[FFmpeg-cvslog] x86: Avoid movs on BUTTERFLYPS when in AVX mode

Vitor Sessak git at videolan.org
Wed May 30 01:51:54 CEST 2012


ffmpeg | branch: master | Vitor Sessak <vitor1001 at gmail.com> | Fri May 11 22:32:08 2012 +0200| [4a301706fd682d3ad43b4239542e6f586ae29811] | committer: Janne Grunau

x86: Avoid movs on BUTTERFLYPS when in AVX mode

Signed-off-by: Janne Grunau <janne-libav at jannau.net>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=4a301706fd682d3ad43b4239542e6f586ae29811
---

 libavutil/x86/x86util.asm |   13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/libavutil/x86/x86util.asm b/libavutil/x86/x86util.asm
index 508f24e..066384b 100644
--- a/libavutil/x86/x86util.asm
+++ b/libavutil/x86/x86util.asm
@@ -84,13 +84,12 @@
 %macro TRANSPOSE4x4PS 5
     SBUTTERFLYPS %1, %2, %5
     SBUTTERFLYPS %3, %4, %5
-    movaps  m%5, m%1
-    movlhps m%1, m%3
-    movhlps m%3, m%5
-    movaps  m%5, m%2
-    movlhps m%2, m%4
-    movhlps m%4, m%5
-    SWAP %2, %3
+    movlhps m%5, m%1, m%3
+    movhlps m%3, m%1
+    SWAP %5, %1
+    movlhps m%5, m%2, m%4
+    movhlps m%4, m%2
+    SWAP %5, %2, %3
 %endmacro
 
 %macro TRANSPOSE8x8W 9-11



More information about the ffmpeg-cvslog mailing list