Port pred8x8l_horizontal_down_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 26150 to svn://svn.ffmpeg.org/ffmpeg/trunk

diff --git a/libavcodec/x86/h264_intrapred.asm b/libavcodec/x86/h264_intrapred.asm
index a99f044b6f2db6e6ddfbf23b1eda993a58e9e0ab..74e604d496dd2e7a16b9b88b206f6577a4fdc52f 100644 (file)
--- a/libavcodec/x86/h264_intrapred.asm
+++ b/libavcodec/x86/h264_intrapred.asm
@@ -2140,6 +2140,127 @@ PRED8x8L_HORIZONTAL_UP mmxext
 PRED8x8L_HORIZONTAL_UP ssse3
 %endif
 
+;-----------------------------------------------------------------------------
+;void pred8x8l_horizontal_down(uint8_t *src, int has_topleft, int has_topright, int stride)
+;-----------------------------------------------------------------------------
+%ifdef CONFIG_GPL
+INIT_MMX
+%define PALIGNR PALIGNR_MMX
+cglobal pred8x8l_horizontal_down_mmxext, 4,5
+    sub          r0, r3
+    lea          r4, [r0+r3*2]
+    movq        mm0, [r0+r3*1-8]
+    punpckhbw   mm0, [r0+r3*0-8]
+    movq        mm1, [r4+r3*1-8]
+    punpckhbw   mm1, [r0+r3*2-8]
+    mov          r4, r0
+    punpckhwd   mm1, mm0
+    lea          r0, [r0+r3*4]
+    movq        mm2, [r0+r3*1-8]
+    punpckhbw   mm2, [r0+r3*0-8]
+    lea          r0, [r0+r3*2]
+    movq        mm3, [r0+r3*1-8]
+    punpckhbw   mm3, [r0+r3*0-8]
+    punpckhwd   mm3, mm2
+    punpckhdq   mm3, mm1
+    lea          r0, [r0+r3*2]
+    movq        mm0, [r0+r3*0-8]
+    movq        mm1, [r4]
+    mov          r0, r4
+    movq        mm4, mm3
+    movq        mm2, mm3
+    PALIGNR     mm4, mm0, 7, mm0
+    PALIGNR     mm1, mm2, 1, mm2
+    test        r1, r1
+    jnz .do_left
+.fix_lt_1:
+    movq        mm5, mm3
+    pxor        mm5, mm4
+    psrlq       mm5, 56
+    psllq       mm5, 48
+    pxor        mm1, mm5
+    jmp .do_left
+.fix_lt_2:
+    movq        mm5, mm3
+    pxor        mm5, mm2
+    psllq       mm5, 56
+    psrlq       mm5, 56
+    pxor        mm2, mm5
+    test         r2, r2
+    jnz .do_top
+.fix_tr_1:
+    movq        mm5, mm3
+    pxor        mm5, mm1
+    psrlq       mm5, 56
+    psllq       mm5, 56
+    pxor        mm1, mm5
+    jmp .do_top
+.do_left:
+    movq        mm0, mm4
+    PRED4x4_LOWPASS mm2, mm1, mm4, mm3, mm5
+    movq        mm4, mm0
+    movq        mm7, mm2
+    movq        mm6, mm2
+    PRED4x4_LOWPASS mm1, mm3, mm0, mm4, mm5
+    psllq       mm1, 56
+    PALIGNR     mm7, mm1, 7, mm3
+    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
+    jz .fix_lt_2
+    test         r2, r2
+    jz .fix_tr_1
+.do_top:
+    PRED4x4_LOWPASS mm4, mm2, mm1, mm3, mm5
+    movq       mm5, mm4
+    lea         r1, [r0+r3*2]
+    psllq      mm7, 56
+    movq       mm2, mm5
+    movq       mm3, mm6
+    movq       mm4, mm2
+    PALIGNR    mm2, mm6, 7, mm5
+    PALIGNR    mm6, mm7, 7, mm0
+    lea         r2, [r1+r3*2]
+    PALIGNR    mm4, mm3, 1, mm7
+    movq       mm5, mm3
+    pavgb      mm3, mm6
+    PRED4x4_LOWPASS mm0, mm4, mm6, mm5, mm7
+    movq       mm4, mm2
+    movq       mm1, mm2
+    lea         r4, [r2+r3*2]
+    psrlq      mm4, 16
+    psrlq      mm1, 8
+    PRED4x4_LOWPASS mm6, mm4, mm2, mm1, mm5
+    movq       mm7, mm3
+    punpcklbw  mm3, mm0
+    punpckhbw  mm7, mm0
+    movq       mm1, mm7
+    movq       mm0, mm7
+    movq       mm4, mm7
+    movq [r4+r3*2], mm3
+    PALIGNR    mm7, mm3, 2, mm5
+    movq [r4+r3*1], mm7
+    PALIGNR    mm1, mm3, 4, mm5
+    movq [r2+r3*2], mm1
+    PALIGNR    mm0, mm3, 6, mm3
+    movq [r2+r3*1], mm0
+    movq       mm2, mm6
+    movq       mm3, mm6
+    movq [r1+r3*2], mm4
+    PALIGNR    mm6, mm4, 2, mm5
+    movq [r1+r3*1], mm6
+    PALIGNR    mm2, mm4, 4, mm5
+    movq [r0+r3*2], mm2
+    PALIGNR    mm3, mm4, 6, mm4
+    movq [r0+r3*1], mm3
+    RET
+%endif
+
 ;-----------------------------------------------------------------------------
 ; void pred4x4_dc_mmxext(uint8_t *src, const uint8_t *topright, int stride)
 ;-----------------------------------------------------------------------------

diff --git a/libavcodec/x86/h264_intrapred_init.c b/libavcodec/x86/h264_intrapred_init.c
index a3f55491fc9758fb663d6c53744848501f8bec84..7c1675ccb1a2e0fe5b7d0d19d9e71304a85f44fc 100644 (file)
--- a/libavcodec/x86/h264_intrapred_init.c
+++ b/libavcodec/x86/h264_intrapred_init.c
@@ -79,6 +79,7 @@ void ff_pred8x8l_vertical_left_sse2(uint8_t *src, int has_topleft, int has_topri
 void ff_pred8x8l_vertical_left_ssse3(uint8_t *src, int has_topleft, int has_topright, int stride);
 void ff_pred8x8l_horizontal_up_mmxext(uint8_t *src, int has_topleft, int has_topright, int stride);
 void ff_pred8x8l_horizontal_up_ssse3(uint8_t *src, int has_topleft, int has_topright, int stride);
+void ff_pred8x8l_horizontal_down_mmxext(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);
@@ -124,6 +125,7 @@ void ff_h264_pred_init_x86(H264PredContext *h, int codec_id)
         h->pred8x8l [DIAG_DOWN_RIGHT_PRED] = ff_pred8x8l_down_right_mmxext;
         h->pred8x8l [VERT_RIGHT_PRED     ] = ff_pred8x8l_vertical_right_mmxext;
         h->pred8x8l [HOR_UP_PRED         ] = ff_pred8x8l_horizontal_up_mmxext;
+        h->pred8x8l [HOR_DOWN_PRED       ] = ff_pred8x8l_horizontal_down_mmxext;
 #endif
         h->pred4x4  [DC_PRED     ] = ff_pred4x4_dc_mmxext;
 #if CONFIG_GPL