Nicholas Tomlinson comments

Results 6 comments of


                                            Nicholas Tomlinson

[Issue]: Memory copy started by dependency signal for pinned memory exhibits race condition

Just a thought that occurred to me: I don't think the bug is specific to the SDMA. It's just more easily reproduced that way. If I put a second GPU...

[Issue]: Memory copy started by dependency signal for pinned memory exhibits race condition

Thanks :) Please let me know if there's anything I can do to help.

[Issue]: Memory copy started by dependency signal for pinned memory exhibits race condition

Hello - I've attached [a core dump, binaries with debug information, and new system information](https://github.com/user-attachments/files/17466372/2024102101-pincopy.tar.gz). - I ran the build and test in the Docker image for Ubuntu 24.04 this...

[Issue]: Memory copy started by dependency signal for pinned memory exhibits race condition

I ran `rocminfo` from the host machine, which is on the ROCm 6.0.2 userspace. That probably accounts for the difference in HSA system attributes you observed. There's a `build.sh` script...

[Issue]: Memory copy started by dependency signal for pinned memory exhibits race condition

I did (by building the parts of ROCm userspace that it uses). It does still happen in that case. I haven't tested bare metal Ubuntu yet though.

[Issue]: Memory copy started by dependency signal for pinned memory exhibits race condition

Just tested with kernel `6.13.1` and it seem to be fixed there :)