Copy2Pts x PointXform [Bug] [SOLVED]

   338   10   1
User Avatar
Member
71 posts
Joined: March 2011
Offline
Hi,

I´m having problems with copy2points in LOPs when copying spot lights.
There is an animated “N” attrib controlling the direction of the light. In the viewport It seems fine, but when rendering, the direction of the light flickers.
I´ve done the same thing using the Point Xform and the render behaves like the viewport, so I´m thinking this is a bug. Can anyone confirm It?

Anyway, is this approach fine? Does It have a big performance hit while creating primitives when compared with just changing their attributes?
Edited by guilhermecasagrandi - Oct. 28, 2020 21:16:40

Attachments:
copy2points_[0001-0024].mp4 (320.9 KB)
ptxform_[0001-0024].mp4 (293.0 KB)
LOP_copy2pts_problem.hiplc (4.5 MB)

User Avatar
Member
3956 posts
Joined: Sept. 2011
Offline
That's really strange. I modified the pointxform side to see if putting the transform on the reference instead of the light produced the same result but it was still normal. The matrices are slightly different, but only after the 5th or 6th sig fig.
User Avatar
Member
71 posts
Joined: March 2011
Offline
jsmack
That's really strange. I modified the pointxform side to see if putting the transform on the reference instead of the light produced the same result but it was still normal. The matrices are slightly different, but only after the 5th or 6th sig fig.
Yes, and the copy2points behaves the same in the viewport, only in the rendered images It behaves differently.
User Avatar
Member
3956 posts
Joined: Sept. 2011
Offline
The length of the row vectors of the copy to points matrix is wacky:
X :0.987061 Y: 1.32959 Z: 0.511131
my mistake, it's a little off, but not crazy:

X: 1.00161 Y: 1.00003 Z: 1.00154
Edited by jsmack - Oct. 26, 2020 17:37:35
User Avatar
Staff
3777 posts
Joined: July 2005
Offline
If you look at the “Local to World Xform” of the light coming out of the Instancer LOP using the spreadsheet view, you can see that it's generating small non-unit scale values (the scale coming out of point xform is exactly 1.0). Even if I turn off “Set Scale” in the Instancer LOP. I'm hoping that's where the discrepancy is coming from, but I'll need to do a bit more digging to figure out why it's ending up with scales in the xform… I'll let you know when/if I make any progress.
User Avatar
Member
3956 posts
Joined: Sept. 2011
Offline
mtucker
If you look at the “Local to World Xform” of the light coming out of the Instancer LOP using the spreadsheet view, you can see that it's generating small non-unit scale values (the scale coming out of point xform is exactly 1.0). Even if I turn off “Set Scale” in the Instancer LOP. I'm hoping that's where the discrepancy is coming from, but I'll need to do a bit more digging to figure out why it's ending up with scales in the xform… I'll let you know when/if I make any progress.

is it a karma bug that 0.1 % scale would cause the spot light size to vary that much?
User Avatar
Staff
3777 posts
Joined: July 2005
Offline
The bug is that there is any scale being set at all. This is happening because we are doing some internal orientation calculations at half-float precision in the Instancer LOP. This problem will be fixed in tomorrow's build.

I don't know how to judge if karma is responding “correctly” to these scale values…
User Avatar
Member
71 posts
Joined: March 2011
Offline
mtucker
The bug is that there is any scale being set at all. This is happening because we are doing some internal orientation calculations at half-float precision in the Instancer LOP. This problem will be fixed in tomorrow's build.

I don't know how to judge if karma is responding “correctly” to these scale values…

Thanks again!

At least in the flipbook, I don´t see the scaling effect, only in the render (with Karma).
User Avatar
Member
3956 posts
Joined: Sept. 2011
Offline
mtucker
The bug is that there is any scale being set at all. This is happening because we are doing some internal orientation calculations at half-float precision in the Instancer LOP. This problem will be fixed in tomorrow's build.

I don't know how to judge if karma is responding “correctly” to these scale values…

I did some tests, and karma seems abnormally sensitive to these scales, although I wouldn't expect scales to affect the spot light effect at all.) A scale of as little as 0.1 % is enough to make the spot radius 10x larger.
User Avatar
Staff
6 posts
Joined: May 2019
Offline
Hi guys
The spotlight transform-scaling bug has been fixed and will be out in the next nightly build.
Thanks!
User Avatar
Member
71 posts
Joined: March 2011
Offline
brians
Hi guys
The spotlight transform-scaling bug has been fixed and will be out in the next nightly build.
Thanks!
Thank you guys for the awesome support!!! (as always).
  • Quick Links