New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Performance regression in mj_copyData (2.3.0 vs 2.1.0 ) #568
Comments
Thanks, we will look into this. In order to help us diagnose, could you please print However, you should know that |
Hi Yuval, thanks for looking into this and also thx for pointing out the more efficient way. Diagnostic information: mujoco version 2.1.0: nbuffer: 5052592 mujoco version 2.3.0: nbuffer: 285568 |
@ThoenigAdrian I find it rather suspicious that Are your numbers reliably reproducible? Would you mind running it against a few different versions in between? |
@saran-t I can understand your suspicion. I thought the same in the beginning. And since the time I let my application run is fixed (around 20 seconds) less time is spent in mj_step (since it's called fewer times) |
I currently don't have enough time to run it against a few different versions. But when I do in the future I'll let you know the results. |
@ThoenigAdrian we will look at the |
@yuvaltassa copying the state and recreating the contents of mjData works fine for me. |
Hi,
I recently migrated a mujoco project from mujoco version 2.1.0 to 2.3.0. I discovered that my application is running way slower. I ran a performance profiler to see why.
Turns out mj_copyData is way slower than before.
For those who are wondering about my motivations and why I'm doing this: I am copying the data, because i have 2 threads one in which i run the simulation and the main thread where i render/visualize.
Reproduce: I can provide a minimal working example if needed. However in this case I think it should be very easy to reproduce and compare. But if it isn't, I can provide something just let me know.
The text was updated successfully, but these errors were encountered: