rapidsai / build-planning

Tracking for RAPIDS-wide build tasks
https://github.com/rapidsai
0 stars 1 forks source link

Update RAPIDS to use `cuda::mr::async_resource_ref` #16

Open harrism opened 5 months ago

harrism commented 5 months ago

This is a cross-RAPIDS tracking issue for the epic described in this document. Please refer to the document for background and details.

### Tasks
- [x] Introduce cuda::mr::{async_}resource_ref to RMM and use in place of device_memory_resource pointers. Introduce conversions between rmm::cuda_stream_view and cuda::stream_ref. Done in rapidsai/rmm#1095 (23.12)
- [x] Require explicit initial pool size in pool_memory_resource. This is because it is difficult to automatically size the pool when the memory is not device memory. Done in rapidsai/rmm#1417
- [ ] https://github.com/rapidsai/rmm/issues/1388
- [ ] https://github.com/rapidsai/rmm/issues/1389
- [ ] https://github.com/rapidsai/rmm/issues/1402
- [ ] https://github.com/rapidsai/rmm/issues/1442
- [ ] https://github.com/rapidsai/rmm/issues/1443
- [ ] https://github.com/rapidsai/rmm/issues/1444
- [ ] https://github.com/rapidsai/cudf/issues/15549
- [ ] https://github.com/rapidsai/cudf/issues/15498
- [ ] https://github.com/rapidsai/raft/issues/2261
- [ ] https://github.com/rapidsai/cugraph/issues/4333
- [ ] https://github.com/rapidsai/cuml/issues/5839
- [ ] https://github.com/rapidsai/cuspatial/issues/1371
- [x] Convert any other RAPIDS repos to use `rmm::mr::device_async_resource_ref` instead of `device_memory_resource` pointers
- [ ] https://github.com/rapidsai/rmm/issues/1445
harrism commented 5 months ago

1443 includes much of this, but I've included all of the work here for cross-repo visibility.

jakirkham commented 1 month ago

Have noted this as "In progress". Please let me know if that is not the case and can change