tempBlured, uint32_t *tempBluredPast, int *maxNoise)
+{
+  const vector signed int zero = vec_splat_s32(0);
+  const vector signed short vsint16_1 = vec_splat_s16(1);
+  vector signed int v_dp = zero;
+  vector signed int v_sysdp = zero;
+  int d, sysd, i;
+  
+  tempBluredPast[127]= maxNoise[0];
+  tempBluredPast[128]= maxNoise[1];
+  tempBluredPast[129]= maxNoise[2];
+
+#define LOAD_LINE(src, i)						\
+  register int j##src##i = i * stride;					\
+  vector unsigned char perm##src##i = vec_lvsl(j##src##i, src);		\
+  const vector unsigned char v_##src##A1##i = vec_ld(j##src##i, src);	\
+  const vector unsigned char v_##src##A2##i = vec_ld(j##src##i + 16, src); \
+  const vector unsigned char v_##src##A##i =				\
+    vec_perm(v_##src##A1##i, v_##src##A2##i, perm##src##i);		\
+  vector signed short v_##src##Ass##i =					\
+    (vector signed short)vec_mergeh((vector signed char)zero,		\
+				    (vector signed char)v_##src##A##i)
+  
+  LOAD_LINE(src, 0);
+  LOAD_LINE(src, 1);
+  LOAD_LINE(src, 2);
+  LOAD_LINE(src, 3);
+  LOAD_LINE(src, 4);
+  LOAD_LINE(src, 5);
+  LOAD_LINE(src, 6);
+  LOAD_LINE(src, 7);
+
+  LOAD_LINE(tempBlured, 0);
+  LOAD_LINE(tempBlured, 1);
+  LOAD_LINE(tempBlured, 2);
+  LOAD_LINE(tempBlured, 3);
+  LOAD_LINE(tempBlured, 4);
+  LOAD_LINE(tempBlured, 5);
+  LOAD_LINE(tempBlured, 6);
+  LOAD_LINE(tempBlured, 7);
+#undef LOAD_LINE
+
+#define ACCUMULATE_DIFFS(i)					\
+  vector signed short v_d##i = vec_sub(v_tempBluredAss##i,	\
+				       v_srcAss##i);		\
+  v_dp = vec_msums(v_d##i, v_d##i, v_dp);			\
+  v_sysdp = vec_msums(v_d##i, vsint16_1, v_sysdp)
+
+  ACCUMULATE_DIFFS(0);
+  ACCUMULATE_DIFFS(1);
+  ACCUMULATE_DIFFS(2);
+  ACCUMULATE_DIFFS(3);
+  ACCUMULATE_DIFFS(4);
+  ACCUMULATE_DIFFS(5);
+  ACCUMULATE_DIFFS(6);
+  ACCUMULATE_DIFFS(7);
+#undef ACCUMULATE_DIFFS
+
+  v_dp = vec_sums(v_dp, zero);
+  v_sysdp = vec_sums(v_sysdp, zero);
+
+  v_dp = vec_splat(v_dp, 3);
+  v_sysdp = vec_splat(v_sysdp, 3);
+  
+  vec_ste(v_dp, 0, &d);
+  vec_ste(v_sysdp, 0, &sysd);
+
+  i = d;
+  d = (4*d
+       +(*(tempBluredPast-256))
+       +(*(tempBluredPast-1))+ (*(tempBluredPast+1))
+       +(*(tempBluredPast+256))
+       +4)>>3;
+
+  *tempBluredPast=i;
+
+  if (d > maxNoise[1]) {
+    if (d < maxNoise[2]) {
+#define OP(i) v_tempBluredAss##i = vec_avg(v_tempBluredAss##i, v_srcAss##i);
+
+      OP(0);
+      OP(1);
+      OP(2);
+      OP(3);
+      OP(4);
+      OP(5);
+      OP(6);
+      OP(7);
+#undef OP
+    } else {
+#define OP(i) v_tempBluredAss##i = v_srcAss##i;
+
+      OP(0);
+      OP(1);
+      OP(2);
+      OP(3);
+      OP(4);
+      OP(5);
+      OP(6);
+      OP(7);
+#undef OP
+    }
+  } else {
+    if (d < maxNoise[0]) {
+      const vector signed short vsint16_7 = vec_splat_s16(7);
+      const vector signed short vsint16_4 = vec_splat_s16(4);
+      const vector unsigned short vuint16_3 = vec_splat_u16(3);
+      
+#define OP(i)								\
+      const vector signed short v_temp##i =				\
+	vec_mladd(v_tempBluredAss##i,					\
+		  vsint16_7, v_srcAss##i);				\
+      const vector signed short v_temp2##i =				\
+	vec_add(v_temp##i, vsint16_4);					\
+      v_tempBluredAss##i = vec_sr(v_temp2##i, vuint16_3)
+
+      OP(0);
+      OP(1);
+      OP(2);
+      OP(3);
+      OP(4);
+      OP(5);
+      OP(6);
+      OP(7);
+#undef OP
+    } else {
+      const vector signed short vsint16_3 = vec_splat_s16(3);
+      const vector signed short vsint16_2 = vec_splat_s16(2);
+      
+#define OP(i)								\
+      const vector signed short v_temp##i =				\
+	vec_mladd(v_tempBluredAss##i,					\
+		  vsint16_3, v_srcAss##i);				\
+      const vector signed short v_temp2##i =				\
+	vec_add(v_temp##i, vsint16_2);					\
+      v_tempBluredAss##i = vec_sr(v_temp2##i, (vector unsigned short)vsint16_2)
+
+      OP(0);
+      OP(1);
+      OP(2);
+      OP(3);
+      OP(4);
+      OP(5);
+      OP(6);
+      OP(7);
+#undef OP
+    }
+  }
+
+  const vector signed char neg1 = vec_splat_s8(-1);
+  const vector unsigned char permHH = (const vector unsigned char)AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+								      0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
+
+#define PACK_AND_STORE(src, i)						\
+  const vector unsigned char perms##src##i =				\
+    vec_lvsr(i * stride, src);						\
+  const vector unsigned char vf##src##i =				\
+    vec_packsu(v_tempBluredAss##i, (vector signed short)zero);		\
+  const vector unsigned char vg##src##i =				\
+    vec_perm(vf##src##i, v_##src##A##i, permHH);			\
+  const vector unsigned char mask##src##i =				\
+    vec_perm((vector unsigned char)zero, (vector unsigned char)neg1, perms##src##i); \
+  const vector unsigned char vg2##src##i =				\
+    vec_perm(vg##src##i, vg##src##i, perms##src##i);			\
+  const vector unsigned char svA##src##i =				\
+    vec_sel(v_##src##A1##i, vg2##src##i, mask##src##i);			\
+  const vector unsigned char svB##src##i =				\
+    vec_sel(vg2##src##i, v_##src##A2##i, mask##src##i);			\
+  vec_st(svA##src##i, i * stride, src);					\
+  vec_st(svB##src##i, i * stride + 16, src)
+
+  PACK_AND_STORE(src, 0);
+  PACK_AND_STORE(src, 1);
+  PACK_AND_STORE(src, 2);
+  PACK_AND_STORE(src, 3);
+  PACK_AND_STORE(src, 4);
+  PACK_AND_STORE(src, 5);
+  PACK_AND_STORE(src, 6);
+  PACK_AND_STORE(src, 7);
+  PACK_AND_STORE(tempBlured, 0);
+  PACK_AND_STORE(tempBlured, 1);
+  PACK_AND_STORE(tempBlured, 2);
+  PACK_AND_STORE(tempBlured, 3);
+  PACK_AND_STORE(tempBlured, 4);
+  PACK_AND_STORE(tempBlured, 5);
+  PACK_AND_STORE(tempBlured, 6);
+  PACK_AND_STORE(tempBlured, 7);
+#undef PACK_AND_STORE
+}
+
+static inline void transpose_16x8_char_toPackedAlign_altivec(unsigned char* dst, unsigned char* src, int stride) {
+  const vector unsigned char zero = vec_splat_u8(0);
+
+#define LOAD_DOUBLE_LINE(i, j)						\
+  vector unsigned char perm1##i = vec_lvsl(i * stride, src);		\
+  vector unsigned char perm2##i = vec_lvsl(j * stride, src);		\
+  vector unsigned char srcA##i = vec_ld(i * stride, src);		\
+  vector unsigned char srcB##i = vec_ld(i * stride + 16, src);          \
+  vector unsigned char srcC##i = vec_ld(j * stride, src);		\
+  vector unsigned char srcD##i = vec_ld(j * stride+ 16, src);           \
+  vector unsigned char src##i = vec_perm(srcA##i, srcB##i, perm1##i);	\
+  vector unsigned char src##j = vec_perm(srcC##i, srcD##i, perm2##i)
+  
+  LOAD_DOUBLE_LINE(0, 1);
+  LOAD_DOUBLE_LINE(2, 3);
+  LOAD_DOUBLE_LINE(4, 5);
+  LOAD_DOUBLE_LINE(6, 7);
+#undef LOAD_DOUBLE_LINE
+
+  vector unsigned char tempA = vec_mergeh(src0, zero);
+  vector unsigned char tempB = vec_mergel(src0, zero);
+  vector unsigned char tempC = vec_mergeh(src1, zero);
+  vector unsigned char tempD = vec_mergel(src1, zero);
+  vector unsigned char tempE = vec_mergeh(src2, zero);
+  vector unsigned char tempF = vec_mergel(src2, zero);
+  vector unsigned char tempG = vec_mergeh(src3, zero);
+  vector unsigned char tempH = vec_mergel(src3, zero);
+  vector unsigned char tempI = vec_mergeh(src4, zero);
+  vector unsigned char tempJ = vec_mergel(src4, zero);
+  vector unsigned char tempK = vec_mergeh(src5, zero);
+  vector unsigned char tempL = vec_mergel(src5, zero);
+  vector unsigned char tempM = vec_mergeh(src6, zero);
+  vector unsigned char tempN = vec_mergel(src6, zero);
+  vector unsigned char tempO = vec_mergeh(src7, zero);
+  vector unsigned char tempP = vec_mergel(src7, zero);
+
+  vector unsigned char temp0 = vec_mergeh(tempA, tempI);
+  vector unsigned char temp1 = vec_mergel(tempA, tempI);
+  vector unsigned char temp2 = vec_mergeh(tempB, tempJ);
+  vector unsigned char temp3 = vec_mergel(tempB, tempJ);
+  vector unsigned char temp4 = vec_mergeh(tempC, tempK);
+  vector unsigned char temp5 = vec_mergel(tempC, tempK);
+  vector unsigned char temp6 = vec_mergeh(tempD, tempL);
+  vector unsigned char temp7 = vec_mergel(tempD, tempL);
+  vector unsigned char temp8 = vec_mergeh(tempE, tempM);
+  vector unsigned char temp9 = vec_mergel(tempE, tempM);
+  vector unsigned char temp10 = vec_mergeh(tempF, tempN);
+  vector unsigned char temp11 = vec_mergel(tempF, tempN);
+  vector unsigned char temp12 = vec_mergeh(tempG, tempO);
+  vector unsigned char temp13 = vec_mergel(tempG, tempO);
+  vector unsigned char temp14 = vec_mergeh(tempH, tempP);
+  vector unsigned char temp15 = vec_mergel(tempH, tempP);
+
+  tempA = vec_mergeh(temp0, temp8);
+  tempB = vec_mergel(temp0, temp8);
+  tempC = vec_mergeh(temp1, temp9);
+  tempD = vec_mergel(temp1, temp9);
+  tempE = vec_mergeh(temp2, temp10);
+  tempF = vec_mergel(temp2, temp10);
+  tempG = vec_mergeh(temp3, temp11);
+  tempH = vec_mergel(temp3, temp11);
+  tempI = vec_mergeh(temp4, temp12);
+  tempJ = vec_mergel(temp4, temp12);
+  tempK = vec_mergeh(temp5, temp13);
+  tempL = vec_mergel(temp5, temp13);
+  tempM = vec_mergeh(temp6, temp14);
+  tempN = vec_mergel(temp6, temp14);
+  tempO = vec_mergeh(temp7, temp15);
+  tempP = vec_mergel(temp7, temp15);
+
+  temp0 = vec_mergeh(tempA, tempI);
+  temp1 = vec_mergel(tempA, tempI);
+  temp2 = vec_mergeh(tempB, tempJ);
+  temp3 = vec_mergel(tempB, tempJ);
+  temp4 = vec_mergeh(tempC, tempK);
+  temp5 = vec_mergel(tempC, tempK);
+  temp6 = vec_mergeh(tempD, tempL);
+  temp7 = vec_mergel(tempD, tempL);
+  temp8 = vec_mergeh(tempE, tempM);
+  temp9 = vec_mergel(tempE, tempM);
+  temp10 = vec_mergeh(tempF, tempN);
+  temp11 = vec_mergel(tempF, tempN);
+  temp12 = vec_mergeh(tempG, tempO);
+  temp13 = vec_mergel(tempG, tempO);
+  temp14 = vec_mergeh(tempH, tempP);
+  temp15 = vec_mergel(tempH, tempP);
+
+  vec_st(temp0, 0, dst);
+  vec_st(temp1, 16, dst);
+  vec_st(temp2, 32, dst);
+  vec_st(temp3, 48, dst);
+  vec_st(temp4, 64, dst);
+  vec_st(temp5, 80, dst);
+  vec_st(temp6, 96, dst);
+  vec_st(temp7, 112, dst);
+  vec_st(temp8, 128, dst);
+  vec_st(temp9, 144, dst);
+  vec_st(temp10, 160, dst);
+  vec_st(temp11, 176, dst);
+  vec_st(temp12, 192, dst);
+  vec_st(temp13, 208, dst);
+  vec_st(temp14, 224, dst);
+  vec_st(temp15, 240, dst);
+}
+
+static inline void transpose_8x16_char_fromPackedAlign_altivec(unsigned char* dst, unsigned char* src, int stride) {
+  const vector unsigned char zero = vec_splat_u8(0);
+  const vector unsigned char magic_perm = (const vector unsigned char)
+    AVV(0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
+	0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F);
+  
+#define LOAD_DOUBLE_LINE(i, j)			    		\
+  vector unsigned char src##i = vec_ld(i * 16, src);		\
+  vector unsigned char src##j = vec_ld(j * 16, src)
+
+  LOAD_DOUBLE_LINE(0, 1);
+  LOAD_DOUBLE_LINE(2, 3);
+  LOAD_DOUBLE_LINE(4, 5);
+  LOAD_DOUBLE_LINE(6, 7);
+  LOAD_DOUBLE_LINE(8, 9);
+  LOAD_DOUBLE_LINE(10, 11);
+  LOAD_DOUBLE_LINE(12, 13);
+  LOAD_DOUBLE_LINE(14, 15);
+#undef LOAD_DOUBLE_LINE
+
+  vector unsigned char tempA = vec_mergeh(src0, src8);
+  vector unsigned char tempB;
+  vector unsigned char tempC = vec_mergeh(src1, src9);
+  vector unsigned char tempD;
+  vector unsigned char tempE = vec_mergeh(src2, src10);
+  vector unsigned char tempG = vec_mergeh(src3, src11);
+  vector unsigned char tempI = vec_mergeh(src4, src12);
+  vector unsigned char tempJ;
+  vector unsigned char tempK = vec_mergeh(src5, src13);
+  vector unsigned char tempL;
+  vector unsigned char tempM = vec_mergeh(src6, src14);
+  vector unsigned char tempO = vec_mergeh(src7, src15);
+
+  vector unsigned char temp0 = vec_mergeh(tempA, tempI);
+  vector unsigned char temp1 = vec_mergel(tempA, tempI);
+  vector unsigned char temp2;
+  vector unsigned char temp3;
+  vector unsigned char temp4 = vec_mergeh(tempC, tempK);
+  vector unsigned char temp5 = vec_mergel(tempC, tempK);
+  vector unsigned char temp6;
+  vector unsigned char temp7;
+  vector unsigned char temp8 = vec_mergeh(tempE, tempM);
+  vector unsigned char temp9 = vec_mergel(tempE, tempM);
+  vector unsigned char temp12 = vec_mergeh(tempG, tempO);
+  vector unsigned char temp13 = vec_mergel(tempG, tempO);
+
+  tempA = vec_mergeh(temp0, temp8);
+  tempB = vec_mergel(temp0, temp8);
+  tempC = vec_mergeh(temp1, temp9);
+  tempD = vec_mergel(temp1, temp9);
+  tempI = vec_mergeh(temp4, temp12);
+  tempJ = vec_mergel(temp4, temp12);
+  tempK = vec_mergeh(temp5, temp13);
+  tempL = vec_mergel(temp5, temp13);
+
+  temp0 = vec_mergeh(tempA, tempI);
+  temp1 = vec_mergel(tempA, tempI);
+  temp2 = vec_mergeh(tempB, tempJ);
+  temp3 = vec_mergel(tempB, tempJ);
+  temp4 = vec_mergeh(tempC, tempK);
+  temp5 = vec_mergel(tempC, tempK);
+  temp6 = vec_mergeh(tempD, tempL);
+  temp7 = vec_mergel(tempD, tempL);
+
+
+  const vector signed char neg1 = vec_splat_s8(-1);
+#define STORE_DOUBLE_LINE(i, j)						\
+  vector unsigned char dstA##i = vec_ld(i * stride, dst);		\
+  vector unsigned char dstB##i = vec_ld(i * stride + 16, dst);		\
+  vector unsigned char dstA##j = vec_ld(j * stride, dst);		\
+  vector unsigned char dstB##j = vec_ld(j * stride+ 16, dst);		\
+  vector unsigned char align##i = vec_lvsr(i * stride, dst);		\
+  vector unsigned char align##j = vec_lvsr(j * stride, dst);		\
+  vector unsigned char mask##i = vec_perm(zero, (vector unsigned char)neg1, align##i);	\
+  vector unsigned char mask##j = vec_perm(zero, (vector unsigned char)neg1, align##j);	\
+  vector unsigned char dstR##i = vec_perm(temp##i, temp##i, align##i);	\
+  vector unsigned char dstR##j = vec_perm(temp##j, temp##j, align##j);	\
+  vector unsigned char dstAF##i = vec_sel(dstA##i, dstR##i, mask##i);	\
+  vector unsigned char dstBF##i = vec_sel(dstR##i, dstB##i, mask##i);	\
+  vector unsigned char dstAF##j = vec_sel(dstA##j, dstR##j, mask##j);	\
+  vector unsigned char dstBF##j = vec_sel(dstR##j, dstB##j, mask##j);	\
+  vec_st(dstAF##i, i * stride, dst);					\
+  vec_st(dstBF##i, i * stride + 16, dst);				\
+  vec_st(dstAF##j, j * stride, dst);					\
+  vec_st(dstBF##j, j * stride + 16, dst)
+
+  STORE_DOUBLE_LINE(0,1);
+  STORE_DOUBLE_LINE(2,3);
+  STORE_DOUBLE_LINE(4,5);
+  STORE_DOUBLE_LINE(6,7);
+}
diff --git a/src/libffmpeg/libavcodec/libpostproc/postprocess_internal.h b/src/libffmpeg/libavcodec/libpostproc/postprocess_internal.h
index db50fa3b5..01d4679ad 100644
--- a/src/libffmpeg/libavcodec/libpostproc/postprocess_internal.h
+++ b/src/libffmpeg/libavcodec/libpostproc/postprocess_internal.h
@@ -37,9 +37,11 @@
 
 // Experimental vertical filters
 #define V_X1_FILTER	0x0200			// 512
+#define V_A_DEBLOCK	0x0400
 
 // Experimental horizontal filters
 #define H_X1_FILTER	0x2000			// 8192
+#define H_A_DEBLOCK	0x4000
 
 /// select between full y range (255-0) or standart one (234-16)
 #define FULL_Y_RANGE	0x8000			// 32768
@@ -158,3 +160,11 @@ typedef struct PPContext{
 } PPContext;
 
 
+static inline void linecpy(void *dest, void *src, int lines, int stride)
+{
+	if (stride > 0) {
+		memcpy(dest, src, lines*stride);
+	} else {
+		memcpy(dest+(lines-1)*stride, src+(lines-1)*stride, -lines*stride);
+	}
+}
diff --git a/src/libffmpeg/libavcodec/libpostproc/postprocess_template.c b/src/libffmpeg/libavcodec/libpostproc/postprocess_template.c
index 4e81bd556..d1307caca 100644
--- a/src/libffmpeg/libavcodec/libpostproc/postprocess_template.c
+++ b/src/libffmpeg/libavcodec/libpostproc/postprocess_template.c
@@ -22,15 +22,37 @@
  */
 
 
+#ifdef ARCH_X86_64
+#  define REGa  rax
+#  define REGc  rcx
+#  define REGd  rdx
+#  define REG_a  "rax"
+#  define REG_c  "rcx"
+#  define REG_d  "rdx"
+#  define REG_SP "rsp"
+#  define ALIGN_MASK "$0xFFFFFFFFFFFFFFF8"
+#else
+#  define REGa  eax
+#  define REGc  ecx
+#  define REGd  edx
+#  define REG_a  "eax"
+#  define REG_c  "ecx"
+#  define REG_d  "edx"
+#  define REG_SP "esp"
+#  define ALIGN_MASK "$0xFFFFFFF8"
+#endif
+
+
 #undef PAVGB
 #undef PMINUB
 #undef PMAXUB
 
 #ifdef HAVE_MMX2
-#define PAVGB(a,b) "pavgb " #a ", " #b " \n\t"
+#define REAL_PAVGB(a,b) "pavgb " #a ", " #b " \n\t"
 #elif defined (HAVE_3DNOW)
-#define PAVGB(a,b) "pavgusb " #a ", " #b " \n\t"
+#define REAL_PAVGB(a,b) "pavgusb " #a ", " #b " \n\t"
 #endif
+#define PAVGB(a,b)  REAL_PAVGB(a,b)
 
 #ifdef HAVE_MMX2
 #define PMINUB(a,b,t) "pminub " #a ", " #b " \n\t"
@@ -64,12 +86,12 @@ asm volatile(
                 );
                 
 asm volatile(
-		"leal (%2, %3), %%eax				\n\t"
+		"lea (%2, %3), %%"REG_a"			\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%1	eax	eax+%2	eax+2%2	%1+4%2	ecx	ecx+%2	ecx+2%2	%1+8%2	ecx+4%2
 
 		"movq (%2), %%mm0				\n\t"
-		"movq (%%eax), %%mm1				\n\t"
+		"movq (%%"REG_a"), %%mm1			\n\t"
                 "movq %%mm0, %%mm3				\n\t"
                 "movq %%mm0, %%mm4				\n\t"
                 PMAXUB(%%mm1, %%mm4)
@@ -78,7 +100,7 @@ asm volatile(
 		"paddb %%mm7, %%mm0				\n\t"
 		"pcmpgtb %%mm6, %%mm0				\n\t"
 
-		"movq (%%eax,%3), %%mm2				\n\t"
+		"movq (%%"REG_a",%3), %%mm2			\n\t"
                 PMAXUB(%%mm2, %%mm4)
                 PMINUB(%%mm2, %%mm3, %%mm5)
 		"psubb %%mm2, %%mm1				\n\t"
@@ -86,7 +108,7 @@ asm volatile(
 		"pcmpgtb %%mm6, %%mm1				\n\t"
 		"paddb %%mm1, %%mm0				\n\t"
 
-		"movq (%%eax, %3, 2), %%mm1			\n\t"
+		"movq (%%"REG_a", %3, 2), %%mm1			\n\t"
                 PMAXUB(%%mm1, %%mm4)
                 PMINUB(%%mm1, %%mm3, %%mm5)
 		"psubb %%mm1, %%mm2				\n\t"
@@ -94,7 +116,7 @@ asm volatile(
 		"pcmpgtb %%mm6, %%mm2				\n\t"
 		"paddb %%mm2, %%mm0				\n\t"
 		
-		"leal (%%eax, %3, 4), %%eax			\n\t"
+		"lea (%%"REG_a", %3, 4), %%"REG_a"		\n\t"
 
 		"movq (%2, %3, 4), %%mm2			\n\t"
                 PMAXUB(%%mm2, %%mm4)
@@ -104,7 +126,7 @@ asm volatile(
 		"pcmpgtb %%mm6, %%mm1				\n\t"
 		"paddb %%mm1, %%mm0				\n\t"
 
-		"movq (%%eax), %%mm1				\n\t"
+		"movq (%%"REG_a"), %%mm1			\n\t"
                 PMAXUB(%%mm1, %%mm4)
                 PMINUB(%%mm1, %%mm3, %%mm5)
 		"psubb %%mm1, %%mm2				\n\t"
@@ -112,7 +134,7 @@ asm volatile(
 		"pcmpgtb %%mm6, %%mm2				\n\t"
 		"paddb %%mm2, %%mm0				\n\t"
 
-		"movq (%%eax, %3), %%mm2			\n\t"
+		"movq (%%"REG_a", %3), %%mm2			\n\t"
                 PMAXUB(%%mm2, %%mm4)
                 PMINUB(%%mm2, %%mm3, %%mm5)
 		"psubb %%mm2, %%mm1				\n\t"
@@ -120,7 +142,7 @@ asm volatile(
 		"pcmpgtb %%mm6, %%mm1				\n\t"
 		"paddb %%mm1, %%mm0				\n\t"
 
-		"movq (%%eax, %3, 2), %%mm1			\n\t"
+		"movq (%%"REG_a", %3, 2), %%mm1			\n\t"
                 PMAXUB(%%mm1, %%mm4)
                 PMINUB(%%mm1, %%mm3, %%mm5)
 		"psubb %%mm1, %%mm2				\n\t"
@@ -152,8 +174,8 @@ asm volatile(
 		"movd %%mm4, %1					\n\t"
 
 		: "=r" (numEq), "=r" (dcOk)
-		: "r" (src), "r" (stride), "m" (c->pQPb)
-		: "%eax"
+		: "r" (src), "r" ((long)stride), "m" (c->pQPb)
+		: "%"REG_a
 		);
 
 	numEq= (-numEq) &0xFF;
@@ -194,10 +216,10 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
 		"por %%mm2, %%mm6				\n\t"// First Line to Filter
 
 		"movq (%0, %1, 8), %%mm5			\n\t"
-		"leal (%0, %1, 4), %%eax			\n\t"
-		"leal (%0, %1, 8), %%ecx			\n\t"
-		"subl %1, %%ecx					\n\t"
-		"addl %1, %0					\n\t" // %0 points to line 1 not 0
+		"lea (%0, %1, 4), %%"REG_a"			\n\t"
+		"lea (%0, %1, 8), %%"REG_c"			\n\t"
+		"sub %1, %%"REG_c"				\n\t"
+		"add %1, %0					\n\t" // %0 points to line 1 not 0
 		"movq (%0, %1, 8), %%mm7			\n\t"
 		"movq %%mm5, %%mm1				\n\t"
 		"movq %%mm7, %%mm2				\n\t"
@@ -225,7 +247,7 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
 
 		"movq (%0, %1, 4), %%mm2			\n\t" //     1
 		"movq %%mm2, %%mm5				\n\t" //     1
-		PAVGB((%%eax), %%mm2)				      //    11	/2
+		PAVGB((%%REGa), %%mm2)				      //    11	/2
 		PAVGB((%0, %1, 2), %%mm2)			      //   211	/4
 		"movq %%mm2, %%mm3				\n\t" //   211	/4
 		"movq (%0), %%mm4				\n\t" // 1
@@ -237,15 +259,15 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
 		PAVGB(%%mm6, %%mm0)				      //1 1	/2
 		"movq %%mm4, %%mm3				\n\t" // 1
 		PAVGB((%0,%1,2), %%mm3)				      // 1 1	/2
-		PAVGB((%%eax,%1,2), %%mm5)			      //     11	/2
-		PAVGB((%%eax), %%mm5)				      //    211 /4
+		PAVGB((%%REGa,%1,2), %%mm5)			      //     11	/2
+		PAVGB((%%REGa), %%mm5)				      //    211 /4
 		PAVGB(%%mm5, %%mm3)				      // 2 2211 /8
 		PAVGB(%%mm0, %%mm3)				      //4242211 /16
 		"movq %%mm3, (%0,%1)				\n\t" //  X
 		// mm1=2 mm2=3(211) mm4=1 mm5=4(211) mm6=0 mm7=9
 		PAVGB(%%mm4, %%mm6)				      //11	/2
-		"movq (%%ecx), %%mm0				\n\t" //       1
-		PAVGB((%%eax, %1, 2), %%mm0)			      //      11/2
+		"movq (%%"REG_c"), %%mm0			\n\t" //       1
+		PAVGB((%%REGa, %1, 2), %%mm0)			      //      11/2
 		"movq %%mm0, %%mm3				\n\t" //      11/2
 		PAVGB(%%mm1, %%mm0)				      //  2   11/4
 		PAVGB(%%mm6, %%mm0)				      //222   11/8
@@ -253,17 +275,17 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
 		"movq (%0, %1, 2), %%mm2			\n\t" //   1
 		"movq %%mm0, (%0, %1, 2)			\n\t" //   X
 		// mm1=2 mm2=3 mm3=6(11) mm4=1 mm5=4(211) mm6=0(11) mm7=9
-		"movq (%%eax, %1, 4), %%mm0			\n\t" //        1
-		PAVGB((%%ecx), %%mm0)				      //       11	/2
+		"movq (%%"REG_a", %1, 4), %%mm0			\n\t" //        1
+		PAVGB((%%REGc), %%mm0)				      //       11	/2
 		PAVGB(%%mm0, %%mm6)				      //11     11	/4
 		PAVGB(%%mm1, %%mm4)				      // 11		/2
 		PAVGB(%%mm2, %%mm1)				      //  11		/2
 		PAVGB(%%mm1, %%mm6)				      //1122   11	/8
 		PAVGB(%%mm5, %%mm6)				      //112242211	/16
-		"movq (%%eax), %%mm5				\n\t" //    1
-		"movq %%mm6, (%%eax)				\n\t" //    X
+		"movq (%%"REG_a"), %%mm5			\n\t" //    1
+		"movq %%mm6, (%%"REG_a")			\n\t" //    X
 		// mm0=7(11) mm1=2(11) mm2=3 mm3=6(11) mm4=1(11) mm5=4 mm7=9
-		"movq (%%eax, %1, 4), %%mm6			\n\t" //        1
+		"movq (%%"REG_a", %1, 4), %%mm6			\n\t" //        1
 		PAVGB(%%mm7, %%mm6)				      //        11	/2
 		PAVGB(%%mm4, %%mm6)				      // 11     11	/4
 		PAVGB(%%mm3, %%mm6)				      // 11   2211	/8
@@ -276,29 +298,29 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
 		PAVGB(%%mm7, %%mm1)				      //  11     2	/4
 		PAVGB(%%mm4, %%mm5)				      //    11		/2
 		PAVGB(%%mm5, %%mm0)				      //    11 11	/4
-		"movq (%%eax, %1, 2), %%mm6			\n\t" //      1
+		"movq (%%"REG_a", %1, 2), %%mm6			\n\t" //      1
 		PAVGB(%%mm6, %%mm1)				      //  11  4  2	/8
 		PAVGB(%%mm0, %%mm1)				      //  11224222	/16
-		"movq %%mm1, (%%eax, %1, 2)			\n\t" //      X
+		"movq %%mm1, (%%"REG_a", %1, 2)			\n\t" //      X
 		// mm2=3(112) mm3=6(11) mm4=5 mm5=4(11) mm6=6 mm7=9
-		PAVGB((%%ecx), %%mm2)				      //   112 4	/8
-		"movq (%%eax, %1, 4), %%mm0			\n\t" //        1
+		PAVGB((%%REGc), %%mm2)				      //   112 4	/8
+		"movq (%%"REG_a", %1, 4), %%mm0			\n\t" //        1
 		PAVGB(%%mm0, %%mm6)				      //      1 1	/2
 		PAVGB(%%mm7, %%mm6)				      //      1 12	/4
 		PAVGB(%%mm2, %%mm6)				      //   1122424	/4
-		"movq %%mm6, (%%ecx)				\n\t" //       X
+		"movq %%mm6, (%%"REG_c")			\n\t" //       X
 		// mm0=8 mm3=6(11) mm4=5 mm5=4(11) mm7=9
 		PAVGB(%%mm7, %%mm5)				      //    11   2	/4
 		PAVGB(%%mm7, %%mm5)				      //    11   6	/8
 
 		PAVGB(%%mm3, %%mm0)				      //      112	/4
 		PAVGB(%%mm0, %%mm5)				      //    112246	/16
-		"movq %%mm5, (%%eax, %1, 4)			\n\t" //        X
-		"subl %1, %0					\n\t"
+		"movq %%mm5, (%%"REG_a", %1, 4)			\n\t" //        X
+		"sub %1, %0					\n\t"
 
 		:
-		: "r" (src), "r" (stride), "m" (c->pQPb)
-		: "%eax", "%ecx"
+		: "r" (src), "r" ((long)stride), "m" (c->pQPb)
+		: "%"REG_a, "%"REG_c
 	);
 #else
 	const int l1= stride;
@@ -317,25 +339,26 @@ static inline void RENAME(doVertLowPass)(uint8_t *src, int stride, PPContext *c)
 		const int first= ABS(src[0] - src[l1]) < c->QP ? src[0] : src[l1];
 		const int last= ABS(src[l8] - src[l9]) < c->QP ? src[l9] : src[l8];
 
-		int sums[9];
-		sums[0] = first + src[l1];
-		sums[1] = src[l1] + src[l2];
-		sums[2] = src[l2] + src[l3];
-		sums[3] = src[l3] + src[l4];
-		sums[4] = src[l4] + src[l5];
-		sums[5] = src[l5] + src[l6];
-		sums[6] = src[l6] + src[l7];
-		sums[7] = src[l7] + src[l8];
-		sums[8] = src[l8] + last;
-
-		src[l1]= ((sums[0]<<2) + ((first + sums[2])<<1) + sums[4] + 8)>>4;
-		src[l2]= ((src[l2]<<2) + ((first + sums[0] + sums[3])<<1) + sums[5] + 8)>>4;
-		src[l3]= ((src[l3]<<2) + ((first + sums[1] + sums[4])<<1) + sums[6] + 8)>>4;
-		src[l4]= ((src[l4]<<2) + ((sums[2] + sums[5])<<1) + sums[0] + sums[7] + 8)>>4;
-		src[l5]= ((src[l5]<<2) + ((sums[3] + sums[6])<<1) + sums[1] + sums[8] + 8)>>4;
-		src[l6]= ((src[l6]<<2) + ((last + sums[7] + sums[4])<<1) + sums[2] + 8)>>4;
-		src[l7]= (((last + src[l7])<<2) + ((src[l8] + sums[5])<<1) + sums[3] + 8)>>4;
-		src[l8]= ((sums[8]<<2) + ((last + sums[6])<<1) + sums[4] + 8)>>4;
+		int sums[10];
+		sums[0] = 4*first + src[l1] + src[l2] + src[l3] + 4;
+		sums[1] = sums[0] - first  + src[l4];
+		sums[2] = sums[1] - first  + src[l5];
+		sums[3] = sums[2] - first  + src[l6];
+		sums[4] = sums[3] - first  + src[l7];
+		sums[5] = sums[4] - src[l1] + src[l8];
+		sums[6] = sums[5] - src[l2] + last;
+		sums[7] = sums[6] - src[l3] + last;
+		sums[8] = sums[7] - src[l4] + last;
+		sums[9] = sums[8] - src[l5] + last;
+
+		src[l1]= (sums[0] + sums[2] + 2*src[l1])>>4;
+		src[l2]= (sums[1] + sums[3] + 2*src[l2])>>4;
+		src[l3]= (sums[2] + sums[4] + 2*src[l3])>>4;
+		src[l4]= (sums[3] + sums[5] + 2*src[l4])>>4;
+		src[l5]= (sums[4] + sums[6] + 2*src[l5])>>4;
+		src[l6]= (sums[5] + sums[7] + 2*src[l6])>>4;
+		src[l7]= (sums[6] + sums[8] + 2*src[l7])>>4;
+		src[l8]= (sums[7] + sums[9] + 2*src[l8])>>4;
 
 		src++;
 	}
@@ -363,8 +386,8 @@ static inline void RENAME(vertRK1Filter)(uint8_t *src, int stride, int QP)
 	asm volatile(
 		"pxor %%mm7, %%mm7				\n\t" // 0
 		"movq "MANGLE(b80)", %%mm6			\n\t" // MIN_SIGNED_BYTE
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%ecx			\n\t"
+		"leal (%0, %1), %%"REG_a"			\n\t"
+		"leal (%%"REG_a", %1, 4), %%"REG_c"		\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	ecx	ecx+%1	ecx+2%1	%0+8%1	ecx+4%1
 		"movq "MANGLE(pQPb)", %%mm0			\n\t" // QP,..., QP
@@ -374,7 +397,7 @@ static inline void RENAME(vertRK1Filter)(uint8_t *src, int stride, int QP)
 		"pand "MANGLE(b3F)", %%mm0			\n\t" // QP/4,..., QP/4
 		"paddusb %%mm1, %%mm0				\n\t" // QP*1.25 ...
 		"movq (%0, %1, 4), %%mm2			\n\t" // line 4
-		"movq (%%ecx), %%mm3				\n\t" // line 5
+		"movq (%%"REG_c"), %%mm3				\n\t" // line 5
 		"movq %%mm2, %%mm4				\n\t" // line 4
 		"pcmpeqb %%mm5, %%mm5				\n\t" // -1
 		"pxor %%mm2, %%mm5				\n\t" // -line 4 - 1
@@ -392,32 +415,32 @@ static inline void RENAME(vertRK1Filter)(uint8_t *src, int stride, int QP)
 //		"psubb %%mm6, %%mm2				\n\t"
 		"movq %%mm2, (%0,%1, 4)				\n\t"
 
-		"movq (%%ecx), %%mm2				\n\t"
+		"movq (%%"REG_c"), %%mm2				\n\t"
 //		"paddb %%mm6, %%mm2				\n\t" // line 5 + 0x80
 		"psubb %%mm5, %%mm2				\n\t"
 //		"psubb %%mm6, %%mm2				\n\t"
-		"movq %%mm2, (%%ecx)				\n\t"
+		"movq %%mm2, (%%"REG_c")				\n\t"
 
 		"paddb %%mm6, %%mm5				\n\t"
 		"psrlw $2, %%mm5				\n\t"
 		"pand "MANGLE(b3F)", %%mm5			\n\t"
 		"psubb "MANGLE(b20)", %%mm5			\n\t" // (l5-l4)/8
 
-		"movq (%%eax, %1, 2), %%mm2			\n\t"
+		"movq (%%"REG_a", %1, 2), %%mm2			\n\t"
 		"paddb %%mm6, %%mm2				\n\t" // line 3 + 0x80
 		"paddsb %%mm5, %%mm2				\n\t"
 		"psubb %%mm6, %%mm2				\n\t"
-		"movq %%mm2, (%%eax, %1, 2)			\n\t"
+		"movq %%mm2, (%%"REG_a", %1, 2)			\n\t"
 
-		"movq (%%ecx, %1), %%mm2			\n\t"
+		"movq (%%"REG_c", %1), %%mm2			\n\t"
 		"paddb %%mm6, %%mm2				\n\t" // line 6 + 0x80
 		"psubsb %%mm5, %%mm2				\n\t"
 		"psubb %%mm6, %%mm2				\n\t"
-		"movq %%mm2, (%%ecx, %1)			\n\t"
+		"movq %%mm2, (%%"REG_c", %1)			\n\t"
 
 		:
-		: "r" (src), "r" (stride)
-		: "%eax", "%ecx"
+		: "r" (src), "r" ((long)stride)
+		: "%"REG_a, "%"REG_c
 	);
 #else
  	const int l1= stride;
@@ -463,18 +486,18 @@ static inline void RENAME(vertX1Filter)(uint8_t *src, int stride, PPContext *co)
 
 	asm volatile(
 		"pxor %%mm7, %%mm7				\n\t" // 0
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%ecx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_c"		\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	ecx	ecx+%1	ecx+2%1	%0+8%1	ecx+4%1
-		"movq (%%eax, %1, 2), %%mm0			\n\t" // line 3
+		"movq (%%"REG_a", %1, 2), %%mm0			\n\t" // line 3
 		"movq (%0, %1, 4), %%mm1			\n\t" // line 4
 		"movq %%mm1, %%mm2				\n\t" // line 4
 		"psubusb %%mm0, %%mm1				\n\t"
 		"psubusb %%mm2, %%mm0				\n\t"
 		"por %%mm1, %%mm0				\n\t" // |l2 - l3|
-		"movq (%%ecx), %%mm3				\n\t" // line 5
-		"movq (%%ecx, %1), %%mm4			\n\t" // line 6
+		"movq (%%"REG_c"), %%mm3				\n\t" // line 5
+		"movq (%%"REG_c", %1), %%mm4			\n\t" // line 6
 		"movq %%mm3, %%mm5				\n\t" // line 5
 		"psubusb %%mm4, %%mm3				\n\t"
 		"psubusb %%mm5, %%mm4				\n\t"
@@ -506,43 +529,43 @@ static inline void RENAME(vertX1Filter)(uint8_t *src, int stride, PPContext *co)
 		"pxor %%mm2, %%mm0				\n\t"
 		"movq %%mm0, (%0, %1, 4)			\n\t" // line 4
 
-		"movq (%%ecx), %%mm0				\n\t" // line 5
+		"movq (%%"REG_c"), %%mm0			\n\t" // line 5
 		"pxor %%mm2, %%mm0				\n\t" //(l4 - l5) <= 0 ? -l5-1 : l5
 		"paddusb %%mm3, %%mm0				\n\t"
 		"pxor %%mm2, %%mm0				\n\t"
-		"movq %%mm0, (%%ecx)				\n\t" // line 5
+		"movq %%mm0, (%%"REG_c")			\n\t" // line 5
 
 		PAVGB(%%mm7, %%mm1)				      // d/4
 
-		"movq (%%eax, %1, 2), %%mm0			\n\t" // line 3
+		"movq (%%"REG_a", %1, 2), %%mm0			\n\t" // line 3
 		"pxor %%mm2, %%mm0				\n\t" //(l4 - l5) <= 0 ? -l4-1 : l4
 		"psubusb %%mm1, %%mm0				\n\t"
 		"pxor %%mm2, %%mm0				\n\t"
-		"movq %%mm0, (%%eax, %1, 2)			\n\t" // line 3
+		"movq %%mm0, (%%"REG_a", %1, 2)			\n\t" // line 3
 
-		"movq (%%ecx, %1), %%mm0			\n\t" // line 6
+		"movq (%%"REG_c", %1), %%mm0			\n\t" // line 6
 		"pxor %%mm2, %%mm0				\n\t" //(l4 - l5) <= 0 ? -l5-1 : l5
 		"paddusb %%mm1, %%mm0				\n\t"
 		"pxor %%mm2, %%mm0				\n\t"
-		"movq %%mm0, (%%ecx, %1)			\n\t" // line 6
+		"movq %%mm0, (%%"REG_c", %1)			\n\t" // line 6
 
 		PAVGB(%%mm7, %%mm1)				      // d/8
 
-		"movq (%%eax, %1), %%mm0			\n\t" // line 2
+		"movq (%%"REG_a", %1), %%mm0			\n\t" // line 2
 		"pxor %%mm2, %%mm0				\n\t" //(l4 - l5) <= 0 ? -l2-1 : l2
 		"psubusb %%mm1, %%mm0				\n\t"
 		"pxor %%mm2, %%mm0				\n\t"
-		"movq %%mm0, (%%eax, %1)			\n\t" // line 2
+		"movq %%mm0, (%%"REG_a", %1)			\n\t" // line 2
 
-		"movq (%%ecx, %1, 2), %%mm0			\n\t" // line 7
+		"movq (%%"REG_c", %1, 2), %%mm0			\n\t" // line 7
 		"pxor %%mm2, %%mm0				\n\t" //(l4 - l5) <= 0 ? -l7-1 : l7
 		"paddusb %%mm1, %%mm0				\n\t"
 		"pxor %%mm2, %%mm0				\n\t"
-		"movq %%mm0, (%%ecx, %1, 2)			\n\t" // line 7
+		"movq %%mm0, (%%"REG_c", %1, 2)			\n\t" // line 7
 
 		:
-		: "r" (src), "r" (stride), "m" (co->pQPb)
-		: "%eax", "%ecx"
+		: "r" (src), "r" ((long)stride), "m" (co->pQPb)
+		: "%"REG_a, "%"REG_c
 	);
 #else
 
@@ -607,8 +630,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 
 #if 0 //sligtly more accurate and slightly slower
 		"pxor %%mm7, %%mm7				\n\t" // 0
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%ecx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_c"		\n\t"
 //	0	1	2	3	4	5	6	7
 //	%0	%0+%1	%0+2%1	eax+2%1	%0+4%1	eax+4%1	ecx+%1	ecx+2%1
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	ecx	ecx+%1	ecx+2%1
@@ -621,8 +644,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		PAVGB(%%mm1, %%mm0)				      // ~(l2 + 2l0)/4
 		PAVGB(%%mm2, %%mm0)				      // ~(5l2 + 2l0)/8
 
-		"movq (%%eax), %%mm1				\n\t" // l1
-		"movq (%%eax, %1, 2), %%mm3			\n\t" // l3
+		"movq (%%"REG_a"), %%mm1			\n\t" // l1
+		"movq (%%"REG_a", %1, 2), %%mm3			\n\t" // l3
 		"movq %%mm1, %%mm4				\n\t" // l1
 		PAVGB(%%mm7, %%mm1)				      // ~l1/2
 		PAVGB(%%mm3, %%mm1)				      // ~(l1 + 2l3)/4
@@ -640,7 +663,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		PAVGB(%%mm2, %%mm0)				      // ~(l4 + 2l2)/4
 		PAVGB(%%mm4, %%mm0)				      // ~(5l4 + 2l2)/8
 
-		"movq (%%ecx), %%mm2				\n\t" // l5
+		"movq (%%"REG_c"), %%mm2			\n\t" // l5
 		"movq %%mm3, %%mm5				\n\t" // l3
 		PAVGB(%%mm7, %%mm3)				      // ~l3/2
 		PAVGB(%%mm2, %%mm3)				      // ~(l3 + 2l5)/4
@@ -653,13 +676,13 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		"pcmpeqb %%mm7, %%mm0				\n\t" // SIGN(2l2 - 5l3 + 5l4 - 2l5)
 // mm0= SIGN(menergy), mm1= |lenergy|, mm2= l5, mm3= |menergy|, mm4=l4, mm5= l3, mm7=0
 
-		"movq (%%ecx, %1), %%mm6			\n\t" // l6
+		"movq (%%"REG_c", %1), %%mm6			\n\t" // l6
 		"movq %%mm6, %%mm5				\n\t" // l6
 		PAVGB(%%mm7, %%mm6)				      // ~l6/2
 		PAVGB(%%mm4, %%mm6)				      // ~(l6 + 2l4)/4
 		PAVGB(%%mm5, %%mm6)				      // ~(5l6 + 2l4)/8
 
-		"movq (%%ecx, %1, 2), %%mm5			\n\t" // l7
+		"movq (%%"REG_c", %1, 2), %%mm5			\n\t" // l7
 		"movq %%mm2, %%mm4				\n\t" // l5
 		PAVGB(%%mm7, %%mm2)				      // ~l5/2
 		PAVGB(%%mm5, %%mm2)				      // ~(l5 + 2l7)/4
@@ -686,7 +709,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		"paddusb %%mm1, %%mm3				\n\t"
 //		"paddusb "MANGLE(b01)", %%mm3			\n\t"
 
-		"movq (%%eax, %1, 2), %%mm6			\n\t" //l3
+		"movq (%%"REG_a", %1, 2), %%mm6			\n\t" //l3
 		"movq (%0, %1, 4), %%mm5			\n\t" //l4
 		"movq (%0, %1, 4), %%mm4			\n\t" //l4
 		"psubusb %%mm6, %%mm5				\n\t"
@@ -700,7 +723,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		"psubusb "MANGLE(b01)", %%mm3			\n\t"
 		PAVGB(%%mm7, %%mm3)
 
-		"movq (%%eax, %1, 2), %%mm0			\n\t"
+		"movq (%%"REG_a", %1, 2), %%mm0			\n\t"
 		"movq (%0, %1, 4), %%mm2			\n\t"
 		"pxor %%mm6, %%mm0				\n\t"
 		"pxor %%mm6, %%mm2				\n\t"
@@ -708,36 +731,36 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		"paddb %%mm3, %%mm2				\n\t"
 		"pxor %%mm6, %%mm0				\n\t"
 		"pxor %%mm6, %%mm2				\n\t"
-		"movq %%mm0, (%%eax, %1, 2)			\n\t"
+		"movq %%mm0, (%%"REG_a", %1, 2)			\n\t"
 		"movq %%mm2, (%0, %1, 4)			\n\t"
 #endif
 
-		"leal (%0, %1), %%eax				\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
 		"pcmpeqb %%mm6, %%mm6				\n\t" // -1
 //	0	1	2	3	4	5	6	7
 //	%0	%0+%1	%0+2%1	eax+2%1	%0+4%1	eax+4%1	ecx+%1	ecx+2%1
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	ecx	ecx+%1	ecx+2%1
 
 
-		"movq (%%eax, %1, 2), %%mm1			\n\t" // l3
+		"movq (%%"REG_a", %1, 2), %%mm1			\n\t" // l3
 		"movq (%0, %1, 4), %%mm0			\n\t" // l4
 		"pxor %%mm6, %%mm1				\n\t" // -l3-1
 		PAVGB(%%mm1, %%mm0)				      // -q+128 = (l4-l3+256)/2
 // mm1=-l3-1, mm0=128-q
 
-		"movq (%%eax, %1, 4), %%mm2			\n\t" // l5
-		"movq (%%eax, %1), %%mm3			\n\t" // l2
+		"movq (%%"REG_a", %1, 4), %%mm2			\n\t" // l5
+		"movq (%%"REG_a", %1), %%mm3			\n\t" // l2
 		"pxor %%mm6, %%mm2				\n\t" // -l5-1
 		"movq %%mm2, %%mm5				\n\t" // -l5-1
 		"movq "MANGLE(b80)", %%mm4			\n\t" // 128
-		"leal (%%eax, %1, 4), %%ecx			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_c"		\n\t"
 		PAVGB(%%mm3, %%mm2)				      // (l2-l5+256)/2
 		PAVGB(%%mm0, %%mm4)				      // ~(l4-l3)/4 + 128
 		PAVGB(%%mm2, %%mm4)				      // ~(l2-l5)/4 +(l4-l3)/8 + 128
 		PAVGB(%%mm0, %%mm4)				      // ~(l2-l5)/8 +5(l4-l3)/16 + 128
 // mm1=-l3-1, mm0=128-q, mm3=l2, mm4=menergy/16 + 128, mm5= -l5-1
 
-		"movq (%%eax), %%mm2				\n\t" // l1
+		"movq (%%"REG_a"), %%mm2			\n\t" // l1
 		"pxor %%mm6, %%mm2				\n\t" // -l1-1
 		PAVGB(%%mm3, %%mm2)				      // (l2-l1+256)/2
 		PAVGB((%0), %%mm1)				      // (l0-l3+256)/2
@@ -747,8 +770,8 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		PAVGB(%%mm2, %%mm3)				      // ~(l0-l3)/8 +5(l2-l1)/16 + 128
 // mm0=128-q, mm3=lenergy/16 + 128, mm4= menergy/16 + 128, mm5= -l5-1
 
-		PAVGB((%%ecx, %1), %%mm5)			      // (l6-l5+256)/2
-		"movq (%%ecx, %1, 2), %%mm1			\n\t" // l7
+		PAVGB((%%REGc, %1), %%mm5)			      // (l6-l5+256)/2
+		"movq (%%"REG_c", %1, 2), %%mm1			\n\t" // l7
 		"pxor %%mm6, %%mm1				\n\t" // -l7-1
 		PAVGB((%0, %1, 4), %%mm1)			      // (l4-l7+256)/2
 		"movq "MANGLE(b80)", %%mm2			\n\t" // 128
@@ -797,7 +820,7 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		"pxor %%mm1, %%mm7				\n\t" // SIGN(d*q)
 
 		"pand %%mm7, %%mm4				\n\t"
-		"movq (%%eax, %1, 2), %%mm0			\n\t"
+		"movq (%%"REG_a", %1, 2), %%mm0			\n\t"
 		"movq (%0, %1, 4), %%mm2			\n\t"
 		"pxor %%mm1, %%mm0				\n\t"
 		"pxor %%mm1, %%mm2				\n\t"
@@ -805,12 +828,12 @@ static inline void RENAME(doVertDefFilter)(uint8_t src[], int stride, PPContext
 		"psubb %%mm4, %%mm2				\n\t"
 		"pxor %%mm1, %%mm0				\n\t"
 		"pxor %%mm1, %%mm2				\n\t"
-		"movq %%mm0, (%%eax, %1, 2)			\n\t"
+		"movq %%mm0, (%%"REG_a", %1, 2)			\n\t"
 		"movq %%mm2, (%0, %1, 4)			\n\t"
 
 		:
-		: "r" (src), "r" (stride), "m" (c->pQPb)
-		: "%eax", "%ecx"
+		: "r" (src), "r" ((long)stride), "m" (c->pQPb)
+		: "%"REG_a, "%"REG_c
 	);
 
 /*
@@ -881,8 +904,8 @@ src-=8;
 	src+= stride*4;
 	asm volatile(
 		"pxor %%mm7, %%mm7				\n\t"
-		"leal -40(%%esp), %%ecx				\n\t" // make space for 4 8-byte vars
-		"andl $0xFFFFFFF8, %%ecx			\n\t" // align
+		"lea -40(%%"REG_SP"), %%"REG_c"			\n\t" // make space for 4 8-byte vars
+		"and "ALIGN_MASK", %%"REG_c"			\n\t" // align
 //	0	1	2	3	4	5	6	7
 //	%0	%0+%1	%0+2%1	eax+2%1	%0+4%1	eax+4%1	edx+%1	edx+2%1
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1
@@ -893,12 +916,12 @@ src-=8;
 		"punpckhbw %%mm7, %%mm1				\n\t" // high part of line 0
 
 		"movq (%0, %1), %%mm2				\n\t"
-		"leal (%0, %1, 2), %%eax			\n\t"
+		"lea (%0, %1, 2), %%"REG_a"			\n\t"
 		"movq %%mm2, %%mm3				\n\t"
 		"punpcklbw %%mm7, %%mm2				\n\t" // low part of line 1
 		"punpckhbw %%mm7, %%mm3				\n\t" // high part of line 1
 
-		"movq (%%eax), %%mm4				\n\t"
+		"movq (%%"REG_a"), %%mm4			\n\t"
 		"movq %%mm4, %%mm5				\n\t"
 		"punpcklbw %%mm7, %%mm4				\n\t" // low part of line 2
 		"punpckhbw %%mm7, %%mm5				\n\t" // high part of line 2
@@ -915,7 +938,7 @@ src-=8;
 		"psubw %%mm2, %%mm0				\n\t" // 2L0 - 5L1 + 5L2
 		"psubw %%mm3, %%mm1				\n\t" // 2H0 - 5H1 + 5H2
 
-		"movq (%%eax, %1), %%mm2			\n\t"
+		"movq (%%"REG_a", %1), %%mm2			\n\t"
 		"movq %%mm2, %%mm3				\n\t"
 		"punpcklbw %%mm7, %%mm2				\n\t" // L3
 		"punpckhbw %%mm7, %%mm3				\n\t" // H3
@@ -924,24 +947,24 @@ src-=8;
 		"psubw %%mm3, %%mm1				\n\t" // 2H0 - 5H1 + 5H2 - H3
 		"psubw %%mm2, %%mm0				\n\t" // 2L0 - 5L1 + 5L2 - 2L3
 		"psubw %%mm3, %%mm1				\n\t" // 2H0 - 5H1 + 5H2 - 2H3
-		"movq %%mm0, (%%ecx)				\n\t" // 2L0 - 5L1 + 5L2 - 2L3
-		"movq %%mm1, 8(%%ecx)				\n\t" // 2H0 - 5H1 + 5H2 - 2H3
+		"movq %%mm0, (%%"REG_c")			\n\t" // 2L0 - 5L1 + 5L2 - 2L3
+		"movq %%mm1, 8(%%"REG_c")			\n\t" // 2H0 - 5H1 + 5H2 - 2H3
 
-		"movq (%%eax, %1, 2), %%mm0			\n\t"
+		"movq (%%"REG_a", %1, 2), %%mm0			\n\t"
 		"movq %%mm0, %%mm1				\n\t"
 		"punpcklbw %%mm7, %%mm0				\n\t" // L4
 		"punpckhbw %%mm7, %%mm1				\n\t" // H4
 
 		"psubw %%mm0, %%mm2				\n\t" // L3 - L4
 		"psubw %%mm1, %%mm3				\n\t" // H3 - H4
-		"movq %%mm2, 16(%%ecx)				\n\t" // L3 - L4
-		"movq %%mm3, 24(%%ecx)				\n\t" // H3 - H4
+		"movq %%mm2, 16(%%"REG_c")			\n\t" // L3 - L4
+		"movq %%mm3, 24(%%"REG_c")			\n\t" // H3 - H4
 		"paddw %%mm4, %%mm4				\n\t" // 2L2
 		"paddw %%mm5, %%mm5				\n\t" // 2H2
 		"psubw %%mm2, %%mm4				\n\t" // 2L2 - L3 + L4
 		"psubw %%mm3, %%mm5				\n\t" // 2H2 - H3 + H4
 
-		"leal (%%eax, %1), %0				\n\t"
+		"lea (%%"REG_a", %1), %0			\n\t"
 		"psllw $2, %%mm2				\n\t" // 4L3 - 4L4
 		"psllw $2, %%mm3				\n\t" // 4H3 - 4H4
 		"psubw %%mm2, %%mm4				\n\t" // 2L2 - 5L3 + 5L4
@@ -956,10 +979,10 @@ src-=8;
 		"psubw %%mm2, %%mm4				\n\t" // 2L2 - 5L3 + 5L4 - 2L5
 		"psubw %%mm3, %%mm5				\n\t" // 2H2 - 5H3 + 5H4 - 2H5
 
-		"movq (%%eax, %1, 4), %%mm6			\n\t"
+		"movq (%%"REG_a", %1, 4), %%mm6			\n\t"
 		"punpcklbw %%mm7, %%mm6				\n\t" // L6
 		"psubw %%mm6, %%mm2				\n\t" // L5 - L6
-		"movq (%%eax, %1, 4), %%mm6			\n\t"
+		"movq (%%"REG_a", %1, 4), %%mm6			\n\t"
 		"punpckhbw %%mm7, %%mm6				\n\t" // H6
 		"psubw %%mm6, %%mm3				\n\t" // H5 - H6
 
@@ -983,8 +1006,8 @@ src-=8;
 		"psubw %%mm2, %%mm0				\n\t" // 2L4 - 5L5 + 5L6 - 2L7
 		"psubw %%mm3, %%mm1				\n\t" // 2H4 - 5H5 + 5H6 - 2H7
 
-		"movq (%%ecx), %%mm2				\n\t" // 2L0 - 5L1 + 5L2 - 2L3
-		"movq 8(%%ecx), %%mm3				\n\t" // 2H0 - 5H1 + 5H2 - 2H3
+		"movq (%%"REG_c"), %%mm2			\n\t" // 2L0 - 5L1 + 5L2 - 2L3
+		"movq 8(%%"REG_c"), %%mm3			\n\t" // 2H0 - 5H1 + 5H2 - 2H3
 
 #ifdef HAVE_MMX2
 		"movq %%mm7, %%mm6				\n\t" // 0
@@ -1030,6 +1053,9 @@ src-=8;
 		"psubw %%mm6, %%mm1				\n\t"
 #endif
 
+		"movd %2, %%mm2					\n\t" // QP
+		"punpcklbw %%mm7, %%mm2				\n\t"
+
 		"movq %%mm7, %%mm6				\n\t" // 0
 		"pcmpgtw %%mm4, %%mm6				\n\t" // sign(2L2 - 5L3 + 5L4 - 2L5)
 		"pxor %%mm6, %%mm4				\n\t"
@@ -1038,7 +1064,6 @@ src-=8;
 		"pxor %%mm7, %%mm5				\n\t"
 		"psubw %%mm7, %%mm5				\n\t" // |2H2 - 5H3 + 5H4 - 2H5|
 // 100 opcodes
-		"movd %2, %%mm2					\n\t" // QP
 		"psllw $3, %%mm2				\n\t" // 8QP
 		"movq %%mm2, %%mm3				\n\t" // 8QP
 		"pcmpgtw %%mm4, %%mm2				\n\t"
@@ -1060,8 +1085,8 @@ src-=8;
 		"psrlw $6, %%mm4				\n\t"
 		"psrlw $6, %%mm5				\n\t"
 
-		"movq 16(%%ecx), %%mm0				\n\t" // L3 - L4
-		"movq 24(%%ecx), %%mm1				\n\t" // H3 - H4
+		"movq 16(%%"REG_c"), %%mm0			\n\t" // L3 - L4
+		"movq 24(%%"REG_c"), %%mm1			\n\t" // H3 - H4
 
 		"pxor %%mm2, %%mm2				\n\t"
 		"pxor %%mm3, %%mm3				\n\t"
@@ -1104,8 +1129,8 @@ src-=8;
 		"movq %%mm0, (%0, %1)				\n\t"
 
 		: "+r" (src)
-		: "r" (stride), "m" (c->pQPb)
-		: "%eax", "%ecx"
+		: "r" ((long)stride), "m" (c->pQPb)
+		: "%"REG_a, "%"REG_c
 	);
 #else
 	const int l1= stride;
@@ -1168,20 +1193,20 @@ static inline void RENAME(dering)(uint8_t src[], int stride, PPContext *c)
 		"packuswb %%mm0, %%mm0				\n\t"
 		"movq %%mm0, %3					\n\t"
 
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%edx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_d"		\n\t"
 		
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1
 
 #undef FIND_MIN_MAX
 #ifdef HAVE_MMX2
-#define FIND_MIN_MAX(addr)\
+#define REAL_FIND_MIN_MAX(addr)\
 		"movq " #addr ", %%mm0				\n\t"\
 		"pminub %%mm0, %%mm7				\n\t"\
 		"pmaxub %%mm0, %%mm6				\n\t"
 #else
-#define FIND_MIN_MAX(addr)\
+#define REAL_FIND_MIN_MAX(addr)\
 		"movq " #addr ", %%mm0				\n\t"\
 		"movq %%mm7, %%mm1				\n\t"\
 		"psubusb %%mm0, %%mm6				\n\t"\
@@ -1189,14 +1214,15 @@ static inline void RENAME(dering)(uint8_t src[], int stride, PPContext *c)
 		"psubusb %%mm0, %%mm1				\n\t"\
 		"psubb %%mm1, %%mm7				\n\t"
 #endif
+#define FIND_MIN_MAX(addr)  REAL_FIND_MIN_MAX(addr)
 
-FIND_MIN_MAX((%%eax))
-FIND_MIN_MAX((%%eax, %1))
-FIND_MIN_MAX((%%eax, %1, 2))
+FIND_MIN_MAX((%%REGa))
+FIND_MIN_MAX((%%REGa, %1))
+FIND_MIN_MAX((%%REGa, %1, 2))
 FIND_MIN_MAX((%0, %1, 4))
-FIND_MIN_MAX((%%edx))
-FIND_MIN_MAX((%%edx, %1))
-FIND_MIN_MAX((%%edx, %1, 2))
+FIND_MIN_MAX((%%REGd))
+FIND_MIN_MAX((%%REGd, %1))
+FIND_MIN_MAX((%%REGd, %1, 2))
 FIND_MIN_MAX((%0, %1, 8))
 
 		"movq %%mm7, %%mm4				\n\t"
@@ -1249,13 +1275,13 @@ FIND_MIN_MAX((%0, %1, 8))
 		"movd %%mm6, %%ecx				\n\t"
 		"cmpb "MANGLE(deringThreshold)", %%cl		\n\t"
 		" jb 1f						\n\t"
-		"leal -24(%%esp), %%ecx				\n\t"
-		"andl $0xFFFFFFF8, %%ecx			\n\t" 
+		"lea -24(%%"REG_SP"), %%"REG_c"			\n\t"
+		"and "ALIGN_MASK", %%"REG_c"			\n\t" 
 		PAVGB(%%mm0, %%mm7)				      // a=(max + min)/2
 		"punpcklbw %%mm7, %%mm7				\n\t"
 		"punpcklbw %%mm7, %%mm7				\n\t"
 		"punpcklbw %%mm7, %%mm7				\n\t"
-		"movq %%mm7, (%%ecx)				\n\t"
+		"movq %%mm7, (%%"REG_c")			\n\t"
 
 		"movq (%0), %%mm0				\n\t" // L10
 		"movq %%mm0, %%mm1				\n\t" // L10
@@ -1280,13 +1306,13 @@ FIND_MIN_MAX((%0, %1, 8))
 		"paddb %%mm2, %%mm0				\n\t"
 		"paddb %%mm3, %%mm0				\n\t"
 
-		"movq (%%eax), %%mm2				\n\t" // L11
+		"movq (%%"REG_a"), %%mm2			\n\t" // L11
 		"movq %%mm2, %%mm3				\n\t" // L11
 		"movq %%mm2, %%mm4				\n\t" // L11
 		"psllq $8, %%mm3				\n\t"
 		"psrlq $8, %%mm4				\n\t"
-		"movd -4(%%eax), %%mm5				\n\t"
-		"movd 8(%%eax), %%mm6				\n\t"
+		"movd -4(%%"REG_a"), %%mm5			\n\t"
+		"movd 8(%%"REG_a"), %%mm6			\n\t"
 		"psrlq $24, %%mm5				\n\t"
 		"psllq $56, %%mm6				\n\t"
 		"por %%mm5, %%mm3				\n\t" // L01
@@ -1303,7 +1329,7 @@ FIND_MIN_MAX((%0, %1, 8))
 		"paddb %%mm4, %%mm2				\n\t"
 		"paddb %%mm5, %%mm2				\n\t"
 // 0, 2, 3, 1
-#define DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1) \
+#define REAL_DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1) \
 		"movq " #src ", " #sx "				\n\t" /* src[0] */\
 		"movq " #sx ", " #lx "				\n\t" /* src[0] */\
 		"movq " #sx ", " #t0 "				\n\t" /* src[0] */\
@@ -1319,8 +1345,8 @@ FIND_MIN_MAX((%0, %1, 8))
 		PAVGB(t0, lx)				              /* (src[-1] + src[+1])/2 */\
 		PAVGB(sx, lx)				      /* (src[-1] + 2src[0] + src[+1])/4 */\
 		PAVGB(lx, pplx)					     \
-		"movq " #lx ", 8(%%ecx)				\n\t"\
-		"movq (%%ecx), " #lx "				\n\t"\
+		"movq " #lx ", 8(%%"REG_c")			\n\t"\
+		"movq (%%"REG_c"), " #lx "			\n\t"\
 		"psubusb " #lx ", " #t1 "			\n\t"\
 		"psubusb " #lx ", " #t0 "			\n\t"\
 		"psubusb " #lx ", " #sx "			\n\t"\
@@ -1347,8 +1373,10 @@ FIND_MIN_MAX((%0, %1, 8))
 		"pandn " #dst ", " #ppsx "			\n\t"\
 		"por " #pplx ", " #ppsx "			\n\t"\
 		"movq " #ppsx ", " #dst "			\n\t"\
-		"movq 8(%%ecx), " #lx "				\n\t"
+		"movq 8(%%"REG_c"), " #lx "			\n\t"
 
+#define DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1) \
+   REAL_DERING_CORE(dst,src,ppsx,psx,sx,pplx,plx,lx,t0,t1)
 /*
 0000000
 1111111
@@ -1365,18 +1393,18 @@ FIND_MIN_MAX((%0, %1, 8))
 
 */
 //DERING_CORE(dst,src                  ,ppsx ,psx  ,sx   ,pplx ,plx  ,lx   ,t0   ,t1)
-DERING_CORE((%%eax),(%%eax, %1)        ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
-DERING_CORE((%%eax, %1),(%%eax, %1, 2) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
-DERING_CORE((%%eax, %1, 2),(%0, %1, 4) ,%%mm4,%%mm0,%%mm2,%%mm5,%%mm1,%%mm3,%%mm6,%%mm7)
-DERING_CORE((%0, %1, 4),(%%edx)        ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
-DERING_CORE((%%edx),(%%edx, %1)        ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
-DERING_CORE((%%edx, %1), (%%edx, %1, 2),%%mm4,%%mm0,%%mm2,%%mm5,%%mm1,%%mm3,%%mm6,%%mm7)
-DERING_CORE((%%edx, %1, 2),(%0, %1, 8) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
-DERING_CORE((%0, %1, 8),(%%edx, %1, 4) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
+DERING_CORE((%%REGa),(%%REGa, %1)        ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
+DERING_CORE((%%REGa, %1),(%%REGa, %1, 2) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
+DERING_CORE((%%REGa, %1, 2),(%0, %1, 4) ,%%mm4,%%mm0,%%mm2,%%mm5,%%mm1,%%mm3,%%mm6,%%mm7)
+DERING_CORE((%0, %1, 4),(%%REGd)        ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
+DERING_CORE((%%REGd),(%%REGd, %1)        ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
+DERING_CORE((%%REGd, %1), (%%REGd, %1, 2),%%mm4,%%mm0,%%mm2,%%mm5,%%mm1,%%mm3,%%mm6,%%mm7)
+DERING_CORE((%%REGd, %1, 2),(%0, %1, 8) ,%%mm0,%%mm2,%%mm4,%%mm1,%%mm3,%%mm5,%%mm6,%%mm7)
+DERING_CORE((%0, %1, 8),(%%REGd, %1, 4) ,%%mm2,%%mm4,%%mm0,%%mm3,%%mm5,%%mm1,%%mm6,%%mm7)
 
 		"1:			\n\t"
-		: : "r" (src), "r" (stride), "m" (c->pQPb), "m"(c->pQPb2)
-		: "%eax", "%edx", "%ecx"
+		: : "r" (src), "r" ((long)stride), "m" (c->pQPb), "m"(c->pQPb2)
+		: "%"REG_a, "%"REG_d, "%"REG_c
 	);
 #else
 	int y;
@@ -1523,27 +1551,27 @@ static inline void RENAME(deInterlaceInterpolateLinear)(uint8_t src[], int strid
 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
 	src+= 4*stride;
 	asm volatile(
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%ecx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_c"		\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	ecx	ecx+%1	ecx+2%1	%0+8%1	ecx+4%1
 
 		"movq (%0), %%mm0				\n\t"
-		"movq (%%eax, %1), %%mm1			\n\t"
+		"movq (%%"REG_a", %1), %%mm1			\n\t"
 		PAVGB(%%mm1, %%mm0)
-		"movq %%mm0, (%%eax)				\n\t"
+		"movq %%mm0, (%%"REG_a")			\n\t"
 		"movq (%0, %1, 4), %%mm0			\n\t"
 		PAVGB(%%mm0, %%mm1)
-		"movq %%mm1, (%%eax, %1, 2)			\n\t"
-		"movq (%%ecx, %1), %%mm1			\n\t"
+		"movq %%mm1, (%%"REG_a", %1, 2)			\n\t"
+		"movq (%%"REG_c", %1), %%mm1			\n\t"
 		PAVGB(%%mm1, %%mm0)
-		"movq %%mm0, (%%ecx)				\n\t"
+		"movq %%mm0, (%%"REG_c")			\n\t"
 		"movq (%0, %1, 8), %%mm0			\n\t"
 		PAVGB(%%mm0, %%mm1)
-		"movq %%mm1, (%%ecx, %1, 2)			\n\t"
+		"movq %%mm1, (%%"REG_c", %1, 2)			\n\t"
 
-		: : "r" (src), "r" (stride)
-		: "%eax", "%ecx"
+		: : "r" (src), "r" ((long)stride)
+		: "%"REG_a, "%"REG_c
 	);
 #else
 	int a, b, x;
@@ -1576,15 +1604,15 @@ static inline void RENAME(deInterlaceInterpolateCubic)(uint8_t src[], int stride
 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
 	src+= stride*3;
 	asm volatile(
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%edx			\n\t"
-		"leal (%%edx, %1, 4), %%ecx			\n\t"
-		"addl %1, %%ecx					\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_d"		\n\t"
+		"lea (%%"REG_d", %1, 4), %%"REG_c"		\n\t"
+		"add %1, %%"REG_c"				\n\t"
 		"pxor %%mm7, %%mm7				\n\t"
 //	0	1	2	3	4	5	6	7	8	9	10
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1 ecx
 
-#define DEINT_CUBIC(a,b,c,d,e)\
+#define REAL_DEINT_CUBIC(a,b,c,d,e)\
 		"movq " #a ", %%mm0				\n\t"\
 		"movq " #b ", %%mm1				\n\t"\
 		"movq " #d ", %%mm2				\n\t"\
@@ -1605,14 +1633,15 @@ static inline void RENAME(deInterlaceInterpolateCubic)(uint8_t src[], int stride
 		"psubw %%mm2, %%mm3				\n\t"	/* H(9b + 9d - a - e)/16 */\
 		"packuswb %%mm3, %%mm1				\n\t"\
 		"movq %%mm1, " #c "				\n\t"
+#define DEINT_CUBIC(a,b,c,d,e)  REAL_DEINT_CUBIC(a,b,c,d,e)
 
-DEINT_CUBIC((%0), (%%eax, %1), (%%eax, %1, 2), (%0, %1, 4), (%%edx, %1))
-DEINT_CUBIC((%%eax, %1), (%0, %1, 4), (%%edx), (%%edx, %1), (%0, %1, 8))
-DEINT_CUBIC((%0, %1, 4), (%%edx, %1), (%%edx, %1, 2), (%0, %1, 8), (%%ecx))
-DEINT_CUBIC((%%edx, %1), (%0, %1, 8), (%%edx, %1, 4), (%%ecx), (%%ecx, %1, 2))
+DEINT_CUBIC((%0), (%%REGa, %1), (%%REGa, %1, 2), (%0, %1, 4), (%%REGd, %1))
+DEINT_CUBIC((%%REGa, %1), (%0, %1, 4), (%%REGd), (%%REGd, %1), (%0, %1, 8))
+DEINT_CUBIC((%0, %1, 4), (%%REGd, %1), (%%REGd, %1, 2), (%0, %1, 8), (%%REGc))
+DEINT_CUBIC((%%REGd, %1), (%0, %1, 8), (%%REGd, %1, 4), (%%REGc), (%%REGc, %1, 2))
 
-		: : "r" (src), "r" (stride)
-		: "%eax", "%edx", "ecx"
+		: : "r" (src), "r" ((long)stride)
+		: "%"REG_a, "%"REG_d, "%"REG_c
 	);
 #else
 	int x;
@@ -1640,14 +1669,14 @@ static inline void RENAME(deInterlaceFF)(uint8_t src[], int stride, uint8_t *tmp
 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
 	src+= stride*4;
 	asm volatile(
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%edx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_d"		\n\t"
 		"pxor %%mm7, %%mm7				\n\t"
 		"movq (%2), %%mm0				\n\t"
 //	0	1	2	3	4	5	6	7	8	9	10
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1 ecx
 
-#define DEINT_FF(a,b,c,d)\
+#define REAL_DEINT_FF(a,b,c,d)\
 		"movq " #a ", %%mm1				\n\t"\
 		"movq " #b ", %%mm2				\n\t"\
 		"movq " #c ", %%mm3				\n\t"\
@@ -1675,14 +1704,16 @@ static inline void RENAME(deInterlaceFF)(uint8_t src[], int stride, uint8_t *tmp
 		"packuswb %%mm4, %%mm1				\n\t"\
 		"movq %%mm1, " #b "				\n\t"\
 
-DEINT_FF((%0)       , (%%eax)       , (%%eax, %1), (%%eax, %1, 2))
-DEINT_FF((%%eax, %1), (%%eax, %1, 2), (%0, %1, 4), (%%edx)       )
-DEINT_FF((%0, %1, 4), (%%edx)       , (%%edx, %1), (%%edx, %1, 2))
-DEINT_FF((%%edx, %1), (%%edx, %1, 2), (%0, %1, 8), (%%edx, %1, 4))
+#define DEINT_FF(a,b,c,d)  REAL_DEINT_FF(a,b,c,d)
+
+DEINT_FF((%0)       ,  (%%REGa)       , (%%REGa, %1), (%%REGa, %1, 2))
+DEINT_FF((%%REGa, %1), (%%REGa, %1, 2), (%0, %1, 4),  (%%REGd)       )
+DEINT_FF((%0, %1, 4),  (%%REGd)       , (%%REGd, %1), (%%REGd, %1, 2))
+DEINT_FF((%%REGd, %1), (%%REGd, %1, 2), (%0, %1, 8),  (%%REGd, %1, 4))
 
 		"movq %%mm0, (%2)				\n\t"
-		: : "r" (src), "r" (stride), "r"(tmp)
-		: "%eax", "%edx"
+		: : "r" (src), "r" ((long)stride), "r"(tmp)
+		: "%"REG_a, "%"REG_d
 	);
 #else
 	int x;
@@ -1718,15 +1749,15 @@ static inline void RENAME(deInterlaceL5)(uint8_t src[], int stride, uint8_t *tmp
 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
 	src+= stride*4;
 	asm volatile(
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%edx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_d"		\n\t"
 		"pxor %%mm7, %%mm7				\n\t"
 		"movq (%2), %%mm0				\n\t"
 		"movq (%3), %%mm1				\n\t"
 //	0	1	2	3	4	5	6	7	8	9	10
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1 ecx
 
-#define DEINT_L5(t1,t2,a,b,c)\
+#define REAL_DEINT_L5(t1,t2,a,b,c)\
 		"movq " #a ", %%mm2				\n\t"\
 		"movq " #b ", %%mm3				\n\t"\
 		"movq " #c ", %%mm4				\n\t"\
@@ -1759,19 +1790,21 @@ static inline void RENAME(deInterlaceL5)(uint8_t src[], int stride, uint8_t *tmp
 		"packuswb %%mm5, %%mm2				\n\t"\
 		"movq %%mm2, " #a "				\n\t"\
 
-DEINT_L5(%%mm0, %%mm1, (%0)          , (%%eax)       , (%%eax, %1)   )
-DEINT_L5(%%mm1, %%mm0, (%%eax)       , (%%eax, %1)   , (%%eax, %1, 2))
-DEINT_L5(%%mm0, %%mm1, (%%eax, %1)   , (%%eax, %1, 2), (%0, %1, 4)   )
-DEINT_L5(%%mm1, %%mm0, (%%eax, %1, 2), (%0, %1, 4)   , (%%edx)       )
-DEINT_L5(%%mm0, %%mm1, (%0, %1, 4)   , (%%edx)       , (%%edx, %1)   )  
-DEINT_L5(%%mm1, %%mm0, (%%edx)       , (%%edx, %1)   , (%%edx, %1, 2))
-DEINT_L5(%%mm0, %%mm1, (%%edx, %1)   , (%%edx, %1, 2), (%0, %1, 8)   )
-DEINT_L5(%%mm1, %%mm0, (%%edx, %1, 2), (%0, %1, 8)   , (%%edx, %1, 4))
+#define DEINT_L5(t1,t2,a,b,c)  REAL_DEINT_L5(t1,t2,a,b,c)
+
+DEINT_L5(%%mm0, %%mm1, (%0)           , (%%REGa)       , (%%REGa, %1)   )
+DEINT_L5(%%mm1, %%mm0, (%%REGa)       , (%%REGa, %1)   , (%%REGa, %1, 2))
+DEINT_L5(%%mm0, %%mm1, (%%REGa, %1)   , (%%REGa, %1, 2), (%0, %1, 4)   )
+DEINT_L5(%%mm1, %%mm0, (%%REGa, %1, 2), (%0, %1, 4)    , (%%REGd)       )
+DEINT_L5(%%mm0, %%mm1, (%0, %1, 4)    , (%%REGd)       , (%%REGd, %1)   )  
+DEINT_L5(%%mm1, %%mm0, (%%REGd)       , (%%REGd, %1)   , (%%REGd, %1, 2))
+DEINT_L5(%%mm0, %%mm1, (%%REGd, %1)   , (%%REGd, %1, 2), (%0, %1, 8)   )
+DEINT_L5(%%mm1, %%mm0, (%%REGd, %1, 2), (%0, %1, 8)    , (%%REGd, %1, 4))
 
 		"movq %%mm0, (%2)				\n\t"
 		"movq %%mm1, (%3)				\n\t"
-		: : "r" (src), "r" (stride), "r"(tmp), "r"(tmp2)
-		: "%eax", "%edx"
+		: : "r" (src), "r" ((long)stride), "r"(tmp), "r"(tmp2)
+		: "%"REG_a, "%"REG_d
 	);
 #else
 	int x;
@@ -1818,49 +1851,49 @@ static inline void RENAME(deInterlaceBlendLinear)(uint8_t src[], int stride, uin
 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
 	src+= 4*stride;
 	asm volatile(
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%edx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_d"		\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1
 
 		"movq (%2), %%mm0				\n\t" // L0
-		"movq (%%eax), %%mm1				\n\t" // L2
+		"movq (%%"REG_a"), %%mm1			\n\t" // L2
 		PAVGB(%%mm1, %%mm0)				      // L0+L2
 		"movq (%0), %%mm2				\n\t" // L1
 		PAVGB(%%mm2, %%mm0)
 		"movq %%mm0, (%0)				\n\t"
-		"movq (%%eax, %1), %%mm0			\n\t" // L3
+		"movq (%%"REG_a", %1), %%mm0			\n\t" // L3
 		PAVGB(%%mm0, %%mm2)				      // L1+L3
 		PAVGB(%%mm1, %%mm2)				      // 2L2 + L1 + L3
-		"movq %%mm2, (%%eax)				\n\t"
-		"movq (%%eax, %1, 2), %%mm2			\n\t" // L4
+		"movq %%mm2, (%%"REG_a")			\n\t"
+		"movq (%%"REG_a", %1, 2), %%mm2			\n\t" // L4
 		PAVGB(%%mm2, %%mm1)				      // L2+L4
 		PAVGB(%%mm0, %%mm1)				      // 2L3 + L2 + L4
-		"movq %%mm1, (%%eax, %1)			\n\t"
+		"movq %%mm1, (%%"REG_a", %1)			\n\t"
 		"movq (%0, %1, 4), %%mm1			\n\t" // L5
 		PAVGB(%%mm1, %%mm0)				      // L3+L5
 		PAVGB(%%mm2, %%mm0)				      // 2L4 + L3 + L5
-		"movq %%mm0, (%%eax, %1, 2)			\n\t"
-		"movq (%%edx), %%mm0				\n\t" // L6
+		"movq %%mm0, (%%"REG_a", %1, 2)			\n\t"
+		"movq (%%"REG_d"), %%mm0			\n\t" // L6
 		PAVGB(%%mm0, %%mm2)				      // L4+L6
 		PAVGB(%%mm1, %%mm2)				      // 2L5 + L4 + L6
 		"movq %%mm2, (%0, %1, 4)			\n\t"
-		"movq (%%edx, %1), %%mm2			\n\t" // L7
+		"movq (%%"REG_d", %1), %%mm2			\n\t" // L7
 		PAVGB(%%mm2, %%mm1)				      // L5+L7
 		PAVGB(%%mm0, %%mm1)				      // 2L6 + L5 + L7
-		"movq %%mm1, (%%edx)				\n\t"
-		"movq (%%edx, %1, 2), %%mm1			\n\t" // L8
+		"movq %%mm1, (%%"REG_d")			\n\t"
+		"movq (%%"REG_d", %1, 2), %%mm1			\n\t" // L8
 		PAVGB(%%mm1, %%mm0)				      // L6+L8
 		PAVGB(%%mm2, %%mm0)				      // 2L7 + L6 + L8
-		"movq %%mm0, (%%edx, %1)			\n\t"
+		"movq %%mm0, (%%"REG_d", %1)			\n\t"
 		"movq (%0, %1, 8), %%mm0			\n\t" // L9
 		PAVGB(%%mm0, %%mm2)				      // L7+L9
 		PAVGB(%%mm1, %%mm2)				      // 2L8 + L7 + L9
-		"movq %%mm2, (%%edx, %1, 2)			\n\t"
+		"movq %%mm2, (%%"REG_d", %1, 2)			\n\t"
 		"movq %%mm1, (%2)				\n\t"
 
-		: : "r" (src), "r" (stride), "r" (tmp)
-		: "%eax", "%edx"
+		: : "r" (src), "r" ((long)stride), "r" (tmp)
+		: "%"REG_a, "%"REG_d
 	);
 #else
 	int a, b, c, x;
@@ -1920,62 +1953,62 @@ static inline void RENAME(deInterlaceMedian)(uint8_t src[], int stride)
 	src+= 4*stride;
 #ifdef HAVE_MMX2
 	asm volatile(
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%edx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_d"		\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1
 
 		"movq (%0), %%mm0				\n\t" //
-		"movq (%%eax, %1), %%mm2			\n\t" //
-		"movq (%%eax), %%mm1				\n\t" //
+		"movq (%%"REG_a", %1), %%mm2			\n\t" //
+		"movq (%%"REG_a"), %%mm1			\n\t" //
 		"movq %%mm0, %%mm3				\n\t"
 		"pmaxub %%mm1, %%mm0				\n\t" //
 		"pminub %%mm3, %%mm1				\n\t" //
 		"pmaxub %%mm2, %%mm1				\n\t" //
 		"pminub %%mm1, %%mm0				\n\t"
-		"movq %%mm0, (%%eax)				\n\t"
+		"movq %%mm0, (%%"REG_a")			\n\t"
 
 		"movq (%0, %1, 4), %%mm0			\n\t" //
-		"movq (%%eax, %1, 2), %%mm1			\n\t" //
+		"movq (%%"REG_a", %1, 2), %%mm1			\n\t" //
 		"movq %%mm2, %%mm3				\n\t"
 		"pmaxub %%mm1, %%mm2				\n\t" //
 		"pminub %%mm3, %%mm1				\n\t" //
 		"pmaxub %%mm0, %%mm1				\n\t" //
 		"pminub %%mm1, %%mm2				\n\t"
-		"movq %%mm2, (%%eax, %1, 2)			\n\t"
+		"movq %%mm2, (%%"REG_a", %1, 2)			\n\t"
 
-		"movq (%%edx), %%mm2				\n\t" //
-		"movq (%%edx, %1), %%mm1			\n\t" //
+		"movq (%%"REG_d"), %%mm2			\n\t" //
+		"movq (%%"REG_d", %1), %%mm1			\n\t" //
 		"movq %%mm2, %%mm3				\n\t"
 		"pmaxub %%mm0, %%mm2				\n\t" //
 		"pminub %%mm3, %%mm0				\n\t" //
 		"pmaxub %%mm1, %%mm0				\n\t" //
 		"pminub %%mm0, %%mm2				\n\t"
-		"movq %%mm2, (%%edx)				\n\t"
+		"movq %%mm2, (%%"REG_d")			\n\t"
 
-		"movq (%%edx, %1, 2), %%mm2			\n\t" //
+		"movq (%%"REG_d", %1, 2), %%mm2			\n\t" //
 		"movq (%0, %1, 8), %%mm0			\n\t" //
 		"movq %%mm2, %%mm3				\n\t"
 		"pmaxub %%mm0, %%mm2				\n\t" //
 		"pminub %%mm3, %%mm0				\n\t" //
 		"pmaxub %%mm1, %%mm0				\n\t" //
 		"pminub %%mm0, %%mm2				\n\t"
-		"movq %%mm2, (%%edx, %1, 2)			\n\t"
+		"movq %%mm2, (%%"REG_d", %1, 2)			\n\t"
 
 
-		: : "r" (src), "r" (stride)
-		: "%eax", "%edx"
+		: : "r" (src), "r" ((long)stride)
+		: "%"REG_a, "%"REG_d
 	);
 
 #else // MMX without MMX2
 	asm volatile(
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%edx			\n\t"
+		"lea (%0, %1), %%"REG_a"			\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_d"		\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1
 		"pxor %%mm7, %%mm7				\n\t"
 
-#define MEDIAN(a,b,c)\
+#define REAL_MEDIAN(a,b,c)\
 		"movq " #a ", %%mm0				\n\t"\
 		"movq " #b ", %%mm2				\n\t"\
 		"movq " #c ", %%mm1				\n\t"\
@@ -1998,14 +2031,15 @@ static inline void RENAME(deInterlaceMedian)(uint8_t src[], int stride)
 		"pand %%mm2, %%mm0				\n\t"\
 		"pand %%mm1, %%mm0				\n\t"\
 		"movq %%mm0, " #b "				\n\t"
+#define MEDIAN(a,b,c)  REAL_MEDIAN(a,b,c)
 
-MEDIAN((%0), (%%eax), (%%eax, %1))
-MEDIAN((%%eax, %1), (%%eax, %1, 2), (%0, %1, 4))
-MEDIAN((%0, %1, 4), (%%edx), (%%edx, %1))
-MEDIAN((%%edx, %1), (%%edx, %1, 2), (%0, %1, 8))
+MEDIAN((%0), (%%REGa), (%%REGa, %1))
+MEDIAN((%%REGa, %1), (%%REGa, %1, 2), (%0, %1, 4))
+MEDIAN((%0, %1, 4), (%%REGd), (%%REGd, %1))
+MEDIAN((%%REGd, %1), (%%REGd, %1, 2), (%0, %1, 8))
 
-		: : "r" (src), "r" (stride)
-		: "%eax", "%edx"
+		: : "r" (src), "r" ((long)stride)
+		: "%"REG_a, "%"REG_d
 	);
 #endif // MMX
 #else
@@ -2039,17 +2073,17 @@ MEDIAN((%%edx, %1), (%%edx, %1, 2), (%0, %1, 8))
 static inline void RENAME(transpose1)(uint8_t *dst1, uint8_t *dst2, uint8_t *src, int srcStride)
 {
 	asm(
-		"leal (%0, %1), %%eax				\n\t"
+		"lea (%0, %1), %%"REG_a"	\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1
 		"movq (%0), %%mm0		\n\t" // 12345678
-		"movq (%%eax), %%mm1		\n\t" // abcdefgh
+		"movq (%%"REG_a"), %%mm1	\n\t" // abcdefgh
 		"movq %%mm0, %%mm2		\n\t" // 12345678
 		"punpcklbw %%mm1, %%mm0		\n\t" // 1a2b3c4d
 		"punpckhbw %%mm1, %%mm2		\n\t" // 5e6f7g8h
 
-		"movq (%%eax, %1), %%mm1	\n\t"
-		"movq (%%eax, %1, 2), %%mm3	\n\t"
+		"movq (%%"REG_a", %1), %%mm1	\n\t"
+		"movq (%%"REG_a", %1, 2), %%mm3	\n\t"
 		"movq %%mm1, %%mm4		\n\t"
 		"punpcklbw %%mm3, %%mm1		\n\t"
 		"punpckhbw %%mm3, %%mm4		\n\t"
@@ -2076,16 +2110,16 @@ static inline void RENAME(transpose1)(uint8_t *dst1, uint8_t *dst2, uint8_t *src
 		"psrlq $32, %%mm1		\n\t"
 		"movd %%mm1, 112(%3)		\n\t"
 
-		"leal (%%eax, %1, 4), %%eax	\n\t"
+		"lea (%%"REG_a", %1, 4), %%"REG_a"	\n\t"
 		
 		"movq (%0, %1, 4), %%mm0	\n\t" // 12345678
-		"movq (%%eax), %%mm1		\n\t" // abcdefgh
+		"movq (%%"REG_a"), %%mm1	\n\t" // abcdefgh
 		"movq %%mm0, %%mm2		\n\t" // 12345678
 		"punpcklbw %%mm1, %%mm0		\n\t" // 1a2b3c4d
 		"punpckhbw %%mm1, %%mm2		\n\t" // 5e6f7g8h
 
-		"movq (%%eax, %1), %%mm1	\n\t"
-		"movq (%%eax, %1, 2), %%mm3	\n\t"
+		"movq (%%"REG_a", %1), %%mm1	\n\t"
+		"movq (%%"REG_a", %1, 2), %%mm3	\n\t"
 		"movq %%mm1, %%mm4		\n\t"
 		"punpcklbw %%mm3, %%mm1		\n\t"
 		"punpckhbw %%mm3, %%mm4		\n\t"
@@ -2113,8 +2147,8 @@ static inline void RENAME(transpose1)(uint8_t *dst1, uint8_t *dst2, uint8_t *src
 		"movd %%mm1, 116(%3)		\n\t"
 
 
-	:: "r" (src), "r" (srcStride), "r" (dst1), "r" (dst2)
-	: "%eax"
+	:: "r" (src), "r" ((long)srcStride), "r" (dst1), "r" (dst2)
+	: "%"REG_a
 	);
 }
 
@@ -2124,8 +2158,8 @@ static inline void RENAME(transpose1)(uint8_t *dst1, uint8_t *dst2, uint8_t *src
 static inline void RENAME(transpose2)(uint8_t *dst, int dstStride, uint8_t *src)
 {
 	asm(
-		"leal (%0, %1), %%eax				\n\t"
-		"leal (%%eax, %1, 4), %%edx			\n\t"
+		"lea (%0, %1), %%"REG_a"	\n\t"
+		"lea (%%"REG_a",%1,4), %%"REG_d"\n\t"
 //	0	1	2	3	4	5	6	7	8	9
 //	%0	eax	eax+%1	eax+2%1	%0+4%1	edx	edx+%1	edx+2%1	%0+8%1	edx+4%1
 		"movq (%2), %%mm0		\n\t" // 12345678
@@ -2149,16 +2183,16 @@ static inline void RENAME(transpose2)(uint8_t *dst, int dstStride, uint8_t *src)
 
 		"movd %%mm0, (%0)		\n\t"
 		"psrlq $32, %%mm0		\n\t"
-		"movd %%mm0, (%%eax)		\n\t"
-		"movd %%mm3, (%%eax, %1)	\n\t"
+		"movd %%mm0, (%%"REG_a")	\n\t"
+		"movd %%mm3, (%%"REG_a", %1)	\n\t"
 		"psrlq $32, %%mm3		\n\t"
-		"movd %%mm3, (%%eax, %1, 2)	\n\t"
+		"movd %%mm3, (%%"REG_a", %1, 2)	\n\t"
 		"movd %%mm2, (%0, %1, 4)	\n\t"
 		"psrlq $32, %%mm2		\n\t"
-		"movd %%mm2, (%%edx)		\n\t"
-		"movd %%mm1, (%%edx, %1)	\n\t"
+		"movd %%mm2, (%%"REG_d")	\n\t"
+		"movd %%mm1, (%%"REG_d", %1)	\n\t"
 		"psrlq $32, %%mm1		\n\t"
-		"movd %%mm1, (%%edx, %1, 2)	\n\t"
+		"movd %%mm1, (%%"REG_d", %1, 2)	\n\t"
 
 
 		"movq 64(%2), %%mm0		\n\t" // 12345678
@@ -2182,24 +2216,25 @@ static inline void RENAME(transpose2)(uint8_t *dst, int dstStride, uint8_t *src)
 
 		"movd %%mm0, 4(%0)		\n\t"
 		"psrlq $32, %%mm0		\n\t"
-		"movd %%mm0, 4(%%eax)		\n\t"
-		"movd %%mm3, 4(%%eax, %1)	\n\t"
+		"movd %%mm0, 4(%%"REG_a")		\n\t"
+		"movd %%mm3, 4(%%"REG_a", %1)	\n\t"
 		"psrlq $32, %%mm3		\n\t"
-		"movd %%mm3, 4(%%eax, %1, 2)	\n\t"
+		"movd %%mm3, 4(%%"REG_a", %1, 2)	\n\t"
 		"movd %%mm2, 4(%0, %1, 4)	\n\t"
 		"psrlq $32, %%mm2		\n\t"
-		"movd %%mm2, 4(%%edx)		\n\t"
-		"movd %%mm1, 4(%%edx, %1)	\n\t"
+		"movd %%mm2, 4(%%"REG_d")		\n\t"
+		"movd %%mm1, 4(%%"REG_d", %1)	\n\t"
 		"psrlq $32, %%mm1		\n\t"
-		"movd %%mm1, 4(%%edx, %1, 2)	\n\t"
+		"movd %%mm1, 4(%%"REG_d", %1, 2)	\n\t"
 
-	:: "r" (dst), "r" (dstStride), "r" (src)
-	: "%eax", "%edx"
+	:: "r" (dst), "r" ((long)dstStride), "r" (src)
+	: "%"REG_a, "%"REG_d
 	);
 }
 #endif
-//static int test=0;
+//static long test=0;
 
+#ifndef HAVE_ALTIVEC
 static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
 				    uint8_t *tempBlured, uint32_t *tempBluredPast, int *maxNoise)
 {
@@ -2212,9 +2247,9 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
 //#define L1_DIFF //u should change the thresholds too if u try that one
 #if defined (HAVE_MMX2) || defined (HAVE_3DNOW)
 	asm volatile(
-		"leal (%2, %2, 2), %%eax			\n\t" // 3*stride
-		"leal (%2, %2, 4), %%edx			\n\t" // 5*stride
-		"leal (%%edx, %2, 2), %%ecx			\n\t" // 7*stride
+		"lea (%2, %2, 2), %%"REG_a"			\n\t" // 3*stride
+		"lea (%2, %2, 4), %%"REG_d"			\n\t" // 5*stride
+		"lea (%%"REG_d", %2, 2), %%"REG_c"		\n\t" // 7*stride
 //	0	1	2	3	4	5	6	7	8	9
 //	%x	%x+%2	%x+2%2	%x+eax	%x+4%2	%x+edx	%x+2eax	%x+ecx	%x+8%2
 //FIXME reorder?
@@ -2225,29 +2260,30 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
 		"psadbw (%1, %2), %%mm1				\n\t" // |L1-R1|
 		"movq (%0, %2, 2), %%mm2			\n\t" // L2
 		"psadbw (%1, %2, 2), %%mm2			\n\t" // |L2-R2|
-		"movq (%0, %%eax), %%mm3			\n\t" // L3
-		"psadbw (%1, %%eax), %%mm3			\n\t" // |L3-R3|
+		"movq (%0, %%"REG_a"), %%mm3			\n\t" // L3
+		"psadbw (%1, %%"REG_a"), %%mm3			\n\t" // |L3-R3|
 
 		"movq (%0, %2, 4), %%mm4			\n\t" // L4
 		"paddw %%mm1, %%mm0				\n\t"
 		"psadbw (%1, %2, 4), %%mm4			\n\t" // |L4-R4|
-		"movq (%0, %%edx), %%mm5			\n\t" // L5
+		"movq (%0, %%"REG_d"), %%mm5			\n\t" // L5
 		"paddw %%mm2, %%mm0				\n\t"
-		"psadbw (%1, %%edx), %%mm5			\n\t" // |L5-R5|
-		"movq (%0, %%eax, 2), %%mm6			\n\t" // L6
+		"psadbw (%1, %%"REG_d"), %%mm5			\n\t" // |L5-R5|
+		"movq (%0, %%"REG_a", 2), %%mm6			\n\t" // L6
 		"paddw %%mm3, %%mm0				\n\t"
-		"psadbw (%1, %%eax, 2), %%mm6			\n\t" // |L6-R6|
-		"movq (%0, %%ecx), %%mm7			\n\t" // L7
+		"psadbw (%1, %%"REG_a", 2), %%mm6		\n\t" // |L6-R6|
+		"movq (%0, %%"REG_c"), %%mm7			\n\t" // L7
 		"paddw %%mm4, %%mm0				\n\t"
-		"psadbw (%1, %%ecx), %%mm7			\n\t" // |L7-R7|
+		"psadbw (%1, %%"REG_c"), %%mm7			\n\t" // |L7-R7|
 		"paddw %%mm5, %%mm6				\n\t"
 		"paddw %%mm7, %%mm6				\n\t"
 		"paddw %%mm6, %%mm0				\n\t"
-#elif defined (FAST_L2_DIFF)
+#else
+#if defined (FAST_L2_DIFF)
 		"pcmpeqb %%mm7, %%mm7				\n\t"
 		"movq "MANGLE(b80)", %%mm6			\n\t"
 		"pxor %%mm0, %%mm0				\n\t"
-#define L2_DIFF_CORE(a, b)\
+#define REAL_L2_DIFF_CORE(a, b)\
 		"movq " #a ", %%mm5				\n\t"\
 		"movq " #b ", %%mm2				\n\t"\
 		"pxor %%mm7, %%mm2				\n\t"\
@@ -2261,19 +2297,10 @@ static inline void RENAME(tempNoiseReducer)(uint8_t *src, int stride,
 		"psrld $14, %%mm5				\n\t"\
 		"paddd %%mm5, %%mm0				\n\t"
 
-L2_DIFF_CORE((%0), (%1))
-L2_DIFF_CORE((%0, %2), (%1, %2))
-L2_DIFF_CORE((%0, %2, 2), (%1, %2, 2))
-L2_DIFF_CORE((%0, %%eax), (%1, %%eax))
-L2_DIFF_CORE((%0, %2, 4), (%1, %2, 4))
-L2_DIFF_CORE((%0, %%edx), (%1, %%edx))
-L2_DIFF_CORE((%0, %%eax,2), (%1, %%eax,2))
-L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
-
 #else
 		"pxor %%mm7, %%mm7				\n\t"
 		"pxor %%mm0, %%mm0				\n\t"
-#define L2_DIFF_CORE(a, b)\
+#define REAL_L2_DIFF_CORE(a, b)\
 		"movq " #a ", %%mm5				\n\t"\
 		"movq " #b ", %%mm2				\n\t"\
 		"movq %%mm5, %%mm1				\n\t"\
@@ -2289,14 +2316,18 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
 		"paddd %%mm1, %%mm5				\n\t"\
 		"paddd %%mm5, %%mm0				\n\t"
 
+#endif
+
+#define L2_DIFF_CORE(a, b)  REAL_L2_DIFF_CORE(a, b)
+
 L2_DIFF_CORE((%0), (%1))
 L2_DIFF_CORE((%0, %2), (%1, %2))
 L2_DIFF_CORE((%0, %2, 2), (%1, %2, 2))
-L2_DIFF_CORE((%0, %%eax), (%1, %%eax))
+L2_DIFF_CORE((%0, %%REGa), (%1, %%REGa))
 L2_DIFF_CORE((%0, %2, 4), (%1, %2, 4))
-L2_DIFF_CORE((%0, %%edx), (%1, %%edx))
-L2_DIFF_CORE((%0, %%eax,2), (%1, %%eax,2))
-L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
+L2_DIFF_CORE((%0, %%REGd), (%1, %%REGd))
+L2_DIFF_CORE((%0, %%REGa,2), (%1, %%REGa,2))
+L2_DIFF_CORE((%0, %%REGc), (%1, %%REGc))
 
 #endif
 
@@ -2305,94 +2336,94 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
 		"paddd %%mm0, %%mm4				\n\t"
 		"movd %%mm4, %%ecx				\n\t"
 		"shll $2, %%ecx					\n\t"
-		"movl %3, %%edx					\n\t"
-		"addl -4(%%edx), %%ecx				\n\t"
-		"addl 4(%%edx), %%ecx				\n\t"
-		"addl -1024(%%edx), %%ecx			\n\t"
+		"mov %3, %%"REG_d"				\n\t"
+		"addl -4(%%"REG_d"), %%ecx			\n\t"
+		"addl 4(%%"REG_d"), %%ecx			\n\t"
+		"addl -1024(%%"REG_d"), %%ecx			\n\t"
 		"addl $4, %%ecx					\n\t"
-		"addl 1024(%%edx), %%ecx			\n\t"
+		"addl 1024(%%"REG_d"), %%ecx			\n\t"
 		"shrl $3, %%ecx					\n\t"
-		"movl %%ecx, (%%edx)				\n\t"
+		"movl %%ecx, (%%"REG_d")			\n\t"
 
-//		"movl %3, %%ecx					\n\t"
-//		"movl %%ecx, test				\n\t"
+//		"mov %3, %%"REG_c"				\n\t"
+//		"mov %%"REG_c", test				\n\t"
 //		"jmp 4f \n\t"
-		"cmpl 512(%%edx), %%ecx				\n\t"
+		"cmpl 512(%%"REG_d"), %%ecx			\n\t"
 		" jb 2f						\n\t"
-		"cmpl 516(%%edx), %%ecx				\n\t"
+		"cmpl 516(%%"REG_d"), %%ecx			\n\t"
 		" jb 1f						\n\t"
 
-		"leal (%%eax, %2, 2), %%edx			\n\t" // 5*stride
-		"leal (%%edx, %2, 2), %%ecx			\n\t" // 7*stride
+		"lea (%%"REG_a", %2, 2), %%"REG_d"		\n\t" // 5*stride
+		"lea (%%"REG_d", %2, 2), %%"REG_c"		\n\t" // 7*stride
 		"movq (%0), %%mm0				\n\t" // L0
 		"movq (%0, %2), %%mm1				\n\t" // L1
 		"movq (%0, %2, 2), %%mm2			\n\t" // L2
-		"movq (%0, %%eax), %%mm3			\n\t" // L3
+		"movq (%0, %%"REG_a"), %%mm3			\n\t" // L3
 		"movq (%0, %2, 4), %%mm4			\n\t" // L4
-		"movq (%0, %%edx), %%mm5			\n\t" // L5
-		"movq (%0, %%eax, 2), %%mm6			\n\t" // L6
-		"movq (%0, %%ecx), %%mm7			\n\t" // L7
+		"movq (%0, %%"REG_d"), %%mm5			\n\t" // L5
+		"movq (%0, %%"REG_a", 2), %%mm6			\n\t" // L6
+		"movq (%0, %%"REG_c"), %%mm7			\n\t" // L7
 		"movq %%mm0, (%1)				\n\t" // L0
 		"movq %%mm1, (%1, %2)				\n\t" // L1
 		"movq %%mm2, (%1, %2, 2)			\n\t" // L2
-		"movq %%mm3, (%1, %%eax)			\n\t" // L3
+		"movq %%mm3, (%1, %%"REG_a")			\n\t" // L3
 		"movq %%mm4, (%1, %2, 4)			\n\t" // L4
-		"movq %%mm5, (%1, %%edx)			\n\t" // L5
-		"movq %%mm6, (%1, %%eax, 2)			\n\t" // L6
-		"movq %%mm7, (%1, %%ecx)			\n\t" // L7
+		"movq %%mm5, (%1, %%"REG_d")			\n\t" // L5
+		"movq %%mm6, (%1, %%"REG_a", 2)			\n\t" // L6
+		"movq %%mm7, (%1, %%"REG_c")			\n\t" // L7
 		"jmp 4f						\n\t"
 
 		"1:						\n\t"
-		"leal (%%eax, %2, 2), %%edx			\n\t" // 5*stride
-		"leal (%%edx, %2, 2), %%ecx			\n\t" // 7*stride
+		"lea (%%"REG_a", %2, 2), %%"REG_d"		\n\t" // 5*stride
+		"lea (%%"REG_d", %2, 2), %%"REG_c"		\n\t" // 7*stride
 		"movq (%0), %%mm0				\n\t" // L0
 		PAVGB((%1), %%mm0)				      // L0
 		"movq (%0, %2), %%mm1				\n\t" // L1
 		PAVGB((%1, %2), %%mm1)				      // L1
 		"movq (%0, %2, 2), %%mm2			\n\t" // L2
 		PAVGB((%1, %2, 2), %%mm2)			      // L2
-		"movq (%0, %%eax), %%mm3			\n\t" // L3
-		PAVGB((%1, %%eax), %%mm3)			      // L3
+		"movq (%0, %%"REG_a"), %%mm3			\n\t" // L3
+		PAVGB((%1, %%REGa), %%mm3)			      // L3
 		"movq (%0, %2, 4), %%mm4			\n\t" // L4
 		PAVGB((%1, %2, 4), %%mm4)			      // L4
-		"movq (%0, %%edx), %%mm5			\n\t" // L5
-		PAVGB((%1, %%edx), %%mm5)			      // L5
-		"movq (%0, %%eax, 2), %%mm6			\n\t" // L6
-		PAVGB((%1, %%eax, 2), %%mm6)			      // L6
-		"movq (%0, %%ecx), %%mm7			\n\t" // L7
-		PAVGB((%1, %%ecx), %%mm7)			      // L7
+		"movq (%0, %%"REG_d"), %%mm5			\n\t" // L5
+		PAVGB((%1, %%REGd), %%mm5)			      // L5
+		"movq (%0, %%"REG_a", 2), %%mm6			\n\t" // L6
+		PAVGB((%1, %%REGa, 2), %%mm6)			      // L6
+		"movq (%0, %%"REG_c"), %%mm7			\n\t" // L7
+		PAVGB((%1, %%REGc), %%mm7)			      // L7
 		"movq %%mm0, (%1)				\n\t" // R0
 		"movq %%mm1, (%1, %2)				\n\t" // R1
 		"movq %%mm2, (%1, %2, 2)			\n\t" // R2
-		"movq %%mm3, (%1, %%eax)			\n\t" // R3
+		"movq %%mm3, (%1, %%"REG_a")			\n\t" // R3
 		"movq %%mm4, (%1, %2, 4)			\n\t" // R4
-		"movq %%mm5, (%1, %%edx)			\n\t" // R5
-		"movq %%mm6, (%1, %%eax, 2)			\n\t" // R6
-		"movq %%mm7, (%1, %%ecx)			\n\t" // R7
+		"movq %%mm5, (%1, %%"REG_d")			\n\t" // R5
+		"movq %%mm6, (%1, %%"REG_a", 2)			\n\t" // R6
+		"movq %%mm7, (%1, %%"REG_c")			\n\t" // R7
 		"movq %%mm0, (%0)				\n\t" // L0
 		"movq %%mm1, (%0, %2)				\n\t" // L1
 		"movq %%mm2, (%0, %2, 2)			\n\t" // L2
-		"movq %%mm3, (%0, %%eax)			\n\t" // L3
+		"movq %%mm3, (%0, %%"REG_a")			\n\t" // L3
 		"movq %%mm4, (%0, %2, 4)			\n\t" // L4
-		"movq %%mm5, (%0, %%edx)			\n\t" // L5
-		"movq %%mm6, (%0, %%eax, 2)			\n\t" // L6
-		"movq %%mm7, (%0, %%ecx)			\n\t" // L7
+		"movq %%mm5, (%0, %%"REG_d")			\n\t" // L5
+		"movq %%mm6, (%0, %%"REG_a", 2)			\n\t" // L6
+		"movq %%mm7, (%0, %%"REG_c")			\n\t" // L7
 		"jmp 4f						\n\t"
 
 		"2:						\n\t"
-		"cmpl 508(%%edx), %%ecx				\n\t"
+		"cmpl 508(%%"REG_d"), %%ecx			\n\t"
 		" jb 3f						\n\t"
 
-		"leal (%%eax, %2, 2), %%edx			\n\t" // 5*stride
-		"leal (%%edx, %2, 2), %%ecx			\n\t" // 7*stride
+		"lea (%%"REG_a", %2, 2), %%"REG_d"		\n\t" // 5*stride
+		"lea (%%"REG_d", %2, 2), %%"REG_c"		\n\t" // 7*stride
 		"movq (%0), %%mm0				\n\t" // L0
 		"movq (%0, %2), %%mm1				\n\t" // L1
 		"movq (%0, %2, 2), %%mm2			\n\t" // L2
-		"movq (%0, %%eax), %%mm3			\n\t" // L3
+		"movq (%0, %%"REG_a"), %%mm3			\n\t" // L3
 		"movq (%1), %%mm4				\n\t" // R0
 		"movq (%1, %2), %%mm5				\n\t" // R1
 		"movq (%1, %2, 2), %%mm6			\n\t" // R2
-		"movq (%1, %%eax), %%mm7			\n\t" // R3
+		"movq (%1, %%"REG_a"), %%mm7			\n\t" // R3
 		PAVGB(%%mm4, %%mm0)
 		PAVGB(%%mm5, %%mm1)
 		PAVGB(%%mm6, %%mm2)
@@ -2404,20 +2435,20 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
 		"movq %%mm0, (%1)				\n\t" // R0
 		"movq %%mm1, (%1, %2)				\n\t" // R1
 		"movq %%mm2, (%1, %2, 2)			\n\t" // R2
-		"movq %%mm3, (%1, %%eax)			\n\t" // R3
+		"movq %%mm3, (%1, %%"REG_a")			\n\t" // R3
 		"movq %%mm0, (%0)				\n\t" // L0
 		"movq %%mm1, (%0, %2)				\n\t" // L1
 		"movq %%mm2, (%0, %2, 2)			\n\t" // L2
-		"movq %%mm3, (%0, %%eax)			\n\t" // L3
+		"movq %%mm3, (%0, %%"REG_a")			\n\t" // L3
 
 		"movq (%0, %2, 4), %%mm0			\n\t" // L4
-		"movq (%0, %%edx), %%mm1			\n\t" // L5
-		"movq (%0, %%eax, 2), %%mm2			\n\t" // L6
-		"movq (%0, %%ecx), %%mm3			\n\t" // L7
+		"movq (%0, %%"REG_d"), %%mm1			\n\t" // L5
+		"movq (%0, %%"REG_a", 2), %%mm2			\n\t" // L6
+		"movq (%0, %%"REG_c"), %%mm3			\n\t" // L7
 		"movq (%1, %2, 4), %%mm4			\n\t" // R4
-		"movq (%1, %%edx), %%mm5			\n\t" // R5
-		"movq (%1, %%eax, 2), %%mm6			\n\t" // R6
-		"movq (%1, %%ecx), %%mm7			\n\t" // R7
+		"movq (%1, %%"REG_d"), %%mm5			\n\t" // R5
+		"movq (%1, %%"REG_a", 2), %%mm6			\n\t" // R6
+		"movq (%1, %%"REG_c"), %%mm7			\n\t" // R7
 		PAVGB(%%mm4, %%mm0)
 		PAVGB(%%mm5, %%mm1)
 		PAVGB(%%mm6, %%mm2)
@@ -2427,26 +2458,26 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
 		PAVGB(%%mm6, %%mm2)
 		PAVGB(%%mm7, %%mm3)
 		"movq %%mm0, (%1, %2, 4)			\n\t" // R4
-		"movq %%mm1, (%1, %%edx)			\n\t" // R5
-		"movq %%mm2, (%1, %%eax, 2)			\n\t" // R6
-		"movq %%mm3, (%1, %%ecx)			\n\t" // R7
+		"movq %%mm1, (%1, %%"REG_d")			\n\t" // R5
+		"movq %%mm2, (%1, %%"REG_a", 2)			\n\t" // R6
+		"movq %%mm3, (%1, %%"REG_c")			\n\t" // R7
 		"movq %%mm0, (%0, %2, 4)			\n\t" // L4
-		"movq %%mm1, (%0, %%edx)			\n\t" // L5
-		"movq %%mm2, (%0, %%eax, 2)			\n\t" // L6
-		"movq %%mm3, (%0, %%ecx)			\n\t" // L7
+		"movq %%mm1, (%0, %%"REG_d")			\n\t" // L5
+		"movq %%mm2, (%0, %%"REG_a", 2)			\n\t" // L6
+		"movq %%mm3, (%0, %%"REG_c")			\n\t" // L7
 		"jmp 4f						\n\t"
 
 		"3:						\n\t"
-		"leal (%%eax, %2, 2), %%edx			\n\t" // 5*stride
-		"leal (%%edx, %2, 2), %%ecx			\n\t" // 7*stride
+		"lea (%%"REG_a", %2, 2), %%"REG_d"		\n\t" // 5*stride
+		"lea (%%"REG_d", %2, 2), %%"REG_c"		\n\t" // 7*stride
 		"movq (%0), %%mm0				\n\t" // L0
 		"movq (%0, %2), %%mm1				\n\t" // L1
 		"movq (%0, %2, 2), %%mm2			\n\t" // L2
-		"movq (%0, %%eax), %%mm3			\n\t" // L3
+		"movq (%0, %%"REG_a"), %%mm3			\n\t" // L3
 		"movq (%1), %%mm4				\n\t" // R0
 		"movq (%1, %2), %%mm5				\n\t" // R1
 		"movq (%1, %2, 2), %%mm6			\n\t" // R2
-		"movq (%1, %%eax), %%mm7			\n\t" // R3
+		"movq (%1, %%"REG_a"), %%mm7			\n\t" // R3
 		PAVGB(%%mm4, %%mm0)
 		PAVGB(%%mm5, %%mm1)
 		PAVGB(%%mm6, %%mm2)
@@ -2462,20 +2493,20 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
 		"movq %%mm0, (%1)				\n\t" // R0
 		"movq %%mm1, (%1, %2)				\n\t" // R1
 		"movq %%mm2, (%1, %2, 2)			\n\t" // R2
-		"movq %%mm3, (%1, %%eax)			\n\t" // R3
+		"movq %%mm3, (%1, %%"REG_a")			\n\t" // R3
 		"movq %%mm0, (%0)				\n\t" // L0
 		"movq %%mm1, (%0, %2)				\n\t" // L1
 		"movq %%mm2, (%0, %2, 2)			\n\t" // L2
-		"movq %%mm3, (%0, %%eax)			\n\t" // L3
+		"movq %%mm3, (%0, %%"REG_a")			\n\t" // L3
 
 		"movq (%0, %2, 4), %%mm0			\n\t" // L4
-		"movq (%0, %%edx), %%mm1			\n\t" // L5
-		"movq (%0, %%eax, 2), %%mm2			\n\t" // L6
-		"movq (%0, %%ecx), %%mm3			\n\t" // L7
+		"movq (%0, %%"REG_d"), %%mm1			\n\t" // L5
+		"movq (%0, %%"REG_a", 2), %%mm2			\n\t" // L6
+		"movq (%0, %%"REG_c"), %%mm3			\n\t" // L7
 		"movq (%1, %2, 4), %%mm4			\n\t" // R4
-		"movq (%1, %%edx), %%mm5			\n\t" // R5
-		"movq (%1, %%eax, 2), %%mm6			\n\t" // R6
-		"movq (%1, %%ecx), %%mm7			\n\t" // R7
+		"movq (%1, %%"REG_d"), %%mm5			\n\t" // R5
+		"movq (%1, %%"REG_a", 2), %%mm6			\n\t" // R6
+		"movq (%1, %%"REG_c"), %%mm7			\n\t" // R7
 		PAVGB(%%mm4, %%mm0)
 		PAVGB(%%mm5, %%mm1)
 		PAVGB(%%mm6, %%mm2)
@@ -2489,25 +2520,25 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
 		PAVGB(%%mm6, %%mm2)
 		PAVGB(%%mm7, %%mm3)
 		"movq %%mm0, (%1, %2, 4)			\n\t" // R4
-		"movq %%mm1, (%1, %%edx)			\n\t" // R5
-		"movq %%mm2, (%1, %%eax, 2)			\n\t" // R6
-		"movq %%mm3, (%1, %%ecx)			\n\t" // R7
+		"movq %%mm1, (%1, %%"REG_d")			\n\t" // R5
+		"movq %%mm2, (%1, %%"REG_a", 2)			\n\t" // R6
+		"movq %%mm3, (%1, %%"REG_c")			\n\t" // R7
 		"movq %%mm0, (%0, %2, 4)			\n\t" // L4
-		"movq %%mm1, (%0, %%edx)			\n\t" // L5
-		"movq %%mm2, (%0, %%eax, 2)			\n\t" // L6
-		"movq %%mm3, (%0, %%ecx)			\n\t" // L7
+		"movq %%mm1, (%0, %%"REG_d")			\n\t" // L5
+		"movq %%mm2, (%0, %%"REG_a", 2)			\n\t" // L6
+		"movq %%mm3, (%0, %%"REG_c")			\n\t" // L7
 
 		"4:						\n\t"
 
-		:: "r" (src), "r" (tempBlured), "r"(stride), "m" (tempBluredPast)
-		: "%eax", "%edx", "%ecx", "memory"
+		:: "r" (src), "r" (tempBlured), "r"((long)stride), "m" (tempBluredPast)
+		: "%"REG_a, "%"REG_d, "%"REG_c, "memory"
 		);
 //printf("%d\n", test);
 #else
 {
 	int y;
 	int d=0;
-	int sysd=0;
+//	int sysd=0;
 	int i;
 
 	for(y=0; y<8; y++)
@@ -2522,7 +2553,7 @@ L2_DIFF_CORE((%0, %%ecx), (%1, %%ecx))
 //			if(y==0 || y==7) d1+= d1>>1;
 //			d+= ABS(d1);
 			d+= d1*d1;
-			sysd+= d1;
+//			sysd+= d1;
 		}
 	}
 	i=d;
@@ -2608,6 +2639,543 @@ Switch between
 }
 #endif
 }
+#endif //HAVE_ALTIVEC
+
+#ifdef HAVE_MMX
+/**
+ * accurate deblock filter
+ */
+static always_inline void RENAME(do_a_deblock)(uint8_t *src, int step, int stride, PPContext *c){
+	int64_t dc_mask, eq_mask;
+	int64_t sums[10*8*2];
+	src+= step*3; // src points to begin of the 8x8 Block
+//START_TIMER
+asm volatile(
+		"movq %0, %%mm7					\n\t" 
+		"movq %1, %%mm6					\n\t" 
+                : : "m" (c->mmxDcOffset[c->nonBQP]),  "m" (c->mmxDcThreshold[c->nonBQP])
+                );
+                
+asm volatile(
+		"lea (%2, %3), %%"REG_a"			\n\t"
+//	0	1	2	3	4	5	6	7	8	9
+//	%1	eax	eax+%2	eax+2%2	%1+4%2	ecx	ecx+%2	ecx+2%2	%1+8%2	ecx+4%2
+
+		"movq (%2), %%mm0				\n\t"
+		"movq (%%"REG_a"), %%mm1			\n\t"
+                "movq %%mm1, %%mm3				\n\t"
+                "movq %%mm1, %%mm4				\n\t"
+		"psubb %%mm1, %%mm0				\n\t" // mm0 = differnece
+		"paddb %%mm7, %%mm0				\n\t"
+		"pcmpgtb %%mm6, %%mm0				\n\t"
+
+		"movq (%%"REG_a",%3), %%mm2			\n\t"
+                PMAXUB(%%mm2, %%mm4)
+                PMINUB(%%mm2, %%mm3, %%mm5)
+		"psubb %%mm2, %%mm1				\n\t"
+		"paddb %%mm7, %%mm1				\n\t"
+		"pcmpgtb %%mm6, %%mm1				\n\t"
+		"paddb %%mm1, %%mm0				\n\t"
+
+		"movq (%%"REG_a", %3, 2), %%mm1			\n\t"
+                PMAXUB(%%mm1, %%mm4)
+                PMINUB(%%mm1, %%mm3, %%mm5)
+		"psubb %%mm1, %%mm2				\n\t"
+		"paddb %%mm7, %%mm2				\n\t"
+		"pcmpgtb %%mm6, %%mm2				\n\t"
+		"paddb %%mm2, %%mm0				\n\t"
+		
+		"lea (%%"REG_a", %3, 4), %%"REG_a"		\n\t"
+
+		"movq (%2, %3, 4), %%mm2			\n\t"
+                PMAXUB(%%mm2, %%mm4)
+                PMINUB(%%mm2, %%mm3, %%mm5)
+		"psubb %%mm2, %%mm1				\n\t"
+		"paddb %%mm7, %%mm1				\n\t"
+		"pcmpgtb %%mm6, %%mm1				\n\t"
+		"paddb %%mm1, %%mm0				\n\t"
+
+		"movq (%%"REG_a"), %%mm1			\n\t"
+                PMAXUB(%%mm1, %%mm4)
+                PMINUB(%%mm1, %%mm3, %%mm5)
+		"psubb %%mm1, %%mm2				\n\t"
+		"paddb %%mm7, %%mm2				\n\t"
+		"pcmpgtb %%mm6, %%mm2				\n\t"
+		"paddb %%mm2, %%mm0				\n\t"
+
+		"movq (%%"REG_a", %3), %%mm2			\n\t"
+                PMAXUB(%%mm2, %%mm4)
+                PMINUB(%%mm2, %%mm3, %%mm5)
+		"psubb %%mm2, %%mm1				\n\t"
+		"paddb %%mm7, %%mm1				\n\t"
+		"pcmpgtb %%mm6, %%mm1				\n\t"
+		"paddb %%mm1, %%mm0				\n\t"
+
+		"movq (%%"REG_a", %3, 2), %%mm1			\n\t"
+                PMAXUB(%%mm1, %%mm4)
+                PMINUB(%%mm1, %%mm3, %%mm5)
+		"psubb %%mm1, %%mm2				\n\t"
+		"paddb %%mm7, %%mm2				\n\t"
+		"pcmpgtb %%mm6, %%mm2				\n\t"
+		"paddb %%mm2, %%mm0				\n\t"
+
+		"movq (%2, %3, 8), %%mm2			\n\t"
+                PMAXUB(%%mm2, %%mm4)
+                PMINUB(%%mm2, %%mm3, %%mm5)
+		"psubb %%mm2, %%mm1				\n\t"
+		"paddb %%mm7, %%mm1				\n\t"
+		"pcmpgtb %%mm6, %%mm1				\n\t"
+		"paddb %%mm1, %%mm0				\n\t"
+
+		"movq (%%"REG_a", %3, 4), %%mm1			\n\t"
+		"psubb %%mm1, %%mm2				\n\t"
+		"paddb %%mm7, %%mm2				\n\t"
+		"pcmpgtb %%mm6, %%mm2				\n\t"
+		"paddb %%mm2, %%mm0				\n\t"
+		"psubusb %%mm3, %%mm4				\n\t"
+
+		"pxor %%mm6, %%mm6				\n\t"
+                "movq %4, %%mm7					\n\t" // QP,..., QP
+		"paddusb %%mm7, %%mm7				\n\t" // 2QP ... 2QP
+		"psubusb %%mm4, %%mm7				\n\t" // Diff >=2QP -> 0
+		"pcmpeqb %%mm6, %%mm7				\n\t" // Diff < 2QP -> 0
+		"pcmpeqb %%mm6, %%mm7				\n\t" // Diff < 2QP -> 0
+		"movq %%mm7, %1					\n\t"
+
+		"movq %5, %%mm7					\n\t"
+		"punpcklbw %%mm7, %%mm7				\n\t"
+		"punpcklbw %%mm7, %%mm7				\n\t"
+		"punpcklbw %%mm7, %%mm7				\n\t"
+		"psubb %%mm0, %%mm6				\n\t"
+		"pcmpgtb %%mm7, %%mm6				\n\t"
+		"movq %%mm6, %0					\n\t"
+
+		: "=m" (eq_mask), "=m" (dc_mask)
+		: "r" (src), "r" ((long)step), "m" (c->pQPb), "m"(c->ppMode.flatnessThreshold)
+		: "%"REG_a
+		);
+
+	if(dc_mask & eq_mask){
+		long offset= -8*step;
+		int64_t *temp_sums= sums;
+
+		asm volatile(
+		"movq %2, %%mm0					\n\t"  // QP,..., QP
+		"pxor %%mm4, %%mm4				\n\t"
+
+		"movq (%0), %%mm6				\n\t"
+		"movq (%0, %1), %%mm5				\n\t"
+		"movq %%mm5, %%mm1				\n\t"
+		"movq %%mm6, %%mm2				\n\t"
+		"psubusb %%mm6, %%mm5				\n\t"
+		"psubusb %%mm1, %%mm2				\n\t"
+		"por %%mm5, %%mm2				\n\t" // ABS Diff of lines
+		"psubusb %%mm2, %%mm0				\n\t" // diff >= QP -> 0
+		"pcmpeqb %%mm4, %%mm0				\n\t" // diff >= QP -> FF
+
+		"pxor %%mm6, %%mm1				\n\t"
+		"pand %%mm0, %%mm1				\n\t"
+		"pxor %%mm1, %%mm6				\n\t"
+		// 0:QP  6:First
+
+		"movq (%0, %1, 8), %%mm5			\n\t"
+		"add %1, %0					\n\t" // %0 points to line 1 not 0
+		"movq (%0, %1, 8), %%mm7			\n\t"
+		"movq %%mm5, %%mm1				\n\t"
+		"movq %%mm7, %%mm2				\n\t"
+		"psubusb %%mm7, %%mm5				\n\t"
+		"psubusb %%mm1, %%mm2				\n\t"
+		"por %%mm5, %%mm2				\n\t" // ABS Diff of lines
+		"movq %2, %%mm0					\n\t"  // QP,..., QP
+		"psubusb %%mm2, %%mm0				\n\t" // diff >= QP -> 0
+		"pcmpeqb %%mm4, %%mm0				\n\t" // diff >= QP -> FF
+
+		"pxor %%mm7, %%mm1				\n\t"
+		"pand %%mm0, %%mm1				\n\t"
+		"pxor %%mm1, %%mm7				\n\t"
+		
+		"movq %%mm6, %%mm5				\n\t"
+		"punpckhbw %%mm4, %%mm6				\n\t"
+		"punpcklbw %%mm4, %%mm5				\n\t"
+		// 4:0 5/6:First 7:Last
+
+		"movq %%mm5, %%mm0				\n\t"
+		"movq %%mm6, %%mm1				\n\t"
+		"psllw $2, %%mm0				\n\t"
+		"psllw $2, %%mm1				\n\t"
+		"paddw "MANGLE(w04)", %%mm0			\n\t"
+		"paddw "MANGLE(w04)", %%mm1			\n\t"
+
+#define NEXT\
+		"movq (%0), %%mm2				\n\t"\
+		"movq (%0), %%mm3				\n\t"\
+		"add %1, %0					\n\t"\
+		"punpcklbw %%mm4, %%mm2				\n\t"\
+		"punpckhbw %%mm4, %%mm3				\n\t"\
+		"paddw %%mm2, %%mm0				\n\t"\
+		"paddw %%mm3, %%mm1				\n\t"
+
+#define PREV\
+		"movq (%0), %%mm2				\n\t"\
+		"movq (%0), %%mm3				\n\t"\
+		"add %1, %0					\n\t"\
+		"punpcklbw %%mm4, %%mm2				\n\t"\
+		"punpckhbw %%mm4, %%mm3				\n\t"\
+		"psubw %%mm2, %%mm0				\n\t"\
+		"psubw %%mm3, %%mm1				\n\t"
+
+				
+		NEXT //0
+		NEXT //1
+		NEXT //2
+		"movq %%mm0, (%3)				\n\t"
+		"movq %%mm1, 8(%3)				\n\t"
+
+		NEXT //3
+		"psubw %%mm5, %%mm0				\n\t"
+		"psubw %%mm6, %%mm1				\n\t"
+		"movq %%mm0, 16(%3)				\n\t"
+		"movq %%mm1, 24(%3)				\n\t"
+
+		NEXT //4
+		"psubw %%mm5, %%mm0				\n\t"
+		"psubw %%mm6, %%mm1				\n\t"
+		"movq %%mm0, 32(%3)				\n\t"
+		"movq %%mm1, 40(%3)				\n\t"
+
+		NEXT //5
+		"psubw %%mm5, %%mm0				\n\t"
+		"psubw %%mm6, %%mm1				\n\t"
+		"movq %%mm0, 48(%3)				\n\t"
+		"movq %%mm1, 56(%3)				\n\t"
+
+		NEXT //6
+		"psubw %%mm5, %%mm0				\n\t"
+		"psubw %%mm6, %%mm1				\n\t"
+		"movq %%mm0, 64(%3)				\n\t"
+		"movq %%mm1, 72(%3)				\n\t"
+
+		"movq %%mm7, %%mm6				\n\t"
+		"punpckhbw %%mm4, %%mm7				\n\t"
+		"punpcklbw %%mm4, %%mm6				\n\t"
+		
+		NEXT //7
+		"mov %4, %0					\n\t"
+		"add %1, %0					\n\t"
+		PREV //0
+		"movq %%mm0, 80(%3)				\n\t"
+		"movq %%mm1, 88(%3)				\n\t"
+
+		PREV //1
+		"paddw %%mm6, %%mm0				\n\t"
+		"paddw %%mm7, %%mm1				\n\t"
+		"movq %%mm0, 96(%3)				\n\t"
+		"movq %%mm1, 104(%3)				\n\t"
+		
+		PREV //2
+		"paddw %%mm6, %%mm0				\n\t"
+		"paddw %%mm7, %%mm1				\n\t"
+		"movq %%mm0, 112(%3)				\n\t"
+		"movq %%mm1, 120(%3)				\n\t"
+
+		PREV //3
+		"paddw %%mm6, %%mm0				\n\t"
+		"paddw %%mm7, %%mm1				\n\t"
+		"movq %%mm0, 128(%3)				\n\t"
+		"movq %%mm1, 136(%3)				\n\t"
+
+		PREV //4
+		"paddw %%mm6, %%mm0				\n\t"
+		"paddw %%mm7, %%mm1				\n\t"
+		"movq %%mm0, 144(%3)				\n\t"
+		"movq %%mm1, 152(%3)				\n\t"
+
+		"mov %4, %0					\n\t" //FIXME
+
+		: "+&r"(src)
+		: "r" ((long)step), "m" (c->pQPb), "r"(sums), "g"(src)
+		);
+
+		src+= step; // src points to begin of the 8x8 Block
+
+		asm volatile(
+		"movq %4, %%mm6					\n\t"
+		"pcmpeqb %%mm5, %%mm5				\n\t"
+		"pxor %%mm6, %%mm5				\n\t"
+		"pxor %%mm7, %%mm7				\n\t"
+
+		"1:						\n\t"
+		"movq (%1), %%mm0				\n\t"
+		"movq 8(%1), %%mm1				\n\t"
+		"paddw 32(%1), %%mm0				\n\t"
+		"paddw 40(%1), %%mm1				\n\t"
+		"movq (%0, %3), %%mm2				\n\t"
+		"movq %%mm2, %%mm3				\n\t"
+		"movq %%mm2, %%mm4				\n\t"
+		"punpcklbw %%mm7, %%mm2				\n\t"
+		"punpckhbw %%mm7, %%mm3				\n\t"
+		"paddw %%mm2, %%mm0				\n\t"
+		"paddw %%mm3, %%mm1				\n\t"
+		"paddw %%mm2, %%mm0				\n\t"
+		"paddw %%mm3, %%mm1				\n\t"
+		"psrlw $4, %%mm0				\n\t"
+		"psrlw $4, %%mm1				\n\t"
+		"packuswb %%mm1, %%mm0				\n\t"
+		"pand %%mm6, %%mm0				\n\t"
+		"pand %%mm5, %%mm4				\n\t"
+		"por %%mm4, %%mm0				\n\t"
+		"movq %%mm0, (%0, %3)				\n\t"
+		"add $16, %1					\n\t"
+		"add %2, %0					\n\t"
+		" js 1b						\n\t"
+
+		: "+r"(offset), "+r"(temp_sums)
+		: "r" ((long)step), "r"(src - offset), "m"(dc_mask & eq_mask)
+		);
+	}else
+		src+= step; // src points to begin of the 8x8 Block
+
+	if(eq_mask != -1LL){
+		uint8_t *temp_src= src;
+		asm volatile(
+		"pxor %%mm7, %%mm7				\n\t"
+		"lea -40(%%"REG_SP"), %%"REG_c"			\n\t" // make space for 4 8-byte vars
+		"and "ALIGN_MASK", %%"REG_c"			\n\t" // align
+//	0	1	2	3	4	5	6	7	8	9
+//	%0	eax	eax+%1	eax+2%1	%0+4%1	ecx	ecx+%1	ecx+2%1	%1+8%1	ecx+4%1
+
+		"movq (%0), %%mm0				\n\t"
+		"movq %%mm0, %%mm1				\n\t"
+		"punpcklbw %%mm7, %%mm0				\n\t" // low part of line 0
+		"punpckhbw %%mm7, %%mm1				\n\t" // high part of line 0
+
+		"movq (%0, %1), %%mm2				\n\t"
+		"lea (%0, %1, 2), %%"REG_a"			\n\t"
+		"movq %%mm2, %%mm3				\n\t"
+		"punpcklbw %%mm7, %%mm2				\n\t" // low part of line 1
+		"punpckhbw %%mm7, %%mm3				\n\t" // high part of line 1
+
+		"movq (%%"REG_a"), %%mm4			\n\t"
+		"movq %%mm4, %%mm5				\n\t"
+		"punpcklbw %%mm7, %%mm4				\n\t" // low part of line 2
+		"punpckhbw %%mm7, %%mm5				\n\t" // high part of line 2
+
+		"paddw %%mm0, %%mm0				\n\t" // 2L0
+		"paddw %%mm1, %%mm1				\n\t" // 2H0
+		"psubw %%mm4, %%mm2				\n\t" // L1 - L2
+		"psubw %%mm5, %%mm3				\n\t" // H1 - H2
+		"psubw %%mm2, %%mm0				\n\t" // 2L0 - L1 + L2
+		"psubw %%mm3, %%mm1				\n\t" // 2H0 - H1 + H2
+
+		"psllw $2, %%mm2				\n\t" // 4L1 - 4L2
+		"psllw $2, %%mm3				\n\t" // 4H1 - 4H2
+		"psubw %%mm2, %%mm0				\n\t" // 2L0 - 5L1 + 5L2
+		"psubw %%mm3, %%mm1				\n\t" // 2H0 - 5H1 + 5H2
+
+		"movq (%%"REG_a", %1), %%mm2			\n\t"
+		"movq %%mm2, %%mm3				\n\t"
+		"punpcklbw %%mm7, %%mm2				\n\t" // L3
+		"punpckhbw %%mm7, %%mm3				\n\t" // H3
+
+		"psubw %%mm2, %%mm0				\n\t" // 2L0 - 5L1 + 5L2 - L3
+		"psubw %%mm3, %%mm1				\n\t" // 2H0 - 5H1 + 5H2 - H3
+		"psubw %%mm2, %%mm0				\n\t" // 2L0 - 5L1 + 5L2 - 2L3
+		"psubw %%mm3, %%mm1				\n\t" // 2H0 - 5H1 + 5H2 - 2H3
+		"movq %%mm0, (%%"REG_c")			\n\t" // 2L0 - 5L1 + 5L2 - 2L3
+		"movq %%mm1, 8(%%"REG_c")			\n\t" // 2H0 - 5H1 + 5H2 - 2H3
+
+		"movq (%%"REG_a", %1, 2), %%mm0			\n\t"
+		"movq %%mm0, %%mm1				\n\t"
+		"punpcklbw %%mm7, %%mm0				\n\t" // L4
+		"punpckhbw %%mm7, %%mm1				\n\t" // H4
+
+		"psubw %%mm0, %%mm2				\n\t" // L3 - L4
+		"psubw %%mm1, %%mm3				\n\t" // H3 - H4
+		"movq %%mm2, 16(%%"REG_c")			\n\t" // L3 - L4
+		"movq %%mm3, 24(%%"REG_c")			\n\t" // H3 - H4
+		"paddw %%mm4, %%mm4				\n\t" // 2L2
+		"paddw %%mm5, %%mm5				\n\t" // 2H2
+		"psubw %%mm2, %%mm4				\n\t" // 2L2 - L3 + L4
+		"psubw %%mm3, %%mm5				\n\t" // 2H2 - H3 + H4
+
+		"lea (%%"REG_a", %1), %0			\n\t"
+		"psllw $2, %%mm2				\n\t" // 4L3 - 4L4
+		"psllw $2, %%mm3				\n\t" // 4H3 - 4H4
+		"psubw %%mm2, %%mm4				\n\t" // 2L2 - 5L3 + 5L4
+		"psubw %%mm3, %%mm5				\n\t" // 2H2 - 5H3 + 5H4
+//50 opcodes so far
+		"movq (%0, %1, 2), %%mm2			\n\t"
+		"movq %%mm2, %%mm3				\n\t"
+		"punpcklbw %%mm7, %%mm2				\n\t" // L5
+		"punpckhbw %%mm7, %%mm3				\n\t" // H5
+		"psubw %%mm2, %%mm4				\n\t" // 2L2 - 5L3 + 5L4 - L5
+		"psubw %%mm3, %%mm5				\n\t" // 2H2 - 5H3 + 5H4 - H5
+		"psubw %%mm2, %%mm4				\n\t" // 2L2 - 5L3 + 5L4 - 2L5
+		"psubw %%mm3, %%mm5				\n\t" // 2H2 - 5H3 + 5H4 - 2H5
+
+		"movq (%%"REG_a", %1, 4), %%mm6			\n\t"
+		"punpcklbw %%mm7, %%mm6				\n\t" // L6
+		"psubw %%mm6, %%mm2				\n\t" // L5 - L6
+		"movq (%%"REG_a", %1, 4), %%mm6			\n\t"
+		"punpckhbw %%mm7, %%mm6				\n\t" // H6
+		"psubw %%mm6, %%mm3				\n\t" // H5 - H6
+
+		"paddw %%mm0, %%mm0				\n\t" // 2L4
+		"paddw %%mm1, %%mm1				\n\t" // 2H4
+		"psubw %%mm2, %%mm0				\n\t" // 2L4 - L5 + L6
+		"psubw %%mm3, %%mm1				\n\t" // 2H4 - H5 + H6
+
+		"psllw $2, %%mm2				\n\t" // 4L5 - 4L6
+		"psllw $2, %%mm3				\n\t" // 4H5 - 4H6
+		"psubw %%mm2, %%mm0				\n\t" // 2L4 - 5L5 + 5L6
+		"psubw %%mm3, %%mm1				\n\t" // 2H4 - 5H5 + 5H6
+
+		"movq (%0, %1, 4), %%mm2			\n\t"
+		"movq %%mm2, %%mm3				\n\t"
+		"punpcklbw %%mm7, %%mm2				\n\t" // L7
+		"punpckhbw %%mm7, %%mm3				\n\t" // H7
+
+		"paddw %%mm2, %%mm2				\n\t" // 2L7
+		"paddw %%mm3, %%mm3				\n\t" // 2H7
+		"psubw %%mm2, %%mm0				\n\t" // 2L4 - 5L5 + 5L6 - 2L7
+		"psubw %%mm3, %%mm1				\n\t" // 2H4 - 5H5 + 5H6 - 2H7
+
+		"movq (%%"REG_c"), %%mm2			\n\t" // 2L0 - 5L1 + 5L2 - 2L3
+		"movq 8(%%"REG_c"), %%mm3			\n\t" // 2H0 - 5H1 + 5H2 - 2H3
+
+#ifdef HAVE_MMX2
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"psubw %%mm0, %%mm6				\n\t"
+		"pmaxsw %%mm6, %%mm0				\n\t" // |2L4 - 5L5 + 5L6 - 2L7|
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"psubw %%mm1, %%mm6				\n\t"
+		"pmaxsw %%mm6, %%mm1				\n\t" // |2H4 - 5H5 + 5H6 - 2H7|
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"psubw %%mm2, %%mm6				\n\t"
+		"pmaxsw %%mm6, %%mm2				\n\t" // |2L0 - 5L1 + 5L2 - 2L3|
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"psubw %%mm3, %%mm6				\n\t"
+		"pmaxsw %%mm6, %%mm3				\n\t" // |2H0 - 5H1 + 5H2 - 2H3|
+#else
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"pcmpgtw %%mm0, %%mm6				\n\t"
+		"pxor %%mm6, %%mm0				\n\t"
+		"psubw %%mm6, %%mm0				\n\t" // |2L4 - 5L5 + 5L6 - 2L7|
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"pcmpgtw %%mm1, %%mm6				\n\t"
+		"pxor %%mm6, %%mm1				\n\t"
+		"psubw %%mm6, %%mm1				\n\t" // |2H4 - 5H5 + 5H6 - 2H7|
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"pcmpgtw %%mm2, %%mm6				\n\t"
+		"pxor %%mm6, %%mm2				\n\t"
+		"psubw %%mm6, %%mm2				\n\t" // |2L0 - 5L1 + 5L2 - 2L3|
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"pcmpgtw %%mm3, %%mm6				\n\t"
+		"pxor %%mm6, %%mm3				\n\t"
+		"psubw %%mm6, %%mm3				\n\t" // |2H0 - 5H1 + 5H2 - 2H3|
+#endif
+
+#ifdef HAVE_MMX2
+		"pminsw %%mm2, %%mm0				\n\t"
+		"pminsw %%mm3, %%mm1				\n\t"
+#else
+		"movq %%mm0, %%mm6				\n\t"
+		"psubusw %%mm2, %%mm6				\n\t"
+		"psubw %%mm6, %%mm0				\n\t"
+		"movq %%mm1, %%mm6				\n\t"
+		"psubusw %%mm3, %%mm6				\n\t"
+		"psubw %%mm6, %%mm1				\n\t"
+#endif
+
+		"movd %2, %%mm2					\n\t" // QP
+		"punpcklbw %%mm7, %%mm2				\n\t"
+
+		"movq %%mm7, %%mm6				\n\t" // 0
+		"pcmpgtw %%mm4, %%mm6				\n\t" // sign(2L2 - 5L3 + 5L4 - 2L5)
+		"pxor %%mm6, %%mm4				\n\t"
+		"psubw %%mm6, %%mm4				\n\t" // |2L2 - 5L3 + 5L4 - 2L5|
+		"pcmpgtw %%mm5, %%mm7				\n\t" // sign(2H2 - 5H3 + 5H4 - 2H5)
+		"pxor %%mm7, %%mm5				\n\t"
+		"psubw %%mm7, %%mm5				\n\t" // |2H2 - 5H3 + 5H4 - 2H5|
+// 100 opcodes
+		"psllw $3, %%mm2				\n\t" // 8QP
+		"movq %%mm2, %%mm3				\n\t" // 8QP
+		"pcmpgtw %%mm4, %%mm2				\n\t"
+		"pcmpgtw %%mm5, %%mm3				\n\t"
+		"pand %%mm2, %%mm4				\n\t"
+		"pand %%mm3, %%mm5				\n\t"
+
+
+		"psubusw %%mm0, %%mm4				\n\t" // hd
+		"psubusw %%mm1, %%mm5				\n\t" // ld
+
+
+		"movq "MANGLE(w05)", %%mm2			\n\t" // 5
+		"pmullw %%mm2, %%mm4				\n\t"
+		"pmullw %%mm2, %%mm5				\n\t"
+		"movq "MANGLE(w20)", %%mm2			\n\t" // 32
+		"paddw %%mm2, %%mm4				\n\t"
+		"paddw %%mm2, %%mm5				\n\t"
+		"psrlw $6, %%mm4				\n\t"
+		"psrlw $6, %%mm5				\n\t"
+
+		"movq 16(%%"REG_c"), %%mm0			\n\t" // L3 - L4
+		"movq 24(%%"REG_c"), %%mm1			\n\t" // H3 - H4
+
+		"pxor %%mm2, %%mm2				\n\t"
+		"pxor %%mm3, %%mm3				\n\t"
+
+		"pcmpgtw %%mm0, %%mm2				\n\t" // sign (L3-L4)
+		"pcmpgtw %%mm1, %%mm3				\n\t" // sign (H3-H4)
+		"pxor %%mm2, %%mm0				\n\t"
+		"pxor %%mm3, %%mm1				\n\t"
+		"psubw %%mm2, %%mm0				\n\t" // |L3-L4|
+		"psubw %%mm3, %%mm1				\n\t" // |H3-H4|
+		"psrlw $1, %%mm0				\n\t" // |L3 - L4|/2
+		"psrlw $1, %%mm1				\n\t" // |H3 - H4|/2
+
+		"pxor %%mm6, %%mm2				\n\t"
+		"pxor %%mm7, %%mm3				\n\t"
+		"pand %%mm2, %%mm4				\n\t"
+		"pand %%mm3, %%mm5				\n\t"
+
+#ifdef HAVE_MMX2
+		"pminsw %%mm0, %%mm4				\n\t"
+		"pminsw %%mm1, %%mm5				\n\t"
+#else
+		"movq %%mm4, %%mm2				\n\t"
+		"psubusw %%mm0, %%mm2				\n\t"
+		"psubw %%mm2, %%mm4				\n\t"
+		"movq %%mm5, %%mm2				\n\t"
+		"psubusw %%mm1, %%mm2				\n\t"
+		"psubw %%mm2, %%mm5				\n\t"
+#endif
+		"pxor %%mm6, %%mm4				\n\t"
+		"pxor %%mm7, %%mm5				\n\t"
+		"psubw %%mm6, %%mm4				\n\t"
+		"psubw %%mm7, %%mm5				\n\t"
+		"packsswb %%mm5, %%mm4				\n\t"
+		"movq %3, %%mm1					\n\t"
+		"pandn %%mm4, %%mm1				\n\t"
+		"movq (%0), %%mm0				\n\t"
+		"paddb   %%mm1, %%mm0				\n\t"
+		"movq %%mm0, (%0)				\n\t"
+		"movq (%0, %1), %%mm0				\n\t"
+		"psubb %%mm1, %%mm0				\n\t"
+		"movq %%mm0, (%0, %1)				\n\t"
+
+		: "+r" (temp_src)
+		: "r" ((long)step), "m" (c->pQPb), "m"(eq_mask)
+		: "%"REG_a, "%"REG_c
+		);
+	}
+/*if(step==16){
+    STOP_TIMER("step16")
+}else{
+    STOP_TIMER("stepX")
+}*/
+}
+#endif //HAVE_MMX
 
 static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int dstStride, int width, int height,
 	QP_STORE_T QPs[], int QPStride, int isColor, PPContext *c);
@@ -2628,13 +3196,13 @@ static inline void RENAME(blockCopy)(uint8_t dst[], int dstStride, uint8_t src[]
 	{
 #ifdef HAVE_MMX
 					asm volatile(
-						"movq (%%eax), %%mm2	\n\t" // packedYOffset
-						"movq 8(%%eax), %%mm3	\n\t" // packedYScale
-						"leal (%2,%4), %%eax	\n\t"
-						"leal (%3,%5), %%edx	\n\t"
+						"movq (%%"REG_a"), %%mm2	\n\t" // packedYOffset
+						"movq 8(%%"REG_a"), %%mm3	\n\t" // packedYScale
+						"lea (%2,%4), %%"REG_a"	\n\t"
+						"lea (%3,%5), %%"REG_d"	\n\t"
 						"pxor %%mm4, %%mm4	\n\t"
 #ifdef HAVE_MMX2
-#define SCALED_CPY(src1, src2, dst1, dst2)					\
+#define REAL_SCALED_CPY(src1, src2, dst1, dst2)					\
 						"movq " #src1 ", %%mm0	\n\t"\
 						"movq " #src1 ", %%mm5	\n\t"\
 						"movq " #src2 ", %%mm1	\n\t"\
@@ -2657,7 +3225,7 @@ static inline void RENAME(blockCopy)(uint8_t dst[], int dstStride, uint8_t src[]
 						"movq %%mm1, " #dst2 "	\n\t"\
 
 #else //HAVE_MMX2
-#define SCALED_CPY(src1, src2, dst1, dst2)					\
+#define REAL_SCALED_CPY(src1, src2, dst1, dst2)					\
 						"movq " #src1 ", %%mm0	\n\t"\
 						"movq " #src1 ", %%mm5	\n\t"\
 						"punpcklbw %%mm4, %%mm0 \n\t"\
@@ -2684,22 +3252,24 @@ static inline void RENAME(blockCopy)(uint8_t dst[], int dstStride, uint8_t src[]
 						"movq %%mm1, " #dst2 "	\n\t"\
 
 #endif //!HAVE_MMX2
+#define SCALED_CPY(src1, src2, dst1, dst2)\
+   REAL_SCALED_CPY(src1, src2, dst1, dst2)
 
 SCALED_CPY((%2)       , (%2, %4)      , (%3)       , (%3, %5))
-SCALED_CPY((%2, %4, 2), (%%eax, %4, 2), (%3, %5, 2), (%%edx, %5, 2))
-SCALED_CPY((%2, %4, 4), (%%eax, %4, 4), (%3, %5, 4), (%%edx, %5, 4))
-						"leal (%%eax,%4,4), %%eax	\n\t"
-						"leal (%%edx,%5,4), %%edx	\n\t"
-SCALED_CPY((%%eax, %4), (%%eax, %4, 2), (%%edx, %5), (%%edx, %5, 2))
+SCALED_CPY((%2, %4, 2), (%%REGa, %4, 2), (%3, %5, 2), (%%REGd, %5, 2))
+SCALED_CPY((%2, %4, 4), (%%REGa, %4, 4), (%3, %5, 4), (%%REGd, %5, 4))
+						"lea (%%"REG_a",%4,4), %%"REG_a"	\n\t"
+						"lea (%%"REG_d",%5,4), %%"REG_d"	\n\t"
+SCALED_CPY((%%REGa, %4), (%%REGa, %4, 2), (%%REGd, %5), (%%REGd, %5, 2))
 
 
 						: "=&a" (packedOffsetAndScale)
 						: "0" (packedOffsetAndScale),
 						"r"(src),
 						"r"(dst),
-						"r" (srcStride),
-						"r" (dstStride)
-						: "%edx"
+						"r" ((long)srcStride),
+						"r" ((long)dstStride)
+						: "%"REG_d
 					);
 #else
 				for(i=0; i<8; i++)
@@ -2711,27 +3281,30 @@ SCALED_CPY((%%eax, %4), (%%eax, %4, 2), (%%edx, %5), (%%edx, %5, 2))
 	{
 #ifdef HAVE_MMX
 					asm volatile(
-						"leal (%0,%2), %%eax	\n\t"
-						"leal (%1,%3), %%edx	\n\t"
+						"lea (%0,%2), %%"REG_a"	\n\t"
+						"lea (%1,%3), %%"REG_d"	\n\t"
 
-#define SIMPLE_CPY(src1, src2, dst1, dst2)				\
+#define REAL_SIMPLE_CPY(src1, src2, dst1, dst2)				\
 						"movq " #src1 ", %%mm0	\n\t"\
 						"movq " #src2 ", %%mm1	\n\t"\
 						"movq %%mm0, " #dst1 "	\n\t"\
 						"movq %%mm1, " #dst2 "	\n\t"\
 
+#define SIMPLE_CPY(src1, src2, dst1, dst2)\
+   REAL_SIMPLE_CPY(src1, src2, dst1, dst2)
+
 SIMPLE_CPY((%0)       , (%0, %2)      , (%1)       , (%1, %3))
-SIMPLE_CPY((%0, %2, 2), (%%eax, %2, 2), (%1, %3, 2), (%%edx, %3, 2))
-SIMPLE_CPY((%0, %2, 4), (%%eax, %2, 4), (%1, %3, 4), (%%edx, %3, 4))
-						"leal (%%eax,%2,4), %%eax	\n\t"
-						"leal (%%edx,%3,4), %%edx	\n\t"
-SIMPLE_CPY((%%eax, %2), (%%eax, %2, 2), (%%edx, %3), (%%edx, %3, 2))
+SIMPLE_CPY((%0, %2, 2), (%%REGa, %2, 2), (%1, %3, 2), (%%REGd, %3, 2))
+SIMPLE_CPY((%0, %2, 4), (%%REGa, %2, 4), (%1, %3, 4), (%%REGd, %3, 4))
+						"lea (%%"REG_a",%2,4), %%"REG_a"	\n\t"
+						"lea (%%"REG_d",%3,4), %%"REG_d"	\n\t"
+SIMPLE_CPY((%%REGa, %2), (%%REGa, %2, 2), (%%REGd, %3), (%%REGd, %3, 2))
 
 						: : "r" (src),
 						"r" (dst),
-						"r" (srcStride),
-						"r" (dstStride)
-						: "%eax", "%edx"
+						"r" ((long)srcStride),
+						"r" ((long)dstStride)
+						: "%"REG_a, "%"REG_d
 					);
 #else
 				for(i=0; i<8; i++)
@@ -2749,12 +3322,12 @@ static inline void RENAME(duplicate)(uint8_t src[], int stride)
 #ifdef HAVE_MMX
 	asm volatile(
 		"movq (%0), %%mm0		\n\t"
-		"addl %1, %0			\n\t"
+		"add %1, %0			\n\t"
 		"movq %%mm0, (%0)		\n\t"
 		"movq %%mm0, (%0, %1)		\n\t"
 		"movq %%mm0, (%0, %1, 2)	\n\t"
 		: "+r" (src)
-		: "r" (-stride)
+		: "r" ((long)-stride)
 	);
 #else
 	int i;
@@ -2793,8 +3366,8 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 
 	//FIXME remove
 	uint64_t * const yHistogram= c.yHistogram;
-	uint8_t * const tempSrc= c.tempSrc;
-	uint8_t * const tempDst= c.tempDst;
+	uint8_t * const tempSrc= srcStride > 0 ? c.tempSrc : c.tempSrc - 23*srcStride;
+	uint8_t * const tempDst= dstStride > 0 ? c.tempDst : c.tempDst - 23*dstStride;
 	//const int mbWidth= isColor ? (width+7)>>3 : (width+15)>>4;
 
 #ifdef HAVE_MMX
@@ -2814,7 +3387,8 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 		|| (mode & LOWPASS5_DEINT_FILTER)) copyAhead=14;
 	else if(   (mode & V_DEBLOCK)
 		|| (mode & LINEAR_IPOL_DEINT_FILTER)
-		|| (mode & MEDIAN_DEINT_FILTER)) copyAhead=13;
+		|| (mode & MEDIAN_DEINT_FILTER)
+		|| (mode & V_A_DEBLOCK)) copyAhead=13;
 	else if(mode & V_X1_FILTER) copyAhead=11;
 //	else if(mode & V_RK1_FILTER) copyAhead=10;
 	else if(mode & DERING) copyAhead=9;
@@ -2905,22 +3479,22 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 */
 
 			asm(
-				"movl %4, %%eax			\n\t"
-				"shrl $2, %%eax			\n\t"
-				"andl $6, %%eax			\n\t"
-				"addl %5, %%eax			\n\t"
-				"movl %%eax, %%edx		\n\t"
-				"imul %1, %%eax			\n\t"
-				"imul %3, %%edx			\n\t"
-				"prefetchnta 32(%%eax, %0)	\n\t"
-				"prefetcht0 32(%%edx, %2)	\n\t"
-				"addl %1, %%eax			\n\t"
-				"addl %3, %%edx			\n\t"
-				"prefetchnta 32(%%eax, %0)	\n\t"
-				"prefetcht0 32(%%edx, %2)	\n\t"
-			:: "r" (srcBlock), "r" (srcStride), "r" (dstBlock), "r" (dstStride),
-			"m" (x), "m" (copyAhead)
-			: "%eax", "%edx"
+				"mov %4, %%"REG_a"		\n\t"
+				"shr $2, %%"REG_a"		\n\t"
+				"and $6, %%"REG_a"		\n\t"
+				"add %5, %%"REG_a"		\n\t"
+				"mov %%"REG_a", %%"REG_d"	\n\t"
+				"imul %1, %%"REG_a"		\n\t"
+				"imul %3, %%"REG_d"		\n\t"
+				"prefetchnta 32(%%"REG_a", %0)	\n\t"
+				"prefetcht0 32(%%"REG_d", %2)	\n\t"
+				"add %1, %%"REG_a"		\n\t"
+				"add %3, %%"REG_d"		\n\t"
+				"prefetchnta 32(%%"REG_a", %0)	\n\t"
+				"prefetcht0 32(%%"REG_d", %2)	\n\t"
+			:: "r" (srcBlock), "r" ((long)srcStride), "r" (dstBlock), "r" ((long)dstStride),
+			"m" ((long)x), "m" ((long)copyAhead)
+			: "%"REG_a, "%"REG_d
 			);
 
 #elif defined(HAVE_3DNOW)
@@ -2955,8 +3529,8 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 			dstBlock+=8;
 			srcBlock+=8;
 		}
-		if(width==dstStride)
-			memcpy(dst, tempDst + 9*dstStride, copyAhead*dstStride);
+		if(width==ABS(dstStride))
+			linecpy(dst, tempDst + 9*dstStride, copyAhead, dstStride);
 		else
 		{
 			int i;
@@ -2978,7 +3552,7 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 		uint8_t *tempBlock2= c.tempBlocks + 8;
 #endif
 		int8_t *QPptr= &QPs[(y>>qpVShift)*QPStride];
-		int8_t *nonBQPptr= &c.nonBQPTable[(y>>qpVShift)*QPStride];
+		int8_t *nonBQPptr= &c.nonBQPTable[(y>>qpVShift)*ABS(QPStride)];
 		int QP=0;
 		/* can we mess with a 8x16 block from srcBlock/dstBlock downwards and 1 line upwards
 		   if not than use a temporary buffer */
@@ -2987,19 +3561,19 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 			int i;
 			/* copy from line (copyAhead) to (copyAhead+7) of src, these will be copied with
 			   blockcopy to dst later */
-			memcpy(tempSrc + srcStride*copyAhead, srcBlock + srcStride*copyAhead,
-				srcStride*MAX(height-y-copyAhead, 0) );
+			linecpy(tempSrc + srcStride*copyAhead, srcBlock + srcStride*copyAhead,
+				MAX(height-y-copyAhead, 0), srcStride);
 
 			/* duplicate last line of src to fill the void upto line (copyAhead+7) */
 			for(i=MAX(height-y, 8); i<copyAhead+8; i++)
-				memcpy(tempSrc + srcStride*i, src + srcStride*(height-1), srcStride);
+				memcpy(tempSrc + srcStride*i, src + srcStride*(height-1), ABS(srcStride));
 
 			/* copy up to (copyAhead+1) lines of dst (line -1 to (copyAhead-1))*/
-			memcpy(tempDst, dstBlock - dstStride, dstStride*MIN(height-y+1, copyAhead+1) );
+			linecpy(tempDst, dstBlock - dstStride, MIN(height-y+1, copyAhead+1), dstStride);
 
 			/* duplicate last line of dst to fill the void upto line (copyAhead) */
 			for(i=height-y+1; i<=copyAhead; i++)
-				memcpy(tempDst + dstStride*i, dst + dstStride*(height-1), dstStride);
+				memcpy(tempDst + dstStride*i, dst + dstStride*(height-1), ABS(dstStride));
 
 			dstBlock= tempDst + dstStride;
 			srcBlock= tempSrc;
@@ -3051,22 +3625,22 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 */
 
 			asm(
-				"movl %4, %%eax			\n\t"
-				"shrl $2, %%eax			\n\t"
-				"andl $6, %%eax			\n\t"
-				"addl %5, %%eax			\n\t"
-				"movl %%eax, %%edx		\n\t"
-				"imul %1, %%eax			\n\t"
-				"imul %3, %%edx			\n\t"
-				"prefetchnta 32(%%eax, %0)	\n\t"
-				"prefetcht0 32(%%edx, %2)	\n\t"
-				"addl %1, %%eax			\n\t"
-				"addl %3, %%edx			\n\t"
-				"prefetchnta 32(%%eax, %0)	\n\t"
-				"prefetcht0 32(%%edx, %2)	\n\t"
-			:: "r" (srcBlock), "r" (srcStride), "r" (dstBlock), "r" (dstStride),
-			"m" (x), "m" (copyAhead)
-			: "%eax", "%edx"
+				"mov %4, %%"REG_a"		\n\t"
+				"shr $2, %%"REG_a"		\n\t"
+				"and $6, %%"REG_a"		\n\t"
+				"add %5, %%"REG_a"		\n\t"
+				"mov %%"REG_a", %%"REG_d"	\n\t"
+				"imul %1, %%"REG_a"		\n\t"
+				"imul %3, %%"REG_d"		\n\t"
+				"prefetchnta 32(%%"REG_a", %0)	\n\t"
+				"prefetcht0 32(%%"REG_d", %2)	\n\t"
+				"add %1, %%"REG_a"		\n\t"
+				"add %3, %%"REG_d"		\n\t"
+				"prefetchnta 32(%%"REG_a", %0)	\n\t"
+				"prefetcht0 32(%%"REG_d", %2)	\n\t"
+			:: "r" (srcBlock), "r" ((long)srcStride), "r" (dstBlock), "r" ((long)dstStride),
+			"m" ((long)x), "m" ((long)copyAhead)
+			: "%"REG_a, "%"REG_d
 			);
 
 #elif defined(HAVE_3DNOW)
@@ -3110,6 +3684,8 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 						RENAME(doVertLowPass)(dstBlock, stride, &c);
 					else if(t==2)
 						RENAME(doVertDefFilter)(dstBlock, stride, &c);
+				}else if(mode & V_A_DEBLOCK){
+					RENAME(do_a_deblock)(dstBlock, stride, 1, &c);
 				}
 			}
 
@@ -3131,6 +3707,8 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 						RENAME(doVertLowPass)(tempBlock1, 16, &c);
 					else if(t==2)
 						RENAME(doVertDefFilter)(tempBlock1, 16, &c);
+				}else if(mode & H_A_DEBLOCK){
+					RENAME(do_a_deblock)(tempBlock1, 16, 1, &c);
 				}
 
 				RENAME(transpose2)(dstBlock-4, dstStride, tempBlock1 + 4*16);
@@ -3140,12 +3718,29 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 					horizX1Filter(dstBlock-4, stride, QP);
 				else if(mode & H_DEBLOCK)
 				{
+#ifdef HAVE_ALTIVEC
+					unsigned char __attribute__ ((aligned(16))) tempBlock[272];
+					transpose_16x8_char_toPackedAlign_altivec(tempBlock, dstBlock - (4 + 1), stride);
+
+					const int t=vertClassify_altivec(tempBlock-48, 16, &c);
+					if(t==1) {
+						doVertLowPass_altivec(tempBlock-48, 16, &c);
+                                                transpose_8x16_char_fromPackedAlign_altivec(dstBlock - (4 + 1), tempBlock, stride);
+                                        }
+					else if(t==2) {
+						doVertDefFilter_altivec(tempBlock-48, 16, &c);
+                                                transpose_8x16_char_fromPackedAlign_altivec(dstBlock - (4 + 1), tempBlock, stride);
+                                        }
+#else
 					const int t= RENAME(horizClassify)(dstBlock-4, stride, &c);
 
 					if(t==1)
 						RENAME(doHorizLowPass)(dstBlock-4, stride, &c);
 					else if(t==2)
 						RENAME(doHorizDefFilter)(dstBlock-4, stride, &c);
+#endif
+				}else if(mode & H_A_DEBLOCK){
+					RENAME(do_a_deblock)(dstBlock-8, 1, stride, &c);
 				}
 #endif
 				if(mode & DERING)
@@ -3190,8 +3785,8 @@ static void RENAME(postProcess)(uint8_t src[], int srcStride, uint8_t dst[], int
 		if(y+15 >= height)
 		{
 			uint8_t *dstBlock= &(dst[y*dstStride]);
-			if(width==dstStride)
-				memcpy(dstBlock, tempDst + dstStride, dstStride*(height-y));
+			if(width==ABS(dstStride))
+				linecpy(dstBlock, tempDst + dstStride, height-y, dstStride);
 			else
 			{
 				int i;
-- 
cgit v1.2.3