[x265] [PATCH] all_angs_pred_8x8: optimize mov instruction

[Steve Borho]

On 01/14, praveen at multicorewareinc.com wrote:
> # HG changeset patch
> # User Praveen Tiwari
> # Date 1421231301 -19800
> # Node ID 2d617be47353f481c6c6e7e235cee1ba58f340f7
> # Parent  7ad510c96fea192a9e834a8340905b54bc49aa9d
> all_angs_pred_8x8: optimize mov instruction

3731c -> 3642c, queued

> diff -r 7ad510c96fea -r 2d617be47353 source/common/x86/intrapred8.asm
> --- a/source/common/x86/intrapred8.asm	Wed Jan 14 12:06:54 2015 +0530
> +++ b/source/common/x86/intrapred8.asm	Wed Jan 14 15:58:21 2015 +0530
> @@ -10659,35 +10659,35 @@
>      ; mode 18 [row 0, 1, 2, 3, 4, 5, 6, 7]
>  
>      movh          m1,          [r2]
> -    movh          [r0 + 1024], m1
>  
>      pslldq        m2,          m1,         1
>      pinsrb        m2,          [r2 + 1 + 16],   0
> -    movh          [r0 + 1032], m2
> +    punpcklqdq    m1,          m2
> +    movu          [r0 + 1024], m1
>  
>      pslldq        m2,          1
>      pinsrb        m2,          [r2 + 2 + 16],   0
> -    movh          [r0 + 1040], m2
> -
> -    pslldq        m2,          1
> -    pinsrb        m2,          [r2 + 3 + 16],   0
> -    movh          [r0 + 1048], m2
> -
> -    pslldq        m2,          1
> -    pinsrb        m2,          [r2 + 4 + 16],   0
> -    movh          [r0 + 1056], m2
> -
> -    pslldq        m2,          1
> +
> +    pslldq        m0,          m2,          1
> +    pinsrb        m0,          [r2 + 3 + 16],   0
> +    punpcklqdq    m2,          m0
> +    movu          [r0 + 1040], m2
> +
> +    pslldq        m0,          1
> +    pinsrb        m0,          [r2 + 4 + 16],   0
> +
> +    pslldq        m2,          m0,              1
>      pinsrb        m2,          [r2 + 5 + 16],   0
> -    movh          [r0 + 1064], m2
> +    punpcklqdq    m0,          m2
> +    movu          [r0 + 1056], m0
>  
>      pslldq        m2,          1
>      pinsrb        m2,          [r2 + 6 + 16],   0
> -    movh          [r0 + 1072], m2
> -
> -    pslldq        m2,          1
> -    pinsrb        m2,          [r2 + 7 + 16],   0
> -    movh          [r0 + 1080], m2
> +
> +    pslldq        m0,           m2,             1
> +    pinsrb        m0,          [r2 + 7 + 16],   0
> +    punpcklqdq    m2,          m0
> +    movu          [r0 + 1072], m2
>  
>      ; mode 19 [row 0, 1]
>  
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel

-- 
Steve Borho