Port pred8x8l_vertical_mmxext/ssse3 (H.264 intra prediction) from x264 to

FFmpeg. Original authors: Holger Lubitz <holger lubitz org>, Jason Garrett-
Glaser <darkshikari gmail com> (approves LGPL relicensing for this code) and
Loren Merritt <lorenm at u dot washington dot edu> (approves LGPL relicensing
for this code). Patch by Daniel Kang <daniel dot d dot kang at gmail com>, as
part of Google's GCI 2010.

Originally committed as revision 26140 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/x86/h264_intrapred.asm

@@ -1322,6 +1322,58 @@ PRED8x8L_HORIZONTAL mmxext
 PRED8x8L_HORIZONTAL ssse3
 %endif
 
+;-----------------------------------------------------------------------------
+; void pred8x8l_vertical(uint8_t *src, int has_topleft, int has_topright, int stride)
+;-----------------------------------------------------------------------------
+%ifdef CONFIG_GPL
+%macro PRED8x8L_VERTICAL 1
+cglobal pred8x8l_vertical_%1, 4,4
+    sub          r0, r3
+    movq        mm0, [r0-8]
+    movq        mm3, [r0]
+    movq        mm1, [r0+8]
+    movq        mm2, mm3
+    movq        mm4, mm3
+    PALIGNR     mm2, mm0, 7, mm0
+    PALIGNR     mm1, mm4, 1, mm4
+    test         r1, r1 ; top_left
+    jz .fix_lt_2
+    test         r2, r2 ; top_right
+    jz .fix_tr_1
+    jmp .body
+.fix_lt_2:
+    movq        mm5, mm3
+    pxor        mm5, mm2
+    psllq       mm5, 56
+    psrlq       mm5, 56
+    pxor        mm2, mm5
+    test         r2, r2 ; top_right
+    jnz .body
+.fix_tr_1:
+    movq        mm5, mm3
+    pxor        mm5, mm1
+    psrlq       mm5, 56
+    psllq       mm5, 56
+    pxor        mm1, mm5
+.body
+    PRED4x4_LOWPASS mm0, mm2, mm1, mm3, mm5
+%rep 3
+    movq [r0+r3*1], mm0
+    movq [r0+r3*2], mm0
+    lea    r0, [r0+r3*2]
+%endrep
+    movq [r0+r3*1], mm0
+    movq [r0+r3*2], mm0
+    RET
+%endmacro
+
+INIT_MMX
+%define PALIGNR PALIGNR_MMX
+PRED8x8L_VERTICAL mmxext
+%define PALIGNR PALIGNR_SSSE3
+PRED8x8L_VERTICAL ssse3
+%endif
+
 ;-----------------------------------------------------------------------------
 ; void pred4x4_dc_mmxext(uint8_t *src, const uint8_t *topright, int stride)
 ;-----------------------------------------------------------------------------

libavcodec/x86/h264_intrapred_init.c

@@ -65,6 +65,8 @@ void ff_pred8x8l_dc_mmxext         (uint8_t *src, int has_topleft, int has_topri
 void ff_pred8x8l_dc_ssse3          (uint8_t *src, int has_topleft, int has_topright, int stride);
 void ff_pred8x8l_horizontal_mmxext (uint8_t *src, int has_topleft, int has_topright, int stride);
 void ff_pred8x8l_horizontal_ssse3  (uint8_t *src, int has_topleft, int has_topright, int stride);
+void ff_pred8x8l_vertical_mmxext   (uint8_t *src, int has_topleft, int has_topright, int stride);
+void ff_pred8x8l_vertical_ssse3    (uint8_t *src, int has_topleft, int has_topright, int stride);
 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_tm_vp8_mmx         (uint8_t *src, const uint8_t *topright, int stride);
@@ -106,6 +108,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
         h->pred8x8l [TOP_DC_PRED ] = ff_pred8x8l_top_dc_mmxext;
         h->pred8x8l [DC_PRED     ] = ff_pred8x8l_dc_mmxext;
         h->pred8x8l [HOR_PRED    ] = ff_pred8x8l_horizontal_mmxext;
+        h->pred8x8l [VERT_PRED   ] = ff_pred8x8l_vertical_mmxext;
 #endif
         h->pred4x4  [DC_PRED     ] = ff_pred4x4_dc_mmxext;
 #if CONFIG_GPL
@@ -163,6 +166,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
         h->pred8x8l [TOP_DC_PRED ] = ff_pred8x8l_top_dc_ssse3;
         h->pred8x8l [DC_PRED     ] = ff_pred8x8l_dc_ssse3;
         h->pred8x8l [HOR_PRED    ] = ff_pred8x8l_horizontal_ssse3;
+        h->pred8x8l [VERT_PRED   ] = ff_pred8x8l_vertical_ssse3;
 #endif
         if (codec_id == CODEC_ID_VP8) {
             h->pred8x8  [PLANE_PRED8x8] = ff_pred8x8_tm_vp8_ssse3;