hello,

it seems that there is something a bit wrong with your math, but I have no clue what exactly.

I tried to send matrix generated in javascript to jit.gl.mesh, and the problem was still there, so it seems not to be nurbs related.

try to draw blob crossection:

for (var l=0; l < 16; l++){

and move camera to 2.27 3.47 2.98

maybe you'll see something that might help...

you could also use jit.expr to do what your javascript peopleMaker does, if you need more speed.

hth, nesa

yes, I also found out that nurbs is not too happy with many points converging in [0, 0] but with [0.0001, 0.0001] there is no problem. Are you suggesting replacing all the math with jit.expr objects ? I did not realize that would be faster than JS Math objects, but I will try!

Thanks, Greg

Comments Welcome: http://cnm.berkeley.edu/

Check out the performance difference between jsglsurface-example.pat (uses jit.expr to evaluate surface) and jsglsurface2-example.pat (uses JS to evaluate surface). You'll notice something like a 20-40x performance difference in your surface calulation by useing jit.expr. However if there aren't too many control points to calulate, then it's not likely to result in overall performance benefits.

Also worth noting, certain types of calculation (like using adjacent input or output cells to calculate the expression) are either a nuisance (for adjacent input cells) or not possible (for adjacent output cells). For adjacent input cells you *can* use one of the spatial objects like jit.rota(in[0], @offset_x -1) in your expression to grab a neighboring cell along the x axis for instance.

Haven't had time to examine your nurbs closing problem, but let us know if you still are unable to track it down and I can try to figure out a solution. The nurbs object closes connections about each axis (connect top to bottom or left to right) and hence is good for closed connections like a cylinder or torus, but doesn't close connections *along* any given side of your mesh, so it can't smoothly make spheres for instance.

Hope this helps.

-Joshua

hello,

besides 'many points in 0,0' thing, i've noticed that there are some surfaces that 'bend' or 'sunk' from the top of your shape towards the bottom, the crossection of the top looks like more like M than like Π , like when top vertex goes from top to bottom... or maybe this was intended?

thanks to Joshua for jit.expr tips!

best, nesa