Document CUB temp storage alignment

[jrhemstad]

Updates the tempstorage alias to clarify that no special alignment is required.

Updates all documentation to use the Doxygen alias.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[jrhemstad]

/ok to test fb52488

[jrhemstad]

/ok to test 7b84060

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: a42cd15f-8263-434d-b172-26728da41e1e

📥 Commits

Reviewing files that changed from the base of the PR and between fb52488 and 93cb78c.

📒 Files selected for processing (21)

• cub/cub/device/device_adjacent_difference.cuh
• cub/cub/device/device_copy.cuh
• cub/cub/device/device_find.cuh
• cub/cub/device/device_for.cuh
• cub/cub/device/device_histogram.cuh
• cub/cub/device/device_memcpy.cuh
• cub/cub/device/device_merge.cuh
• cub/cub/device/device_merge_sort.cuh
• cub/cub/device/device_partition.cuh
• cub/cub/device/device_radix_sort.cuh
• cub/cub/device/device_reduce.cuh
• cub/cub/device/device_run_length_encode.cuh
• cub/cub/device/device_scan.cuh
• cub/cub/device/device_segmented_radix_sort.cuh
• cub/cub/device/device_segmented_reduce.cuh
• cub/cub/device/device_segmented_scan.cuh
• cub/cub/device/device_segmented_sort.cuh
• cub/cub/device/device_select.cuh
• cub/cub/device/device_topk.cuh
• docs/cub/Doxyfile
• docs/cub/api_docs/device_wide.rst

✅ Files skipped from review due to trivial changes (19)

• docs/cub/api_docs/device_wide.rst
• cub/cub/device/device_memcpy.cuh
• docs/cub/Doxyfile
• cub/cub/device/device_find.cuh
• cub/cub/device/device_adjacent_difference.cuh
• cub/cub/device/device_topk.cuh
• cub/cub/device/device_segmented_radix_sort.cuh
• cub/cub/device/device_reduce.cuh
• cub/cub/device/device_segmented_reduce.cuh
• cub/cub/device/device_merge_sort.cuh
• cub/cub/device/device_copy.cuh
• cub/cub/device/device_for.cuh
• cub/cub/device/device_select.cuh
• cub/cub/device/device_segmented_scan.cuh
• cub/cub/device/device_radix_sort.cuh
• cub/cub/device/device_merge.cuh
• cub/cub/device/device_segmented_sort.cuh
• cub/cub/device/device_scan.cuh
• cub/cub/device/device_histogram.cuh

🚧 Files skipped from review as they are similar to previous changes (1)

• cub/cub/device/device_partition.cuh

---

Note: CodeRabbit is enabled on this repository as a convenience for maintainers
and contributors. Use your best judgment when considering its review comments and
suggestions — a suggested change may be inadequate, unnecessary, or safe to ignore.
Contributors are not expected to address every comment. Human reviews are what
ultimately matter for merging.

Overview

This PR updates CUB's documentation to standardize how temporary storage parameters are documented and clarifies that no special alignment is required. The changes are documentation-only and do not modify any functional code or API signatures.

Changes

Doxygen Alias Update (docs/cub/Doxyfile)

• Refined the devicestorage alias to clarify that no special alignment is required for d_temp_storage and to document the requirement for a pointer-sized allocation during execution

API Documentation Updates (cub/cub/device/*.cuh)
Updated Doxygen documentation across 22 device headers to replace verbose descriptive text about temporary storage allocation with a standardized @devicestorage Doxygen tag. This change was applied to:

• DeviceAdjacentDifference, DeviceCopy, DeviceFind, DeviceFor
• DeviceHistogram, DeviceMemcpy, DeviceMerge, DeviceMergeSort
• DevicePartition, DeviceRadixSort, DeviceReduce, DeviceRunLengthEncode
• DeviceScan, DeviceSegmentedRadixSort, DeviceSegmentedReduce, DeviceSegmentedScan
• DeviceSegmentedSort, DeviceSelect, DeviceTopK

API Documentation Update (docs/cub/api_docs/device_wide.rst)

• Clarified the temporary storage allocation step by rewording to explicitly state that allocating device-accessible memory is required and that no special alignment is required

Impact

• Lines changed: ~124 insertions, ~283 deletions across all files
• Scope: Documentation only; no functional code changes
• Review effort: Low

Walkthrough

This PR standardizes Doxygen documentation for temporary storage parameters across 16 CUB device algorithm headers. Verbose multi-line descriptions of d_temp_storage allocation behavior are replaced with a centralized @devicestorage tag, defined in the Doxyfile and clarified in user-facing API docs.

Changes

Doxygen Temporary Storage Documentation Standardization

Layer / File(s)	Summary
Define @devicestorage tag and user guidance docs/cub/Doxyfile, docs/cub/api_docs/device_wide.rst	The @devicestorage Doxygen alias is updated to document device-accessible memory requirements, execution-time pointer validity, and clarify that no special alignment is needed. User-facing API documentation is updated to specify device-accessible memory allocation and note the lack of alignment requirements.
Apply @devicestorage tag across device APIs cub/cub/device/device_adjacent_difference.cuh, device_copy.cuh, device_find.cuh, device_for.cuh, device_histogram.cuh, device_memcpy.cuh, device_merge.cuh, device_merge_sort.cuh, device_partition.cuh, device_radix_sort.cuh, device_reduce.cuh, device_run_length_encode.cuh, device_scan.cuh, device_segmented_radix_sort.cuh, device_segmented_reduce.cuh, device_segmented_scan.cuh, device_segmented_sort.cuh, device_select.cuh, device_topk.cuh	Replace verbose d_temp_storage parameter documentation (describing allocation behavior and nullptr handling) with the standardized @devicestorage tag across 156 individual overloads and variants.

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[jrhemstad]

/ok to test 93cb78c