Deprecation Notice: warp.sim Module

[momo-van]

As we introduce Newton, our next-generation physics engine built on Warp, we will begin deprecating the warp.sim module.

Deprecation timeline
We follow a two-release deprecation policy (approximately 4 months) to give users time to transition smoothly:

• July 1, 2025 – warp.sim marked deprecated in Warp 1.8. Importing it will raise a DeprecationWarning.
  • Newton will be available via PyPI and other channels in early June
  • A migration guide will be included in the Warp 1.8 docs
• September 1, 2025 – Warp 1.9 release
• October 1, 2025 – Warp 1.9.1 patch release, warp.sim removed from the main branch
• November 3, 2025 – Warp 1.10 release:
  • warp.sim is fully removed
  • Importing it will result in a ModuleNotFoundError

If you're using warp.sim, we recommend starting your migration to Newton as soon as it is available.

Looking ahead
Warp continues to evolve as the high-performance Python framework powering Newton and other applications across simulation, geometry processing, and AI. As Newton grows, so does the foundation beneath it.

We’d love to hear from you
As we navigate this transition, please use this thread, open a new discussion, or file an issue here on GitHub to share your thoughts and feedback.

Thanks for building with Warp!

The Warp Team

[shi-eric]

FYI @CliffieVanR @JonathanKuelz @jc211 @etaoxing @maria-korosteleva @Hantao-Ye

[CliffieVanR]

Thanks for heads-up Eric!

[maria-korosteleva]

Thank you for the info!

[OmidGhotbi]

Thank you for the heads up.

[cadop]

Thanks for posting this openly!

1. Is this removing only the warp.sim module, or are other parts also being removed?
2. Is FEM staying?
3. Is the optimization part staying/continue to develop?
4. Are any of the components going to move to other areas? (or would you consider more examples)?

For example, one of my favorite aspects of this warp library has been in the small code snippets that I can use for my own application.
I don't use the sim module, but things like small functions in particles, I will pull out or reference.
This is just 20 lines, but it makes developing much easier.

warp/warp/sim/particles.py

Line 22 in e8e7733

def particle_force(n: wp.vec3, v: wp.vec3, c: float, k_n: float, k_d: float, k_f: float, k_mu: float):

• It would be nice to expand the examples, so these types of reference functions could be maintained. E.g. I don't need a sim module with an integrator, but it would still be nice for an example to have a 20 line kernel doing Euler integration, with ~100 lines defining points on a soft-body.

Warp continues to evolve as the high-performance Python framework powering Newton and other applications across simulation, geometry processing,

What will warp evolve into? If it is to power other applications that do simulation and geometry, does that mean it will, for example, not have raycasting, or mesh point query in the future?

---

I guess it depends on how it will develop or be deprecated, but if it is going more back to early stages of warp, where there wasn't as much focus on building a full simulator, and more about 'helper' functions for geometry processing, spatial queries, kernel translations, etc. that would be nice.

[shi-eric]

Here are my answers to your questions:

1. Yes, only the warp.sim module is being moved.
2. Yes, warp.fem is staying.
3. Optimization is staying. Yes, it's in need of some refactoring and improvements as it hasn't received much attention, can't say how this ranks against the other priorities.
4. No plans for moving the other Warp modules.

What will warp evolve into? If it is to power other applications that do simulation and geometry, does that mean it will, for example, not have raycasting, or mesh point query in the future?

Warp's position as a framework for developers/researchers to build applications on top of does not imply that mesh queries will be removed. They're important building blocks whose functionality needs to be maintained at the CUDA C level, e.g. something Warp should provide.

[cadop]

Thanks Eric, for q4, I am wondering if any parts of sim will get moved into other modules , or examples. Or is this a cut and dry, warp.sim folder gets deleted, nothing else changes (at least got this specific topic)

[shi-eric]

It's more like the latter, but the plan for examples that rely on warp.sim is to move them into Newton rather than creating a warp.sim-less version of them.