[x265] [PATCH] asm: fix crash due to unaligned memory access

chen chenm003 at 163.com
Wed Sep 2 15:27:50 CEST 2015


Please check input buffer, we need found reason, it is better than change mova to movu
we DCT input uffer may alignment for more performance.

At 2015-09-02 19:38:27,dnyaneshwar at multicorewareinc.com wrote:
># HG changeset patch
># User Dnyaneshwar G <dnyaneshwar at multicorewareinc.com>
># Date 1441193878 -19800
>#      Wed Sep 02 17:07:58 2015 +0530
># Node ID 8295bff6f2b70202af8aa99397767e21ff8ab5e4
># Parent  86e9bd7dd19278fceef65fc93a06dc8746ec9daf
>asm: fix crash due to unaligned memory access
>
>diff -r 86e9bd7dd192 -r 8295bff6f2b7 source/common/x86/dct8.asm
>--- a/source/common/x86/dct8.asm Tue Sep 01 17:06:05 2015 +0530
>+++ b/source/common/x86/dct8.asm Wed Sep 02 17:07:58 2015 +0530
>@@ -2112,25 +2112,25 @@
>     pxor     m5,  m5
>     shr      r3d, 3
> .loop:
>-    mova     m0, [r0]
>+    movu     m0, [r0]
>     pabsw    m1, m0
> 
>-    mova     m2, [r1]
>+    movu     m2, [r1]
>     pmovsxwd m3, m1
>     paddd    m2, m3
>-    mova     [r1], m2
>-    mova     m2, [r1 + 16]
>+    movu     [r1], m2
>+    movu     m2, [r1 + 16]
>     psrldq   m3, m1, 8
>     pmovsxwd m4, m3
>     paddd    m2, m4
>-    mova     [r1 + 16], m2
>+    movu     [r1 + 16], m2
> 
>     movu     m3, [r2]
>     psubusw  m1, m3
>     pcmpgtw  m4, m1, m5
>     pand     m1, m4
>     psignw   m1, m0
>-    mova     [r0], m1
>+    movu     [r0], m1
>     add      r0, 16
>     add      r1, 32
>     add      r2, 16
>_______________________________________________
>x265-devel mailing list
>x265-devel at videolan.org
>https://mailman.videolan.org/listinfo/x265-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20150902/3b7a78fc/attachment.html>


More information about the x265-devel mailing list