Remove rmm::detail::polyfill in favor of CCCL concepts, require CCCL >=3.3

[bdice]

Description

Closes #2282

Remove the rmm::detail::polyfill namespace and delete rmm/detail/cuda_memory_resource.hpp. All usages are replaced with the CCCL concepts directly (cuda::mr::synchronous_resource, cuda::mr::resource, etc.). The CCCL version check is relocated to rmm/detail/export.hpp and updated to require CCCL >=3.3. (xref: rapidsai/rapids-cmake#981 -- also note that the CCCL 3.3 pre-release was the same as the final release)

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[coderabbitai]

📝 Walkthrough

Walkthrough

Removes the rmm::detail::polyfill header and replaces all uses of its traits with direct CCCL cuda::mr:: concepts; relocates the CCCL version check into export.hpp and updates includes and static_asserts across headers and tests (≈13 files).

Changes

Cohort / File(s)	Summary
Polyfill removal & export guard cpp/include/rmm/detail/cuda_memory_resource.hpp, cpp/include/rmm/detail/cccl_adaptors.hpp, cpp/include/rmm/detail/export.hpp	Deleted the polyfill header; removed its include from cccl_adaptors; added <cuda/version> and a CCCL >= 3.3 compile-time check in export.hpp.
Core MR headers cpp/include/rmm/mr/device_memory_resource.hpp, cpp/include/rmm/mr/system_memory_resource.hpp, cpp/include/rmm/mr/pinned_host_memory_resource.hpp, cpp/include/rmm/mr/cuda_async_managed_memory_resource.hpp	Replaced rmm::detail::polyfill::... trait checks with equivalent cuda::mr::... concepts and adjusted includes (added <cuda/memory_resource> where required).
Adaptor / trait utilities cpp/include/rmm/mr/is_resource_adaptor.hpp, cpp/include/rmm/mr/thrust_allocator_adaptor.hpp	Removed polyfill include; switched trait detection to cuda::mr:: concepts and added/ reordered CUDA memory-resource include(s).
Tests - core & mocks cpp/tests/device_scalar_tests.cpp, cpp/tests/mock_resource.hpp	Removed polyfill include, updated static_assert predicates to cuda::mr:: traits, and updated copyright years.
Tests - MR suites cpp/tests/mr/adaptor_tests.cpp, cpp/tests/mr/arena_mr_tests.cpp, cpp/tests/mr/cuda_async_mr_tests.cpp, cpp/tests/mr/cuda_async_view_mr_tests.cpp, cpp/tests/mr/host_mr_ref_tests.cpp, cpp/tests/mr/pool_mr_tests.cpp, cpp/tests/mr/system_mr_tests.cu	Systematically replaced rmm::detail::polyfill::resource* / async_resource* usages with cuda::mr::synchronous_resource* / cuda::mr::resource* equivalents and removed polyfill includes; minor header year updates and formatting tweaks.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Suggested labels

3 - Ready for review

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly describes the main change: removing the polyfill namespace and requiring a newer CCCL version.
Linked Issues check	✅ Passed	The PR fully addresses issue #2282 objectives: deletes the polyfill header, replaces all polyfill usages with CCCL equivalents across headers and tests, updates version requirement to CCCL >=3.3.
Out of Scope Changes check	✅ Passed	All changes are in-scope and directly support the primary objective of removing the polyfill namespace and requiring CCCL >=3.3.
Description check	✅ Passed	The PR description clearly describes the objective to remove the rmm::detail::polyfill namespace and replace all usages with direct CCCL concepts, matching the changeset.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

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

[coderabbitai]

🧹 Nitpick comments (1)

cpp/include/rmm/mr/system_memory_resource.hpp (1)

157-158: Add #include <cuda/memory_resource> to make this public header self-contained.

This header directly uses cuda::mr::device_accessible, cuda::mr::host_accessible, and cuda::mr::resource_with in its public interface (lines 146, 149, 153, 157–158) but relies on transitive inclusion from <rmm/device_memory_resource.hpp>. Making the dependency explicit prevents brittleness if includes are reorganized.

Proposed fix

 `#include` <rmm/detail/format.hpp>
 `#include` <rmm/mr/device_memory_resource.hpp>
 
+#include <cuda/memory_resource>
+
 `#include` <cstddef>
 `#include` <string>

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@cpp/include/rmm/mr/system_memory_resource.hpp` around lines 157 - 158, The
header is missing an explicit include for cuda::mr types used in its public API
(static_asserts referencing cuda::mr::resource_with,
cuda::mr::device_accessible, cuda::mr::host_accessible and the
system_memory_resource type); add `#include` <cuda/memory_resource> at the top of
the header so the declarations using resource_with, device_accessible, and
host_accessible are resolved without relying on transitive includes (ensuring
the public header is self-contained).

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@cpp/include/rmm/mr/system_memory_resource.hpp`:
- Around line 157-158: The header is missing an explicit include for cuda::mr
types used in its public API (static_asserts referencing
cuda::mr::resource_with, cuda::mr::device_accessible, cuda::mr::host_accessible
and the system_memory_resource type); add `#include` <cuda/memory_resource> at the
top of the header so the declarations using resource_with, device_accessible,
and host_accessible are resolved without relying on transitive includes
(ensuring the public header is self-contained).

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 4f4c3a58-7cd8-4b1e-ad2e-78556db69ed4

📥 Commits

Reviewing files that changed from the base of the PR and between 2ba0900 and 94f2ade.

📒 Files selected for processing (18)

• cpp/include/rmm/detail/cccl_adaptors.hpp
• cpp/include/rmm/detail/cuda_memory_resource.hpp
• cpp/include/rmm/detail/export.hpp
• cpp/include/rmm/mr/cuda_async_managed_memory_resource.hpp
• cpp/include/rmm/mr/device_memory_resource.hpp
• cpp/include/rmm/mr/is_resource_adaptor.hpp
• cpp/include/rmm/mr/pinned_host_memory_resource.hpp
• cpp/include/rmm/mr/system_memory_resource.hpp
• cpp/include/rmm/mr/thrust_allocator_adaptor.hpp
• cpp/tests/device_scalar_tests.cpp
• cpp/tests/mock_resource.hpp
• cpp/tests/mr/adaptor_tests.cpp
• cpp/tests/mr/arena_mr_tests.cpp
• cpp/tests/mr/cuda_async_mr_tests.cpp
• cpp/tests/mr/cuda_async_view_mr_tests.cpp
• cpp/tests/mr/host_mr_ref_tests.cpp
• cpp/tests/mr/pool_mr_tests.cpp
• cpp/tests/mr/system_mr_tests.cu

💤 Files with no reviewable changes (2)

• cpp/include/rmm/detail/cuda_memory_resource.hpp
• cpp/include/rmm/detail/cccl_adaptors.hpp

[bdice]

/merge