Scene optimisation

Started by cyphyr, April 19, 2019, 03:11:00 AM

Previous topic - Next topic

cyphyr

I have a theory question about scene optimisation.

Which is quicker to render? Less nodes or less node connections?

I know that if I have multiple populations that use the same object it is quicker and "lighter" to load the object once and reference it many times rather than have each population load the same object again and again.
But what about nodes?
If an identical node, say a power fractal is used multiple times is it more efficient to have multiple copies of the same node or to have multiple connections to a single node?

I am thinking specifically about the internal "Tint defuse colour" on a population but I am sure there are many other instances where a similar question can be asked.
www.richardfraservfx.com
https://www.facebook.com/RichardFraserVFX/
/|\

Ryzen 9 5950X OC@4Ghz, 64Gb (TG4 benchmark 4:13)

bobbystahr

How exactly do you do the referencing thing Richard...never been able to suss that?
something borrowed,
something Blue.
Ring out the Old.
Bring in the New
Bobby Stahr, Paracosmologist

cyphyr

Quote from: bobbystahr on April 19, 2019, 11:13:45 AM
How exactly do you do the referencing thing Richard...never been able to suss that?

You just delete the object node inside the population and create a new object node outside of the relevant type (tgo, obj or lwo) then copy the name of the new object node into the "Object marker" box in the population node.
www.richardfraservfx.com
https://www.facebook.com/RichardFraserVFX/
/|\

Ryzen 9 5950X OC@4Ghz, 64Gb (TG4 benchmark 4:13)

bobbystahr

Quote from: cyphyr on April 19, 2019, 12:42:58 PM
Quote from: bobbystahr on April 19, 2019, 11:13:45 AM
How exactly do you do the referencing thing Richard...never been able to suss that?

You just delete the object node inside the population and create a new object node outside of the relevant type (tgo, obj or lwo) then copy the name of the new object node into the "Object marker" box in the population node.

D'oh...should have figured that out...more senior's moments than I'd like lately
something borrowed,
something Blue.
Ring out the Old.
Bring in the New
Bobby Stahr, Paracosmologist

Matt

If you construct a shader network where two branches of the network reference the same fractal while shading the same point in space, this might be slower if you use a duplicate node instead, because the fractal caches some of its evaluations for reuse by nearby points. This doesn't happen in your population tinting example, but it might be able to share data from the largest features in the fractal. I doubt it would make much difference but you could try it.

For most nodes I don't think you'll benefit from reusing the same shader. But I don't know for sure, so some real world testing might show some surprising results.

Just because milk is white doesn't mean that clouds are made of milk.