refactor index buffers creation #431
Conversation
Trass3r
force-pushed
the
indexbuffers
branch
3 times, most recently
from
ab7e912
to
ae191a9
Compare
| } | ||
|
|
||
| // remove trailing empty LODs |
There was a problem hiding this comment.
There was a problem hiding this comment.
Gem holder looks normal in the model viewer so not sure if it breaks something there.
There was a problem hiding this comment.
Though thinking again about how that code expects a sorted set of numbers I'm not sure why it works at all, resp. doesn't fail horribly.
use i8 buffers instead of i32 to save memory avoid creating empty LOD buffers use a List instead of a HashMap
|
Looks good to me. Is this only to fix the tangent generation warning (and bad data)? Was there anything else to this? |
|
It started simple by noticing we create int32 index buffers even though the source data is only int8. One more thing we could add to the naming improvements is taking the mesh name from the filename instead of the string saved in the kmf which is off for many MESH files. But we can also just leave it for now. |
No description provided.