0
$\begingroup$

I have a Geo Node housekeeping question today that I can't quite answer with what I've read/found.

Disregarding specific use cases, which method is more expensive to isolate a one or two axes along a node formula?

a.) using vector math to multiply the axis by 1.0 or

b.) using a combine xyz to socket into the chosen axis

Maybe I'm asking if the math is calculated on all the empty sockets/fields for each node, I'm not really certain. Are they both the same in cost? (I'd like to know if they are different even if the difference in calculation is insignificant.)

Thanks for your time and thoughts!

$\endgroup$

1 Answer 1

2
$\begingroup$

(Using Blender 3.6.8)

The following GeometryNodes modifier is creating a $10^6$ points cloud.


In a first test, points are projected along the X axis either by multiplying their Position by the vector (1,0,0), either by separating& combining their X coordinate. Switching several times, to get some statistical trend, shows that both approaches run at about the same speed.

Multiply node: Multiply timing

Separate XYZ and Combine XYZ nodes: Combine timing


In a second test, identity transformation is applied either by multiplying Position by the vector (1,1,1), either by separating& combining the three coordinates. Switching several times, to get some statistical trend, shows that the speed of the Multiply approach is unaffected whereas the time increases significantly for the Separate & Combine approach.

Multiply node: Multiply timing

Separate XYZ and Combine XYZ nodes: Combine timing


Resources:

$\endgroup$
1
  • $\begingroup$ Interesting! Thank you! I don’t think I fully understand, so I would guess that some of the increase is contributed by the use of the additional node separating the coordinates. Is there a way to isolate combine xyz vs vector math (mult) for a similar test? Or am I missing the point (no pun intended, but ha, ha) entirely? $\endgroup$
    – sevens
    Commented May 25 at 16:34

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .