x86util: XOP optimized HADDD

diff --git a/common/x86/x86util.asm b/common/x86/x86util.asm
index 79ff34f3703684349c7acdfae82f5c24737092bf..b18a1d10c3b9115e10ca0673143ff6f1564ee57e 100644 (file)
--- a/common/x86/x86util.asm
+++ b/common/x86/x86util.asm
@@ -298,11 +298,16 @@
     paddd   %1, %2
 %endif
 %if mmsize >= 16
+%if cpuflag(xop) && sizeof%1 == 16
+    vphadddq %1, %1
+%endif
     movhlps %2, %1
     paddd   %1, %2
 %endif
+%if notcpuflag(xop) || sizeof%1 != 16
     PSHUFLW %2, %1, q0032
     paddd   %1, %2
+%endif
 %undef %1
 %undef %2
 %endmacro