[stir]

Hey i got a issue in my latest scene.
I have tried debugging it for some time now, but seems like i need to call for help.

Here is a link to the 3ds max file if anyone is intrested in checking it out.

http://we.tl/TAXbmCa4Jn

EDIT:
Here is a MXS file of the scene.

http://we.tl/s6VDHbqqu4


The problem seems to occure around hairfarm and mxLights in the 3ds max scene.

I have cleaned up this scene to only contain the core items.

1 - Head (without displacement)
1 - Hairfarm mesh (Eye brows)
2 - MxLights

The scene is originaly lighted with a HDRI but i don't have the rights to share that. So in the uploaded file its switched to Physical sky, this does not affect the freezing at SL 1 at 99%


The problem is, when i'm hitting render with all lights and all objects visible and on. Max will send the scene to maxwell render and start rendering. the render will start voxelating without any problems and then start rendering. When the rendering starts you can see SL:1 and the rendering bar go from 0-> 23->45->60-> and end up at 99% there it will stop. and anyway how long i wait it will stay there.

There is 3 solutions to the problem.

1 - Hide the hairfarm mesh (now it renders perfectly fine, obviously a bad solution, because i want the hair)

2 - Lower the amount of hairs generated in the hairfarm mesh to something redicilusly low ( now it renders fine, but again, a bad asolution as now the hair looks shit.)

3 - Hide/turn of the mxlights (This is where things gets strange, turning of or hiding the mxlights seems to fix the freezing, only rendering with the physicalsky/HDRI)


So this is what i got so far. There is something between the lights placed in the scene and the hairfarm mesh.


SPEC:

3ds max 2015 SP3
Maxwell plugin 3.1.8
Maxwell render 3.1.1.0
Hairfarm 2.4.5 (build 168)


Best regards.
-Mats

[Mihnea Balta]

I think this is a bug in the ray navigation code inside the engine. The rays get stuck in some part of the hair geometry so the render never makes any progress. The local lights can trigger such problems, since they are sampled differently compared to the environment. I've notified the engine guys about the problem, they'll take a look at your scene.

[Mihnea Balta]

One more thing: I don't have HairFarm installed on the machine I'm currently using. Could you please export a MXS file for your scene with pack'n'go and share that too? That will make it easier for the engine guys to debug the problem.

[stir]

Thanks for the quick reply.
I have edited the first post to contain a link to the MXS file.

If you think it has something to do with ray's the material might be affecting it to?
The material used on the hair is maxwell assistant Preset "HAIR" with a brownish color.





Cheers
-Mats

[stir]

Hope you guys keep us posted if you stumble over a solution, or if you dont get the same problem at all, meaning something is wrong on our end.

Still some time until we are at the render stage of this project. But the earlier the better, so we can planout what to do if we cant use Maxwell

Thanks for helping
-Mats

[Mihnea Balta]

We were able to reproduce the problem, but I haven't heard anything yet from the guy in charge or fixing it. I'll keep you posted when we figure it out.

[Mihnea Balta]

The problem will be fixed in the next Maxwell (engine) build. I don't have a release date for it yet, but it's probably a couple of weeks or so from now.

[stir]

Amazing! That's some great work Mihnea, thank you!
2-3 weeks will be fine, will just fake the still images that we send to the customer, and tell em will get the final look on the final animation renders

Out of curiosity, what seemed to be the problem?

Cheers
-Mats

[Mihnea Balta]

The bug was in the hair extension, during a preprocessing step which cleans up invalid hairs (e.g. zero length, coincident points etc.). In some cases it was not using the root/tip width set by the user, so instead of cleaning up the hairs, it broke the data.

[T0M0]

I'm very sorry for hijacking this thread, but please consider my little request - Grass/Hair in Globals :)
It would be really handy when tuning scene, so in cases with a lot of objects with hair/grass modifier applied, wouldn't be needed to turn on/off it for each object. Thank you.

[Mihnea Balta]

That's sort of above my pay grade at the moment, but if I were to offer an opinion, I'd say it's a bit hacky, since hair is just another extension, like the particle renderer, mesher, instancer, Alembic importer and any other extension which will be added in the future by Next Limit or a third party. Why would hair get a special checkbox? Also, if a 3rd party (say, Yeti) adds a proprietary hair extension in the future, that checkbox will not apply to it.

A less hacky approach would be to add a mechanism for disabling extensions by name. We may be able to do that, but it requires a bit of planning, because for some types of extensions it's not just a matter of skipping them when the renderer is initialized. I'm directly involved in overhauling the extensions system and I'll take this feature into account when the time comes, but this is a mid-term project so it will be a while until it will make it in a beta build.

[stir]

sorry to bug you again Mihnea Balta.
You got any headsup, on when we can expect to see the new engine version?

cheers
-Mats

[stir]

hey again.
Whats the chances of seeing the update sometime this week?


We also figured the SSS is a part of the problem. Because when we check override material and just pick a default grey maxwell material it renders all the hair fine, and fast.

Its when we add the SSS to the character it stalls. That can then relate to my first post, mentioning turning on and off the different lights. Because i assume SSS is light sensitive.

So it seems it a combination of SSS character, Direct lights and a high amount of hair that stalls it. But this might all boil down to the same problem as you mentioned " ray navigation code"

i thought i just add the information just in case.

-Mats

[Mihnea Balta]

The bug is fixed in the 3.2 branch, but I don't think it will be backported to 3.1 (and even if it was, I don't think it's likely to see a 3.1 patch release before the 3.2 release). Unfortunately I can't give you an ETA for the 3.2 beta, it's outside of my hands anyway. All I know if that everyone is working towards it, so it should be out soon, but I don't know how soon.

[Mihnea Balta]

Well, it turns out the release date for the Maxwell 3.2 beta was much sooner than I thought - it was just uploaded to the early builds site. I'll upload a compatible Max plug-in tomorrow.