Billy Biggs vektor at dumbterm.net
Thu Dec 18 21:38:50 EET 2003

Keith Packard (keithp at keithp.com):

> Around 13 o'clock on Dec 18, Billy Biggs wrote:
> >   Why not make 0x80000000 (or above) be opaque, as you still have lots
> > of precision, and this would make the multiplication math easier.
> Ah, there's no precision requirement in this spec.  My implementation takes
> the value, stuffs it into a picture and uses it as the mask in the 
> compositing operation, so I'm only using 8 bits.  Changing Render to make 
> those operations less precise is a different question; I'm not sure that's 
> entirely necessary assuming either hardware acceleration or at least 
> MMX-like software.

  Less precise?  I am sorry, I don't follow what you're saying at all.
I see how you're using the value though, and yeah, you would have to do
a little more work in my scheme, but it's just one value.

  My argument is that if I have a window with an alpha channel, and I
have to multiply that by a constant value before compositing it, why not
let me divide by 32768 (easy) instead of by 65535 (hard).


More information about the xdg mailing list