Re: [AGG] Re: Small bug in agg_alpha_mask_u8.h?

[Maxim S. <mcs...@ya...>]

> byte f(byte alpha, byte cover) 
> {
>   int q=alpha*cover+128; 
>   return (q+(q>>8))>>8;  
> }
> 
> exact for 0< alpha,cover,f() < 256 and not terribly slower than
> (alpha*cover + 255) >> 8. 

Thanks! I have checked it, and it works well even for 16 bits, and even when
alpha=16bit, cover=8bit. But the latest, 
((65535*255+128) + ((65535*255+128)>>8))>>8 gives us 65534. But that's fine I
suppose, because I always check for "cover < 255".

McSeem


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com