PhysX VS PhysX

I know you mean PhysX how it was WAYYY back (& GPU accel), but it's been open source for ages.

Unity’s default physics engine is literally PhysX (https://github.com/NVIDIAGameWorks/PhysX).

It all gets petty hairy.. which is why writing physics engines is Hard, and you should be pretty skeptical when someone is like: "Hay gauis I made a new phYsics engine!! you should totally use it!" Even the big guys thrash around in this space.. as someone mentioned in another comment... although there is sometimes a case to be made for rolling your own, if you're valve for instance.. All that said, I'm really partial to Bullet just because it's so damn battle tested (unity/blender and countless other engines) But I'm generally pro havok/physX and also the newer Jolt physics library. https://github.com/NVIDIAGameWorks/PhysX https://github.com/ashconnell/physx-js https://jrouwe.github.io/JoltPhysics/

Then what about this? https://github.com/NVIDIAGameWorks/PhysX/tree/4.1/physx

PhysX has gone open source quite a while ago as well: https://github.com/NVIDIAGameWorks/PhysX https://github.com/NVIDIAGameWorks/UnityPhysXPlugin

The pull request can be found here: https://github.com/NVIDIAGameWorks/PhysX/pull/577

For typical game physics engines... not that much. Math libraries like Eigen or Blaze use lots of template metaprogramming techniques under the hood that can help when you're doing large batched matrix multiplications (since it can remove temporary allocations at compile-time and can also fuse operations efficiently, as well as applying various SIMD optimizations), but it doesn't really help when you need lots of small operations (with mat3 / mat4 / vec3 / quat / etc.). Typical game physics engines tend to use iterative algorithms for their solvers (Gauss-Seidel, PBD, etc...) instead of batched "matrix"-oriented ones, so you'll get less benefits out of Eigen / Blaze compared to what you typically see in deep learning / scientific computing workloads.

The codebases I've seen in many game physics engines seem to all roll their own math libraries for these stuff, or even just use SIMD (SSE / AVX) intrinsics directly. Examples: PhysX (https://github.com/NVIDIA-Omniverse/PhysX), Box2D (https://github.com/erincatto/box2d), Bullet (https://github.com/bulletphysics/bullet3)...

NVIDIA PhysX GitHub repo: NVIDIA-Omniverse/PhysX

I don’t know of any books but here is the official documentation which includes an API reference. You can also find a lot of sample code snippets and examples on their GitHub.

On the other hand, PhysX is available under the MIT License and is both heavily used in games and also well documented with tons of sample code.

The CPU implementation of PhysX is too, no reason you couldn't port that to run on AMD GPUs: https://github.com/NVIDIA-Omniverse/PhysX

The LICENSE.md file on the repo doesn't mention BSD-3-Clause at all, it's just a copywrite notice that reads more like CC-BY than a software license. That's why I said it was weird.

Very! I just picked a random file: https://github.com/NVIDIA-Omniverse/PhysX/blob/release/104.0/physx/source/physx/src/NpArticulationJointReducedCoordinate.cpp