chromas converstion: optimize MMX/SSE acceleration even more by leveraging out...

chromas converstion: optimize MMX/SSE acceleration even more by leveraging out of order execution, basically avoid CPU stalls by using more registers and by spacing memory fetch and use as much as possible (CPU will run available instructions while waiting for memory fecth to complete)

modules/video_chroma/i420_yuy2.c

@@ -158,8 +158,17 @@ static int Activate( vlc_object_t *p_this )
     return 0;
 }
 
-/* Following functions are local */
+#if 0
+static inline unsigned long long read_cycles(void)
+{
+    unsigned long long v;
+    __asm__ __volatile__("rdtsc" : "=A" (v): );
+
+    return v;
+}
+#endif
 
+/* Following functions are local */
 /*****************************************************************************
  * I420_YUY2: planar YUV 4:2:0 to packed YUYV 4:2:2
  *****************************************************************************/
@@ -306,9 +315,10 @@ static void I420_YUY2( vout_thread_t *p_vout, picture_t *p_source,
 
 #else // defined(MODULE_NAME_IS_i420_yuy2_sse2)
     /*
-    ** SSE2 128 bytes fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster 
     ** if memory access is 16 bytes aligned
     */
+
     if( 0 == (15 & (p_source->p[Y_PLANE].i_pitch|p_dest->p->i_pitch|
         ((int)p_line2|(int)p_y2))) )
     {
@@ -366,6 +376,7 @@ static void I420_YUY2( vout_thread_t *p_vout, picture_t *p_source,
             p_line2 += i_dest_margin;
         }
     }
+
 #endif // defined(MODULE_NAME_IS_i420_yuy2_sse2)
 }
 
@@ -489,6 +500,10 @@ static void I420_YVYU( vout_thread_t *p_vout, picture_t *p_source,
             MMX_CALL( MMX_YUV420_YVYU );
 #endif
         }
+        for( i_x = ( p_vout->render.i_width % 8 ) / 2; i_x-- ; )
+        {
+            C_YUV420_YVYU( );
+        }
 
         p_y1 += i_source_margin;
         p_y2 += i_source_margin;
@@ -508,7 +523,7 @@ static void I420_YVYU( vout_thread_t *p_vout, picture_t *p_source,
 
 #else // defined(MODULE_NAME_IS_i420_yuy2_sse2)
     /*
-    ** SSE2 128 bytes fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster 
     ** if memory access is 16 bytes aligned
     */
     if( 0 == (15 & (p_source->p[Y_PLANE].i_pitch|p_dest->p->i_pitch|
@@ -714,7 +729,7 @@ static void I420_UYVY( vout_thread_t *p_vout, picture_t *p_source,
 
 #else // defined(MODULE_NAME_IS_i420_yuy2_sse2)
     /*
-    ** SSE2 128 bytes fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster 
     ** if memory access is 16 bytes aligned
     */
     if( 0 == (15 & (p_source->p[Y_PLANE].i_pitch|p_dest->p->i_pitch|
@@ -832,6 +847,10 @@ static void I420_cyuv( vout_thread_t *p_vout, picture_t *p_source,
             MMX_CALL( MMX_YUV420_UYVY );
 #endif
         }
+        for( i_x = ( p_vout->render.i_width % 8 ) / 2; i_x-- ; )
+        {
+            C_YUV420_UYVY( );
+        }
 
         p_y1 += i_source_margin;
         p_y2 += i_source_margin;
@@ -847,7 +866,7 @@ static void I420_cyuv( vout_thread_t *p_vout, picture_t *p_source,
 
 #else // defined(MODULE_NAME_IS_i420_yuy2_sse2)
     /*
-    ** SSE2 128 bytes fetch/store instructions are faster 
+    ** SSE2 128 bits fetch/store instructions are faster 
     ** if memory access is 16 bytes aligned
     */
     if( 0 == (15 & (p_source->p[Y_PLANE].i_pitch|p_dest->p->i_pitch|