setting up multi seeded particle caching on the farm

[borisb2]

Hi.

 

At the moment I am in the process of including houdini in my FX-pipeline (finally ). I have a small farm working, managed by a render-manager (RenderPal).

 

Concerning particle caching:

With maya and partio I am used to cache out multi-seeded particles as 1-frame jobs, where a mel-script changes the seed of the particle-object according to the frame-number. As an example, if I want to have 15 caches, it will be a 15-frame job submitted to all machines as chunks of 1 frame each. So far so good.

 

I was reading, with houdini you typically use a wedge-rop to drive the seed. But how would I submit that to the render-manager? Is it a common technique to submit 1-frame-jobs and feed the Wedge-ROP somehow with the current frame-number (of the chunk)?

 

About rendering:

in Peter Quints videos he creates for every cache one delayed-load procedural. Is that still the up-to-date workflow?  Is there maybe a way to define just one procedural and somehow copy/stamp the file-path according to number of caches?

 

Thanks a lot

Edited August 6, 2014 by borisb2

[Artem Smirnov]

Well, it is quite dependent on the renderfarm maneger. For example, I intergrated houdini to pixars tractor recently. Tractor is based on the alfred scripts. When the wedge job is spooled, renderscript detects it by type of a node and write a sequence of subtasks to the alfred script. Each subtask with it's own wedge num. The resulting caches get unique filenames by inserting "wedgenum" and "spooled time" tokens.

 

For render you can use point instance procedural. Just generate as much points as you have wedges and write "instancefile" attrib on them.

[borisb2]

I'm on windows using RenderPal render manager. That submits system-commands to the various executables. Normally (for other sim-jobs like through maya) I write small mel-files to disk that get called through the "render iob"

Not sure how I would pass wedgenum or something similar to the hython-script (without writing as many hython-scripts to disk as I need wedges). But I guess with subtasks you mean in my case as many "render jobs" as I need wedges?

[kleer001]

I'm super cheap & easy. I use a little hscript slimyness when I want multi-seed sims. It's a *great* technique.

 

I put the following chunk in the Pre-Render Script:

 

opparm /obj/AutoDopNetwork/flipfluidobject1/ jitterseed `opdigits(".")`

 

The Jitter seed points to the flipfluid object and the opdigits depends on the ROP having a number. Setting the former to the latter. That's pretty clear, right?

[kleer001]

Ahahahaha, just searched the archives here for the answer to wedging out renders... and low and behold I answered it myself last years. Thanks Past-me. And thanks BorisB2 for asking!

 

('O_O)