Hi all,

POV-Ray for Windows version 3.6.1a.icl8.win32 
Windows XP sp2

I believe I have discovered a bug with #while loop processing when
comparing conditional values that are less than one.  When checking
the "Loop" variable for less than one, the #while block is still
processed when the value one is reached.

The bug can be best explained and reproduced with the following code.

>>> snip >>>

// demonstration of possible bug in POV-Ray for Windows
// Version 3.6.1a.icl8.win32
camera {
  location <0, 0.5, -2>
  look_at <0, 0.5, 0>
}


// bug - should not enter while block when "Loop" = 1
#declare Loop = 0.0;
#while ( Loop < 1 )
  sphere {
    <-0.1, Loop, 0>, 0.05 
    pigment { rgb <1, 0, 0> } 
    finish { ambient 1 } 
  }
  #declare Loop = Loop + 0.1;
#end

// example for comparison  
#declare Loop = 0.0;
#while ( Loop < 0.99 )
  sphere {
    <0.1, Loop, 0>, 0.05 
    pigment { rgb <0, 1, 0> } 
    finish { ambient 1 } 
  }
  #declare Loop = Loop + 0.1;
#end

>>> snip >>>

The #while loop that produces the red spheres is executed for the
"Loop" value of one when it should stop at 0.90, similar to "Loop <=
1".  The #while loop that produces the green spheres is there for
comparison.

The #while loop processes properly if the comparison is changed to a
higher value, such as "Loop < 10".  It also fails for lower values,
such as "Loop < 0.90", so the problem seems to be for values of one or
less.

Unless I'm really missing something obvious, this seems to be a bug.

Thanks to everyone involved for such a great product!

Best Regards,

Kyle Johnson

Post a reply to this message

hob### [at] gatenet wrote:
>   #declare Loop = Loop + 0.1;

  Not a bug. A limitation of floating point numbers.

  The number 0.1 is pathological in that it cannot be represented
accurately in the base-2 floating point format of the FPU (for the
same reason that for example 1/3 cannot be represented accurately
with a base-10 floating point representation).

  When you add 0.1 like this, the error caused by this accumulates
in such way that after you have added it ten times the result is
not 1 but a tiny fraction less.

  If you use integers only (ie. "#declare Loop = Loop + 1;") then
there's no such problem (because floating point numbers can
represent integers accurately up to a quite big value).

-- 
#macro M(A,N,D,L)plane{-z,-9pigment{mandel L*9translate N color_map{[0rgb x]
[1rgb 9]}scale<D,D*3D>*1e3}rotate y*A*8}#end M(-3<1.206434.28623>70,7)M(
-1<.7438.1795>1,20)M(1<.77595.13699>30,20)M(3<.75923.07145>80,99)// - Warp -

Post a reply to this message

"Warp" <war### [at] tagpovrayorg> wrote in message
news:425c5415@news.povray.org...
> hob### [at] gatenet wrote:
> >   #declare Loop = Loop + 0.1;
>
>   Not a bug. A limitation of floating point numbers.
>

The following link may be of interest

http://tinyurl.com/272fa

Post a reply to this message

Thanks to Warp and Tom.  This clears things up quite a bit.  The link
was very helpful.

Kyle


On Wed, 13 Apr 2005 09:46:15 +0100, "Tom Melly"
<pov### [at] tomandlucouk> wrote:

>"Warp" <war### [at] tagpovrayorg> wrote in message
>news:425c5415@news.povray.org...
>> hob### [at] gatenet wrote:
>> >   #declare Loop = Loop + 0.1;
>>
>>   Not a bug. A limitation of floating point numbers.
>>
>
>The following link may be of interest
>
>http://tinyurl.com/272fa
>

Post a reply to this message