[x265] [PATCH] asm: Replaced SSE4 instructions with SSE2 and general purpose instructions for chroma vsp filter functions

[Nabajit Deka]

Okay. I will keep the earlier SSE4 code.


On Wed, Nov 13, 2013 at 4:59 PM, chen <chenm003 at 163.com> wrote:

> At 2013-11-13 14:53:27,nabajit at multicorewareinc.com wrote:
>
> ># HG changeset patch
> ># User Nabajit Deka
> ># Date 1384325281 -19800
> >#      Wed Nov 13 12:18:01 2013 +0530
> ># Node ID 017763dc543d091170082eccf7b42a0c47c453ff
> ># Parent  c4ca80d19105ccf1ba2ec14dd65915f2820a660d
>
> >asm: Replaced SSE4 instructions with SSE2 and general purpose instructions for chroma vsp filter functions.
> >
> >diff -r c4ca80d19105 -r 017763dc543d source/common/x86/ipfilter8.asm
> >--- a/source/common/x86/ipfilter8.asm Tue Nov 12 19:10:23 2013 +0530
> >+++ b/source/common/x86/ipfilter8.asm Wed Nov 13 12:18:01 2013 +0530
> >@@ -3183,11 +3183,16 @@
> >     packssdw  m0, m2
> >     packuswb  m0, m0
> >
> >-    pextrw    [r2], m0, 0
> >-    pextrw    [r2 + r3], m0, 1
> >-    pextrw    [r2 + 2 * r3], m0, 2
> >+    movd      r5d, m0
> >+    mov       [r2], r5w
> >+    shr       r5d, 16
> >+    mov       [r2 + r3], r5w
> >+    pshufd    m0, m0, 1
> >+    movd      r5d, m0
> pshufd + movd is slower than pextrd
>
> >+    mov       [r2 + 2 * r3], r5w
> >+    shr       r5d, 16
> >     lea       r2, [r2 + 2 * r3]
> >-    pextrw    [r2 + r3], m0, 3
> >+    mov       [r2 + r3], r5w
> >
> >     lea       r2, [r2 + 2 * r3]
> >
>
> _______________________________________________
> 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/20131113/02fd0ddc/attachment.html>