[x264-devel] ppc: Use the vec_xst_len for partial stores

[Luca Barbato]

x264 | branch: master | Luca Barbato <lu_zero at gentoo.org> | Thu Aug 23 08:30:37 2018 +0000| [40688108dd13fc0bf1847a6dfc1cf86a728654fb] | committer: Anton Mitrofanov

ppc: Use the vec_xst_len for partial stores

Seems to give about a 1-2% overall speedup on --slow.

> http://git.videolan.org/gitweb.cgi/x264.git/?a=commit;h=40688108dd13fc0bf1847a6dfc1cf86a728654fb
---

 common/ppc/ppccommon.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/common/ppc/ppccommon.h b/common/ppc/ppccommon.h
index 51936e0a..8df69380 100644
--- a/common/ppc/ppccommon.h
+++ b/common/ppc/ppccommon.h
@@ -148,8 +148,12 @@ typedef union {
 /***********************************************************************
  * VEC_STORE##n:  stores n bytes from vector v to address p
  **********************************************************************/
+#ifndef __POWER9_VECTOR__
 #define VEC_STORE8( v, p ) \
     vec_vsx_st( vec_xxpermdi( v, vec_vsx_ld( 0, p ), 1 ), 0, p )
+#else
+#define VEC_STORE8( v, p ) vec_xst_len( v, p, 8 )
+#endif
 
 /***********************************************************************
  * VEC_TRANSPOSE_8