[x264-devel] x86util: XOP optimized HADDD

[James Almer]

x264 | branch: master | James Almer <jamrial at gmail.com> | Wed Apr  9 03:33:04 2014 -0300| [ee43d33b1ef8897eda84117af10f8334f97b297d] | committer: Jason Garrett-Glaser

x86util: XOP optimized HADDD

> http://git.videolan.org/gitweb.cgi/x264.git/?a=commit;h=ee43d33b1ef8897eda84117af10f8334f97b297d
---

 common/x86/x86util.asm |    5 +++++
 1 file changed, 5 insertions(+)

diff --git a/common/x86/x86util.asm b/common/x86/x86util.asm
index 79ff34f..b18a1d1 100644
--- 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