Crash on fetchResult - Unhandled Exception (Physx 3.2.3)


i have some problems with creating and simulating particles. I properly use the same SimulationStepper as implemented in the example.

At first i initialize physx successfully and create a static solid object. (Simulation run in idle loop. And i am connected to the PVD.)
No Problems here.

Then i create Particles (at least 8.000 - 80.000) (doing that with ParticleSystem from example code), the first simulation circle succeded. I also can readout all particle buffer (position, velocity, etc) and see my solid object and all particles in the PVD.
BUT if the second simulation circle starts, the sync::wait CRASH’s. I also tested it with the DebugStepper. The same problem as already mentioned on call PxScene::fetchResults. (Physx3CHECKED_x86.dll exception).

The crash apears in the PxWorker0? Thread, in Debug and Release Version as well as x86 or x64 bit Version.

Does anybody knows such a kind of problem. Is their any solution for that problem?


I’d say it is very difficult to determine the cause of the problem without additional information. Are there any warning messages on the error stream in the checked build? Are you running the particles on CPU or GPU?

Thanks for your reply.

There are no warning or error messages. The particles run on CPU.

I create a PxRigidStatic and cook some meshes (with 126000 or more triangles) and add it to the actor. Are there some restriction of simulating particles and rigids? Do i need to extend allocating some memory?


I suspect that the size of the mesh may have something to do with it. Have you found that it works with smaller meshes? I’ll bring this to the attention of one of the particle system specialists, he will know the latest limitations.


For testing I deprecate all meshes with more than 65535 triangles now …
… and it works. Specially the maximum number of triangles are exactly 65535, that might be allowed to simulate!

I split the meshes into blocks of 65535 triangles, to show the complete geometry of the actor. If I add one more triangle (set the maximum to 65536), the simulation crashs.

Thanks for your report. There is no known issue with meshes that have more triangles than 65535 and we should support such meshes. I will investigate. Sorry for the inconvenience!