Jorrit Rouwe
@jrouwe
Followers
2,458
Following
145
Media
22
Statuses
226
Lead Game Tech Programmer at Guerrilla Games, creator of the Jolt Physics library. @jrouwe @mastodon .gamedev.place
Joined November 2009
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
#TGS2024
• 170748 Tweets
Tigers
• 103271 Tweets
Miami
• 79815 Tweets
Knicks
• 67573 Tweets
Towns
• 64649 Tweets
Minnesota
• 63816 Tweets
#SmackDown
• 53662 Tweets
Randle
• 48205 Tweets
Wolves
• 34983 Tweets
Rudy
• 34560 Tweets
Royals
• 27266 Tweets
Hart
• 22474 Tweets
#DesafioXX
• 21379 Tweets
भगत सिंह
• 21148 Tweets
Asharamji Bapu Case
• 18228 Tweets
日本破壊クソメガネ
• 16671 Tweets
Equality Exploited
• 15666 Tweets
Cam Ward
• 15401 Tweets
Celtics
• 11244 Tweets
v4.0.0 of Jolt Physics was released. Most important new features:
* Added support for soft bodies (still in development).
* Support 2D simulations
* Support for surface velocity
* Support non-power of 2 height fields
8
91
761
Jolt Physics reached 4k github stars! Thanks to everyone who starred it! (and if you haven't starred it, please do!). Also thanks to Unity for helping speed up Jolt's adoption rate by changing business model.
10
48
646
Been working on multi threading soft bodies for Jolt Physics. This 10K vertex cloth + 121 boxes can now be multi threaded and runs in ~2ms (3x faster than before), when you have multiple soft bodies the speed up can be 4x and there's more to come.
9
68
612
Jolt Physics got soft body physics today! There's still a lot of performance optimizations to be done, but you can start playing around with it!
10
77
474
I just released Jolt Physics v3.0.0. It includes double precision simulation, performance optimizations for large islands and experimental motor cycle support.
4
47
249
Jolt Physics now supports splitting large islands across multiple CPU cores. This can speed up the simulation of large piles of rigid bodies up to a factor of 4!
6
38
208
Jolt Physics now supports setting the surface velocity of a body. Conveyor belts here we come!
3
21
213
I've added experimental motorcycle support to Jolt Physics. It will probably require a lot of tweaking / hacks to include it in a game. But it's fun to drive around!
1
16
198
Jolt Physics is now deterministic across platforms! Amazed at how compiling in 'precise' mode makes floating point calculations exactly the same across platforms (Windows vs Linux vs MacOS) and even compilers (MSVC vs clang)! Only trigonometric functions needed fixing.
6
14
152
Have been playing around with deformable heightfields for Jolt Physics this weekend. Was surprisingly hard given that the whole thing is block compressed and has a hierarchical acceleration structure for fast queries.
2
10
132
I created an early version of WebAssembly support for Jolt Physics: . Let me know if you are interested.
@SebAaltonen
@anders_breakin
3
13
75
Mihe just created the first release of Godot Jolt, this is so cool to see! Let's hope many games will be made with it!
@thatmihe
2
15
71
A major milestone for Jolt Physics! It can now be compiled with support for double precision positions to support much larger worlds.
0
9
63
Thanks to
@phoenixillusion
we now have cars, motorcycles and tanks in JoltPhysics.js (JavaScript). See the demos here:
2
4
55
X4 has switched over to Jolt Physics in this update!
Time to head out for your long-awaited reunion with the Boron 🪐
#X4KingdomEnd
and X4 Update 6.00 are now available!
@Steam
➡️
@GOGcom
➡️
#GameRelease
#GameDev
#SpaceGames
#SandboxGames
12
49
202
1
1
42
1
11
26
Thanks for making Jolt part of the potato meme!😁
You like your potato physics baked in high quality?
Go watch my new clip on YouTube!:
#Starfield
#gamedev
#GodotEngine
#indiedev
#solodev
#madeingodot
#screenshotsaturday
0
1
32
2
1
39
@AustinFacey
I enjoy making things and I enjoy it even more when other people want to use the things I make. For me it was a no-brainer to give it away for free.
2
0
35
@mnenoff
The first problem has been fixed:
This is long overdue, but I'm happy to say that we finally have Android and iOS support in Godot Jolt as of version 0.10.0.
Android support in particular should be of interest to you
#VR
people out there.
#GodotEngine
3
24
154
2
1
22
@AlmostMikhail
Can you send me that godot scene? I'd like to profile it. Jolt should easily handle 2000 bodies, the video below shows that it has no problems rendering 2000 bodies at 60 fps (this is from the Jolt sample app).
3
0
21
Missed this one. Cool to see Jolt being used!
Light Tracer Render 2.4 is out
The GPU renderer now uses Guerrilla Games' Jolt physics library, making it possible to place objects in a scene and have them settle naturally under gravity
More details:
#rendering
#visualization
#gamedev
0
7
20
0
1
19
@SebAaltonen
If you have something working, I'd be interested in taking a look, maybe even supporting it officially. To get around the SIMD issue I could make a SIMD-less version of Vec3/Mat44 etc. Note that I have zero experience with WASM at the moment.
1
0
11
@HugoLocurcio
I haven't done any comparisons. Algorithms in 2D are a lot simpler than in 3D so I would expect an optimized/multithreaded 2D physics engine to beat Jolt in 2D, but Jolt will allow you to mix 2D and 3D or have multiple 2D layers at different depths in a single world.
0
0
9
We've opened up 2 position on the Game Tech team at Guerrilla! Interested in animation programming or interested in working on a framework to collect lots of data from the game?
0
0
8
@erin_catto
No it's still cross platform deterministic and I didn't notice a difference in stability either.
0
0
6
@lateasusual_
You could probably use both a wheeled AND tracked vehicle on the same body to make a half track
1
0
2
@SubstrataVr
Triangle mesh vs triangle mesh collision detection is quite expensive. Normally you would convert the triangle mesh of the dynamic object into a StaticCompoundShape of ConvexHullShapes using convex decomposition (see e.g. )
0
0
3
@phys_ballsocket
@SheriefFYI
I knew it was wrong on Linux ARM but it seemed to work fine on x86. In any case: the function is only used for profiling, so for normal users of the lib it doesn't matter what the function returns (eg WASM returns 1).
2
0
3
@raiganburns
@santagada
@reduzio
@nvidia
Personally I think that there are so many approximations in any physics engine that I doubt that SAT is going to make it much more accurate than EPA, but I'm more than willing to implement SAT to find out.
1
0
3
@SheriefFYI
Fancy tool! I rearranged some code in ObjectStream and got the library to compile 20% faster!
0
0
2
@3dflashlo
Yes, JavaScript version is single threaded currently, although if I read the docs correctly it should be a matter of passing a compiler flag
0
0
2
@ian_snyder
@thatmihe
Cool project! I never looked at the Unreal code, but from what I have heard there was a period that it supported PhysX and Chaos through an abstract interface. I think PhysX support got removed together with the abstract interface. If this is true integrating Jolt is difficult.
1
0
1
@SebAaltonen
@anders_breakin
For now it always compiles without SIMD so should work in safari too. With a little bit more work we could compile 2 versions.
0
0
1
0
0
2
@anders_breakin
@MichalZiulek
I'm currently looking into supporting wasm myself (just starting, no promises yet as I have never worked with it).
1
0
2
@Timsecond2
Most of it already is, Godot Jolt tracks the master branch of Jolt, so it gets updated every week or so while I only release a new official version infrequently.
0
0
2
@raiganburns
@santagada
@reduzio
@nvidia
Note that a Godot plugin for Jolt is already in the making:
0
0
2
@PierreTerdiman
Just send me the PEEL source code and give me the plugin code you have so far (PhysX plugin would be nice as an example but not needed). I'll finish it up for you.
0
0
2
@MichalZiulek
Looking good! Some small remarks: AssertFailedImpl is unused (you're compiling without JPH_ENABLE_ASSERTS so asserts are not availble). Also I would static_assert all the constants (e.g. JPH_SHAPE_TYPE_CONVEX) as I can't guarantee they will be stable.
2
0
2
@DavitMasia
The only comparison I've seen is:
Jolt Physics in PlayCanvas! averaging at 130ms frame time with 1770 objects at 600kb filesize! This beast is powering physics in Horizon Forbidden West. Many thanks to
@jrouwe
for the help.
#webgl
@playcanvas
#javascript
#physics
#gamedevelopment
1
1
14
1
0
1
@BattleAxeVR
@SergeyMakeev
I spend a lot of time answering github questions (and I think I answer them all), which questions are we talking about?
1
0
1
@PierreTerdiman
Trying to make a unit test to repro this, but I can't make it happen. Can you try changing the code to this? (and if that doesn't work give me inPolygonToClip, inPlaneOrigin and inPlaneNormal).
1
0
1
@MichalZiulek
Also I'm wondering if something like JPH_BodyCreationSettings_InitDefault() { JPH_BodyCreationSettings cs; ::new (&cs) JPH::BodyCreationSettings; return cs; } would be safer (less repeating of constants from inside Jolt).
0
0
1
@lateasusual_
Yes you can disable stuff on the second constraint. Since you have 2 engines with 2 vehicle constraints, you probably want to set the transmission to manual for the front wheels and set it to neutral always so that all propulsion comes from the tracks.
0
0
1
@AmerKoleci
I don't know (code is the same), perhaps they embed the debug symbols? Try removing -g in CMakeFiles.txt
1
0
1
@gustafamalada
MeshShape vs MeshShape or MeshShape vs HeightFieldShape collisions are not supported, so you can have a kinematic mesh as long as it doesn't collide with any of those shapes. Where did you find this in the documentation? (this used to be true in an older version of the library)
1
0
1
@gustafamalada
The volume of an arbitrary mesh cannot be computed so the mass cannot be calculated so it needs to be provided
1
0
1
@SubstrataVr
@Lo_Rezzed
You can make this work by doing a convex decomposition of the torus (make a StaticCompoundShape of torus segments). Make sure that the segments don't overlap because overlapping volumes will mess with the buoyancy calculations.
1
0
1
@thatmihe
@Bimbam_tm
But I don't know of any generic solution that would allow the bottom bodies to sleep while the top ones are active.
0
0
1
@0430agi
@mux213
@reduzio
@john_clayjohn
@RandomPedroJ
@qbieshay
@dariosamo
@HugoLocurcio
Can you post a link to that ticket?
1
0
1
@BattleAxeVR
@SergeyMakeev
You only need to implement these functions: . If you leave DrawText3D empty then it should be quite simple to map them to a vertex buffer and render them.
1
0
1
@PierreTerdiman
raycasts see meshes as double-sided -> see RayCastSettings::mBackFaceMode
is there a way to accurately compute memory usage -> already replied in other tweet
s there a way to share shapes -> yes, shapes are refcounted you can attach them to as many bodies as you like
2
0
1
@thatmihe
@Bimbam_tm
I'm guessing that if you have some very specific domain knowledge (like knowing they are in a vault and that the bottom of the pile cannot move) that you could mark certain bodies as static until the layers on top of them start accumulating velocity.
1
0
1
@BattleAxeVR
@SergeyMakeev
What we did is to override the debug renderer to render using our own render engine. That way you can see the physics view and the game view at the same time. I've worked with VD's before but imo nothing beats in game view.
1
0
1
@andreas_longva
I've added some documentation on how it works here: (section Solve Velocity Constraints)
0
0
1
@CynicatPro
No, it's because of the block compression algorithm and because each block depends on its surrounding blocks (you need triangles to connect to the next block).
1
0
1
@anders_breakin
I'm currently building a version without simd and Jolt can easily run on 1 thread. I was going to try emscripten but as said have 0 experience so don't know what I am going to run into.
0
0
1
@PierreTerdiman
Hey, good to see that you're integrating this into PEEL! In both your stress tests everything is basically in a single simulation island. For my use cases this never happens, so I did not spend any time trying to split a simulation island across multiple CPU cores.
1
0
1
@twgage
@phoenixillusion
We have tried running it multithreaded, but there appears to be a bug in Safari that prevents it from working (other browsers work). See: and
0
0
1