Wasted my time.
by Bill Vee · 10/21/2009 (1:37 pm) · 3 comments
With the release of TGEA 1.8.2 I was happy that the atlas rendering problems I had run into were apparently fixed. Boy was I wrong.
I should have known after I diffed the source with 1.8.1 and found virtually no changes to the atlas code. I assumed it was a dependent class that was causing the problem.
As far as I have been able to figure out the problem is in the AtlasClipMapBatcher or the clipmap code.
The problem occurs when the ClipMap::recenter code gets called and the clipmap has to jump a large distance say from the center to the side or from one side to the aother in a single frame.
This can happen more often than you think.
For instance when you first load a atlas file the clipmap is set to some arbitrary position, the center I think, then when the atlas file starts to render it recenters the clipmap to the correct position and you get this corrupted clipmap. This can also happen if you "leave" a clipmap from one side and re-enter from another.
Now why this doesn't happen to the MegaTerrain I haven't figured out yet.
It has to be the fact the atlas is loading and unloading image data from the atlas file and MegaTerrain doesn't.
I will just have to resume with the 1.7.1 code base.
DayOfWar.net
I should have known after I diffed the source with 1.8.1 and found virtually no changes to the atlas code. I assumed it was a dependent class that was causing the problem.
As far as I have been able to figure out the problem is in the AtlasClipMapBatcher or the clipmap code.
The problem occurs when the ClipMap::recenter code gets called and the clipmap has to jump a large distance say from the center to the side or from one side to the aother in a single frame.
This can happen more often than you think.
For instance when you first load a atlas file the clipmap is set to some arbitrary position, the center I think, then when the atlas file starts to render it recenters the clipmap to the correct position and you get this corrupted clipmap. This can also happen if you "leave" a clipmap from one side and re-enter from another.
Now why this doesn't happen to the MegaTerrain I haven't figured out yet.
It has to be the fact the atlas is loading and unloading image data from the atlas file and MegaTerrain doesn't.
I will just have to resume with the 1.7.1 code base.
DayOfWar.net
About the author
#2
10/25/2009 (2:36 am)
If 1.7.1 is rendering correctly, perhaps a dif against 1.8.2 will show the critical section....
#3
I was even doing that last night till 1:00 AM.
I just did it again for the last 2 hours.
There is very little changes to the atlas code and what changes there are I think I have thoroughly explored for the cause.
It is really frustrating.
My number one suspect I haven't been able to rule out is the stateblock system introduced in 1.8.0.
I am still learning how the stateblock system works and how it could impact atlas terrain.
10/25/2009 (2:05 pm)
Been there done that.I was even doing that last night till 1:00 AM.
I just did it again for the last 2 hours.
There is very little changes to the atlas code and what changes there are I think I have thoroughly explored for the cause.
It is really frustrating.
My number one suspect I haven't been able to rule out is the stateblock system introduced in 1.8.0.
I am still learning how the stateblock system works and how it could impact atlas terrain.
Torque 3D Owner Marton Szucs
lol , I didnt see the 1.8.2 is released :)