[FFmpeg-devel] [PATCH] Faster ff_sqrt()

[Vitor Sessak]

Hi

Michael Niedermayer wrote:
> On Sun, Jan 13, 2008 at 08:13:00PM +0100, Michael Niedermayer wrote:
>> On Sun, Jan 13, 2008 at 05:50:52PM +0100, Vitor Sessak wrote:

[...]

>> for a less idiotic and simpler variant i wrote a few years ago see
>> http://guru.multimedia.cx/fast-integer-square-root/
>>
>> if anyone can simplify this further that would be welcome :)
> 
> noone? well heres a simpler variant, though possibly slower (didnt benchmark)
> 
> static inline unsigned int sqrt4(unsigned int a)
> {
>     unsigned int b;
> 
>     if(a<(1<<16)){
>         if(a<(1<<10)-3)      b=sqrt_tab[(a+ 3)>>2 ]>>3;
>         else{
>             if(a<(1<<14)-28) b=sqrt_tab[(a+28)>>6 ]>>1;
>             else             b=sqrt_tab[ a    >>8 ];
>         }
>     }else{
>         int s= (av_log2(a)-12)>>1;
>         b= sqrt_tab[a>>(2*s+6)];
>         b= (FASTDIV(a,b)>>s) + (b<<(s-2));
>     }
> 
>     if(a<b*b) b--;
> 
>     return b;
> }
> 

It is indeed simpler and slightly faster than the patch I proposed...

-Vitor