Page 2 of 11

Re: new Physx

PostPosted: Thursday, 26.June 2014, 19:41
by CybeREX
Thx guys! I use web interface to download zip file on github. Don't use any client or software for it =)

Re: new Physx

PostPosted: Thursday, 26.June 2014, 20:15
by ikam
thanks. see my notes.

Skinnytorus wrote:Or even better: allow RigidBoby to have just one Collider input if a primitive collider is output (box, sphere, plane)


yes good idea, and easy, but need to inverse the operator inputs order because optional cannot be first, ok for that.

Re: new Physx

PostPosted: Thursday, 26.June 2014, 21:06
by Skinnytorus
1)
yes good idea, and easy, but need to inverse the operator inputs order because optional cannot be first, ok for that.

Great, but I think I have a more radical proposal. See below.

2) I think I know what's wrong about Physx. Collider operator seems to be redundant. IMHO it might be cooler if it was combined with RigidBody operator thereby facilitating visual perception of the entire physx stack. What do you think? Too radical? ;)

3) Bug? I noticed that physx stack doesn't work for some time if built from scratch (the first showing of the PhysX operator causes no simulation - only some time after playing with params or even after reloading the saved file. Can't pin that down yet...

Re: new Physx

PostPosted: Thursday, 26.June 2014, 21:20
by ikam
2) you mean, no collider operators, but some colliding parameters inside rigidbody ? but it will limit collider to only one collider, interest of building stack colliders is to generate precise colliders around a complex mesh, for a simple mesh like cube or sphere it's ok, but imagine a really complex mesh, goal is to build bounding shapes with basic colliders geometry to match the overall shape of the complex mesh.

3) no simulation ? or crash or freeze ?

[edit] other idea, it is possible to create another rigidbody operator (simple collider body) that will take only a mesh on input and generate only one collider, a bit like the physx debris operator but without needed a chunked mesh and with a choice for the collider geometry instead of the hull by default for shards in debris.

Re: new Physx

PostPosted: Thursday, 26.June 2014, 21:29
by Skinnytorus
2) I think I see what you mean: more flexibility in building colliders. Fair enough - since mesh and mesh hull collision modes have limitations, it makes sense. So, OK... That was just an idea. ;)

3) No simulation. Dynamic rigid body won't move. No crash or freeze.

Re: new Physx

PostPosted: Friday, 27.June 2014, 05:35
by CybeREX
Maybe allow "E" xchange operators like transform and multiply from green to physX with some hotkey?
Like now we can "E" on green transform or multiply to yellow operators.

Re: new Physx

PostPosted: Friday, 27.June 2014, 06:40
by ikam
yes very good idea.
you can now exchange with shift+e (exchange2) colliders <-> scene (yellow, because you can put other things that mesh, but you can also use 'e' and 'shift+e' to convert from mesh to scene, and scene to collider...)
thanks ;)

note : you need a full rebuild to get wire shortcut works correctly. (and colliders operators have changed, so create new ones)

Re: new Physx

PostPosted: Friday, 27.June 2014, 08:36
by Skinnytorus
[edit] other idea, it is possible to create another rigidbody operator (simple collider body) that will take only a mesh on input and generate only one collider, a bit like the physx debris operator but without needed a chunked mesh and with a choice for the collider geometry instead of the hull by default for shards in debris.

Yep. A very cool idea! That would simplify stack thus making things much clearer. I would be happy if you could add such SimpleCollider (SimpleRB) operator. :D
Edit: SimpleRB seems a more adequate name.

Re: new Physx

PostPosted: Friday, 27.June 2014, 09:05
by ikam
in practice the large number of parameters (collider + rigidbody) give this operator not really friendly to use, so don't know if it is such a good idea. however as it is just a design issue, I prefer to work on more usefull features for now like joints and after particles. I'll see later for this potential operator...

Re: new Physx

PostPosted: Friday, 27.June 2014, 09:31
by Skinnytorus
OK, OK... This is all up to you. There's no rush at all. :)

Re: new Physx

PostPosted: Friday, 27.June 2014, 20:42
by Skinnytorus
Using physx Debris now crashes wz4. :(

Re: new Physx

PostPosted: Friday, 27.June 2014, 21:52
by ikam
fixed.
warning, your sliced mesh have too many shards and generate too many physx actors in the same small area that collides at the same times (it dies perf), play with min and max size in sliceandice to reduce nb of debris

Re: new Physx

PostPosted: Saturday, 28.June 2014, 06:49
by ikam
a little scene with most of elements workings in last update

Re: new Physx

PostPosted: Saturday, 28.June 2014, 07:48
by Skinnytorus
OK. I built a small scene from scratch. Physx functioning seems still unstable - occasionally simulation won't run on pressing F6 at all. After reopening the file several times, it may still work. Can't pin down the culprit. Physx stops to playback sim unexpectedly... Typically, sim doesn't work before saving the file for the first time. Once it stopped simulation after I pressed one of the Desc params in PhysX op. It's not clear what's the matter...

Re: new Physx

PostPosted: Saturday, 28.June 2014, 07:57
by Skinnytorus
And another weird thing: Scrubbing the timeline seems to add acceleration to physx kinematic bodies. An Interesting effect, but do we actually need it?
To repro try to scrub the timeline in my above example.
Edit: On the other hand I realize that physics is calculated in real time, and probably nothing can be done about it...