[davebayer]

This PR reworks the group_as mapping to take a viewable range instead of a contiguous piece of memory. This make it much more flexible and easier to use.

[coderabbitai]

Important

Review skipped

Review was skipped as selected files did not have any reviewable changes.

Run configuration

• Source mode: Auto
• Repository: NVIDIA/cccl
• Branch: main
• File: .coderabbit.yaml
• Commit: 036ecd2
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: 5dd99b76-b9ed-4313-86c3-712718478d0e

Commits

Reviewing files that changed from the base of the PR and between fffca4e and 0913bf8.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

📝 Summary

Summary by CodeRabbit

• New Features
  • group_as accepts sized ranges of group counts, including C arrays, cuda::std::array, and cuda::std::span, to define group mappings.
  • Group mappings use the supplied range’s size and values, including values determined at runtime.
• Compatibility
  • group_as no longer accepts index_sequence arguments; use a sized range instead.
• Bug Fixes
  • Exhaustive mappings validate that group counts sum to the available units. Non-exhaustive mappings mark ranks beyond that sum as invalid.

group_as now accepts sized ranges whose values convert to uint32_t. It derives group counts and mapping validity from range values. Tests cover C arrays, cuda::std::array, and cuda::std::span; existing test call sites use arrays instead of index_sequence.

Changes: Group mapping

Layer / File(s)	Summary
Range-based group_as mapping cudax/include/cuda/experimental/coop/__group/fwd.cuh, cudax/include/cuda/experimental/coop/__group/mapping/group_as.cuh, cudax/include/cuda/experimental/coop/__group/synchronizer/interwarp_synchronizer.cuh	group_as stores a sized range whose values convert to uint32_t. It derives group counts and mapping validity from the range. Forward declarations and the interwarp synchronizer diagnostic use the _RangeView parameter name.
Range mapping test coverage cudax/test/coop/group/mapping/group_as.cu, cudax/test/coop/group/mapping/composite_mapping.cu	Tests cover exhaustive and non-exhaustive mappings with C arrays, cuda::std::array, and cuda::std::span. Composite mapping tests use array-backed counts and expect dynamic group counts.
Array-based group_as call sites cudax/test/coop/group/group_stacking.cu, cudax/test/coop/group/virtual_group_stacking.cu, cudax/test/coop/group/make_group.cu	These tests pass cuda::std::array values to group_as instead of index_sequence values.

Possibly related PRs

• NVIDIA/cccl#12229: Both changes apply a range-view template contract to cudax::coop synchronizer components.

Priority: Normal

Change: Feature

Merge Risk: Low · up to fffca

Several narrow range-input cases can fail to compile or produce incorrect group mappings; address them before relying on the new flexibility.

✨ Finishing Touches: 💡 1

🛠️ Fix failing CI checks: 💡

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

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

[coderabbitai]

Actionable comments posted: 1

Caution

Some comments are outside the diff and can’t be posted inline due to GitHub limitations.

⚠️ Outside diff range comments (1)

🟠 Major · important: Cast __n to uint32_t before the braced _MappingResult… · group_as.cuh:114-125

cudax/include/cuda/experimental/coop/__group/mapping/group_as.cuh:114-125
🎯 Functional Correctness | 🟠 Major | ⚡ Quick win

important: Cast __n to uint32_t before the braced _MappingResult initialization.

__n keeps the range element type. The _MappingResult aggregate initializes its uint32_t __unit_count_ field from __n in braces. When the range element type is int or size_t, that conversion narrows. Narrowing in list-initialization is ill-formed when the source value is not a constant expression, and __i_count is not one. The class static_assert admits any element type convertible to uint32_t, so group_as accepts these ranges, but __do_map cannot compile for them.

Suggested fix

-        const auto __n          = __i_count;
+        const auto __n          = static_cast<::cuda::std::uint32_t>(__i_count);

---

ℹ️ Review info

⚙️ Run configuration

• Source mode: Auto
• Repository: NVIDIA/cccl
• Branch: main
• File: .coderabbit.yaml
• Commit: 72a288d
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: 372375f7-8deb-4f21-a1c5-064901c0cdcb

📥 Commits

Reviewing files that changed from the base of the PR and between 72a288d and 4aab0d7.

📒 Files selected for processing: 8

• cudax/include/cuda/experimental/coop/__group/fwd.cuh
• cudax/include/cuda/experimental/coop/__group/mapping/group_as.cuh
• cudax/include/cuda/experimental/coop/__group/synchronizer/interwarp_synchronizer.cuh
• cudax/test/coop/group/group_stacking.cu
• cudax/test/coop/group/make_group.cu
• cudax/test/coop/group/mapping/composite_mapping.cu
• cudax/test/coop/group/mapping/group_as.cu
• cudax/test/coop/group/virtual_group_stacking.cu

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.

[coderabbitai]

Actionable comments posted: 1

---

ℹ️ Review info

⚙️ Run configuration

• Source mode: Auto
• Repository: NVIDIA/cccl
• Branch: main
• File: .coderabbit.yaml
• Commit: 4811a1c
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: c7d01e7a-5ed4-4fc0-9eb8-d5ed3112e252

📥 Commits

Reviewing files that changed from the base of the PR and between 4aab0d7 and ed846e1.

📒 Files selected for processing: 1

• cudax/include/cuda/experimental/coop/__group/mapping/group_as.cuh

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 9 remain after this review.

[coderabbitai]

Actionable comments posted: 1

---

ℹ️ Review info

⚙️ Run configuration

• Source mode: Auto
• Repository: NVIDIA/cccl
• Branch: main
• File: .coderabbit.yaml
• Commit: 4811a1c
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: e08bc12a-f117-4422-a40b-59b0a8de063d

📥 Commits

Reviewing files that changed from the base of the PR and between ed846e1 and fffca4e.

📒 Files selected for processing: 1

• cudax/include/cuda/experimental/coop/__group/mapping/group_as.cuh

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 8 remain after this review.

[github-actions]

🥳 CI Workflow Results

🟩 Finished in 1h 31m: Pass: 100%/73 | Total: 1d 02h | Max: 1h 14m | Hits: 34%/136590

See results here.