This might just be testament to the size of our material networks, but we are finding that Solaris will hang for upwards of a minute if the user jumps inside a Vray Material Builder - even when the active Hydra delegate is HoudiniGL and the cook mode is set to Manual. Could this be considered normal? Ordinarily being in Manual mode negates almost all lag when diving into or out of networks.

We are using 19.0.657 Python 2, on CentOS 7.

That seems suspicious. Usually the delay is due to translating VOPs to USD. But, indeed, the manual mode should avoid that.

Yeah, usually with the information bar saying "evaluating python..."

So, this is very consistent across the board. We are using Vray (I'm not sure if that matters), but the delay/cook happens even when simply selecting a material builder in Manual mode. We have a number of Solaris shows in-house and this is high on the list of complaints from artists (along with the amount of cook time that material builders create in general).

We're not using Karma, but can anyone who is confirm whether this happens to them when the material contains a couple hundred nodes?

We had a (maybe?) similar problem, that when sometimes just entering a material library lop with some materialx and karma nodes inside, Houdini would crash and also just before crashing one could see “evaluating python” in the info bar.

It didn’t happen on every workstation though and was with houdini 19.0 on Windows.

If I remember correctly the problem stopped when we removed an external python installation from the windows user directory.

Cheers
Michael

I encountered something similar. Using a sop import to bring in many particles with packed geo on them. By default it would import them as native instances which meant 8 million descendants in the scene graph. This caused it to hang for each action. But setting the update mode to Manual didn’t help. Each action still froze Houdini for a minute as it was single threadedly working it’s way through the list.

Is there a way to force materials/the usd stage not to update? Material Libraries are hanging forever when in manual mode and diving into and out of material builders or even just selecting them will cause the 'evaluating python' cook to drag on for 20-30 seconds.

I've tried closing the scene viewer so hydra isn't active, I've tried pausing the viewer updates, I've tried loading other desktop layouts, disabling the nodes and turning off the orange flag... nothing stops the endless cooking/python evaluation when trying to author materials in Manual update mode.

Bypassing the Material Library should disable its cooking and therefore the translation.

Just to update this thread - disabling the material library itself doesn't solve the issue. I am thinking this may be a VRay issue but haven't had the time to re-build our materials in Karma to test.

For anyone using Vray, we've worked out that this is happening due to the Parameters pane. Delete any parameter panes/tabs/windows and the lag in Manual mode does not occur. Will be submitting a bug to Chaos and also SideFX for reference.

Submitted a bug to SideFX: #128540

Was there ever any update to this? Or somewhere we can see the status of the above bug?
I'm at a new studio using 19.5.805 and this is still an issue (not just material nodes, and otls as well).

You can contact support@sidefx and ask to be added to that issue ticket to see the status.