sfence is not a valid MMX instruction.

provide a new MMXEXT version. about 15% faster here.

CVS patchset: 5875
CVS date: 2003/12/08 17:06:53

1 files changed, 152 insertions, 3 deletions

diff --git a/src/post/deinterlace/plugins/linearblend.c b/src/post/deinterlace/plugins/linearblend.c
index ef7ab52bc..977efe2b3 100644
--- a/src/post/deinterlace/plugins/linearblend.c
+++ b/src/post/deinterlace/plugins/linearblend.c
@@ -86,6 +86,7 @@ static void deinterlace_scanline_linear_blend( uint8_t *output,
 
         movd_r2m( mm2, *output );
         movd_r2m( mm5, *(output+4) );
+
         output += 8;
         t0 += 8;
         b0 += 8;
@@ -94,7 +95,6 @@ static void deinterlace_scanline_linear_blend( uint8_t *output,
     while( width-- ) {
         *output++ = (*t0++ + *b0++ + (*m1++ << 1)) >> 2;
     }
-    sfence();
     emms();
 #else
     width *= 2;
@@ -151,6 +151,7 @@ static void deinterlace_scanline_linear_blend2( uint8_t *output,
 
         movd_r2m( mm2, *output );
         movd_r2m( mm5, *(output+4) );
+
         output += 8;
         t1 += 8;
         b1 += 8;
@@ -159,7 +160,6 @@ static void deinterlace_scanline_linear_blend2( uint8_t *output,
     while( width-- ) {
         *output++ = (*t1++ + *b1++ + (*m0++ << 1)) >> 2;
     }
-    sfence();
     emms();
 #else
     width *= 2;
@@ -169,6 +169,150 @@ static void deinterlace_scanline_linear_blend2( uint8_t *output,
 #endif
 }
 
+#ifdef ARCH_X86
+
+/* MMXEXT version is about 15% faster with Athlon XP [MF] */
+
+static void deinterlace_scanline_linear_blend_mmxext( uint8_t *output,
+                                               deinterlace_scanline_data_t *data,
+                                               int width )
+{
+    uint8_t *t0 = data->t0;
+    uint8_t *b0 = data->b0;
+    uint8_t *m1 = data->m1;
+    int i;
+    static mmx_t high_mask = {ub:{0xff,0xff,0xff,0xff,0,0,0,0}};
+
+    // Get width in bytes.
+    width *= 2;
+    i = width / 8;
+    width -= i * 8;
+
+    movd_m2r( high_mask, mm6 );
+    pxor_r2r( mm7, mm7 );
+    while( i-- ) {
+        movd_m2r( *t0, mm0 );
+        movd_m2r( *b0, mm1 );
+        movd_m2r( *m1, mm2 );
+
+        movd_m2r( *(t0+4), mm3 );
+        movd_m2r( *(b0+4), mm4 );
+        movd_m2r( *(m1+4), mm5 );
+
+        punpcklbw_r2r( mm7, mm0 );
+        punpcklbw_r2r( mm7, mm1 );
+        punpcklbw_r2r( mm7, mm2 );
+
+        punpcklbw_r2r( mm7, mm3 );
+        punpcklbw_r2r( mm7, mm4 );
+        punpcklbw_r2r( mm7, mm5 );
+
+        psllw_i2r( 1, mm2 );
+        psllw_i2r( 1, mm5 );
+        paddw_r2r( mm0, mm2 );
+        paddw_r2r( mm3, mm5 );
+        paddw_r2r( mm1, mm2 );
+        paddw_r2r( mm4, mm5 );
+        psrlw_i2r( 2, mm2 );
+        psrlw_i2r( 2, mm5 );
+        packuswb_r2r( mm2, mm2 );
+        packuswb_r2r( mm5, mm5 );
+
+        psllq_i2r( 32, mm5 );
+        pand_r2r( mm6, mm2 );
+        por_r2r( mm2, mm5 );
+        movntq_r2m( mm5, *output );
+
+        output += 8;
+        t0 += 8;
+        b0 += 8;
+        m1 += 8;
+    }
+    while( width-- ) {
+        *output++ = (*t0++ + *b0++ + (*m1++ << 1)) >> 2;
+    }
+    sfence();
+    emms();
+}
+
+static void deinterlace_scanline_linear_blend2_mmxext( uint8_t *output,
+                                                deinterlace_scanline_data_t *data,
+                                                int width )
+{
+    uint8_t *m0 = data->m0;
+    uint8_t *t1 = data->t1;
+    uint8_t *b1 = data->b1;
+
+    int i;
+
+    // Get width in bytes.
+    width *= 2;
+    i = width / 8;
+    width -= i * 8;
+
+    pxor_r2r( mm7, mm7 );
+    while( i-- ) {
+        movd_m2r( *t1, mm0 );
+        movd_m2r( *b1, mm1 );
+        movd_m2r( *m0, mm2 );
+
+        movd_m2r( *(t1+4), mm3 );
+        movd_m2r( *(b1+4), mm4 );
+        movd_m2r( *(m0+4), mm5 );
+
+        punpcklbw_r2r( mm7, mm0 );
+        punpcklbw_r2r( mm7, mm1 );
+        punpcklbw_r2r( mm7, mm2 );
+
+        punpcklbw_r2r( mm7, mm3 );
+        punpcklbw_r2r( mm7, mm4 );
+        punpcklbw_r2r( mm7, mm5 );
+
+        psllw_i2r( 1, mm2 );
+        psllw_i2r( 1, mm5 );
+        paddw_r2r( mm0, mm2 );
+        paddw_r2r( mm3, mm5 );
+        paddw_r2r( mm1, mm2 );
+        paddw_r2r( mm4, mm5 );
+        psrlw_i2r( 2, mm2 );
+        psrlw_i2r( 2, mm5 );
+        packuswb_r2r( mm2, mm2 );
+        packuswb_r2r( mm5, mm5 );
+
+        psllq_i2r( 32, mm5 );
+        pand_r2r( mm6, mm2 );
+        por_r2r( mm2, mm5 );
+        movntq_r2m( mm5, *output );
+
+        output += 8;
+        t1 += 8;
+        b1 += 8;
+        m0 += 8;
+    }
+    while( width-- ) {
+        *output++ = (*t1++ + *b1++ + (*m0++ << 1)) >> 2;
+    }
+    sfence();
+    emms();
+}
+
+static deinterlace_method_t linearblendmethod_mmxext =
+{
+    DEINTERLACE_PLUGIN_API_VERSION,
+    "mplayer: Linear Blend",
+    "LinearBlend",
+    2,
+    MM_ACCEL_X86_MMXEXT,
+    0,
+    0,
+    0,
+    1,
+    deinterlace_scanline_linear_blend_mmxext,
+    deinterlace_scanline_linear_blend2_mmxext,
+    0
+};
+
+#endif
 
 static deinterlace_method_t linearblendmethod =
 {
@@ -196,5 +340,10 @@ void deinterlace_plugin_init( void )
 void linearblend_plugin_init( void )
 #endif
 {
-    register_deinterlace_method( &linearblendmethod );
+#ifdef ARCH_X86
+    if( xine_mm_accel() & MM_ACCEL_X86_MMXEXT )
+      register_deinterlace_method( &linearblendmethod_mmxext );
+    else
+#endif
+      register_deinterlace_method( &linearblendmethod );
 }