[x265] [PATCH] blockfill_s_8x8 sse2 asm code optimization

[Praveen Tiwari]

Sent updated patch. Thanks.

Regards,
Praveen

On Mon, Feb 2, 2015 at 4:39 PM, chen <chenm003 at 163.com> wrote:

>
>
>
>
>
> At 2015-02-02 16:55:16,praveen at multicorewareinc.com wrote:
> ># HG changeset patch
> ># User Praveen Tiwari
> ># Date 1422867249 -19800
> ># Branch stable
> ># Node ID 2618352a21d5917ee8c1f79bcc159e858dd19daa
> ># Parent  e2c958ff874e2bf8992ba22605e993530e8a2d8c
> >blockfill_s_8x8 sse2 asm code optimization
> >
> >improved, 100.04c -> 90.05c
> >
> >diff -r e2c958ff874e -r 2618352a21d5 source/common/x86/blockcopy8.asm
> >--- a/source/common/x86/blockcopy8.asm	Sat Jan 31 13:48:34 2015 -0600
> >+++ b/source/common/x86/blockcopy8.asm	Mon Feb 02 14:24:09 2015 +0530
> >@@ -1748,9 +1748,10 @@
> > ; void blockfill_s_8x8(int16_t* dst, intptr_t dstride, int16_t val)
> > ;-----------------------------------------------------------------------------
> > INIT_XMM sse2
> >-cglobal blockfill_s_8x8, 3, 3, 1, dst, dstStride, val
> >+cglobal blockfill_s_8x8, 3, 4, 1, dst, dstStride, val
> >
> > add        r1,            r1
> >+lea        r3,            [3 * r1]
> >
> > movd       m0,            r2d
> > pshuflw    m0,            m0,         0
> >@@ -1760,17 +1761,13 @@
> > movu       [r0 + r1],     m0
> > movu       [r0 + 2 * r1], m0
> >
> >-lea        r0,            [r0 + 2 * r1]
> >+movu       [r0 + r3],     m0
> >+movu       [r0 + 4 * r1], m0
> >+
> >+lea        r0,            [r0 + 4 * r1]
> swap LEA and above movu, you will get less bytes on binary code
>
>
> > movu       [r0 + r1],     m0
> > movu       [r0 + 2 * r1], m0
> >-
> >-lea        r0,            [r0 + 2 * r1]
> >-movu       [r0 + r1],     m0
> >-movu       [r0 + 2 * r1], m0
> >-
> >-lea        r0,            [r0 + 2 * r1]
> >-movu       [r0 + r1],     m0
> >-
> >+movu       [r0 + r3],     m0
> > RET
> >
> > ;-----------------------------------------------------------------------------
> >_______________________________________________
> >x265-devel mailing list
> >x265-devel at videolan.org
> >https://mailman.videolan.org/listinfo/x265-devel
>
>
> _______________________________________________
> 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/20150202/28de94e0/attachment-0001.html>