[FFmpeg-devel] [PATCH 5/5] x86: hpeldsp: implement SSE2 put_pixels16_xy2

[Christophe Gisquet]

This is obviously equivalent to the avg version, without the avg.

3223(mmx) -> 2006(sse2)
---
 libavcodec/x86/hpeldsp.asm    | 21 +++++++++++++++------
 libavcodec/x86/hpeldsp_init.c |  3 +++
 2 files changed, 18 insertions(+), 6 deletions(-)

diff --git a/libavcodec/x86/hpeldsp.asm b/libavcodec/x86/hpeldsp.asm
index 1d26c45..4af423a 100644
--- a/libavcodec/x86/hpeldsp.asm
+++ b/libavcodec/x86/hpeldsp.asm
@@ -551,11 +551,11 @@ AVG_APPROX_PIXELS8_XY2
 
 
 ; void ff_avg_pixels16_xy2(uint8_t *block, const uint8_t *pixels, ptrdiff_t line_size, int h)
-%macro AVG_PIXELS_XY2 0
+%macro SET_PIXELS_XY2 1
 %if cpuflag(sse2)
-cglobal avg_pixels16_xy2, 4,5,8
+cglobal %1_pixels16_xy2, 4,5,8
 %else
-cglobal avg_pixels8_xy2, 4,5
+cglobal %1_pixels8_xy2, 4,5
 %endif
     pxor        m7, m7
     mova        m6, [pw_2]
@@ -588,9 +588,13 @@ cglobal avg_pixels8_xy2, 4,5
     paddusw     m5, m1
     psrlw       m4, 2
     psrlw       m5, 2
+%ifidn %1, avg
     mova        m3, [r0+r4]
     packuswb    m4, m5
     PAVGB       m4, m3
+%else
+    packuswb    m4, m5
+%endif
     mova   [r0+r4], m4
     add         r4, r2
 
@@ -610,9 +614,13 @@ cglobal avg_pixels8_xy2, 4,5
     paddusw     m1, m5
     psrlw       m0, 2
     psrlw       m1, 2
+%ifidn %1, avg
     mova        m3, [r0+r4]
     packuswb    m0, m1
     PAVGB       m0, m3
+%else
+    packuswb    m0, m1
+%endif
     mova   [r0+r4], m0
     add         r4, r2
     sub        r3d, 2
@@ -621,8 +629,9 @@ cglobal avg_pixels8_xy2, 4,5
 %endmacro
 
 INIT_MMX mmxext
-AVG_PIXELS_XY2
+SET_PIXELS_XY2 avg
 INIT_MMX 3dnow
-AVG_PIXELS_XY2
+SET_PIXELS_XY2 avg
 INIT_XMM sse2
-AVG_PIXELS_XY2
+SET_PIXELS_XY2 put
+SET_PIXELS_XY2 avg
diff --git a/libavcodec/x86/hpeldsp_init.c b/libavcodec/x86/hpeldsp_init.c
index 05bd561..cda16dc 100644
--- a/libavcodec/x86/hpeldsp_init.c
+++ b/libavcodec/x86/hpeldsp_init.c
@@ -48,6 +48,8 @@ void ff_put_pixels16_y2_sse2(uint8_t *block, const uint8_t *pixels,
                              ptrdiff_t line_size, int h);
 void ff_avg_pixels16_y2_sse2(uint8_t *block, const uint8_t *pixels,
                              ptrdiff_t line_size, int h);
+void ff_put_pixels16_xy2_sse2(uint8_t *block, const uint8_t *pixels,
+                              ptrdiff_t line_size, int h);
 void ff_avg_pixels16_xy2_sse2(uint8_t *block, const uint8_t *pixels,
                               ptrdiff_t line_size, int h);
 void ff_put_no_rnd_pixels8_x2_mmxext(uint8_t *block, const uint8_t *pixels,
@@ -296,6 +298,7 @@ static void hpeldsp_init_sse2(HpelDSPContext *c, int flags, int cpu_flags)
         c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2;
         c->put_pixels_tab[0][1]        = ff_put_pixels16_x2_sse2;
         c->put_pixels_tab[0][2]        = ff_put_pixels16_y2_sse2;
+        c->put_pixels_tab[0][3]        = ff_put_pixels16_xy2_sse2;
         c->avg_pixels_tab[0][0]        = ff_avg_pixels16_sse2;
         c->avg_pixels_tab[0][1]        = ff_avg_pixels16_x2_sse2;
         c->avg_pixels_tab[0][2]        = ff_avg_pixels16_y2_sse2;
-- 
1.8.0.msysgit.0