|
meyendlesss
???????????
|
I noticed something weird while plugging a Free Gradient into a Divide component...
What's causing the line to show up?  |
| Posted: May 6, 2010 12:14 pm |
Details
E-Mail
|
|
meyendlesss
???????????
|
Invert the gradient and the line is gone...  |
| Posted: May 6, 2010 12:15 pm |
Details
E-Mail
|
|
meyendlesss
???????????
|
Rotating also gets rid of it...  |
| Posted: May 6, 2010 12:15 pm |
Details
E-Mail
|
|
meyendlesss
???????????
|
This time I typed in 0.50000000000001 (which showed still as 0.5) for the 'start X' and got a single pixel in the corner...  |
| Posted: May 6, 2010 12:19 pm |
Details
E-Mail
|
|
meyendlesss
???????????
|
Is this some kind of bug in the gradient or divide components?
Or is it something that I just don't understand?
|
| Posted: May 6, 2010 12:22 pm |
Details
E-Mail
|
|
Totte
Ãœbernerd
|
Isn't that divide by zero or something? - I never expected the Spanish inquisition |
| Posted: May 6, 2010 1:19 pm |
Details
E-Mail
|
Vladimir Golovin
Administrator
|
Pink is error color (see the inputs of the Divide component). If you see pink in the output of Divide, it means you're dividing by zero in that region.
|
| Posted: May 6, 2010 1:28 pm |
Details
E-Mail
|
|
meyendlesss
???????????
|
I kind of asked the original question in a hurry...
I understand what the error color is.
I'm still curious why it's not there if I invert or rotate the gradient.
Wouldn't it still be dividing by 0 and still show the error color?
|
| Posted: May 19, 2010 2:28 pm |
Details
E-Mail
|
|
tigerAspect
|
This could be a symptom of the weird gradient bug where white is black. I know it exists in Free Gradient, so I don't see why it wouldn't show up in an Angular one. Granted, it could be in Elevation gradient, so this may be unrelated.
|
| Posted: May 19, 2010 8:49 pm |
Details
E-Mail
|
|
Sphinx.
Filter Optimizer
|
It is due to floating point precision and the way floating point values are represented bitwise (approximations).
In your first example, the lowest value is 0, but the highest is 0.9992274. If you invert this, you will not get an exact zero, but something close to.
Small changes (like rotate) can change the exact representation slightly which is why you see different results.
Perhaps it traces back to the precision of the internal angular gradient function, dunno. |
| Posted: May 20, 2010 2:37 am |
Details
E-Mail
|
