Skip to content

Conversation

micmelesse
Copy link
Collaborator

If tensors are on a specific device such device 7 and triton launches a kernel on the default device such as device 0. We get a segfault. I add asserts to catch this bug. I also add a bunch of other asserts to insure we are working well formed inputs.

@micmelesse
Copy link
Collaborator Author

Main suite passes

image

@micmelesse
Copy link
Collaborator Author

Internal tests pass.

image

@micmelesse
Copy link
Collaborator Author

No change in perf
image

@micmelesse micmelesse marked this pull request as ready for review July 14, 2025 18:50
@micmelesse micmelesse merged commit 5db9d18 into main_perf Jul 14, 2025
1 check failed
micmelesse added a commit that referenced this pull request Jul 18, 2025
* Compress segfault work

fix backward segfault

rework offset

ignore .profile

ignore .analysis

save

* assert the kernel launch device and tensor devices are the same

* fix failing asserts

* add asserts to fwd
micmelesse added a commit that referenced this pull request Jul 19, 2025
* Compress segfault work

fix backward segfault

rework offset

ignore .profile

ignore .analysis

save

* assert the kernel launch device and tensor devices are the same

* fix failing asserts

* add asserts to fwd
@micmelesse micmelesse deleted the micmelesse/varlen_bwd_segfault branch July 21, 2025 16:02
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

Successfully merging this pull request may close these issues.

1 participant