--- xvid.c	2002/07/21 23:34:08	1.32
+++ xvid.c	2002/10/05 21:33:39	1.33.2.6
@@ -37,7 +37,7 @@
  *  - 22.12.2001  API change: added xvid_init() - Isibaar
  *  - 16.12.2001	inital version; (c)2001 peter ross <pross@cs.rmit.edu.au>
  *
- *  $Id: xvid.c,v 1.32 2002/07/21 23:34:08 chl Exp $
+ *  $Id: xvid.c,v 1.33.2.6 2002/10/05 21:33:39 Isibaar Exp $
  *
  ****************************************************************************/
 
@@ -50,6 +50,7 @@
 #include "image/colorspace.h"
 #include "image/interpolate8x8.h"
 #include "utils/mem_transfer.h"
+#include "utils/mbfunctions.h"
 #include "quant/quant_h263.h"
 #include "quant/quant_mpeg4.h"
 #include "motion/motion.h"
@@ -216,11 +217,24 @@
 	transfer_16to8add  = transfer_16to8add_c;
 	transfer8x8_copy   = transfer8x8_copy_c;
 
+	/* Interlacing functions */
+	MBFieldTest = MBFieldTest_c;
+
 	/* Image interpolation related functions */
 	interpolate8x8_halfpel_h  = interpolate8x8_halfpel_h_c;
 	interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_c;
 	interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c;
 
+	interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c;
+	interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c;
+	interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c;
+
+	interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c;
+	interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c;
+
+	interpolate8x8_avg2 = interpolate8x8_avg2_c;
+	interpolate8x8_avg4 = interpolate8x8_avg4_c;
+
 	/* Initialize internal colorspace transformation tables */
 	colorspace_init();
 
@@ -245,11 +259,13 @@
 	/* Functions used in motion estimation algorithms */
 	calc_cbp = calc_cbp_c;
 	sad16    = sad16_c;
-	sad16bi  = sad16bi_c;
 	sad8     = sad8_c;
-	dev16    = dev16_c;
+	sad16bi  = sad16bi_c;
 	sad8bi   = sad8bi_c;
-	Halfpel8_Refine = Halfpel8_Refine_c;
+	dev16    = dev16_c;
+	sad16v	 = sad16v_c;
+	
+//	Halfpel8_Refine = Halfpel8_Refine_c;
 
 #ifdef ARCH_X86
 	if ((cpu_flags & XVID_CPU_MMX) > 0) {
@@ -280,12 +296,20 @@
 		transfer_16to8add  = transfer_16to8add_mmx;
 		transfer8x8_copy   = transfer8x8_copy_mmx;
 
+		/* Interlacing Functions */
+		MBFieldTest = MBFieldTest_mmx;
 
 		/* Image Interpolation related functions */
 		interpolate8x8_halfpel_h  = interpolate8x8_halfpel_h_mmx;
 		interpolate8x8_halfpel_v  = interpolate8x8_halfpel_v_mmx;
 		interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx;
 
+		interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx;
+		interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx;
+
+		interpolate8x8_avg2 = interpolate8x8_avg2_mmx;
+		interpolate8x8_avg4 = interpolate8x8_avg4_mmx;
+
 		/* Image RGB->YV12 related functions */
 		rgb24_to_yv12 = rgb24_to_yv12_mmx;
 		rgb32_to_yv12 = rgb32_to_yv12_mmx;
@@ -303,10 +327,22 @@
 		calc_cbp = calc_cbp_mmx;
 		sad16    = sad16_mmx;
 		sad8     = sad8_mmx;
+		sad16bi = sad16bi_mmx;
+		sad8bi  = sad8bi_mmx;
 		dev16    = dev16_mmx;
+		sad16v	 = sad16v_mmx;
 
 	}
 
+	/* these 3dnow functions are faster than mmx, but slower than xmm. */
+	if ((cpu_flags & XVID_CPU_3DNOW) > 0) {
+
+		/* ME functions */
+		sad16bi = sad16bi_3dn;
+		sad8bi  = sad8bi_3dn;
+	}
+
+
 	if ((cpu_flags & XVID_CPU_MMXEXT) > 0) {
 
 		/* Inverse DCT */
@@ -329,9 +365,12 @@
 
 		/* ME functions */
 		sad16 = sad16_xmm;
-		sad16bi = sad16bi_xmm;
 		sad8  = sad8_xmm;
+		sad16bi = sad16bi_xmm;
+		sad8bi  = sad8bi_xmm;
 		dev16 = dev16_xmm;
+		sad16v	 = sad16v_xmm;
+		fprintf(stderr,"sad16v=XMM\n");
 
 	}
 
@@ -378,7 +417,7 @@
 	  sad16bi = sad16bi_ia64;
 	  sad8 = sad8_ia64;
 	  dev16 = dev16_ia64;
-	  Halfpel8_Refine = Halfpel8_Refine_ia64;
+//	  Halfpel8_Refine = Halfpel8_Refine_ia64;
 	  quant_intra = quant_intra_ia64;
 	  dequant_intra = dequant_intra_ia64;
 	  quant_inter = quant_inter_ia64;