altivec yuv->rgb converter

orginal patch  by (Marc Hoffman <mmh at pleasantst dot com>)
critical fixes by (Reza Jelveh <reza.jelveh at tu-harburg dot de>)

known bugs/issues, which should be fixed ASAP by someone who has a ppc:
0..255 vs. 16..235
unneeded recalculation of tables
general cleaup, like removing double initalizing of variables

Originally committed as revision 12699 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc

7 files changed, 878 insertions, 2 deletions

diff --git a/postproc/Makefile b/postproc/Makefile
index b030c72854..06caa9ad1c 100644
--- a/postproc/Makefile
+++ b/postproc/Makefile
@@ -4,8 +4,13 @@ include ../config.mak
 SWSLIB = libswscale.a
 
 SWSSRCS=swscale.c rgb2rgb.c yuv2rgb.c
+
 SWSOBJS=$(SWSSRCS:.c=.o)
 
+ifeq ($(TARGET_ALTIVEC),yes)
+SWSOBJS +=  yuv2rgb_altivec.o
+endif
+
 CFLAGS  = $(OPTFLAGS) $(MLIB_INC) -I. -I.. $(EXTRA_INC)
 # -I/usr/X11R6/include/
 
@@ -44,3 +49,4 @@ swscale-example: swscale-example.o $(SWSLIB)
 ifneq ($(wildcard .depend),)
 include .depend
 endif
+
diff --git a/postproc/cs_test.c b/postproc/cs_test.c
index aad26e236a..0e5fe4a26e 100644
--- a/postproc/cs_test.c
+++ b/postproc/cs_test.c
@@ -28,6 +28,10 @@
 #define srcByte 0x55
 #define dstByte 0xBB
 
+#ifdef __APPLE_CC__
+#define memalign(x,y) malloc(y)
+#endif
+
 static int get_sws_cpuflags()
 {
     return (gCpuCaps.hasMMX ? SWS_CPU_CAPS_MMX : 0) |
@@ -281,4 +285,4 @@ main(int argc, char **argv)
 	
 	printf("%d converters passed, %d converters randomly overwrote memory\n", passedNum, failedNum);
 	return failedNum;
-}
\ No newline at end of file
+}
diff --git a/postproc/swscale.c b/postproc/swscale.c
index f1c9bdb25d..67a75a93af 100644
--- a/postproc/swscale.c
+++ b/postproc/swscale.c
@@ -1726,7 +1726,10 @@ int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange
 
 	yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
 	//FIXME factorize
-	
+
+#ifdef HAVE_ALTIVEC
+	yuv2rgb_altivec_init_tables (c, inv_table);
+#endif	
 	return 0;
 }
 
diff --git a/postproc/swscale_internal.h b/postproc/swscale_internal.h
index 63cc9cd8d8..c6d5a2f974 100644
--- a/postproc/swscale_internal.h
+++ b/postproc/swscale_internal.h
@@ -132,6 +132,19 @@ typedef struct SwsContext{
 	int dstW;
 	int esp;
 	uint64_t vRounder     __attribute__((aligned(8)));
+
+#ifdef HAVE_ALTIVEC
+
+  vector signed short   CY;
+  vector signed short   CRV;
+  vector signed short   CBU;
+  vector signed short   CGU;
+  vector signed short   CGV;
+  vector signed short   OY;
+  vector unsigned short CSHIFT;
+
+#endif
+
 } SwsContext;
 //FIXME check init (where 0)
 
diff --git a/postproc/swscale_template.c b/postproc/swscale_template.c
index 2d266f5655..000f2e291e 100644
--- a/postproc/swscale_template.c
+++ b/postproc/swscale_template.c
@@ -943,9 +943,15 @@ static inline void RENAME(yuv2packedX)(SwsContext *c, int16_t *lumFilter, int16_
 		break;
 #endif
 	default:
+#ifdef HAVE_ALTIVEC
+		altivec_yuv2packedX (c, lumFilter, lumSrc, lumFilterSize,
+			    chrFilter, chrSrc, chrFilterSize,
+			    dest, dstW, dstY);
+#else
 		yuv2packedXinC(c, lumFilter, lumSrc, lumFilterSize,
 			    chrFilter, chrSrc, chrFilterSize,
 			    dest, dstW, dstY);
+#endif
 		break;
 	}
 }
diff --git a/postproc/yuv2rgb.c b/postproc/yuv2rgb.c
index a8acec5f19..317ada954e 100644
--- a/postproc/yuv2rgb.c
+++ b/postproc/yuv2rgb.c
@@ -607,6 +607,14 @@ SwsFunc yuv2rgb_get_func_ptr (SwsContext *c)
 	if(t) return t;
     }
 #endif
+#ifdef HAVE_ALTIVEC
+    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
+    {
+	SwsFunc t = yuv2rgb_init_altivec(c);
+	if(t) return t;
+    }
+#endif
+
     MSG_WARN("No accelerated colorspace conversion found\n");
 
     switch(c->dstFormat){
diff --git a/postproc/yuv2rgb_altivec.c b/postproc/yuv2rgb_altivec.c
new file mode 100644
index 0000000000..2ccbfa5b33
--- /dev/null
+++ b/postproc/yuv2rgb_altivec.c
@@ -0,0 +1,836 @@
+/*
+  marc.hoffman@analog.com    March 8, 2004
+
+  Altivec Acceleration for Color Space Conversion revision 0.2
+
+  convert I420 YV12 to RGB in various formats,
+    it rejects images that are not in 420 formats
+    it rejects images that don't have widths of multiples of 16
+    it rejects images that don't have heights of multiples of 2
+  reject defers to C simulation codes.
+
+  lots of optimizations to be done here
+
+  1. need to fix saturation code, I just couldn't get it to fly with packs and adds.
+     so we currently use max min to clip
+
+  2. the inefficient use of chroma loading needs a bit of brushing up
+
+  3. analysis of pipeline stalls needs to be done, use shark to identify pipeline stalls
+
+
+  MODIFIED to calculate coeffs from currently selected color space.
+  MODIFIED core to be a macro which you spec the output format.
+  ADDED UYVY conversion which is never called due to some thing in SWSCALE.
+  CORRECTED algorithim selection to be strict on input formats.
+  ADDED runtime detection of altivec.
+
+  ADDED altivec_yuv2packedX vertical scl + RGB converter
+
+  March 27,2004
+  PERFORMANCE ANALYSIS
+
+  The C version use 25% of the processor or ~250Mips for D1 video rawvideo used as test
+  The ALTIVEC version uses 10% of the processor or ~100Mips for D1 video same sequence
+
+  720*480*30  ~10MPS
+
+  so we have roughly 10clocks per pixel this is too high something has to be wrong.
+
+  OPTIMIZED clip codes to utilize vec_max and vec_packs removing the need for vec_min.
+
+  OPTIMIZED DST OUTPUT cache/dma controls. we are pretty much
+  guaranteed to have the input video frame it was just decompressed so
+  it probably resides in L1 caches.  However we are creating the
+  output video stream this needs to use the DSTST instruction to
+  optimize for the cache.  We couple this with the fact that we are
+  not going to be visiting the input buffer again so we mark it Least
+  Recently Used.  This shaves 25% of the processor cycles off.
+
+  Now MEMCPY is the largest mips consumer in the system, probably due
+  to the inefficient X11 stuff.
+
+  GL libraries seem to be very slow on this machine 1.33Ghz PB running
+  Jaguar, this is not the case for my 1Ghz PB.  I thought it might be
+  a versioning issues, however i have libGL.1.2.dylib for both
+  machines. ((We need to figure this out now))
+
+  GL2 libraries work now with patch for RGB32
+
+  NOTE quartz vo driver ARGB32_to_RGB24 consumes 30% of the processor
+
+  Integrated luma prescaling adjustment for saturation/contrast/brightness adjustment. 
+
+*/
+#include <stdio.h>
+#include <stdlib.h>
+#include <inttypes.h>
+#include <assert.h>
+#include "config.h"
+#include "rgb2rgb.h"
+#include "swscale.h"
+#include "swscale_internal.h"
+#include "../mangle.h"
+#include "../libvo/img_format.h" //FIXME try to reduce dependency of such stuff
+
+#undef PROFILE_THE_BEAST
+#undef INC_SCALING
+
+typedef unsigned char ubyte;
+typedef signed char   sbyte;
+
+
+/* RGB interleaver, 16 planar pels 8-bit samples per channel in
+   homogeneous vector registers x0,x1,x2 are interleaved with the
+   following technique:
+
+      o0 = vec_mergeh (x0,x1);
+      o1 = vec_perm (o0, x2, perm_rgb_0);
+      o2 = vec_perm (o0, x2, perm_rgb_1);
+      o3 = vec_mergel (x0,x1);
+      o4 = vec_perm (o3,o2,perm_rgb_2);
+      o5 = vec_perm (o3,o2,perm_rgb_3);
+
+  perm_rgb_0:   o0(RG).h v1(B) --> o1*
+              0   1  2   3   4
+             rgbr|gbrg|brgb|rgbr
+             0010 0100 1001 0010
+             0102 3145 2673 894A
+
+  perm_rgb_1:   o0(RG).h v1(B) --> o2
+              0   1  2   3   4
+             gbrg|brgb|bbbb|bbbb
+             0100 1001 1111 1111
+             B5CD 6EF7 89AB CDEF
+
+  perm_rgb_2:   o3(RG).l o2(rgbB.l) --> o4*
+              0   1  2   3   4
+             gbrg|brgb|rgbr|gbrg
+             1111 1111 0010 0100
+             89AB CDEF 0182 3945
+
+  perm_rgb_2:   o3(RG).l o2(rgbB.l) ---> o5*
+              0   1  2   3   4
+             brgb|rgbr|gbrg|brgb
+             1001 0010 0100 1001
+             a67b 89cA BdCD eEFf
+
+*/
+static
+const vector unsigned char
+  perm_rgb_0 = (vector unsigned char)(0x00,0x01,0x10,0x02,0x03,0x11,0x04,0x05,
+				      0x12,0x06,0x07,0x13,0x08,0x09,0x14,0x0a),
+  perm_rgb_1 = (vector unsigned char)(0x0b,0x15,0x0c,0x0d,0x16,0x0e,0x0f,0x17,
+				      0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f),
+  perm_rgb_2 = (vector unsigned char)(0x10,0x11,0x12,0x13,0x14,0x15,0x16,0x17,
+				      0x00,0x01,0x18,0x02,0x03,0x19,0x04,0x05),
+  perm_rgb_3 = (vector unsigned char)(0x1a,0x06,0x07,0x1b,0x08,0x09,0x1c,0x0a,
+				      0x0b,0x1d,0x0c,0x0d,0x1e,0x0e,0x0f,0x1f);
+
+#define vec_merge3(x2,x1,x0,y0,y1,y2)    \
+do {					 \
+  typeof(x0) o0,o2,o3;			 \
+      o0 = vec_mergeh (x0,x1);		 \
+      y0 = vec_perm (o0, x2, perm_rgb_0);\
+      o2 = vec_perm (o0, x2, perm_rgb_1);\
+      o3 = vec_mergel (x0,x1);		 \
+      y1 = vec_perm (o3,o2,perm_rgb_2);	 \
+      y2 = vec_perm (o3,o2,perm_rgb_3);	 \
+} while(0)
+
+#define vec_mstrgb24(x0,x1,x2,ptr)        \
+do {					 \
+  typeof(x0) _0,_1,_2;			 \
+  vec_merge3 (x0,x1,x2,_0,_1,_2);	 \
+  vec_st (_0, 0, ptr++);		 \
+  vec_st (_1, 0, ptr++);		 \
+  vec_st (_2, 0, ptr++);		 \
+}  while (0);
+
+#define vec_mstbgr24(x0,x1,x2,ptr)       \
+do {					 \
+  typeof(x0) _0,_1,_2;			 \
+  vec_merge3 (x2,x1,x0,_0,_1,_2);	 \
+  vec_st (_0, 0, ptr++);		 \
+  vec_st (_1, 0, ptr++);		 \
+  vec_st (_2, 0, ptr++);		 \
+}  while (0);
+
+/* pack the pixels in rgb0 format
+   msb R
+   lsb 0
+*/
+#define vec_mstrgb32(T,x0,x1,x2,x3,ptr)						       \
+do {										       \
+  T _0,_1,_2,_3;								       \
+  _0 = vec_mergeh (x0,x1);							       \
+  _1 = vec_mergeh (x2,x3);					    	  	       \
+  _2 = (T)vec_mergeh ((vector unsigned short)_0,(vector unsigned short)_1);            \
+  _3 = (T)vec_mergel ((vector unsigned short)_0,(vector unsigned short)_1);            \
+  vec_st (_2, 0*16, (T *)ptr);						               \
+  vec_st (_3, 1*16, (T *)ptr);						       	       \
+  _0 = vec_mergel (x0,x1);							       \
+  _1 = vec_mergel (x2,x3);						       	       \
+  _2 = (T)vec_mergeh ((vector unsigned short)_0,(vector unsigned short)_1); 	       \
+  _3 = (T)vec_mergel ((vector unsigned short)_0,(vector unsigned short)_1); 	       \
+  vec_st (_2, 2*16, (T *)ptr);						       	       \
+  vec_st (_3, 3*16, (T *)ptr);						       	       \
+  ptr += 4;									       \
+}  while (0);
+
+/*
+
+  | 1     0       1.4021   | | Y |
+  | 1    -0.3441 -0.7142   |x| Cb|
+  | 1     1.7718  0	   | | Cr|
+
+
+  Y:      [-128 127]
+  Cb/Cr : [-128 127]
+
+  typical yuv conversion work on Y: 0-255 this version has been optimized for jpeg decode.
+
+*/
+
+
+
+
+#define vec_unh(x) \
+  (vector signed short) \
+    vec_perm(x,(typeof(x))(0),\
+             (vector unsigned char)(0x10,0x00,0x10,0x01,0x10,0x02,0x10,0x03,\
+                                    0x10,0x04,0x10,0x05,0x10,0x06,0x10,0x07))
+#define vec_unl(x) \
+  (vector signed short) \
+    vec_perm(x,(typeof(x))(0),\
+             (vector unsigned char)(0x10,0x08,0x10,0x09,0x10,0x0A,0x10,0x0B,\
+                                    0x10,0x0C,0x10,0x0D,0x10,0x0E,0x10,0x0F))
+
+#define vec_clip(x) \
+  vec_max (vec_min (x, (typeof(x))(255)), (typeof(x))(0))
+
+#define vec_packclp_a(x,y) \
+  (vector unsigned char)vec_pack (vec_clip (x), vec_clip (y))
+
+#define vec_packclp(x,y) \
+  (vector unsigned char)vec_packs \
+      ((vector unsigned short)vec_max (x,(vector signed short) (0)), \
+       (vector unsigned short)vec_max (y,(vector signed short) (0)))
+
+//#define out_pixels(a,b,c,ptr) vec_mstrgb32(typeof(a),((typeof (a))(0)),a,a,a,ptr)
+
+
+static inline cvtyuvtoRGB (SwsContext *c,
+			   vector signed short Y, vector signed short U, vector signed short V,
+			   vector signed short *R, vector signed short *G, vector signed short *B)
+{
+  vector signed   short vx,ux,uvx;
+
+  Y = vec_mradds (Y, c->CY, c->OY);
+
+  U = vec_sub (U,(vector signed short)(128));
+  V = vec_sub (V,(vector signed short)(128));
+
+  //   ux  = (CBU*(u<<c->CSHIFT)+0x4000)>>15;
+  ux = vec_sl (U, c->CSHIFT);
+  *B = vec_mradds (ux, c->CBU, Y);
+
+  // vx  = (CRV*(v<<c->CSHIFT)+0x4000)>>15;
+  vx = vec_sl (V, c->CSHIFT);
+  *R = vec_mradds (vx, c->CRV, Y);
+
+  // uvx = ((CGU*u) + (CGV*v))>>15;
+  uvx = vec_mradds (U, c->CGU, Y);
+  *G = vec_mradds (V, c->CGV, uvx);
+}
+
+
+/*
+  ------------------------------------------------------------------------------
+  CS converters
+  ------------------------------------------------------------------------------
+*/
+
+
+#define DEFCSP420_CVT(name,out_pixels)                                     \
+static int altivec_##name (SwsContext *c,                                  \
+				unsigned char **in, int *instrides,	   \
+				int srcSliceY,	int srcSliceH,		   \
+				unsigned char **oplanes, int *outstrides)  \
+{									   \
+  int w = c->srcW;							   \
+  int h = srcSliceH;							   \
+  int i,j;								   \
+  int instrides_scl[3];							   \
+  vector unsigned char y0,y1;						   \
+									   \
+  vector signed char  u,v;						   \
+									   \
+  vector signed short Y0,Y1,Y2,Y3;					   \
+  vector signed short U,V;						   \
+  vector signed short vx,ux,uvx;					   \
+  vector signed short vx0,ux0,uvx0;					   \
+  vector signed short vx1,ux1,uvx1;					   \
+  vector signed short R0,G0,B0;						   \
+  vector signed short R1,G1,B1;						   \
+  vector unsigned char R,G,B;						   \
+									   \
+  vector unsigned char *uivP, *vivP;			   		   \
+  vector unsigned char align_perm;					   \
+									   \
+  vector signed short 							   \
+    lCY  = c->CY,							   \
+    lOY  = c->OY,							   \
+    lCRV = c->CRV,							   \
+    lCBU = c->CBU,							   \
+    lCGU = c->CGU,							   \
+    lCGV = c->CGV;							   \
+									   \
+  vector unsigned short lCSHIFT = c->CSHIFT;				   \
+									   \
+  ubyte *y1i   = in[0];							   \
+  ubyte *y2i   = in[0]+w;						   \
+  ubyte *ui    = in[1];							   \
+  ubyte *vi    = in[2];							   \
+									   \
+  vector unsigned char *oute						   \
+    = (vector unsigned char *)						   \
+        (oplanes[0]+srcSliceY*outstrides[0]);				   \
+  vector unsigned char *outo						   \
+    = (vector unsigned char *)						   \
+        (oplanes[0]+srcSliceY*outstrides[0]+outstrides[0]);		   \
+									   \
+									   \
+  instrides_scl[0] = instrides[0];					   \
+  instrides_scl[1] = instrides[1]-w/2;  /* the loop moves ui by w/2 */	   \
+  instrides_scl[2] = instrides[2]-w/2;  /* the loop moves vi by w/2 */	   \
+									   \
+									   \
+  for (i=0;i<h/2;i++) {							   \
+    vec_dstst (outo, (0x02000002|(((w*3+32)/32)<<16)), 0);                 \
+    vec_dstst (oute, (0x02000002|(((w*3+32)/32)<<16)), 1);                 \
+									   \
+    for (j=0;j<w/16;j++) {						   \
+									   \
+      y0 = vec_ldl (0,y1i);						   \
+      y1 = vec_ldl (0,y2i);						   \
+      uivP = (vector unsigned char *)ui;				   \
+      vivP = (vector unsigned char *)vi;				   \
+									   \
+      align_perm = vec_lvsl (0, ui);					   \
+      u = (vector signed char)vec_perm (uivP[0], uivP[1], align_perm);	   \
+									   \
+      align_perm = vec_lvsl (0, vi);					   \
+      v = (vector signed char)vec_perm (vivP[0], vivP[1], align_perm);	   \
+									   \
+      u  = (vector signed char)vec_sub (u, (vector signed char)(128));	   \
+      v  = (vector signed char)vec_sub (v, (vector signed char)(128));	   \
+      U  = vec_unpackh (u);						   \
+      V  = vec_unpackh (v);						   \
+									   \
+									   \
+	Y0 = vec_unh (y0);						   \
+	Y1 = vec_unl (y0);						   \
+	Y2 = vec_unh (y1);						   \
+	Y3 = vec_unl (y1);						   \
+									   \
+        Y0 = vec_mradds (Y0, lCY, lOY);					   \
+        Y1 = vec_mradds (Y1, lCY, lOY);					   \
+        Y2 = vec_mradds (Y2, lCY, lOY);					   \
+        Y3 = vec_mradds (Y3, lCY, lOY);					   \
+									   \
+	/*   ux  = (CBU*(u<<CSHIFT)+0x4000)>>15 */			   \
+	ux = vec_sl (U, lCSHIFT);					   \
+	ux = vec_mradds (ux, lCBU, (vector signed short)(0));		   \
+	ux0  = vec_mergeh (ux,ux);					   \
+	ux1  = vec_mergel (ux,ux);					   \
+									   \
+	/* vx  = (CRV*(v<<CSHIFT)+0x4000)>>15;	*/			   \
+	vx = vec_sl (V, lCSHIFT);					   \
+	vx = vec_mradds (vx, lCRV, (vector signed short)(0));		   \
+	vx0  = vec_mergeh (vx,vx);					   \
+	vx1  = vec_mergel (vx,vx);					   \
+									   \
+	/* uvx = ((CGU*u) + (CGV*v))>>15 */				   \
+	uvx = vec_mradds (U, lCGU, (vector signed short)(0));		   \
+	uvx = vec_mradds (V, lCGV, uvx);				   \
+	uvx0 = vec_mergeh (uvx,uvx);					   \
+	uvx1 = vec_mergel (uvx,uvx);					   \
+									   \
+	R0 = vec_add (Y0,vx0);						   \
+	G0 = vec_add (Y0,uvx0);						   \
+	B0 = vec_add (Y0,ux0);						   \
+	R1 = vec_add (Y1,vx1);						   \
+	G1 = vec_add (Y1,uvx1);						   \
+	B1 = vec_add (Y1,ux1);						   \
+									   \
+	R  = vec_packclp (R0,R1);					   \
+	G  = vec_packclp (G0,G1);					   \
+	B  = vec_packclp (B0,B1);					   \
+									   \
+	out_pixels(R,G,B,oute);						   \
+									   \
+	R0 = vec_add (Y2,vx0);						   \
+	G0 = vec_add (Y2,uvx0);						   \
+	B0 = vec_add (Y2,ux0);						   \
+	R1 = vec_add (Y3,vx1);						   \
+	G1 = vec_add (Y3,uvx1);						   \
+	B1 = vec_add (Y3,ux1);						   \
+	R  = vec_packclp (R0,R1);					   \
+	G  = vec_packclp (G0,G1);					   \
+	B  = vec_packclp (B0,B1);					   \
+									   \
+									   \
+	out_pixels(R,G,B,outo);						   \
+									   \
+      y1i  += 16;							   \
+      y2i  += 16;							   \
+      ui   += 8;							   \
+      vi   += 8;							   \
+									   \
+    }									   \
+									   \
+    outo += (outstrides[0])>>4;					           \
+    oute += (outstrides[0])>>4;					           \
+									   \
+    ui    += instrides_scl[1];						   \
+    vi    += instrides_scl[2];						   \
+    y1i   += instrides_scl[0];						   \
+    y2i   += instrides_scl[0];						   \
+  }									   \
+  return srcSliceH;							   \
+}
+
+
+#define out_abgr(a,b,c,ptr)  vec_mstrgb32(typeof(a),((typeof (a))(0)),c,b,a,ptr)
+#define out_bgra(a,b,c,ptr)  vec_mstrgb32(typeof(a),c,b,a,((typeof (a))(0)),ptr)
+#define out_rgba(a,b,c,ptr)  vec_mstrgb32(typeof(a),a,b,c,((typeof (a))(0)),ptr)
+#define out_argb(a,b,c,ptr)  vec_mstrgb32(typeof(a),((typeof (a))(0)),a,b,c,ptr)
+#define out_rgb24(a,b,c,ptr) vec_mstrgb24(a,b,c,ptr)
+#define out_bgr24(a,b,c,ptr) vec_mstrgb24(c,b,a,ptr)
+
+DEFCSP420_CVT (yuv2_abgr32, out_abgr)
+DEFCSP420_CVT (yuv2_bgra32, out_argb)
+DEFCSP420_CVT (yuv2_rgba32, out_rgba)
+DEFCSP420_CVT (yuv2_argb32, out_argb)
+DEFCSP420_CVT (yuv2_rgb24,  out_rgb24)
+DEFCSP420_CVT (yuv2_bgr24,  out_bgr24)
+
+
+// uyvy|uyvy|uyvy|uyvy
+// 0123 4567 89ab cdef
+static
+const vector unsigned char
+  demux_u = (vector unsigned char)(0x10,0x00,0x10,0x00,
+				   0x10,0x04,0x10,0x04,
+				   0x10,0x08,0x10,0x08,
+				   0x10,0x0c,0x10,0x0c),
+  demux_v = (vector unsigned char)(0x10,0x02,0x10,0x02,
+				   0x10,0x06,0x10,0x06,
+				   0x10,0x0A,0x10,0x0A,
+				   0x10,0x0E,0x10,0x0E),
+  demux_y = (vector unsigned char)(0x10,0x01,0x10,0x03,
+				   0x10,0x05,0x10,0x07,
+				   0x10,0x09,0x10,0x0B,
+				   0x10,0x0D,0x10,0x0F);
+
+/*
+  this is so I can play live CCIR raw video
+*/
+static int altivec_uyvy_rgb32 (SwsContext *c,
+			       unsigned char **in, int *instrides,
+			       int srcSliceY,	int srcSliceH,
+			       unsigned char **oplanes, int *outstrides)
+{
+  int w = c->srcW;
+  int h = srcSliceH;
+  int i,j;
+  vector unsigned char uyvy;
+  vector signed   short Y,U,V;
+  vector signed   short vx,ux,uvx;
+  vector signed   short R0,G0,B0,R1,G1,B1;
+  vector unsigned char  R,G,B;
+  vector unsigned char *out;
+  ubyte *img;
+
+  img = in[0];
+  out = (vector unsigned char *)(oplanes[0]+srcSliceY*outstrides[0]);
+
+  for (i=0;i<h;i++) {
+    for (j=0;j<w/16;j++) {
+      uyvy = vec_ld (0, img);
+      U = (vector signed short)
+	vec_perm (uyvy, (vector unsigned char)(0), demux_u);
+
+      V = (vector signed short)
+	vec_perm (uyvy, (vector unsigned char)(0), demux_v);
+
+      Y = (vector signed short)
+	vec_perm (uyvy, (vector unsigned char)(0), demux_y);
+
+      cvtyuvtoRGB (c, Y,U,V,&R0,&G0,&B0);
+
+      uyvy = vec_ld (16, img);
+      U = (vector signed short)
+	vec_perm (uyvy, (vector unsigned char)(0), demux_u);
+
+      V = (vector signed short)
+	vec_perm (uyvy, (vector unsigned char)(0), demux_v);
+
+      Y = (vector signed short)
+	vec_perm (uyvy, (vector unsigned char)(0), demux_y);
+
+      cvtyuvtoRGB (c, Y,U,V,&R1,&G1,&B1);
+
+      R  = vec_packclp (R0,R1);
+      G  = vec_packclp (G0,G1);
+      B  = vec_packclp (B0,B1);
+
+      //      vec_mstbgr24 (R,G,B, out);
+      out_rgba (R,G,B,out);
+
+      img += 32;
+    }
+  }
+}
+
+
+
+/* Ok currently the acceleration routine only supports
+   inputs of widths a multiple of 16
+   and heights a multiple 2
+
+   So we just fall back to the C codes for this.
+*/
+SwsFunc yuv2rgb_init_altivec (SwsContext *c)
+{
+  if (!(c->flags & SWS_CPU_CAPS_ALTIVEC))    
+    return NULL;
+
+  /*
+    and this seems not to matter too much I tried a bunch of 
+    videos with abnormal widths and mplayer crashes else where.
+    mplayer -vo x11 -rawvideo on:w=350:h=240 raw-350x240.eyuv 
+    boom with X11 bad match.
+    
+  */
+  if ((c->srcW & 0xf) != 0)    return NULL;
+
+  switch (c->srcFormat) {
+  case IMGFMT_YVU9:
+  case IMGFMT_IF09:
+  case IMGFMT_YV12:
+  case IMGFMT_I420:
+  case IMGFMT_IYUV:
+  case IMGFMT_CLPL:
+  case IMGFMT_Y800:
+  case IMGFMT_Y8:
+  case IMGFMT_NV12:
+  case IMGFMT_NV21:
+    if ((c->srcH & 0x1) != 0)
+      return NULL;
+
+    switch(c->dstFormat){
+    case IMGFMT_RGB24:
+      MSG_WARN("ALTIVEC: Color Space RGB24\n");
+      return altivec_yuv2_rgb24;
+    case IMGFMT_BGR24:
+      MSG_WARN("ALTIVEC: Color Space BGR24\n");
+      return altivec_yuv2_bgr24;
+    case IMGFMT_RGB32:
+      MSG_WARN("ALTIVEC: Color Space ARGB32\n");
+      return altivec_yuv2_argb32;
+    case IMGFMT_BGR32:
+      MSG_WARN("ALTIVEC: Color Space BGRA32\n");
+      //      return profile_altivec_bgra32;
+
+      return altivec_yuv2_bgra32;
+    default: return NULL;
+    }
+    break;
+
+  case IMGFMT_UYVY:
+    switch(c->dstFormat){
+    case IMGFMT_RGB32:
+      MSG_WARN("ALTIVEC: Color Space UYVY -> RGB32\n");
+      return altivec_uyvy_rgb32;
+    case IMGFMT_RGB24:
+    case IMGFMT_BGR32:
+
+    default: return NULL;
+    }
+    break;
+
+  }
+  return NULL;
+}
+
+
+int yuv2rgb_altivec_init_tables (SwsContext *c, const int inv_table[4])
+{
+
+  vector signed short
+    CY  = (vector signed short)(0x7fff),
+    CRV = (vector signed short)(22972),
+    CBU = (vector signed short)(29029),
+    CGU = (vector signed short)(-11276),
+    CGV = (vector signed short)(-23400),
+    OY;
+
+  vector unsigned short CSHIFT = (vector unsigned short)(1);
+
+  vector signed short Y0;
+  int brightness = c->brightness,  contrast = c->contrast,  saturation = c->saturation;
+  int64_t crv  __attribute__ ((aligned(16)));
+  int64_t cbu  __attribute__ ((aligned(16)));
+  int64_t cgu  __attribute__ ((aligned(16)));
+  int64_t cgv  __attribute__ ((aligned(16)));
+  short tmp __attribute__ ((aligned(16)));
+
+  int64_t cy  = (1<<16)-1;
+  int64_t oy  = 0;
+
+  if ((c->flags & SWS_CPU_CAPS_ALTIVEC) == 0)
+    return;
+
+  crv = inv_table[0];
+  cbu = inv_table[1];
+  cgu = inv_table[2];
+  cgv = inv_table[3];
+
+#if 0
+  printf ("crv: %hvx\n", CRV);
+  printf ("cbu: %hvx\n", CBU);
+  printf ("cgv: %hvx\n", CGV);
+  printf ("cgu: %hvx\n", CGU);
+
+  printf ("contrast: %d,   brightness: %d, saturation: %d\n", contrast, brightness, saturation);
+
+  printf("%lld %lld %lld %lld %lld\n", cy, crv, cbu, cgu, cgv);
+#endif
+
+  cy = (cy *contrast             )>>17;
+  crv= (crv*contrast * saturation)>>32;
+  cbu= (cbu*contrast * saturation)>>32;
+  cgu= (cgu*contrast * saturation)>>32;
+  cgv= (cgv*contrast * saturation)>>32;
+
+  oy -= 256*brightness;
+
+
+  //printf("%llx %llx %llx %llx %llx\n", cy, crv, cbu, cgu, cgv);
+
+  //  vector signed short CBU,CRV,CGU,CGY,CY;
+  tmp = cy;
+  CY = vec_lde (0, &tmp);
+  CY  = vec_splat (CY, 0);
+
+  tmp = oy;
+  OY = vec_lde (0, &tmp);
+  OY  = vec_splat (OY, 0);
+
+  tmp = crv>>3;
+  CRV = vec_lde (0, &tmp);
+  CRV  = vec_splat (CRV, 0);
+  tmp = cbu>>3;
+  CBU = vec_lde (0, &tmp);
+  CBU  = vec_splat (CBU, 0);
+
+  tmp = -(cgu>>1);
+  CGU = vec_lde (0, &tmp);
+  CGU  = vec_splat (CGU, 0);
+  tmp = -(cgv>>1);
+  CGV = vec_lde (0, &tmp);
+  CGV  = vec_splat (CGV, 0);
+
+  CSHIFT = (vector unsigned short)(2);
+#if 1
+  c->CSHIFT = CSHIFT;
+  c->CY = CY;
+  c->OY = OY;
+  c->CRV = CRV;
+  c->CBU = CBU;
+  c->CGU = CGU;
+  c->CGV = CGV;
+#endif
+#if 1
+  printf ("cy:  %hvx\n", CY);
+  printf ("oy:  %hvx\n", OY);
+  printf ("crv: %hvx\n", CRV);
+  printf ("cbu: %hvx\n", CBU);
+  printf ("cgv: %hvx\n", CGV);
+  printf ("cgu: %hvx\n", CGU);
+#endif
+}
+
+
+void
+altivec_yuv2packedX (SwsContext *c,
+		       int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
+		       int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
+		       uint8_t *dest, int dstW, int dstY)
+{
+  int i,j;
+  short tmp __attribute__((aligned (16)));
+  short *p;
+  short *f;
+  vector signed short X,X0,X1,Y0,U0,V0,Y1,U1,V1,U,V;
+  vector signed short R0,G0,B0,R1,G1,B1;
+
+  vector unsigned char R,G,B,pels[3];
+  vector unsigned char *out,*nout;
+  vector signed short   RND = (vector signed short)(1<<3);
+  vector unsigned short SCL = (vector unsigned short)(4);
+  unsigned long scratch[16] __attribute__ ((aligned (16)));
+
+  vector signed short *vYCoeffsBank, *vCCoeffsBank;
+
+  vector signed short *YCoeffs, *CCoeffs;
+
+  vYCoeffsBank = malloc (sizeof (vector signed short)*lumFilterSize*dstW);
+  vCCoeffsBank = malloc (sizeof (vector signed short)*chrFilterSize*dstW);
+
+  for (i=0;i<lumFilterSize*dstW;i++) {
+    tmp = c->vLumFilter[i];
+    p = &vYCoeffsBank[i];
+    for (j=0;j<8;j++)
+      p[j] = tmp;
+  }
+
+  for (i=0;i<chrFilterSize*dstW;i++) {
+    tmp = c->vChrFilter[i];
+    p = &vCCoeffsBank[i];
+    for (j=0;j<8;j++)
+      p[j] = tmp;
+  }
+
+  YCoeffs = vYCoeffsBank+dstY*lumFilterSize;
+  CCoeffs = vCCoeffsBank+dstY*chrFilterSize;
+
+  out = (vector unsigned char *)dest;
+
+  for(i=0; i<dstW; i+=16){
+    Y0 = RND;
+    Y1 = RND;
+    /* extract 16 coeffs from lumSrc */
+    for(j=0; j<lumFilterSize; j++) {
+      X0 = vec_ld (0,  &lumSrc[j][i]);
+      X1 = vec_ld (16, &lumSrc[j][i]);
+      Y0 = vec_mradds (X0, YCoeffs[j], Y0);
+      Y1 = vec_mradds (X1, YCoeffs[j], Y1);
+    }
+
+    U = RND;
+    V = RND;
+    /* extract 8 coeffs from U,V */
+    for(j=0; j<chrFilterSize; j++) {
+      X  = vec_ld (0, &chrSrc[j][i/2]);
+      U  = vec_mradds (X, CCoeffs[j], U);
+      X  = vec_ld (0, &chrSrc[j][i/2+2048]);
+      V  = vec_mradds (X, CCoeffs[j], V);
+    }
+
+    /* scale and clip signals */
+    Y0 = vec_sra (Y0, SCL);
+    Y1 = vec_sra (Y1, SCL);
+    U  = vec_sra (U,  SCL);
+    V  = vec_sra (V,  SCL);
+
+    Y0 = vec_clip (Y0);
+    Y1 = vec_clip (Y1);
+    U  = vec_clip (U);
+    V  = vec_clip (V);
+
+    /* now we have
+      Y0= y0 y1 y2 y3 y4 y5 y6 y7     Y1= y8 y9 y10 y11 y12 y13 y14 y15
+      U= u0 u1 u2 u3 u4 u5 u6 u7      V= v0 v1 v2 v3 v4 v5 v6 v7
+
+      Y0= y0 y1 y2 y3 y4 y5 y6 y7    Y1= y8 y9 y10 y11 y12 y13 y14 y15
+      U0= u0 u0 u1 u1 u2 u2 u3 u3    U1= u4 u4 u5 u5 u6 u6 u7 u7
+      V0= v0 v0 v1 v1 v2 v2 v3 v3    V1= v4 v4 v5 v5 v6 v6 v7 v7
+    */
+
+    U0 = vec_mergeh (U,U);
+    V0 = vec_mergeh (V,V);
+
+    U1 = vec_mergel (U,U);
+    V1 = vec_mergel (V,V);
+
+    cvtyuvtoRGB (c, Y0,U0,V0,&R0,&G0,&B0);
+    cvtyuvtoRGB (c, Y1,U1,V1,&R1,&G1,&B1);
+
+    R  = vec_packclp (R0,R1);
+    G  = vec_packclp (G0,G1);
+    B  = vec_packclp (B0,B1);
+
+    out_rgba (R,G,B,out);
+  }
+
+  if (i < dstW) {
+    i -= 16;
+
+    Y0 = RND;
+    Y1 = RND;
+    /* extract 16 coeffs from lumSrc */
+    for(j=0; j<lumFilterSize; j++) {
+      X0 = vec_ld (0,  &lumSrc[j][i]);
+      X1 = vec_ld (16, &lumSrc[j][i]);
+      Y0 = vec_mradds (X0, YCoeffs[j], Y0);
+      Y1 = vec_mradds (X1, YCoeffs[j], Y1);
+    }
+
+    U = RND;
+    V = RND;
+    /* extract 8 coeffs from U,V */
+    for(j=0; j<chrFilterSize; j++) {
+      X  = vec_ld (0, &chrSrc[j][i/2]);
+      U  = vec_mradds (X, CCoeffs[j], U);
+      X  = vec_ld (0, &chrSrc[j][i/2+2048]);
+      V  = vec_mradds (X, CCoeffs[j], V);
+    }
+
+    /* scale and clip signals */
+    Y0 = vec_sra (Y0, SCL);
+    Y1 = vec_sra (Y1, SCL);
+    U  = vec_sra (U,  SCL);
+    V  = vec_sra (V,  SCL);
+
+    Y0 = vec_clip (Y0);
+    Y1 = vec_clip (Y1);
+    U  = vec_clip (U);
+    V  = vec_clip (V);
+
+    /* now we have
+       Y0= y0 y1 y2 y3 y4 y5 y6 y7     Y1= y8 y9 y10 y11 y12 y13 y14 y15
+       U= u0 u1 u2 u3 u4 u5 u6 u7      V= v0 v1 v2 v3 v4 v5 v6 v7
+
+       Y0= y0 y1 y2 y3 y4 y5 y6 y7    Y1= y8 y9 y10 y11 y12 y13 y14 y15
+       U0= u0 u0 u1 u1 u2 u2 u3 u3    U1= u4 u4 u5 u5 u6 u6 u7 u7
+       V0= v0 v0 v1 v1 v2 v2 v3 v3    V1= v4 v4 v5 v5 v6 v6 v7 v7
+    */
+
+    U0 = vec_mergeh (U,U);
+    V0 = vec_mergeh (V,V);
+
+    U1 = vec_mergel (U,U);
+    V1 = vec_mergel (V,V);
+
+    cvtyuvtoRGB (c, Y0,U0,V0,&R0,&G0,&B0);
+    cvtyuvtoRGB (c, Y1,U1,V1,&R1,&G1,&B1);
+
+    R  = vec_packclp (R0,R1);
+    G  = vec_packclp (G0,G1);
+    B  = vec_packclp (B0,B1);
+
+    nout = (vector unsigned char *)scratch;
+    out_rgba (R,G,B,nout);
+
+    memcpy (&((uint32_t*)dest)[i], scratch, (dstW-i)/4);
+  }
+
+  if (vYCoeffsBank) free (vYCoeffsBank);
+  if (vCCoeffsBank) free (vCCoeffsBank);
+
+}
+
+