[FFmpeg-cvslog] x86/imdct36: use extractps inside the STORE macro

[James Almer]

ffmpeg | branch: master | James Almer <jamrial at gmail.com> | Thu Jan 28 13:32:53 2016 -0300| [c79252897096b89376bcf17a5bca6cdf8d21b6a0] | committer: James Almer

x86/imdct36: use extractps inside the STORE macro

Reviewed-by: Michael Niedermayer <michael at niedermayer.cc>
Reviewed-by: Henrik Gramner <henrik at gramner.com>
Signed-off-by: James Almer <jamrial at gmail.com>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=c79252897096b89376bcf17a5bca6cdf8d21b6a0
---

 libavcodec/x86/imdct36.asm |    7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/libavcodec/x86/imdct36.asm b/libavcodec/x86/imdct36.asm
index 7218f06..409b2c5 100644
--- a/libavcodec/x86/imdct36.asm
+++ b/libavcodec/x86/imdct36.asm
@@ -143,6 +143,12 @@ SECTION .text
 %endmacro
 
 %macro STORE 4
+%if cpuflag(sse4)
+    movss     [%3       ], %1
+    extractps [%3 +   %4], %1, 1
+    extractps [%3 + 2*%4], %1, 2
+    extractps [%3 + 3*%4], %1, 3
+%else
     movhlps %2, %1
     movss   [%3       ], %1
     movss   [%3 + 2*%4], %2
@@ -150,6 +156,7 @@ SECTION .text
     movss   [%3 +   %4], %1
     movhlps %2, %1
     movss   [%3 + 3*%4], %2
+%endif
 %endmacro
 
 %macro LOAD 4