Skip to content
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

🐛 [BUG] - Significant difference in the values of the Hessian for high-valued points between GPU and CPU #1785

Open
eyalshimony opened this issue Jan 28, 2025 · 0 comments
Labels

Comments

@eyalshimony
Copy link

Description

A significant difference in the values of the Hessian for high-valued points is observed between simulation run on GPU and on CPU. Thus, while the median, 99th, and 99.9th percentile are still the same, a difference is already observed at the 99.99th percentile (2%) and keeps growing from there - 16% at the 99.999th percentile, and 11 order of magnitude difference between the absolute maximum of the two. This difference was not observed for the sensitivity kernels for alpha, beta and rho. While this issue can be overcome by clipping, it still seems like a problem.

Affected SPECFEM3D version

Latest development version

Your software and hardware environment

gcc 10.1.0, slurm 24.05.05, cluster with gpu

Reproduction steps

Model: 80x80x80 elements, 20 km x 20 km x 20 km, dt = 3 ms, NSTEP = 5000.
CPU run on 50 cores, GPU on 2 GPUs.

Screenshots

![DESCRIPTION](LINK.png)

Logs

OS

Linux

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

No branches or pull requests

1 participant