4 Commits

Author SHA1 Message Date
Jiaming Yuan
3fbb221fec
[coll] Implement shutdown for tracker and comm. (#10208)
- Force shutdown the tracker.
- Implement shutdown notice for error handling thread in comm.
2024-04-20 04:08:17 +08:00
Jiaming Yuan
8bad677c2f
Update collective implementation. (#10152)
* Update collective implementation.

- Cleanup resource during `Finalize` to avoid handling threads in destructor.
- Calculate the size for allgather automatically.
- Use simple allgather for small (smaller than the number of worker) allreduce.
2024-03-30 18:57:31 +08:00
Jiaming Yuan
0715ab3c10
Use dlopen to load NCCL. (#9796)
This PR adds optional support for loading nccl with `dlopen` as an alternative of compile time linking. This is to address the size bloat issue with the PyPI binary release.
- Add CMake option to load `nccl` at runtime.
- Add an NCCL stub.

After this, `nccl` will be fetched from PyPI when using pip to install XGBoost, either by a user or by `pyproject.toml`. Others who want to link the nccl at compile time can continue to do so without any change.

At the moment, this is Linux only since we only support MNMG on Linux.
2023-11-22 19:27:31 +08:00
Jiaming Yuan
6c0a190f6d
[coll] Add comm group. (#9759)
- Implement `CommGroup` for double dispatching.
- Small cleanup to tracker for handling abort.
2023-11-07 11:12:31 +08:00