Feedback Obi 8: what's coming up

[josemendez]

Hey @josemendez
Any updates on the development of Obi 8? Really excited to try making some custom constraints 

It's around 80% done. I'm currently porting volume constraints in the compute backend to use the new constraints API, after that compute shape matching and collision constraints remain to be ported, then overall testing/polish and writing documentation. Should be out near end of May/ start of June.

[Qriva0]

Hi! Is Obi8 just around the corner?

[josemendez]

Hi! Is Obi8 just around the corner?

Hi!

It's taking a bit longer than anticipated, new estimation is mid-august. Looking trough the code to allow for custom constraints has exposed quite a lot of things that could be improved, and I'm tackling most of them. Want to get it right on the first try, not have it be a downgrade from Obi 7 in any aspect.

Biggest one is I've rewritten the entirety of the collision detection system (both particle vs collider and particle vs particles) to:

- get rid of any contention points (we had a single atomic counter to append contacts against colliders, for instance, now there's no atomics at all in CPU and shared memory atomics in GPU)

- include a midphase that prunes out individual contacts using swept-sphere tests, this results in fewer speculative contacts.

- take advantage of incremental parallel graph coloring. As a consequence graph coloring is almost 5 times faster, both at runtime for particle contacts and during blueprint generation for all other constraints.

If you're interested I can give you beta access as soon as we get there. This way you can use it asap and we can get your feedback into the engine before it's publicly released.

kind regards,

[Qriva0]

Hi!

It's taking a bit longer than anticipated, new estimation is mid-august. Looking trough the code to allow for custom constraints has exposed quite a lot of things that could be improved, and I'm tackling most of them. Want to get it right on the first try, not have it be a downgrade from Obi 7 in any aspect.

Totally understand, it's huge refactor, I suspected this is the main reason for a delay.

- include a midphase that prunes out individual contacts using swept-sphere tests, this results in fewer speculative contacts.

This is good, I noticed sometimes this part produced a lot of overhead for example when cloth surface was near any other object thing, or their own particles, it had no collisions, but had to check a lot of them, so any improvement here is welcome.

If you're interested I can give you beta access as soon as we get there. This way you can use it asap and we can get your feedback into the engine before it's publicly released.

Actually I am interested in . I plan larger refactor when new version is available, so it would allow me to get familiar with the new code faster and I could test against potential bugs or other problems.

Thanks for the update and have a great day!