I just wanted to follow up that my input geometry had a discontinuity (a "hole" in one contour). When the discontinuity is removed, the OGL tessellator finishes in a reasonable time and space. arg!! sorry everyone ... bad data.

btw I'd appreciate any pointers to good 2d tessellators ... I am familiar delaunay triangulation and trimming out interiors ... what does OGL use? thanks!

hi all,

I don't have much experience with the OGL tessellator, but does anyone have a rule of thumb on how it should scale with input geometry? I'm putting in some pretty large polygons and it just keeps growing on the heap, over 512mib ... is this expected? My polygons have ~8 contours each, with perimeter < 2000px each. Edge counts are ~1000 per contour.

The offending object is the "GLUhalfedge" (approximate name ... I can't remember the exact name), which consumes most of the space.

Has anyone had problems with tessellating a complex polygon?

I'm using a work-around that is to break up the polygon into pieces, but this isn't optimal (just more computation time).  I'd like to get feedback from more experienced people ... I feel the amount of memory being used is way too much ... that is, I think I'm missing something.

My goal is to tessellate in 32mb of heap ... is that reasonable?

hi all,

If I want to do picking often (maybe 20 times a second) for a 2d scene, would it be better to store IDs in the depth buffer than to use the standard selection approach (a render per pick)? That is, read the ID from the depth buffer using glReadPixels? Rendering is much faster than picking in the CPU for my case.

thanks

ah, yes, GL has access to all the goods, with pointers being replaced by Buffer objects.

hi all,

Does anyone know if the usage enum for buffer objects is supported in JOGL? Does anyone have any insight into whether this hint affects performance? e.g. GL_STATIC_DRAW for a shape that is repeatedly rendered.

Regards, Brien