[FFmpeg-devel] [PATCH 2/4] avcodec/x86: put_pixels16_y2_sse2

[Michael Niedermayer]

about 1% faster P frame motion compensation for matrixbench on i7

Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
---
 libavcodec/x86/dsputil_mmx.c |    3 +++
 libavcodec/x86/hpeldsp.asm   |   14 ++++++++------
 2 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index 29d87a1..047504e 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -1525,6 +1525,8 @@ void ff_put_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
                           int line_size, int h);
 void ff_put_pixels16_x2_sse2(uint8_t *block, const uint8_t *pixels,
                               int line_size, int h);
+void ff_put_pixels16_y2_sse2(uint8_t *block, const uint8_t *pixels,
+                              int line_size, int h);
 void ff_avg_pixels16_sse2(uint8_t *block, const uint8_t *pixels,
                           int line_size, int h);
 
@@ -2037,6 +2039,7 @@ static void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
         if (!high_bit_depth) {
             c->put_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_no_rnd_pixels_tab[0][0] = ff_put_pixels16_sse2;
             c->avg_pixels_tab[0][0]        = ff_avg_pixels16_sse2;
diff --git a/libavcodec/x86/hpeldsp.asm b/libavcodec/x86/hpeldsp.asm
index 81b6901..088f811 100644
--- a/libavcodec/x86/hpeldsp.asm
+++ b/libavcodec/x86/hpeldsp.asm
@@ -202,21 +202,21 @@ PUT_NO_RND_PIXELS8_X2_EXACT
 
 ; put_pixels8_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h)
 %macro PUT_PIXELS8_Y2 0
-cglobal put_pixels8_y2, 4,5
+cglobal put_pixels %+ mmsize %+ _y2, 4, 5, 3
     movsxdifnidn r2, r2d
     lea          r4, [r2*2]
-    mova         m0, [r1]
+    movu         m0, [r1]
     sub          r0, r2
 .loop:
-    mova         m1, [r1+r2]
-    mova         m2, [r1+r4]
+    movu         m1, [r1+r2]
+    movu         m2, [r1+r4]
     add          r1, r4
     PAVGB        m0, m1
     PAVGB        m1, m2
     mova    [r0+r2], m0
     mova    [r0+r4], m1
-    mova         m1, [r1+r2]
-    mova         m0, [r1+r4]
+    movu         m1, [r1+r2]
+    movu         m0, [r1+r4]
     add          r0, r4
     add          r1, r4
     PAVGB        m2, m1
@@ -233,6 +233,8 @@ INIT_MMX mmxext
 PUT_PIXELS8_Y2
 INIT_MMX 3dnow
 PUT_PIXELS8_Y2
+INIT_XMM sse2
+PUT_PIXELS8_Y2
 
 
 ; put_no_rnd_pixels8_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h)
-- 
1.7.9.5