Real-Time Physics Simulation Forum

After some debugging, it seems that the MT solver does not like connecting 2 springs with objects of infinite mass (though this is a redundant case).
I am still verifying at my end as to how this happened.

Hi, I am running a simulation with about 40K objects, Bullet3 is compiled to use MT. I am also using about 150K springs to connect these objects via btGeneric6DofSpringConstraint My application crashes: tSimulationIslandManagerMt::Island* btSimulationIslandManagerMt::getIsland(int): Assertion `id >=...

cat build_cmake_pybullet_double.sh cmake -DBT_THREADSAFE=1 -DBUILD_PYBULLET=ON -DBUILD_PYBULLET_NUMPY=ON -DUSE_DOUBLE_PRECISION=ON -DBT_USE_EGL=ON -DCMAKE_BUILD_TYPE=Debug .. || exit 1 I added BT_THREADSAFE=1 , with this, btCreateDefaultTaskScheduler() gives me a non-null object adding this to cmake...

using VS017, I select all projects and add BT_THREADSAFE as a preprocessor to all of them.
The simulation runs fine on window, I need something similar on Linux

I am trying to use the default threaded scheduler, cout<<"btCreateDefaultTaskScheduler:"<<btCreateDefaultTaskScheduler()<<endl; cout<<"btGetOpenMPTaskScheduler:"<<btGetOpenMPTaskScheduler()<<endl; both cases return NULL. Is there some flag to be set to while building bullet3 to g...

looking at the code // create an OpenMP task scheduler (if available, otherwise returns null) 786 btITaskScheduler* btGetOpenMPTaskScheduler() 787 { 788 #if BT_USE_OPENMP && BT_THREADSAFE 789 static btTaskSchedulerOpenMP sTaskScheduler; 790 return &sTaskScheduler; 791 #else 792 return NU...

any update....?

btGetOpenMPTaskScheduler and btCreateDefaultTaskScheduler always return null I would like t use this... btCollisionDispatcherMt throws an error stating that a task sheduler is not set (if i set it btSetTaskScheduler(btGetSequentialTaskScheduler()) , it works fine, but I would want to use multicpu) P...