lavr: x86: improve non-SSE4 version of S16_TO_S32_SX macro

Removes a false dependency on existing contents of the 2nd dst register,
giving better performance for OOE.

diff --git a/libavresample/x86/util.asm b/libavresample/x86/util.asm
index 501f662..ca7fde5 100644 (file)
--- a/libavresample/x86/util.asm
+++ b/libavresample/x86/util.asm
@@ -26,7 +26,8 @@
     pmovsxwd     m%1, m%1
     SWAP %1, %2
 %else
-    punpckhwd    m%2, m%1
+    mova         m%2, m%1
+    punpckhwd    m%2, m%2
     punpcklwd    m%1, m%1
     psrad        m%2, 16
     psrad        m%1, 16