Plumb `cuda::execution::runs_on` through various CUB algorithms
Open
- Dominant language
- C++
- Stars
- 2.5k
- Forks
- 486
- Avg merge
- 2d 6h
- Merged PRs (30d)
- 295
Description
Once `cuda::execution::runs_on` is implemented, allow CUB algorithms to query it instead of doing runtime GPU queries for memory estimation. It remains to be seen if this should also be used when doing kernel selection.
- `cub::DeviceRadixSort::SortPairs` / `SortPairsDescending`
- `cub::DeviceSegmentedRadixSort::SortPairs` / `SortPairsDescending`
- `cub::DeviceRadixSort::SortKeys` / `SortKeysDescending`
- `cub::DeviceSegmentedRadixSort::SortKeys` / `SortKeysDescending`
Contributor guide
Assessment
This issue has not been assessed yet.