[Mplayer-cvslog] CVS: main/drivers/radeon README,1.14,1.15

[Michael Niedermayer]

Hi

On Friday 04 January 2002 08:40, Nick Kurshev wrote:
[...]
> > u could use fixpoint math
                ^^^^^^^^
> > 1<<24 == "1.0"
> > int64_t a,b,c
> > ADD(a,b) a + b
> > MUL(a,b) (a*b + (1<<23))>>24
> > DIV(a,b) (a<<24)/b 	//not correctly rounded ...
>
> I never studied bitfield representation of floating types in depth but as I
> remeber: -1 = 111111111b for integer
> -1.0 == 110000000b for  floats
> (roughly)
> so probably your a+b won't work propertly.
fixpoint math not floatingpoint ...

[...]

Michael