Fix some intra pred MMX functions that used MMXEXT instructions

Also add predict_4x4_dc MMXEXT function for vp8/h264.

Originally committed as revision 23873 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/x86/h264_intrapred.asm

@@ -138,12 +138,7 @@ cglobal pred16x16_dc_%1, 2,7
     add       r5d, r6d
     lea       r2d, [r2+r5+16]
     shr       r2d, 5
-%ifidn %1, mmx
+%ifidn %1, mmxext
-    movd       m0, r2d
-    punpcklbw  m0, m0
-    punpcklwd  m0, m0
-    punpckldq  m0, m0
-%elifidn %1, mmxext
     movd       m0, r2d
     punpcklbw  m0, m0
     pshufw     m0, m0, 0
@@ -185,7 +180,6 @@ cglobal pred16x16_dc_%1, 2,7
 %endmacro
 
 INIT_MMX
-PRED16x16_DC    mmx, movq
 PRED16x16_DC mmxext, movq
 INIT_XMM
 PRED16x16_DC    sse, movaps
@@ -337,8 +331,7 @@ PRED8x8_H ssse3
 ; void pred8x8_dc_rv40(uint8_t *src, int stride)
 ;-----------------------------------------------------------------------------
 
-%macro PRED8x8_DC 1
+cglobal pred8x8_dc_rv40_mmxext, 2,7
-cglobal pred8x8_dc_rv40_%1, 2,7
     mov       r4, r0
     sub       r0, r1
     pxor      mm0, mm0
@@ -358,16 +351,9 @@ cglobal pred8x8_dc_rv40_%1, 2,7
     add       r5d, r6d
     lea       r2d, [r2+r5+8]
     shr       r2d, 4
-%ifidn %1, mmx
-    movd      mm0, r2d
-    punpcklbw mm0, mm0
-    punpcklwd mm0, mm0
-    punpckldq mm0, mm0
-%else
     movd      mm0, r2d
     punpcklbw mm0, mm0
     pshufw    mm0, mm0, 0
-%endif
     mov       r3d, 4
 .loop:
     movq [r4+r1*0], mm0
@@ -376,11 +362,6 @@ cglobal pred8x8_dc_rv40_%1, 2,7
     dec   r3d
     jg .loop
     REP_RET
-%endmacro
-
-
-PRED8x8_DC mmx
-PRED8x8_DC mmxext
 
 ;-----------------------------------------------------------------------------
 ; void pred8x8_tm_vp8(uint8_t *src, int stride)
@@ -484,3 +465,28 @@ cglobal pred8x8_tm_vp8_ssse3, 2,3,6
     dec         r2d
     jg .loop
     REP_RET
+
+cglobal pred4x4_dc_mmxext, 3,5
+    pxor   mm7, mm7
+    mov     r4, r0
+    sub     r0, r2
+    movd   mm0, [r0]
+    psadbw mm0, mm7
+    movzx  r1d, byte [r0+r2*1-1]
+    movd   r3d, mm0
+    add    r3d, r1d
+    movzx  r1d, byte [r0+r2*2-1]
+    lea     r0, [r0+r2*2]
+    add    r3d, r1d
+    movzx  r1d, byte [r0+r2*1-1]
+    add    r3d, r1d
+    movzx  r1d, byte [r0+r2*2-1]
+    add    r3d, r1d
+    add    r3d, 4
+    shr    r3d, 3
+    imul   r3d, 0x01010101
+    mov   [r4+r2*0], r3d
+    mov   [r0+r2*0], r3d
+    mov   [r0+r2*1], r3d
+    mov   [r0+r2*2], r3d
+    RET

libavcodec/x86/h264dsp_mmx.c

@@ -2328,7 +2328,6 @@ void ff_pred16x16_vertical_sse     (uint8_t *src, int stride);
 void ff_pred16x16_horizontal_mmx   (uint8_t *src, int stride);
 void ff_pred16x16_horizontal_mmxext(uint8_t *src, int stride);
 void ff_pred16x16_horizontal_ssse3 (uint8_t *src, int stride);
-void ff_pred16x16_dc_mmx           (uint8_t *src, int stride);
 void ff_pred16x16_dc_mmxext        (uint8_t *src, int stride);
 void ff_pred16x16_dc_sse           (uint8_t *src, int stride);
 void ff_pred16x16_dc_sse2          (uint8_t *src, int stride);
@@ -2336,7 +2335,6 @@ void ff_pred16x16_dc_ssse3         (uint8_t *src, int stride);
 void ff_pred16x16_tm_vp8_mmx       (uint8_t *src, int stride);
 void ff_pred16x16_tm_vp8_mmxext    (uint8_t *src, int stride);
 void ff_pred16x16_tm_vp8_sse2      (uint8_t *src, int stride);
-void ff_pred8x8_dc_rv40_mmx        (uint8_t *src, int stride);
 void ff_pred8x8_dc_rv40_mmxext     (uint8_t *src, int stride);
 void ff_pred8x8_vertical_mmx       (uint8_t *src, int stride);
 void ff_pred8x8_horizontal_mmx     (uint8_t *src, int stride);
@@ -2346,6 +2344,7 @@ void ff_pred8x8_tm_vp8_mmx         (uint8_t *src, int stride);
 void ff_pred8x8_tm_vp8_mmxext      (uint8_t *src, int stride);
 void ff_pred8x8_tm_vp8_sse2        (uint8_t *src, int stride);
 void ff_pred8x8_tm_vp8_ssse3       (uint8_t *src, int stride);
+void ff_pred4x4_dc_mmxext          (uint8_t *src, const uint8_t *topright, int stride);
 
 #if CONFIG_H264DSP
 void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
@@ -2354,12 +2353,10 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
     if (mm_flags & FF_MM_MMX) {
         h->pred16x16[VERT_PRED8x8] = ff_pred16x16_vertical_mmx;
         h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_mmx;
-        h->pred16x16[DC_PRED8x8  ] = ff_pred16x16_dc_mmx;
         h->pred8x8  [VERT_PRED8x8] = ff_pred8x8_vertical_mmx;
         h->pred8x8  [HOR_PRED8x8 ] = ff_pred8x8_horizontal_mmx;
         if (codec_id == CODEC_ID_VP8) {
             h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_tm_vp8_mmx;
-            h->pred8x8  [DC_PRED8x8   ] = ff_pred8x8_dc_rv40_mmx;
             h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_tm_vp8_mmx;
         }
     }
@@ -2368,6 +2365,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
         h->pred16x16[HOR_PRED8x8 ] = ff_pred16x16_horizontal_mmxext;
         h->pred16x16[DC_PRED8x8  ] = ff_pred16x16_dc_mmxext;
         h->pred8x8  [HOR_PRED8x8 ] = ff_pred8x8_horizontal_mmxext;
+        h->pred4x4  [DC_PRED     ] = ff_pred4x4_dc_mmxext;
         if (codec_id == CODEC_ID_VP8) {
             h->pred16x16[PLANE_PRED8x8] = ff_pred16x16_tm_vp8_mmxext;
             h->pred8x8  [DC_PRED8x8   ] = ff_pred8x8_dc_rv40_mmxext;