Chaos Corona Forum
Chaos Corona for 3ds Max => [Max] Bug Reporting => Topic started by: wilbertvandenbroek on 2018-07-19, 15:25:54
-
Simple setup with a few boxes and polygons. Corona camera is animated to a curve via path constraint.
When I open the vfb without rendering and scrubbing the timeline max crashes. Also when interactive rendering.
No textures, plain CoronaMtl. Corona sun and sky for lighting.
See attached scene.
-
Works fine here, both VFB open no IR, and IR - note that there is some sort of "pre render script" that the scene wants to run, but that I don't have, so not sure if that is affecting things on your side.
Max 2019, Corona 2 (would be good to know the versions you were using, and if you are getting a crash, the mini dump is always always always very useful - thanks!)
-
Ok thanks for testing, I'm using Max 2017 with the latest sp.
The minidump you are refering to how can I find it?
The pre-render script should not have any influence on the IR rendering.
-
Here's a guide how to collect minidumps: https://coronarenderer.freshdesk.com/support/solutions/articles/5000524006-how-to-report-issues-3ds-max#crashes1
-
I'll look in to it in a moment. I also tried a different machine and user, empty scene with a teapot and coronaMTL, start interactive rendering in CoronaCam, the moment you scrub the timeline it crashes.
-
I was waiting for Max to crash after it froze up again to gain acces to the minidump but after waiting for about 10 minutes the scene was suddenly working just fine. I can now scrub as much as I like.
However still the first initial scrub action will freeze the application.
-
However still the first initial scrub action will freeze the application.
A small "lag" is to expect when changing frame the first time. But not ongoing freezes. How long does it last? And how does memory consumption of the max process behaves during the freeze?
Do you have any material libraries linked in the SME when it happens? I'm asking because recently I discovered a nasty behaviour, it goes like this:
- Have the (relinked) Corona material library in the "Material Map Browser" of SME
- Change frame
Result: Max seems to load the complete library for some reason, all bitmaps. This ends with a teapot scene eating 20GB of ram. Do this with a heavy scene (already at ram limit) and max crahes. Still I have no idea what is causing this, my stomach has just only a sneaking suspicion.
Good Luck
-
Hey Frood, you're absolutly right! Just checked the simple teapot scene. With the slate editor closed there is just a small 'freeze' after which I can scrub the timeline just fine.
I opened the slate editer in which we have a corona mat lib with custom filepaths and the whole scene just froze for like 5 minutes. The ram being used for this simple scene went op to 10gb.
I removed the custom lib from the slate editor which solved the problem for now. I will try a few heavier scenes today.
-
It's getting weird, adding some more information (Max2016 SP4 | Corona 2.0):
- Happens also sometimes when loading another scene from recent file list (and one has the impression of a scene loading error)
- Converted all to Max bitmap in library -> did not happen
- Converted all back to CoronaBitmap in library -> same effect again
- It's enough to have the library in the library list - no material editor needs to be even open
- As soon as you remove the library, ram is released
- Viewport lags come along with this (at least on my box)
- Happens also if all material icon previews are already rendered and saved in the lib
It seems that CoronaBitmap(s) in the library somehow trigger Max to load all maps(?). When I have some time I will convert another lib to CoronaBitmaps and check if it behaves equally. I hope this does not happen with any library making heavy use of CoronaBitmap (if it's the cause at all).
Good Luck
-
I always keep mat/map browser closed, so i never even knew about this issue. It's horrible! For me it happens no matter if there's max bitmaps or Corona ones. Why 3ds max has to even render library thumbnails?! What's the logic behind this? Anyway, turning rendering off in Corona settings or in SME itself, solves the issue, as well as keeping matlib closed or mat/map browser hidden (preferred method for me). But what a mess from autodesk side, yet again :/
-
Hmm, here it only happened with CoronaBitmaps (?)
Thumbnails: If you store the lib after rendering them, they are just there when opening the liib, so why not. And usually Max loads the bitmaps for that one material, renders the preview and releases the bitmap (+memory) again. The issue seems completely independent from thumbnail creation, I created and saved them explicitly and also had renderqueue off while reproducing.
Good Luck
-
Just tested v4 daily 19-01-10 (because of new "Enable CoronaBitmap auto update" checkbox) and it is still the same. I get 24GB ram usage rendering a teapot :)
As mentioned, it does not happen when all maps in the library are converted to standard max bitmaps.
Good Luck
-
Guys, I tried reproducing the two issues from this thread:
-Crash when scrubbing the timeline in the user reported scene while IR is running or while VFB is opened
-Freeze when having a custom material library loaded in tho the Material/Map Browser of Slate Material Editor and scrubbing the timeline (with IR running, VFB opened)
...but I was not able to reproduce either of them.
Can you please provide simple steps how to reproduce those things? Maybe a video recording if I'm not asking for too much? :)
-
I just uploaded a screen capture of this issue.
-
Setup is described here:
https://forum.corona-renderer.com/index.php?topic=21189.msg130891#msg130891
You "need" a material library with corona bitmaps inside, where assets point to some network location. In this case I used the Corona (2.0) library and relinked every asset to a central network source. See attachment "Corona_Library_2-0_Relinked.jpg" and note the bitmap path.
And it's not about a freeze but the fact that all assets seem to be loaded by max - even if you use a teapot without any material applied if you change frame.
Images:
1:
Basic scene, everything is ok. Just a teapot. See memory, all ok.
2:
The library with corona bitmaps inside is loaded into SME, no material from it is used in the scene (actually there is no material in the scene :). Still all ok (IR is running).
3:
Now I changed frame in timeline and you see how ram usage starts to increase.
4:
All bitmaps loaded now. See memory line in task manager.
5:
Closing the material library...
6:
... releases memory. You can still see the peak working set in task manager.
Hope it helps?
Good Luck
-
Setup is described here:
https://forum.corona-renderer.com/index.php?topic=21189.msg130891#msg130891
You "need" a material library with corona bitmaps inside, where assets point to some network location. In this case I used the Corona (2.0) library and relinked every asset to a central network source. See attachment "Corona_Library_2-0_Relinked.jpg" and note the bitmap path.
And it's not about a freeze but the fact that all assets seem to be loaded by max - even if you use a teapot without any material applied if you change frame.
Images:
1:
Basic scene, everything is ok. Just a teapot. See memory, all ok.
2:
The library with corona bitmaps inside is loaded into SME, no material from it is used in the scene (actually there is no material in the scene :). Still all ok (IR is running).
3:
Now I changed frame in timeline and you see how ram usage starts to increase.
4:
All bitmaps loaded now. See memory line in task manager.
5:
Closing the material library...
6:
... releases memory. You can still see the peak working set in task manager.
Hope it helps?
Good Luck
Resurrecting this to look into it, thanks
-
Reported - Second issue regarding Custom Libraries and memory.
(internal id=507977242)
-
Setup is described here:
https://forum.corona-renderer.com/index.php?topic=21189.msg130891#msg130891
You "need" a material library with corona bitmaps inside, where assets point to some network location. In this case I used the Corona (2.0) library and relinked every asset to a central network source. See attachment "Corona_Library_2-0_Relinked.jpg" and note the bitmap path.
And it's not about a freeze but the fact that all assets seem to be loaded by max - even if you use a teapot without any material applied if you change frame.
Images:
1:
Basic scene, everything is ok. Just a teapot. See memory, all ok.
2:
The library with corona bitmaps inside is loaded into SME, no material from it is used in the scene (actually there is no material in the scene :). Still all ok (IR is running).
3:
Now I changed frame in timeline and you see how ram usage starts to increase.
4:
All bitmaps loaded now. See memory line in task manager.
5:
Closing the material library...
6:
... releases memory. You can still see the peak working set in task manager.
Hope it helps?
Good Luck
In regards to the increased memory consumption for coronabitmaps when being used in a custom corona library, it's something that would require a large number of changes on how Corona renderer manages textures. This however will be addressed and partially improved by some optimizations that we have planned in the future.
For the moment, some simple solutions to the issue are to use a native bitmap library and simply use corona converter when all prefered materials are deployed within the SME. Or close temporarily any custom libraries that are not being actively used and open per request.
-
Update:
Here is what I get currently (Corona v9/v10):
Max 2023: no change, all assets are pulled from the network (-> freeze), memory usage increases to 20GB+ (empty scene)
Max 2024: all assets are pulled from the network (-> freeze), but memory usage does not increase any more
Btw: OutOfCoreCache entries (if not disabled) are generated for all assets of the material library additionally.
Good Luck