In the physics consideration I think you are missing many things, like rigid bodies, granular solvers (that are not exactly particles) and FEM elements.
Just wanted to mention this, because those are very important features, specially rigid bodies, and interaction between systems is capital too.
IMHO from a design standpoint, if these are the foundations, have to be there
Maybe you already have this in mind, but I don’t want to miss the opportunity to mention it, better safe than sorry
@silex for the time being at least you can check Storm for granular simulations, the only thing left in cycles to properly render those is the point cloud rendering feature (that will probably come with Cycles-X) and the habilita to turn Alembic into a point cloud (which I think it was near).
I think something related to granular has been tested at some point, related to Mantaflow up to my knowledge, but I’m not entirely sure, time will tell.
My only concern here is that those systems are not lost in time because they are kind of simpler, like rigid bodies, or kind of complex, like granular or FEM
Particle / Physics nodes is a very important feature for Blender, specially if it allows systems interoperability with some kind of attribute translate or intermediate simulation system, but that’s yet to be seen
Let’s not forget about one more important missing simulation element, which is an up to date fracturing algorithm, right now Blender users only have two choices:
An outdated cell fracture add-on stuck in version 0.2 for several years with the slowest algorithm ever seen and many missing features (ex. no way to hide pre-fractured chunks).
A fracture modifier branch which is faster but stuck on blender 2.79 for several years and the cumbersome exporting/importing workflow that comes with it.
Both of which shows no sign of active development to support the newest official blender versions (just pick your poison…).
And while you are at it, it should be about time to move away from the clunky bullet solver and move towards more newer and versatile ones like MPM, FEM, PBD, etc…
I agree with you, but just as a side note, you have a new addon (it’s commercial), RBDLab, on Blender Market.
Regarding Bullet, Bullet it’s a pretty powerful system, it’s used in other widely used physics systems for destruction, like Thinking Particles, the problem is not bullet but it’s current implementation, it’s limited and it does not leverage what bullet can provide.
I would love to be able to make my “point instanced” objects falling to the ground with RBD. Or even better… make the instances accommodate better colliding with each other…
Yep, that’s a cool thing, and I think that’s exactly what that workshop they are doing is about, what you ask for is Rigid Bodies support, and they are talking about how to implement simulation solvers in general, let’s give them time
This. Honestly with how improved the boolean speed got enhanced between 2.xx to 3.00 I have been considering just using boolean hacks to produce fractures lol. But a fully featured and functioning fracture node would make geometry nodes and Blender into the future.
Last time I checked, bullet doesn’t support GPU acceleration, that alone makes it a thing of the past in today’s GPU everything accelerated world. there was some work done when the developer was backed by AMD for a couple of months with promises of GPU acceleration in bullet 3.x, but it didn’t go anywhere, now it’s Google that is backing the dev, for past couple of years to do some robotic related stuff, non of which related to GPU acceleration.
If you google:
What’s up with Bullet 3 and GPU support?
the first result is from the bullet forum, the developer himself is clearly stating the following:
Now if Blender 3.x + is planning for the next 10+ years with things like Cycles X and Geometry nodes, it should be logical to move away from old standards (like bullet, first released back in 2003) and adopt more up to date GPU accelerated ones, like PBD (Position Based Dynamics) for example.
I mean, just look at this resent 2020 work by “Matthias Müller-Fischer, et al.”:
All that stuff described down there is running in REAL TIME !!
The guys at Interactive Computer Graphics have released an open source PBD library under the MIT license back in 2015.
As well as the SPlisHSPlasH open source Library back in 2017 with the same type of license.
They have even made an add-on to port their particle data to blender:
What’s more ? they are based in Germany ^^, (some kind of meeting with them, wouldn’t hurt to help make some killer physics system in blender, finally !)
Some of their implementations are dating back to 2011 like this one (also real time…):
and they look and perform better then what we have currently in blender using the bullet engine.
Sorry if it seems like I just exploded or something, but I kept following the development of Geo nodes silently (most of the time…), but now that things are moving towards physics, I can’t hold my excitement as well as my fears for what to come.
I hope the team covers a wide range of possibilites, because even though the idea of letting the community step in sharing or selling add-ons, one can’t hope for some third party to create something really “big” for industry work.
