x86: SSSE3 implementation of pixel_sad_x3 and pixel_sad_x4

diff --git a/common/pixel.c b/common/pixel.c
index 7a1e95502ee9aa68e5c8788ee71a0ce893d53685..b1cf7a3f9c51862024710608b9c568fd3979b612 100644 (file)
--- a/common/pixel.c
+++ b/common/pixel.c
@@ -1195,9 +1195,8 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         }
         else
         {
-            pixf->sad_x4[PIXEL_8x4] = x264_pixel_sad_x4_8x4_ssse3;
-            pixf->sad_x4[PIXEL_8x8] = x264_pixel_sad_x4_8x8_ssse3;
-            pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_ssse3;
+            INIT2( sad_x3, _ssse3 );
+            INIT5( sad_x4, _ssse3 );
         }
         if( (cpu&X264_CPU_SLOW_ATOM) || (cpu&X264_CPU_SLOW_SHUFFLE) )
         {

diff --git a/common/x86/sad-a.asm b/common/x86/sad-a.asm
index 45f4a10666a9ed9b0b3eb484ae064fa6d5703f4f..2270011b1b4b9390de115f99a1ae89e6b4cc2e38 100644 (file)
--- a/common/x86/sad-a.asm
+++ b/common/x86/sad-a.asm
@@ -1241,21 +1241,34 @@ SAD_X 4,  4,  4
 %endmacro
 
 %macro SAD_X3_END_SSE2 0
+    movifnidn r5, r5mp
+%if cpuflag(ssse3)
+    packssdw m0, m1
+    packssdw m2, m2
+    phaddd   m0, m2
+    mova   [r5], m0
+%else
     movhlps  m3, m0
     movhlps  m4, m1
     movhlps  m5, m2
     paddw    m0, m3
     paddw    m1, m4
     paddw    m2, m5
-    movifnidn r5, r5mp
     movd [r5+0], m0
     movd [r5+4], m1
     movd [r5+8], m2
+%endif
     RET
 %endmacro
 
 %macro SAD_X4_END_SSE2 0
     mov      r0, r6mp
+%if cpuflag(ssse3)
+    packssdw m0, m1
+    packssdw m2, m3
+    phaddd   m0, m2
+    mova   [r0], m0
+%else
     psllq    m1, 32
     psllq    m3, 32
     paddw    m0, m1
@@ -1266,6 +1279,7 @@ SAD_X 4,  4,  4
     paddw    m2, m3
     movq [r0+0], m0
     movq [r0+8], m2
+%endif
     RET
 %endmacro
 
@@ -1504,9 +1518,13 @@ cglobal pixel_sad_x%1_%2x%3, 2+%1,3+%1,8
 %endmacro
 
 INIT_XMM ssse3
-SAD_X_SSSE3 4, 8, 16
-SAD_X_SSSE3 4, 8,  8
-SAD_X_SSSE3 4, 8,  4
+SAD_X_SSE2  3, 16, 16, 7
+SAD_X_SSE2  3, 16,  8, 7
+SAD_X_SSE2  4, 16, 16, 7
+SAD_X_SSE2  4, 16,  8, 7
+SAD_X_SSSE3 4,  8, 16
+SAD_X_SSSE3 4,  8,  8
+SAD_X_SSSE3 4,  8,  4
 
 INIT_XMM avx
 SAD_X_SSE2 3, 16, 16, 6