Page MenuHome

Mantaflow Crash | Whitewater Particles Baking
Closed, ResolvedPublic

Description

System Information
Operating system: Windows-10-10.0.18362-SP0 64 Bits
Graphics card: GeForce RTX 2080/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 442.92

Blender Version
Broken: version: 2.90.0 Alpha, branch: master, commit date: 2020-06-22 22:06, hash: rB51bf04383aa0
Worked: Crashes also in blender 2.82 and 2.83

Short description of error
Blender crashes during mantaflow particles baking in liquid simulation.

Exact steps for others to reproduce the error

  1. Bake liquid simulation
  2. Try to bake whitewater partiles -> leads to crash

[Based on the attached .blend file (as simple as possible)]

crash error
Error : EXCEPTION_ACCESS_VIOLATION
Address : 0x00007FF627DCE8AE
Module : blender.exe
Thread : 0000a604

Blender file:

Event Timeline

Hi,
I will later try it on another computer but on my main machine it crashes every time (video in the link below)
https://youtu.be/H-xcRka2GHM
I've also tried to clean all blender data cuz I thought maybe it was some addon compatibility issue but it also did not work.

Here is the crash txt file generated by blender:

Richard Antalik (ISS) changed the task status from Needs Triage to Needs Information from User.Thu, Jun 25, 1:08 PM

Also got no crash.

Does this happen if you click on File > Defaults > Load Factory Settings, open file and rebake whole simulation?

Yes, with factory settings it also crashes.
I have also tried it on my laptop with the newest blender 2.90.0 and it's the same thing :(

I will try maybe to bake it on my fathers PC with fewer CPU cores, maybe thats the problem

Thank you for your time

Yes, with factory settings it also crashes.
I have also tried it on my laptop with the newest blender 2.90.0 and it's the same thing :(

I will try maybe to bake it on my fathers PC with fewer CPU cores, maybe thats the problem

You can force Blender to use reduced number of threads by running from console with -t argument. See https://docs.blender.org/manual/en/latest/advanced/command_line/arguments.html
I don't really think this is the issue though

Richard Antalik (ISS) changed the task status from Needs Information from User to Needs Developer to Reproduce.Thu, Jun 25, 1:40 PM

OK, I've tested it on next 2 computers with fresh Windows and Blender installation (there were no blender installed on them before).

  • On the computer with intel core i5 3rd gen with 4 core 4 threads liquid simulation goes fine, whitewater particles baking crashes immidietaly after launching with the same error.
  • On second computer with Ryzen 3990X 64c/128th also liquid simulation goes fine, whitewater particles crashes on about frame 56 (still same error message).

Reducing threads on my main machine did not help also as you said.
All computers are running latest blender 2.90.0 version and latest Windows

I'll try maybe with the linux machine but I don't know if it would help

On Ubuntu 19.10 it works without a problem, so it crashes only on Windows machines

Ok, there was in fact an issue that would trigger a crash. Should be fixed in 84e1e0cf95da.

Also note that since the cache recently changed a bit (995185894289) there is a small caveat: When baking with the "Modular" cache, it has to be "Resumable" in order to be able to bake the particles.
This usability issue is similar to the one from T78272. I am working on a solution for that.

It works fine now :)

Thank you for your work

Hi guys, I guess the above mentioned fix is not in 2.83.1-LTS , right ?

I have a similiar problem : Blender crashes on opening (!) a file on a RTX-2070 machine , which was copied from another one with GTX 1080 on which I created and succesfully rendered a mantaflow simulation

No, the fix is not in 2.83.1. I'll make sure it gets ported in to the next LTS release.
If you think your crash is not related to this issue, feel free to open a new report - I can then take a closer look.

I'm not sure about it, I created another simulation from scratch on the RTX machine, it didn't crash but I got a strange behaviour after baking: As I rendered the fluid was not visible.
But to open another bug report I should build a simpler scene than mine.

By the way, I just made a "make update" on 2.9 alpha this evening and tried to load the file there, the same issue.
Is the fix in this branch ? (I'm not a developer, just build blender myself, thus maybe this is a silly question)

Yes, with "make update" it sounds like you should have the version with the fix. If you can reproduce the scene in a simple file and submit it in a report, that would be great!

Regarding the simulation buiIt from scratch on the RTX machine, it was my fault: while playing around with settings, I accidently disabled rendering of the Domain object. Rendering surprinsingly worked as expected after enabling it again :-)

Furthermore as I loaded factory settings on the 2.9-alpha build from yesterday, the file which worked on the RTX machine, did'nt crash on the RTX machine anymore and rendering works fine as well.