Closed Hungary-Dude closed 10 years ago
Libgdx uses signed short indices, which makes it impossible to index (and therefor use) more than 32k vertices within a mesh. Technically it could be possible for fbx-conv to allow more than 32k vertices, but it would not be possible to render those with libgdx (hence the warning). The string "64k" is not an accepted number by atoi, I'm not sure why you would expect it to be so. However, I think that it would be appreciated if we can accept that number formatting. So please send a pull request if possible.
I expected *k numbers like 32k to be accepted because "32k" was listed in the help printout as the "default." So I assumed it was an accepted value.
Is it still the case that libgdx uses signed short indices? Is there no way to have more than 32k triangles in a mesh?
So this explains why I am not able to load complex models... @xoppa, why does libgdx use signed short indices? Is there any workaround so I can load models like this one: http://tf3dm.com/3d-model/sa342-gazelle-97281.html ? Thanks!
Because java doesn't support unsigned short. Don't expect random high poly models to work. Use low poly models instead. See https://github.com/libgdx/libgdx/wiki/Importing-Blender-models-in-LibGDX#maximum-vertices
Thanks @xoppa, and Merry Christmas!
Hi guys, I started use libGDX for our project, but limitation 32k is too small for us.
If I fork engine and change all short values for example to int does it work good?
Or short value is system restriction?
Can somebody explain me please?
After running:
FBX-conv gives me a warning that there are too many mesh vertices. So I try setting the maximum allowed indices with this command:
It prints the warning again and tells me the max is now "64." If FBX-conv does not accept *k numbers then please do not put them in the help printout. So I ran:
Now what does it do, it says "Error, maximum vertex count must be between 0 and 32k" This is annoying, as now I have to go back into my model editor and split up the node.
So in conclusion, please make FBX-conv accept *k numbers, and please make it accept larger vertex counts if possible.