diff x86/vp8dsp-init.c @ 12227:d07e6037846d libavcodec

Use pextrw for SSE4 mbedge filter result writing, speedup 5-10cycles on CPUs supporting it.
author rbultje
date Thu, 22 Jul 2010 19:59:34 +0000
parents 657d353cd515
children e08d65897115
line wrap: on
line diff
--- a/x86/vp8dsp-init.c	Thu Jul 22 12:35:32 2010 +0000
+++ b/x86/vp8dsp-init.c	Thu Jul 22 19:59:34 2010 +0000
@@ -247,6 +247,7 @@
 DECLARE_LOOP_FILTER(mmxext)
 DECLARE_LOOP_FILTER(sse2)
 DECLARE_LOOP_FILTER(ssse3)
+DECLARE_LOOP_FILTER(sse4)
 
 #endif
 
@@ -379,6 +380,9 @@
 
     if (mm_flags & FF_MM_SSE4) {
         c->vp8_idct_dc_add                  = ff_vp8_idct_dc_add_sse4;
+
+        c->vp8_h_loop_filter16y       = ff_vp8_h_loop_filter16y_mbedge_sse4;
+        c->vp8_h_loop_filter8uv       = ff_vp8_h_loop_filter8uv_mbedge_sse4;
     }
 #endif
 }