-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

IanM wrote:
> Hi guys:

don't be so sexist :-)
[Snip an interesting idea]
> Am I missing something?

I say, but I have absolutly no authority : "
 - You take the source,
 - make a reference image-set with original source code for all the
pov-scene you can find (start with the ones delivered with the source:
allanim and allscene (as well as portfolio...)), at a decent size.
 - make your patch to use your simpler solver
 - recomputes the whole image-set with your new code.
Then compares if there is any difference in the pictures (imagemagick
has tools for that), and what is the gain or loss of performance
 (both on a specific scene and on average)
IF and ONLY IF there is no difference on the images, and there is a real
time-gain, then you might want to push your faster code in think like
megapov... and make a lot of advertisement about it so it get taken by
the pov-team..."

Now I'm always afraid of mathematical ideas applied to computational
performance. (the notion of precision that can be achieved on a computer
is something that most abstract mathematicians do not understand.
For instance, 0.1 x 10 is NOT 1.0000 !
(well in fact, there is no way to represent 0.1 exactly with classical
floating point convention). So beware of idea that works fine on paper,
but fails for some computational values.

Moreover, when you said that 2 of the 3 roots are not needed, a full
revue of the relevant code should prove that. Same goes for the cubic
term being always 1... make a web page (with two columns of code) to
explain the changes you are suggesting.


- --
Eifersucht ist die Leidenschaft, die mit Eifer sucht, was Leiden schafft.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (MingW32)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFDHuEDs/YJ43cSjHIRAkPGAJ9aggqoJTsF7aMgll622DHyGZQnHQCfQElh
uhD4G1Td8KYoAfUs1rcd7aE=
=TCZ7
-----END PGP SIGNATURE-----

Post a reply to this message