Port pred4x4_horizontal_up_mmxext (H.264 intra prediction) from x264 (authors:

Jason, Loren, Holger) to FFmpeg. Patch by Daniel Kang <daniel dot d dot kang
at gmail com>, as part of Google's GCI 2010.

Originally committed as revision 26156 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/x86/h264_intrapred.asm

@@ -2561,4 +2561,37 @@ cglobal pred4x4_vertical_left_mmxext, 3,3
     movh      [r1+r2*1], m4
     movh      [r1+r2*2], m0
     RET
+
+;-----------------------------------------------------------------------------
+; void pred4x4_horizontal_up_mmxext(uint8_t *src, const uint8_t *topright, int stride)
+;-----------------------------------------------------------------------------
+
+INIT_MMX
+cglobal pred4x4_horizontal_up_mmxext, 3,3
+    sub       r0, r2
+    lea       r1, [r0+r2*2]
+    movq      m0, [r0+r2*1-8]
+    punpckhbw m0, [r0+r2*2-8]
+    movq      m1, [r1+r2*1-8]
+    punpckhbw m1, [r1+r2*2-8]
+    punpckhwd m0, m1
+    movq      m1, m0
+    punpckhbw m1, m1
+    pshufw    m1, m1, 0xFF
+    punpckhdq m0, m1
+    movq      m2, m0
+    movq      m3, m0
+    movq      m7, m0
+    psrlq     m2, 16
+    psrlq     m3, 8
+    pavgb     m7, m3
+    PRED4x4_LOWPASS m4, m0, m2, m3, m5
+    punpcklbw m7, m4
+    movd    [r0+r2*1], m7
+    psrlq    m7, 16
+    movd    [r0+r2*2], m7
+    psrlq    m7, 16
+    movd    [r1+r2*1], m7
+    movd    [r1+r2*2], m1
+    RET
 %endif

libavcodec/x86/h264_intrapred_init.c

@@ -85,6 +85,7 @@ void ff_pred8x8l_horizontal_down_ssse3(uint8_t *src, int has_topleft, int has_to
 void ff_pred4x4_dc_mmxext          (uint8_t *src, const uint8_t *topright, int stride);
 void ff_pred4x4_down_left_mmxext   (uint8_t *src, const uint8_t *topright, int stride);
 void ff_pred4x4_vertical_left_mmxext(uint8_t *src, const uint8_t *topright, int stride);
+void ff_pred4x4_horizontal_up_mmxext(uint8_t *src, const uint8_t *topright, int stride);
 void ff_pred4x4_tm_vp8_mmx         (uint8_t *src, const uint8_t *topright, int stride);
 void ff_pred4x4_tm_vp8_mmxext      (uint8_t *src, const uint8_t *topright, int stride);
 void ff_pred4x4_tm_vp8_ssse3       (uint8_t *src, const uint8_t *topright, int stride);
@@ -136,6 +137,9 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
             h->pred4x4  [DIAG_DOWN_LEFT_PRED ] = ff_pred4x4_down_left_mmxext;
         if (codec_id == CODEC_ID_SVQ3 || codec_id == CODEC_ID_H264)
             h->pred4x4  [VERT_LEFT_PRED      ] = ff_pred4x4_vertical_left_mmxext;
+        if (codec_id != CODEC_ID_RV40) {
+            h->pred4x4  [HOR_UP_PRED         ] = ff_pred4x4_horizontal_up_mmxext;
+        }
         if (codec_id == CODEC_ID_SVQ3 || codec_id == CODEC_ID_H264) {
             h->pred8x8  [TOP_DC_PRED8x8      ] = ff_pred8x8_top_dc_mmxext;
             h->pred8x8  [DC_PRED8x8          ] = ff_pred8x8_dc_mmxext;