Help Performance spikes in BeginStep and Substep - Printable Version +- Obi Official Forum (https://obi.virtualmethodstudio.com/forum) +-- Forum: Obi Users Category (https://obi.virtualmethodstudio.com/forum/forum-1.html) +--- Forum: General (https://obi.virtualmethodstudio.com/forum/forum-5.html) +--- Thread: Help Performance spikes in BeginStep and Substep (/thread-1518.html) |
Performance spikes in BeginStep and Substep - jp715 - 27-11-2019 Hi, I'm using Obi Cloth 5.0 with Unity 2018.4.9. I'm profiling a scene with an Ambient Force Zone, Late Fixed Updater, and a Skinned Cloth. I seem to be getting some pretty massive spikes in RunLateFixedUpdate(), specifically in Substep and BeginStep, going from around 0.5ms up to 15-30ms. Any ideas on why this could be or how to resolve it? The solver is using only Distance and Skin constraints, with 1 iteration each. The spikes persist even when I disable the force zone and skinned cloth and have nothing in my scene enabled except for the Solver and Updater. I've also tried commenting out the lines UpdateColliders(), UpdateAllRigidbodies(), UpdateAllVelocities() in Substep() as I'm not using any colliders or rigidbodies, with no effect. RE: Performance spikes in BeginStep and Substep - josemendez - 29-11-2019 (27-11-2019, 07:35 PM)jp715 Wrote: Hi, I'm using Obi Cloth 5.0 with Unity 2018.4.9. I'm profiling a scene with an Ambient Force Zone, Late Fixed Updater, and a Skinned Cloth. I seem to be getting some pretty massive spikes in RunLateFixedUpdate(), specifically in Substep and BeginStep, going from around 0.5ms up to 15-30ms. Any ideas on why this could be or how to resolve it? The solver is using only Distance and Skin constraints, with 1 iteration each. The spikes persist even when I disable the force zone and skinned cloth and have nothing in my scene enabled except for the Solver and Updater. I've also tried commenting out the lines UpdateColliders(), UpdateAllRigidbodies(), UpdateAllVelocities() in Substep() as I'm not using any colliders or rigidbodies, with no effect. Hi there, We had another user reporting these spikes in 2018, but as of now we've been unable to reproduce them anywhere (2018 using 4.1 or 2019 using 5.0, in Windows or macOS). Would you be willing to share your project with us? maybe we'll have an easier time reproducing this under the exact same conditions it has been known to happen. If you can, send the project to support(at)virtualmethodstudio.com and we'll take a look. RE: Performance spikes in BeginStep and Substep - jp715 - 29-11-2019 (29-11-2019, 08:26 AM)josemendez Wrote: Hi there, Sure thing! I'll send that to you now |