--- xvid.c	2003/10/07 13:02:35	1.45.2.16
+++ xvid.c	2003/12/17 17:07:38	1.45.2.25
@@ -19,7 +19,7 @@
  *  along with this program ; if not, write to the Free Software
  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  *
- * $Id: xvid.c,v 1.45.2.16 2003/10/07 13:02:35 edgomez Exp $
+ * $Id: xvid.c,v 1.45.2.25 2003/12/17 17:07:38 Isibaar Exp $
  *
  ****************************************************************************/
 
@@ -46,6 +46,7 @@
 #include "utils/timer.h"
 #include "bitstream/mbcoding.h"
 #include "image/qpel.h"
+#include "image/postprocessing.h"
 
 #if defined(_DEBUG)
 unsigned int xvid_debug = 0; /* xvid debug mask */
@@ -280,15 +281,14 @@
 	yv12_to_uyvyi   = yv12_to_uyvyi_c;
 
 	/* Functions used in motion estimation algorithms */
-	calc_cbp = calc_cbp_c;
-	sad16    = sad16_c;
-	sad8     = sad8_c;
-	sad16bi  = sad16bi_c;
-	sad8bi   = sad8bi_c;
-	dev16    = dev16_c;
-	sad16v	 = sad16v_c;
-
-/*	Halfpel8_Refine = Halfpel8_Refine_c; */
+	calc_cbp   = calc_cbp_c;
+	sad16      = sad16_c;
+	sad8       = sad8_c;
+	sad16bi    = sad16bi_c;
+	sad8bi     = sad8bi_c;
+	dev16      = dev16_c;
+	sad16v	   = sad16v_c;
+	sse8_16bit = sse8_16bit_c;
 
 #if defined(ARCH_IS_IA32)
 
@@ -308,7 +308,7 @@
 	if ((cpu_flags & XVID_CPU_MMX)) {
 
 		/* Forward and Inverse Discrete Cosine Transformation functions */
-		fdct = fdct_mmx;
+		fdct = fdct_mmx_skal;
 		idct = idct_mmx;
 
 		/* Qpel stuff */
@@ -380,6 +380,7 @@
 		sad8bi  = sad8bi_mmx;
 		dev16    = dev16_mmx;
 		sad16v	 = sad16v_mmx;
+		sse8_16bit = sse8_16bit_mmx;
 	}
 
 	/* these 3dnow functions are faster than mmx, but slower than xmm. */
@@ -398,7 +399,8 @@
 
 	if ((cpu_flags & XVID_CPU_MMXEXT)) {
 
-		/* Inverse DCT */
+		/* DCT */
+		fdct = fdct_xmm_skal;
 		idct = idct_xmm;
 
 		/* Interpolation */
@@ -478,8 +480,6 @@
 		interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne;
 	}
 
-#if defined(EXPERIMENTAL_SSE2_CODE) /* mark the whole SSE2 stuff as experimental. At least on
-									   my P4, it crashes... */
 	if ((cpu_flags & XVID_CPU_SSE2)) {
 
 		calc_cbp = calc_cbp_sse2;
@@ -490,17 +490,15 @@
 		dequant_h263_intra = dequant_h263_intra_sse2;
 		dequant_h263_inter = dequant_h263_inter_sse2;
 
-		/* ME; slower than xmm */
+		/* SAD operators */
 		sad16    = sad16_sse2;
 		dev16    = dev16_sse2;
-		/* Forward and Inverse DCT */
-#if 0 /* Both function are known to be unprecise, better keep them deactivated */
-		idct  = idct_sse2;
-		fdct = fdct_sse2;
-#endif
+
+		/* DCT operators
+		 * no iDCT because it's not "Walken matching" */
+		fdct = fdct_sse2_skal;
 	}
-#endif
-#endif
+#endif /* ARCH_IS_IA32 */
 
 #if defined(ARCH_IS_IA64)
 	if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */
@@ -561,7 +559,7 @@
 		return XVID_ERR_VERSION;
 
 	info->actual_version = XVID_VERSION;
-	info->build = "dev-api-4";
+	info->build = "xvid-1.0.0";
 	info->cpu_flags = detect_cpu_flags();
 
 #if defined(_SMP) && defined(WIN32)