IPR crashes, but PV renders fine

[Eldar_Tokhtamyshev]

Hello,

I've just upgraded to 3.2 and experiencing a rendering issue with a file that previously rendered fine. I tried to figure it out and Google it myself, but couldn't narrow it down. I'd much appreciate if anyone can help me out or direct the right way.

PV renders fine. IPR crashes the whole application in a second or two. Even on low resolution and lowest render settings. I ran some tests and it seems to me that the complexity of the scene is the issue. There are four lights and a sky object. And IPR doesn't crash on lowest settings if one or two lights are off.

The geometry is pretty bulky, too, geometry.jpg — one of the older renders. And the bug report _BugReport.txt.

From what I could gather, I would think that the scene is just too complex for my computer, but it rendered fine on previous version and it still renders fine in PV, even on high settings.

Mac 10.15.7, 16Gb, R22.

Thank you.

[lee_griggs]

What does the log file say?

[peter.horvath6V6K3]

Since it renders fine in previous versions, it sound like a bug rather then an issue with the scene or hardware configuration. Would it be possible to share the scene with us? Maybe simplify it (e.g. less dense geo, etc.)?

[Eldar_Tokhtamyshev]

@Peter Horvath, sure, here's the simplified version: (http://eldart.com/x/img/Sphere_IPR_Crashing_SIMPLIFIED.c4d). Minimized the geo, stripped all the non-basic textures. Still causes the crash. Thank you for looking into this, Peter.

I also got an email notification of another comment (regarding log file), but I can't see it here now. Went through the log file and I see several warnings. The first few don't look crucial:

00:00:02 1981MB WARNING | [polymesh] /1_1/Stand/Legs/Leg_1: mesh has 1 free-floating vertex

00:00:03 2142MB WARNING | Rendering CPU utilization was only 24%. Your render may be bound by a single threaded process or I/O.

00:00:06 2156MB WARNING | Scene creation was a significant amount of total time (26%). Consider optimizing this process first.

But this last line in the file looks to me like the culprit:

00:00:07 1804MB ERROR | signal caught: SIGSEGV -- Invalid memory reference

[peter.horvath6V6K3 · Accepted solution]

Thanks for the scene. I could reproduce the issue with it and found the problem. It's caused by one of the enhancements introduced in Arnold 6.1 / C4DtoA 3.2.0. Until we fix it, you can disable the enhancement in Render Setting -> Diagnostics -> User options, by typing enable_fast_ipr off in the field.

[Eldar_Tokhtamyshev]

Solved!

Peter, you are a gentleman and a scholar. Thank you very much!

[thiago.ize]

The just released Arnold 6.1.0.1 (C4DtoA 3.2.1) should have a fix for this so that you can re-enable fast ipr. This was fixed by:
core#10140 Crash with quad_light and IPR