--- mem_transfer_mmx.asm	2004/07/24 11:46:08	1.13
+++ mem_transfer_mmx.asm	2005/09/13 12:12:15	1.17
@@ -21,7 +21,7 @@
 ; *  along with this program ; if not, write to the Free Software
 ; *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
 ; *
-; * $Id: mem_transfer_mmx.asm,v 1.13 2004/07/24 11:46:08 edgomez Exp $
+; * $Id: mem_transfer_mmx.asm,v 1.17 2005/09/13 12:12:15 suxen_drol Exp $
 ; *
 ; ***************************************************************************/
 
@@ -29,10 +29,19 @@
 
 %macro cglobal 1
 	%ifdef PREFIX
-		global _%1
-		%define %1 _%1
+		%ifdef MARK_FUNCS
+			global _%1:function %1.endfunc-%1
+			%define %1 _%1:function %1.endfunc-%1
+		%else
+			global _%1
+			%define %1 _%1
+		%endif
 	%else
-		global %1
+		%ifdef MARK_FUNCS
+			global %1:function %1.endfunc-%1
+		%else
+			global %1
+		%endif
 	%endif
 %endmacro
 
@@ -62,8 +71,10 @@
 cglobal transfer_8to16subro_mmx
 cglobal transfer_8to16sub2_mmx
 cglobal transfer_8to16sub2_xmm
+cglobal transfer_8to16sub2ro_xmm
 cglobal transfer_16to8add_mmx
 cglobal transfer8x8_copy_mmx
+cglobal transfer8x4_copy_mmx
 
 ;-----------------------------------------------------------------------------
 ;
@@ -102,6 +113,7 @@
   COPY_8_TO_16 2
   COPY_8_TO_16 3
   ret
+.endfunc
 
 ;-----------------------------------------------------------------------------
 ;
@@ -137,6 +149,7 @@
   lea ecx,[ecx+2*edx]
   COPY_16_TO_8 3
   ret
+.endfunc
 
 ;-----------------------------------------------------------------------------
 ;
@@ -200,6 +213,7 @@
 
   pop ebx
   ret
+.endfunc
 
 
 ALIGN 16
@@ -218,6 +232,7 @@
 
   pop ebx
   ret
+.endfunc
 
 
 ;-----------------------------------------------------------------------------
@@ -316,6 +331,7 @@
   pop esi
   pop ebx
   ret
+.endfunc
 
 ;-----------------------------------------------------------------------------
 ;
@@ -383,6 +399,75 @@
   pop esi
   pop ebx
   ret
+.endfunc
+
+
+;-----------------------------------------------------------------------------
+;
+; void transfer_8to16sub2ro_xmm(int16_t * const dct,
+;				const uint8_t * const cur,
+;				const uint8_t * ref1,
+;				const uint8_t * ref2,
+;				const uint32_t stride)
+;
+;-----------------------------------------------------------------------------
+
+%macro COPY_8_TO_16_SUB2RO_SSE 1
+  movq mm0, [eax]      ; cur
+  movq mm2, [eax+edx]
+  movq mm1, mm0
+  movq mm3, mm2
+
+  punpcklbw mm0, mm7
+  punpcklbw mm2, mm7
+  movq mm4, [ebx]     ; ref1
+  pavgb mm4, [esi]     ; ref2
+  punpckhbw mm1, mm7
+  punpckhbw mm3, mm7
+  movq mm5, [ebx+edx] ; ref
+  pavgb mm5, [esi+edx] ; ref2
+
+  movq mm6, mm4
+  punpcklbw mm4, mm7
+  punpckhbw mm6, mm7
+  psubsw mm0, mm4
+  psubsw mm1, mm6
+  lea esi, [esi+2*edx]
+  movq mm6, mm5
+  punpcklbw mm5, mm7
+  punpckhbw mm6, mm7
+  psubsw mm2, mm5
+  lea eax, [eax+2*edx]
+  psubsw mm3, mm6
+  lea ebx, [ebx+2*edx]
+
+  movq [ecx+%1*32+ 0], mm0 ; dst
+  movq [ecx+%1*32+ 8], mm1
+  movq [ecx+%1*32+16], mm2
+  movq [ecx+%1*32+24], mm3
+%endmacro
+
+ALIGN 16
+transfer_8to16sub2ro_xmm:
+  pxor mm7, mm7
+  mov ecx, [esp  + 4] ; Dst
+  mov eax, [esp  + 8] ; Cur
+  push ebx
+  mov ebx, [esp+4+12] ; Ref1
+  push esi
+  mov esi, [esp+8+16] ; Ref2
+  mov edx, [esp+8+20] ; Stride
+
+  COPY_8_TO_16_SUB2RO_SSE 0
+  COPY_8_TO_16_SUB2RO_SSE 1
+  COPY_8_TO_16_SUB2RO_SSE 2
+  COPY_8_TO_16_SUB2RO_SSE 3
+
+  pop esi
+  pop ebx
+  ret
+.endfunc
+
 
 ;-----------------------------------------------------------------------------
 ;
@@ -427,6 +512,7 @@
   lea ecx,[ecx+2*edx]
   COPY_16_TO_8_ADD 3
   ret
+.endfunc
 
 ;-----------------------------------------------------------------------------
 ;
@@ -459,3 +545,26 @@
   lea ecx,[ecx+2*edx]
   COPY_8_TO_8
   ret
+.endfunc
+
+;-----------------------------------------------------------------------------
+;
+; void transfer8x4_copy_mmx(uint8_t * const dst,
+;					const uint8_t * const src,
+;					const uint32_t stride);
+;
+;
+;-----------------------------------------------------------------------------
+
+ALIGN 16
+transfer8x4_copy_mmx:
+  mov ecx, [esp+ 4] ; Dst
+  mov eax, [esp+ 8] ; Src
+  mov edx, [esp+12] ; Stride
+
+  COPY_8_TO_8
+  lea ecx,[ecx+2*edx]
+  COPY_8_TO_8
+  ret
+.endfunc
+