no… my code is working right. if you do multiply vert pos on objecttransform your code makes the same face number as mine.
Yes, I tried my original algorithm again. Appearantly there is no memory problems with max 2008, or they are orders of magnitude smaller than for 2009. But the result in terms of triangle culling is the same for both versions. This is strange, because I’m pretty sure I had this working in max 2009. I’ll have a look at it tomorrow at work.
Well, back at work. Although your rewrite is clearly better in many details, it still suffers the same insane memory leak when used against my real data in MAX 2009. It also suffers the same slowdown when I try it in MAX 2010 (really, REALLY slow…).
However, the triangle culling problems are nowhere to be seen in 2009 or 2010… but as you mentioned, I was doing it wrong with baseobject vertex coords, just didn’t notice when my test objects where placed at 0,0.
I’m going to try MAX 9 next, or some 32-bit versions.