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

intercept lib launched through LD_PRELOAD cannot intercept cuda driver API calls with pytorch version >=1.10 #114

Open
Fizzbb opened this issue Mar 30, 2022 · 1 comment
Assignees

Comments

@Fizzbb
Copy link
Collaborator

Fizzbb commented Mar 30, 2022

cuda call did not get intercepted for pytorch 1.11.0, test script based on this
https://github.com/CentaurusInfra/alnair/blob/main/test/scripts/pytorch.Dockerfile

@Fizzbb Fizzbb changed the title intercept lib do support all pytorch version intercept lib doesn't support all pytorch version Jun 24, 2022
@Fizzbb Fizzbb changed the title intercept lib doesn't support all pytorch version intercept lib launched through LD_PRELOAD cannot intercept cuda driver API calls with pytorch version >=1.10 Jun 24, 2022
@pint1022
Copy link
Collaborator

The issue appears after the cuda 11.4. The cuda API stops using dlsym to look for the driver functions, instead it calls cuGetProcAddress. All hooks implemented on top of dlsym will fail to get called.

The solution can be found from here: https://forums.developer.nvidia.com/t/how-to-hook-cuda-runtime-api-in-cuda-11-4/190050/8

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

No branches or pull requests

3 participants