Rubikon : Valve's physics engine for Source 2

[RBD]

Could someone (Dirk? ) please point us to more detailed information about Rubikon, Valve's physics engine for Source 2?
Thanks.

[Dirk Gregorius]

Rubikon is our custom rigid body engine we use in Source 2 now. There aren't really any secrets there. What are you questions?

[H3g3m0n]

Rubikon is our custom rigid body engine we use in Source 2 now. There aren't really any secrets there. What are you questions?

I would be interested to know if it's has (or is getting) Vulkan support, or any other GPGPU stuff.

[Dirk Gregorius]

No, everything is running on the CPU. Physics is a big part of gameplay, so I don't see any physics simulation running on the GPU in the near future.

[RBD]

The engine has been referred to for a while, but I haven't yet come across much information, and since Source 2 is now official, I was thinking maybe there would be some doc on Rubikon somewhere. I thought I'm not looking in the right places or perhaps just need to be more patient.
Sorry to bother, just curious in general. Possibly just some high level info then, if there is a URL with an overview of the engine components, or simply Rn* rigid and soft body shapes and joint types, etc. (but no need to list here)

I wasn't privy to your GDC contact solving talk last week, but appears some info might have been provided (e.g. no box primitive, use RnHull).

I understood that Source 2's physics would still be optionally augmented with PhysX (now apparently free to all with source code BTW) for extra SFX, or is that title specific?

Thank you.

[Dirk Gregorius]

I gave three talks over the last three years at GDC which describe essentially the Rubikon collision pipeline. I am thinking to give a talk on mesh and continuous contacts next year and then close with a talk about the solver following this. The solver is essentially SI with post-projection, but there are some nice implementation details I would like to share (e.g. central friction, quaternion constraints, etc). Sergiy gave a talk about physics debugging at GDC in 2014 which also shares a bunch of information about Rubikon.

Personally I think we miss actually more talks about physics authoring. The technical information is all out there. E.g. we have a complete Maya integration which we used to bake a bunch of the destruction shots in the VR demo. These large scale shots were a great stress test for the engine.

Some classes in Source 2 have PhysX in their name which I agree might be misleading, but Source 2 doesn't use any PhysX.

Here is the link to the 2014 and 2015 talks (both Sergiy and me):
http://box2d.org/downloads/

You can download my 2013 talk from Google here:
https://code.google.com/p/box2d/downloads/list

Cheers,
-Dirk

[RBD]

Excellent, thank you for the link to the GDC talk PDFs, that's perfect.

Yes, I mistakenly made assumptions from all of those "PhysX" names, thanks for clarifying.

[Yavga]

Bumping an old topic but the question seems relevant.

I cannot really find anything on "Rubikon's fluid simulation" in Source 2.

I was wondering if, to any extend, the second Source engine is able to handle realistic fluid simulation.

[Dirk Gregorius]

We have basic buoyancy and some tricks, but not a full fluid solver with rigid body coupling. I am not seeing any uses cases for this at the moment.