niemeyer

Is it my math or something in the nurbs object that will not allow this 

Cassini Oval to close? I want to model mitosis...Thanks for any help!

Dear Jitter List,
Is it my math or something in the nurbs object that will not allow this 
Cassini Oval to close? I want to model mitosis...Thanks for any help!

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


cassini-oval-nurbs-wont-close

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

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.

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

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

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.

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

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!

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!

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 

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.

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?

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?

Cassini Oval, NURBS won't close