Probleme mit Fix64

Hi Rolf, hello Avra,

using Fix64 quite a while now, I found the following problem: The Source for "FloatToFix64" above and the build in function in the compiler give erratic results when working with very small negative values:

The code

Var
lFl1 : Float;
..
Begin
lFl1 := -0.00000000001;
SLIP_WriteStr('TXT|lFl1: '+FloatToStr(lFl1));
SLIP_WriteStr('TXT|FloatToFix64(lFl1): '+Fix64ToStr(FloatToFix64(lFl1)) );
...
End;


Will give the following result:

lFl1: -9.999992E-12
FloatToFix64(lFl1): -1.000000000

The FloatToFix64 conversion gives -1 !! I checked, it is not the Fix64ToStr Function. So, all calculation with very small negative Values, that are converted from FloatToFix64 go wrong.

Can you have a look? If you fix the issue, I would be glad, if you could post the corrected code for FloatToFix64 here, so I can prevent the "uses uFix64" clause (making my code too big)

Hope you can help!

Thanks, Michael

Hi Rolf,

Yes, I think you are right. I did some testing and now I have changed the Function a little bit to prevent the false -1 results. However, Avra should have a look, because the internal AVRco routine in uFix64 gives the erratic results too..

Cheers, Michael

function FloatToFix64(const a: float): fix64;
var
Tmp: TFix64Overlay;
begin
If (a > -0.0000000009) And (a < 0) Then
a := 0;
EndIf;
Tmp.i := Trunc(a);
if (a < 0) and (Frac(a) <> 0) then
Dec(Tmp.i);
endif;
Tmp.f := Trunc(Frac(a) * (float($FFFFFFFF) + 1));
return(Tmp.fix);
end;

Hello Avra,
here a code snippet which shows the problem:
rolf

Hello Rolf,

Your solution is in the right direction, but it has a slight error if a=0. This should be the real fix:


I will send you the corrected unit once I return to my hometown.

Michael, Thank you for noticing this bug!