09-12-2019, 12:22 AM
(This post was last modified: 09-12-2019, 12:26 AM by josemendez.)
Hi!
Never measure performance in fps, always use milliseconds. A drop of from 500 fps to 420 fps (80 fps less) is 0.3 ms, so barely any drop in performance. Dropping from 420 to 120 fps (8 ms) is however a much larger hit.
This shouldn't happen if your rope physics was being updated once per frame. However your profiler data indicates it's being updated 12 times per frame. This should not be the case if your max fixed timestep is as close to your fixed timestep value as you indicate. Increasing the max fixed timestep value can only worsen performance, never improve it. Try setting your max fixed timestep to the same value as your fixed timestep, report back results.
Judging from the data your provided, my guess is that you have 6 solvers in your scene. Is that the case?
Quote:I am developing a VR application for Oculus Quest. But the performance issue is same on a PC. Inside the editor, without any of the 6 ropes active, FPS is around 400-500. Activating one rope drops it by 80-100. If all 6 are activated, FPS is around 70-120.
Never measure performance in fps, always use milliseconds. A drop of from 500 fps to 420 fps (80 fps less) is 0.3 ms, so barely any drop in performance. Dropping from 420 to 120 fps (8 ms) is however a much larger hit.
This shouldn't happen if your rope physics was being updated once per frame. However your profiler data indicates it's being updated 12 times per frame. This should not be the case if your max fixed timestep is as close to your fixed timestep value as you indicate. Increasing the max fixed timestep value can only worsen performance, never improve it. Try setting your max fixed timestep to the same value as your fixed timestep, report back results.
Judging from the data your provided, my guess is that you have 6 solvers in your scene. Is that the case?