[FFmpeg-cvslog] x86: fix overflow in h264 8x8 planar prediction

[Mans Rullgard]

ffmpeg | branch: master | Mans Rullgard <mans at mansr.com> | Mon Jan 24 23:24:28 2011 +0000| [e243ed656c445e504d2e1fa83ad28c7f61209b8b] | committer: Michael Niedermayer

x86: fix overflow in h264 8x8 planar prediction

Signed-off-by: Mans Rullgard <mans at mansr.com>
(cherry picked from commit 80944df720da98d6e5ee0e355db5814735914ec9)

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e243ed656c445e504d2e1fa83ad28c7f61209b8b
---

 libavcodec/x86/h264_intrapred.asm |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/libavcodec/x86/h264_intrapred.asm b/libavcodec/x86/h264_intrapred.asm
index dbe6b8a..28c2f39 100644
--- a/libavcodec/x86/h264_intrapred.asm
+++ b/libavcodec/x86/h264_intrapred.asm
@@ -637,10 +637,6 @@ cglobal pred8x8_plane_%1, 2, 7, %2
 %endif
     paddw        m0, m1           ; sum of H coefficients
 
-    pmullw       m0, [pw_17]
-    paddw        m0, [pw_16]
-    psraw        m0, 5
-
     lea          r4, [r0+r2*4-1]
     lea          r3, [r0     -1]
     add          r4, r2
@@ -694,6 +690,10 @@ cglobal pred8x8_plane_%1, 2, 7, %2
     shl          r3, 4
     movd        r1d, m0
     movsx       r1d, r1w
+    imul        r1d, 17
+    add         r1d, 16
+    sar         r1d, 5
+    movd         m0, r1d
     add         r1d, r5d
     sub         r3d, r1d
     add         r1d, r1d