efficiency of nodes
ghostship last edited by
Some initial testing tonight is leading me to think that the older Poser nodes run slower in Superfly than Cycles nodes that do the same thing.
My initial test involved an older hair shader that I made with Poser nodes that was pretty simple and comparing that with my freebee shader that only uses Cycles nodes. The Poser node thing was about 5 or 6 nodes and the Cycles thing is huge compared to it yet rendered faster.
It's late here and I need to go to bed, so I'll look into this more tomorrow. If anybody want's to look at this in the meantime feel free.
Make sure whatever nodes you use, you also plug them into Cycles root, measure, and then into Poser Surface root, and measure again. I have found the SAME nodes perform differently depending on which root you used. I'm talking about plugging Cycles nodes into Poser Surface Alternate_Diffuse with all the other stuff in it turned off.
Also, try a couple "simple" shaders with just Poser_Surface builtin specular, vs. a hand-assembled one into Alt Diffuse like I always do, using Diffuse, Blinn, Blender, and Fresnel_Blend. The built-in specular + diffuse + translucence + transparency + ambient is actually a big shader tree, and if you're not using all its parts they are wasteful.
I have also found that the GlossyBsdf is slightly faster than the legacy Blinn or Specular nodes, so I clearly see evidence that node-for-node there are some performance differences.
However having found a small difference (I'm talking .5 seconds added to 28 seconds base) it would not pay to spend, say, 15 minutes replacing Blinn nodes. I'd have to render 900 times before that would pay for itself.
phdubrov last edited by phdubrov
The built-in specular + diffuse + translucence + transparency + ambient is actually a big shader tree, and if you're not using all its parts they are wasteful.
I experimented with big pure Cycles shaders with the parts of the nodertree mixed with zero weight vs. deleted. Maximum difference was ~5%.
I checked the source code. There are optimizations where a MixClosure with 0% or 100% factor will eliminate the Mixture. There are no such optimizations for AddClosure. If you add a Diffuse at 0% plus Emission at 0% plus Translucent at 0% etc. they will all be there.