 |
|
|
|
|
|
| |
| |
|
|
From: Jörg 'Yadgar' Bleimann
Subject: Tweaking PoV-Ray's epsilon value?
Date: 16 Oct 2003 15:22:25
Message: <3F8F0EF0.83A66CBE@gmx.de>
|
|
|
| |
| |
|
|
High!
Experimenting with worlds covering a large range of orders of magnitude
(for example the Solar System with 1 unit = 1 km), I pretty soon
discovered the limits of PoV-Ray... nothing smaller than 0.0001 or
larger/further away than 10,000,000 units is rendered correctly. I
remember having brought up that issue years ago in
comp.graphics.rendering.raytracing, back then I was told of the ominous
"epsilon". What is this "epsilon" precisely?
Is there a not too complicated way to modify the range of orders of
magnitude to, let's say 1e-8 to 1e13, so that everything from amoeba to
Pluto's orbit can be rendered within one scene, by hacking the code?
See you in Khyberspace!
Yadgar
Post a reply to this message
|
|
| |
| |
|
|
From: Tim Nikias v2 0
Subject: Re: Tweaking PoV-Ray's epsilon value?
Date: 16 Oct 2003 17:31:37
Message: <3f8f0e39@news.povray.org>
|
|
|
| |
| |
|
|
The problem lies with several issues:
we only have a floating-point CPU to a certain degree of detail. We cannot
circumvent that due to compability reasons between the different platforms.
We further need a range a User cannot specify, so that the options he may
specify will be rendered correctly within the given precision boundaries.
And when do you actually need to render an amoeba with Saturn in one image,
where you can't fake either one with textures or variable scaling of
distances?
--
Tim Nikias v2.0
Homepage: http://www.digitaltwilight.de/no_lights
Email: tim.nikias (@) gmx.de
>
> High!
>
> Experimenting with worlds covering a large range of orders of magnitude
> (for example the Solar System with 1 unit = 1 km), I pretty soon
> discovered the limits of PoV-Ray... nothing smaller than 0.0001 or
> larger/further away than 10,000,000 units is rendered correctly. I
> remember having brought up that issue years ago in
> comp.graphics.rendering.raytracing, back then I was told of the ominous
> "epsilon". What is this "epsilon" precisely?
> Is there a not too complicated way to modify the range of orders of
> magnitude to, let's say 1e-8 to 1e13, so that everything from amoeba to
> Pluto's orbit can be rendered within one scene, by hacking the code?
>
> See you in Khyberspace!
>
> Yadgar
>
---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.525 / Virus Database: 322 - Release Date: 09.10.2003
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>
> High!
Hi!
> Is there a not too complicated way to modify the range of orders of
> magnitude to, let's say 1e-8 to 1e13, so that everything from amoeba to
> Pluto's orbit can be rendered within one scene, by hacking the code?
See -
http://tag.povray.org/povQandT/languageQandT.html#largescaleproblems
--
Ken Tyler
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
In article <3F8F0EF0.83A66CBE@gmx.de>,
> Is there a not too complicated way to modify the range of orders of
> magnitude to, let's say 1e-8 to 1e13, so that everything from amoeba to
> Pluto's orbit can be rendered within one scene, by hacking the code?
No. Not generally, anyway. The epsilon and infinity limits could
probably be loosened a little without problems, but in general are
needed. Remember that intermediate values for the computations can go
quite a bit higher than the parameters you specify for the objects, and
can easily exceed the precision of the computer. If you really need this
range of precision, you should probably render the large-scale parts
separately from the small-scale parts and composite them together...but
I can't imagine why you would need this. In any image where a structure
the size of Pluto is visible, you aren't likely to see anything the size
of an ameboa.
--
Christopher James Huff <cja### [at] earthlink net>
http://home.earthlink.net/~cjameshuff/
POV-Ray TAG: chr### [at] tagpovrayorg
http://tag.povray.org/
Post a reply to this message
|
|
| |
| |
|
|
From: Xplo Eristotle
Subject: Re: Tweaking PoV-Ray's epsilon value?
Date: 19 Oct 2003 14:39:37
Message: <3f92da69@news.povray.org>
|
|
|
| |
| |
|
|
Christopher James Huff wrote:
>
> No. Not generally, anyway. The epsilon and infinity limits could
> probably be loosened a little without problems, but in general are
> needed. Remember that intermediate values for the computations can go
> quite a bit higher than the parameters you specify for the objects, and
> can easily exceed the precision of the computer. If you really need this
> range of precision, you should probably render the large-scale parts
> separately from the small-scale parts and composite them together...but
> I can't imagine why you would need this. In any image where a structure
> the size of Pluto is visible, you aren't likely to see anything the size
> of an ameboa.
Judging from the number of posts I've seen about the "problem", I would
say there's a fair number of people who run up against the limits. I
think the last time it happened to me, I was trying to do a render on a
"realistic" earth and atmosphere.. perhaps unnecessary, but hardly as
absurd as rendering Pluto and an amoeba at the same time.
If the problem is one of precision, could greater limits be supported on
64-bit platforms?
-Xplo
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
In article <3f92da69@news.povray.org> , Xplo Eristotle <xpl### [at] infomagicnet>
wrote:
> If the problem is one of precision, could greater limits be supported on
> 64-bit platforms?
If you had read the answer to the faq pointed out in this thread, you would
know how absolutely nonsensical this question is...
Thorsten
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
