h->mc.copy()

git-svn-id: svn://svn.videolan.org/x264/trunk@442 df754926-b1dd-0310-bc7b-ec298dee348c

diff --git a/common/amd64/mc-a.asm b/common/amd64/mc-a.asm
index debb6aef5c6577c3f875bdf63ae9cfe66be285cf..c6f2c820c9d58c169e1bbfd9b7b4dcc7e796d5ad 100644 (file)
--- a/common/amd64/mc-a.asm
+++ b/common/amd64/mc-a.asm
@@ -288,20 +288,20 @@ x264_pixel_avg_weight_4x4_mmxext:
 
 ALIGN 16
 ;-----------------------------------------------------------------------------
-;  void x264_mc_copy_w4_mmxext( uint8_t *src, int i_src_stride,
-;                               uint8_t *dst, int i_dst_stride, int i_height )
+;  void x264_mc_copy_w4_mmxext( uint8_t *dst, int i_dst_stride,
+;                               uint8_t *src, int i_src_stride, int i_height )
 ;-----------------------------------------------------------------------------
 x264_mc_copy_w4_mmxext:
     mov     eax, parm5d         ; i_height
     
 ALIGN 4
 .height_loop
-    mov     r10d, [parm1q]
-    mov     r11d, [parm1q+parm2q]
-    mov     [parm3q], r10d
-    mov     [parm3q+parm4q], r11d
-    lea     parm1q, [parm1q+parm2q*2]
+    mov     r10d, [parm3q]
+    mov     r11d, [parm3q+parm4q]
+    mov     [parm1q], r10d
+    mov     [parm1q+parm2q], r11d
     lea     parm3q, [parm3q+parm4q*2]
+    lea     parm1q, [parm1q+parm2q*2]
     dec     eax
     dec     eax
     jne     .height_loop
@@ -310,27 +310,27 @@ ALIGN 4
 
 ALIGN 16
 ;-----------------------------------------------------------------------------
-;   void x264_mc_copy_w8_mmxext( uint8_t *src, int i_src_stride,
-;                                uint8_t *dst, int i_dst_stride, int i_height )
+;   void x264_mc_copy_w8_mmxext( uint8_t *dst, int i_dst_stride,
+;                                uint8_t *src, int i_src_stride, int i_height )
 ;-----------------------------------------------------------------------------
 x264_mc_copy_w8_mmxext:
     mov     eax, parm5d         ; i_height
 
-    lea     r10, [parm2q+parm2q*2] ; 3 * i_src_stride
-    lea     r11, [parm4q+parm4q*2] ; 3 * i_dst_stride
+    lea     r10, [parm4q+parm4q*2] ; 3 * i_src_stride
+    lea     r11, [parm2q+parm2q*2] ; 3 * i_dst_stride
 
 ALIGN 4
 .height_loop
-    movq    mm0, [parm1q]
-    movq    mm1, [parm1q+parm2q]
-    movq    mm2, [parm1q+parm2q*2]
-    movq    mm3, [parm1q+r10]
-    movq    [parm3q], mm0
-    movq    [parm3q+parm4q], mm1
-    movq    [parm3q+parm4q*2], mm2
-    movq    [parm3q+r11], mm3
-    lea     parm1q, [parm1q+parm2q*4]
+    movq    mm0, [parm3q]
+    movq    mm1, [parm3q+parm4q]
+    movq    mm2, [parm3q+parm4q*2]
+    movq    mm3, [parm3q+r10]
+    movq    [parm1q], mm0
+    movq    [parm1q+parm2q], mm1
+    movq    [parm1q+parm2q*2], mm2
+    movq    [parm1q+r11], mm3
     lea     parm3q, [parm3q+parm4q*4]
+    lea     parm1q, [parm1q+parm2q*4]
     
     sub     eax, byte 4
     jnz     .height_loop
@@ -339,35 +339,35 @@ ALIGN 4
 
 ALIGN 16
 ;-----------------------------------------------------------------------------
-;   void x264_mc_copy_w16_mmxext( uint8_t *src, int i_src_stride,
-;                                 uint8_t *dst, int i_dst_stride, int i_height )
+;   void x264_mc_copy_w16_mmxext( uint8_t *dst, int i_dst_stride,
+;                                 uint8_t *src, int i_src_stride, int i_height )
 ;-----------------------------------------------------------------------------
 x264_mc_copy_w16_mmxext:
     mov     eax, parm5d         ; i_height
     
-    lea     r10, [parm2q+parm2q*2] ; 3 * i_src_stride
-    lea     r11, [parm4q+parm4q*2] ; 3 * i_dst_stride
+    lea     r10, [parm4q+parm4q*2] ; 3 * i_src_stride
+    lea     r11, [parm2q+parm2q*2] ; 3 * i_dst_stride
 
 ALIGN 4
 .height_loop
-    movq    mm0, [parm1q]
-    movq    mm1, [parm1q+8]
-    movq    mm2, [parm1q+parm2q]
-    movq    mm3, [parm1q+parm2q+8]
-    movq    mm4, [parm1q+parm2q*2]
-    movq    mm5, [parm1q+parm2q*2+8]
-    movq    mm6, [parm1q+r10]
-    movq    mm7, [parm1q+r10+8]
-    movq    [parm3q], mm0
-    movq    [parm3q+8], mm1
-    movq    [parm3q+parm4q], mm2
-    movq    [parm3q+parm4q+8], mm3
-    movq    [parm3q+parm4q*2], mm4
-    movq    [parm3q+parm4q*2+8], mm5
-    movq    [parm3q+r11], mm6
-    movq    [parm3q+r11+8], mm7
-    lea     parm1q, [parm1q+parm2q*4]
+    movq    mm0, [parm3q]
+    movq    mm1, [parm3q+8]
+    movq    mm2, [parm3q+parm4q]
+    movq    mm3, [parm3q+parm4q+8]
+    movq    mm4, [parm3q+parm4q*2]
+    movq    mm5, [parm3q+parm4q*2+8]
+    movq    mm6, [parm3q+r10]
+    movq    mm7, [parm3q+r10+8]
+    movq    [parm1q], mm0
+    movq    [parm1q+8], mm1
+    movq    [parm1q+parm2q], mm2
+    movq    [parm1q+parm2q+8], mm3
+    movq    [parm1q+parm2q*2], mm4
+    movq    [parm1q+parm2q*2+8], mm5
+    movq    [parm1q+r11], mm6
+    movq    [parm1q+r11+8], mm7
     lea     parm3q, [parm3q+parm4q*4]
+    lea     parm1q, [parm1q+parm2q*4]
     sub     eax, byte 4
     jnz     .height_loop
     
@@ -376,21 +376,21 @@ ALIGN 4
 
 ALIGN 16
 ;-----------------------------------------------------------------------------
-;   void x264_mc_copy_w16_sse2( uint8_t *src, int i_src_stride, uint8_t *dst, int i_dst_stride, int i_height )
+;   void x264_mc_copy_w16_sse2( uint8_t *dst, int i_dst_stride, uint8_t *src, int i_src_stride, int i_height )
 ;-----------------------------------------------------------------------------
 x264_mc_copy_w16_sse2:
     mov     eax, parm5d         ; i_height
 
 ALIGN 4
 .height_loop
-    movdqu  xmm0, [parm1q]
-    movdqu  xmm1, [parm1q+parm2q]
-    movdqu  [parm3q], xmm0
-    movdqu  [parm3q+parm4q], xmm1
+    movdqu  xmm0, [parm3q]
+    movdqu  xmm1, [parm3q+parm4q]
+    movdqu  [parm1q], xmm0
+    movdqu  [parm1q+parm2q], xmm1
     dec     eax
     dec     eax
-    lea     parm1q, [parm1q+parm2q*2]
     lea     parm3q, [parm3q+parm4q*2]
+    lea     parm1q, [parm1q+parm2q*2]
     jnz     .height_loop
     
     ret

diff --git a/common/i386/mc-a.asm b/common/i386/mc-a.asm
index 96b3ac2a0d75761239d0fa8f5f402675dd23ef89..5a67295b34e0ddbc8c92e89564c72478fe0bbaf8 100644 (file)
--- a/common/i386/mc-a.asm
+++ b/common/i386/mc-a.asm
@@ -357,10 +357,10 @@ x264_mc_copy_w4_mmxext:
     push    esi
     push    edi
 
-    mov     esi, [esp+16]       ; src
-    mov     edi, [esp+24]       ; dst
-    mov     ebx, [esp+20]       ; i_src_stride
-    mov     edx, [esp+28]       ; i_dst_stride
+    mov     esi, [esp+24]       ; src
+    mov     edi, [esp+16]       ; dst
+    mov     ebx, [esp+28]       ; i_src_stride
+    mov     edx, [esp+20]       ; i_dst_stride
     mov     ecx, [esp+32]       ; i_height
 ALIGN 4
 .height_loop
@@ -389,10 +389,10 @@ x264_mc_copy_w8_mmxext:
     push    esi
     push    edi
 
-    mov     esi, [esp+16]       ; src
-    mov     edi, [esp+24]       ; dst
-    mov     ebx, [esp+20]       ; i_src_stride
-    mov     edx, [esp+28]       ; i_dst_stride
+    mov     esi, [esp+24]       ; src
+    mov     edi, [esp+16]       ; dst
+    mov     ebx, [esp+28]       ; i_src_stride
+    mov     edx, [esp+20]       ; i_dst_stride
     mov     ecx, [esp+32]       ; i_height
 ALIGN 4
 .height_loop
@@ -427,10 +427,10 @@ x264_mc_copy_w16_mmxext:
     push    esi
     push    edi
 
-    mov     esi, [esp+16]       ; src
-    mov     edi, [esp+24]       ; dst
-    mov     ebx, [esp+20]       ; i_src_stride
-    mov     edx, [esp+28]       ; i_dst_stride
+    mov     esi, [esp+24]       ; src
+    mov     edi, [esp+16]       ; dst
+    mov     ebx, [esp+28]       ; i_src_stride
+    mov     edx, [esp+20]       ; i_dst_stride
     mov     ecx, [esp+32]       ; i_height
 
 ALIGN 4
@@ -473,10 +473,10 @@ x264_mc_copy_w16_sse2:
     push    esi
     push    edi
 
-    mov     esi, [esp+16]       ; src
-    mov     edi, [esp+24]       ; dst
-    mov     ebx, [esp+20]       ; i_src_stride
-    mov     edx, [esp+28]       ; i_dst_stride
+    mov     esi, [esp+24]       ; src
+    mov     edi, [esp+16]       ; dst
+    mov     ebx, [esp+28]       ; i_src_stride
+    mov     edx, [esp+20]       ; i_dst_stride
     mov     ecx, [esp+32]       ; i_height
 
 ALIGN 4

diff --git a/common/i386/mc-c.c b/common/i386/mc-c.c
index 900ab7575833f22c26a87fb2217656dcb1312831..c163a008fda84732215e6279e5038c7bbc719a1f 100644 (file)
--- a/common/i386/mc-c.c
+++ b/common/i386/mc-c.c
@@ -106,13 +106,13 @@ void mc_luma_mmx( uint8_t *src[4], int i_src_stride,
     {
         switch(i_width) {
         case 4:
-            x264_mc_copy_w4_mmxext( src1, i_src_stride, dst, i_dst_stride, i_height );
+            x264_mc_copy_w4_mmxext( dst, i_dst_stride, src1, i_src_stride, i_height );
             break;
         case 8:
-            x264_mc_copy_w8_mmxext( src1, i_src_stride, dst, i_dst_stride, i_height );
+            x264_mc_copy_w8_mmxext( dst, i_dst_stride, src1, i_src_stride, i_height );
             break;
         case 16:
-            x264_mc_copy_w16_mmxext( src1, i_src_stride, dst, i_dst_stride, i_height );
+            x264_mc_copy_w16_mmxext( dst, i_dst_stride, src1, i_src_stride, i_height );
             break;
         }
     }
@@ -186,6 +186,10 @@ void x264_mc_mmxext_init( x264_mc_functions_t *pf )
     pf->avg_weight[PIXEL_8x4]   = x264_pixel_avg_weight_8x4_mmxext;
     pf->avg_weight[PIXEL_4x4]   = x264_pixel_avg_weight_4x4_mmxext;
     // avg_weight_4x8 is rare and 4x2 is not used
+
+    pf->copy[PIXEL_16x16] = x264_mc_copy_w16_mmxext;
+    pf->copy[PIXEL_8x8]   = x264_mc_copy_w8_mmxext;
+    pf->copy[PIXEL_4x4]   = x264_mc_copy_w4_mmxext;
 }
 void x264_mc_sse2_init( x264_mc_functions_t *pf )
 {

diff --git a/common/mc.c b/common/mc.c
index 9750818ae855a2ce0bedab6e07f1a4c3dd11a3ec..d2842399d3c1336d3d7a381e65a91a2f84cf303c 100644 (file)
--- a/common/mc.c
+++ b/common/mc.c
@@ -343,6 +343,15 @@ static void motion_compensation_chroma_mmxext( uint8_t *src, int i_src_stride,
 }
 #endif
 
+#define MC_COPY(W) \
+static void mc_copy_w##W( uint8_t *dst, int i_dst, uint8_t *src, int i_src, int i_height ) \
+{ \
+    mc_copy( src, i_src, dst, i_dst, W, i_height ); \
+}
+MC_COPY( 16 )
+MC_COPY( 8 )
+MC_COPY( 4 )
+
 void x264_mc_init( int cpu, x264_mc_functions_t *pf )
 {
     pf->mc_luma   = mc_luma;
@@ -371,6 +380,10 @@ void x264_mc_init( int cpu, x264_mc_functions_t *pf )
     pf->avg_weight[PIXEL_2x4]  = pixel_avg_weight_2x4;
     pf->avg_weight[PIXEL_2x2]  = pixel_avg_weight_2x2;
 
+    pf->copy[PIXEL_16x16] = mc_copy_w16;
+    pf->copy[PIXEL_8x8]   = mc_copy_w8;
+    pf->copy[PIXEL_4x4]   = mc_copy_w4;
+
 #ifdef HAVE_MMXEXT
     if( cpu&X264_CPU_MMXEXT ) {
         x264_mc_mmxext_init( pf );

diff --git a/common/mc.h b/common/mc.h
index 88943d47a3c4c8ad7d7c3d105fc28febfa80ebeb..6efe0b2936f36368b8ebb299227d85e60667213b 100644 (file)
--- a/common/mc.h
+++ b/common/mc.h
@@ -47,6 +47,9 @@ typedef struct
 
     void (*avg[10])( uint8_t *dst, int, uint8_t *src, int );
     void (*avg_weight[10])( uint8_t *dst, int, uint8_t *src, int, int i_weight );
+
+    /* only 16x16, 8x8, and 4x4 defined */
+    void (*copy[7])( uint8_t *dst, int, uint8_t *src, int, int i_height );
 } x264_mc_functions_t;
 
 void x264_mc_init( int cpu, x264_mc_functions_t *pf );