This repository was archived by the owner on Mar 21, 2024. It is now read-only.
File tree Expand file tree Collapse file tree
Expand file tree Collapse file tree Original file line number Diff line number Diff line change 1- # CUB 1.15.0 (NVIDIA HPC SDK 21.11)
1+ # CUB 1.15.0
22
33## Summary
44
5- CUB 1.15.0 accompanies the NVIDIA HPC SDK 21.11 release. It includes a
6- new ` cub::DeviceSegmentedSort ` algorithm, which demonstrates up to 5000x speedup
7- compared to ` cub::DeviceSegmentedRadixSort ` when sorting a large number of small
8- segments. A new ` cub::FutureValue<T> ` helper allows the ` cub::DeviceScan `
9- algorithms to lazily load the ` initial_value ` from a pointer. ` cub::DeviceScan `
10- also added ` ScanByKey ` functionality.
5+ CUB 1.15.0 includes a new ` cub::DeviceSegmentedSort ` algorithm, which
6+ demonstrates up to 5000x speedup compared to ` cub::DeviceSegmentedRadixSort `
7+ when sorting a large number of small segments. A new ` cub::FutureValue<T> `
8+ helper allows the ` cub::DeviceScan ` algorithms to lazily load the
9+ ` initial_value ` from a pointer. ` cub::DeviceScan ` also added ` ScanByKey `
10+ functionality.
1111
1212The new ` DeviceSegmentedSort ` algorithm partitions segments into size groups.
1313Each group is processed with specialized kernels using a variety of sorting
@@ -84,6 +84,8 @@ kernel launches.
8484 (@matt-stack) for this contribution.
8585- NVIDIA/cub#388: Fix debug assertion on MSVC when using
8686 `cub::CachingDeviceAllocator`.
87+ - NVIDIA/cub#395: Support building with `__CUDA_NO_HALF_CONVERSIONS__`. Thanks
88+ to Xiang Gao (@zasdfgbnm) for this contribution.
8789
8890# CUB 1.14.0 (NVIDIA HPC SDK 21.9)
8991
Original file line number Diff line number Diff line change @@ -100,7 +100,7 @@ See the [changelog](CHANGELOG.md) for details about specific releases.
100100
101101| CUB Release | Included In |
102102| ------------------------- | --------------------------------------- |
103- | 1.15.0 | NVIDIA HPC SDK 21.11 |
103+ | 1.15.0 | TBD |
104104| 1.14.0 | NVIDIA HPC SDK 21.9 |
105105| 1.13.1 | CUDA Toolkit 11.5 |
106106| 1.13.0 | NVIDIA HPC SDK 21.7 |
You can’t perform that action at this time.
0 commit comments