It seems like instancing large amounts of geo while using a 3rd party render engine creates a severe bottleneck with translation times. I have had this issue using Redshift and Megascans and today I stumbled over a facebook post from a user having the same issue with renderman. He found a solution, splitting up the point cloud into chunks of sub 50k. Doing this process manually is just not a very user friendly solution. Is there a way to make this chunking happen under the hood?
It would be really great if this issue could be resolved since it makes rendering large scenes extremely cumbersome with 3rd party render engines.
Here is the link to the FB post.
https://www.facebook.com/groups/HoudiniArtists/permalink/2883299385119180/ [
www.facebook.com]
And here is the text of that post in case you are not part of that FB group:
Little Tipp to every user which deals with large amounts of Instances and 3rd party Renderengines like Renderman.: Do chunking. In my current scene i have to render 21 billion polys only for grass. Which where instanced as packed primitives and variant copy to point system. When i did that i had translation times of up to 50 Minutes also because i am exporting the motionchannels for denoising. The tradional instancer of Houdini is useless and had translation times up to 3 hours or crashes. Now i have translation times of only 2 Minutes and 34 seconds.. to achive that you only need chunk up your point cloud..Because Renderengines doing scene translations only on single core.So keep the amount of instances per chunk under 50k per chunk. on 51k you will wait “ages” until your render starts to render. FREE File (without geometry) here:
https://drive.google.com/file/d/1D4IxxDRKIpw6qgUeI7vshfWUlBqzURm4/view?usp=sharing [
drive.google.com]
i build a semi automatic system which creates automatic chunkings.. and will also automaticly rename the cache paths for the point cloud patches. The results are 100 chunks in 100 folders consisting more than 29 thousand bgeo.sc files. you be able to read out the last digits of your node name and can use this expression in your file path's to automaticly create all dependencys only by dublicating the nodes… the expression is: opdigits($OS) Houdini is automaticly renaming the nodes sorted by numbers this can be used.. its easy to setup and will save you alot of time. Special thanks to Timucin Ozger and Ihab Ali for providing tipps and ideas to resolve this 4 day journey..