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

Replay Unable to Track Shared Memory Allocators #930

Open
kab163 opened this issue Jan 31, 2025 · 0 comments
Open

Replay Unable to Track Shared Memory Allocators #930

kab163 opened this issue Jan 31, 2025 · 0 comments
Assignees

Comments

@kab163
Copy link
Contributor

kab163 commented Jan 31, 2025

Describe the bug

When gathering a replay trace of some program that uses shared memory allocators, the replay executable is unable to replay that file and no .ult file is dumped. It just seg faults.

To Reproduce

Steps to reproduce the behavior:
Run any program that uses umpire shared memory with the UMPIRE_REPLAY=On env variable. Get the .stats file.
Build and make a separate umpire build with tools enabled (and hip/cuda enabled if appliable, shared memory if applicable, etc.)
If you run the replay executable with the -d option to get a .ult file dumped out, it will seg fault.

Expected behavior

Replay should be able to handle shared memory allocators especially as interest in using shared memory increases...

Compilers & Libraries (please complete the following information):
I ran a marbl triple pt run with the device memory on and it happens to use some shared memory allocators.
I ran this on rzadams with hip enabled, rocm/6.0.3, cmake/3.23.1, and shared memory

Additional context

The problem is likely that Replay is not tracking the traits that shared memory allocators use. It really just needs to know about the size of the shared memory allocator. Once that is added, that should be enough to get things working.

@kab163 kab163 self-assigned this Jan 31, 2025
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

1 participant