[ivanmilevtues]

Motivation

Clang 23 adds __builtin_bitreverseg, which reverses unsigned integers without selecting a width-specific builtin. This implements the host-execution portion of the real upstream request NVIDIA#7779.

Change

• Prefer the generic builtin in the shared cuda::std::bit_reverse host implementation when feature detection reports it available. cuda::bit_reverse delegates to the same implementation.
• Preserve existing width-specific and portable fallbacks for other compilers. Device intrinsics and the constant-evaluation path are unchanged.
• Extend the existing public-API test with every single-bit position and its complement, including 128-bit integers where supported. Expectations come from an independent bit-by-bit reference.

Validation

• Compiled and ran both the standard and CUDA bit-reversal API tests with Clang 23 and GCC, in C++17 and C++20: eight passing combinations, including runtime and static assertions.
• Confirmed Clang 23 enables the generic-builtin feature macro.
• Repository pre-commit checks passed after formatting.
• No GPU execution, NVCC validation, full CCCL build, or performance benchmark was performed. This does not claim a speedup.

Showcase scope

This PR targets the CodeBoarding fork, not NVIDIA upstream. It is a real, independently reviewable implementation change prepared for a later CodeBoarding demonstration. Automated C++ PR analysis is deferred until the analyzer is released; no new analysis has been generated for this PR.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-03T21:02:26.775972Z	4d8e4ef	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[Svilen-Stefanov]

I don't see an analysis on this PR, can we add one and maybe do something like eShop here for C++ users?

[ivanmilevtues]

/codeboarding

[ivanmilevtues]

/codeboarding

[github-actions]

CodeBoarding review

Status: 0 changed components (no analysed file changed)

See the full change in CodeBoarding.

graph LR
    n_CUDA_Standard_C_Library["CUDA Standard C++ Library"]
    n_Experimental_CUDA_C_Extensions["Experimental CUDA C++ Extensions"]
    n_Parallel_Algorithm_Template_Library["Parallel Algorithm Template Library"]
    n_GPU_Collective_Compute_Primitives["GPU Collective Compute Primitives"]
    n_Python_Compute_and_Task_Bindings["Python Compute and Task Bindings"]
    n_C_Runtime_and_JIT_Interface["C Runtime and JIT Interface"]
    n_Continuous_Integration_and_Documentation["Continuous Integration and Documentation"]
    n_Build_and_Static_Analysis_Tooling["Build and Static Analysis Tooling"]
    n_CUDA_Standard_C_Library -- "references execution disposition types" --> n_Experimental_CUDA_C_Extensions
    n_CUDA_Standard_C_Library -- "invokes parallel algorithms" --> n_Parallel_Algorithm_Template_Library
    n_CUDA_Standard_C_Library -- "uses device runtime primitives" --> n_C_Runtime_and_JIT_Interface
    n_Experimental_CUDA_C_Extensions -- "uses synchronization barriers and views" --> n_CUDA_Standard_C_Library
    n_Experimental_CUDA_C_Extensions -- "dispatches parallel query operations" --> n_Parallel_Algorithm_Template_Library
    n_Experimental_CUDA_C_Extensions -- "invokes device intrinsics and state helpers" --> n_C_Runtime_and_JIT_Interface
    n_Parallel_Algorithm_Template_Library -- "delegates error termination and iterators" --> n_CUDA_Standard_C_Library
    n_Parallel_Algorithm_Template_Library -- "routes memory error handling" --> n_Experimental_CUDA_C_Extensions
    n_GPU_Collective_Compute_Primitives -- "queries device and execution context" --> n_Experimental_CUDA_C_Extensions
    n_GPU_Collective_Compute_Primitives -- "dispatches parallel kernel invocations" --> n_C_Runtime_and_JIT_Interface
    n_C_Runtime_and_JIT_Interface -- "uses standard memory management utilities" --> n_CUDA_Standard_C_Library
    n_C_Runtime_and_JIT_Interface -- "controls stream and task execution" --> n_Experimental_CUDA_C_Extensions
    n_C_Runtime_and_JIT_Interface -- "manages dynamic buffer storage" --> n_Parallel_Algorithm_Template_Library
    n_C_Runtime_and_JIT_Interface -- "applies type trait modernization rules" --> n_Build_and_Static_Analysis_Tooling
    classDef added fill:#1f883d,stroke:#0b5d23,color:#ffffff;
    classDef modified fill:#bf8700,stroke:#7d4e00,color:#ffffff;
    classDef deleted fill:#cf222e,stroke:#82071e,color:#ffffff,stroke-dasharray:5 3;

Loading

_{download artifacts · run 42613379981}

[github-actions]

CodeBoarding review

Status: 5 changed components

See the full change in CodeBoarding.

graph LR
    n_CUDA_Standard_C_Library["CUDA Standard C++ Library"]
    n_Experimental_CUDA_C_Extensions["Experimental CUDA C++ Extensions"]
    n_Parallel_Algorithm_Template_Library["Parallel Algorithm Template Library"]
    n_GPU_Collective_Compute_Primitives["GPU Collective Compute Primitives"]
    n_Python_Compute_and_Task_Bindings["Python Compute and Task Bindings"]
    n_C_Runtime_and_JIT_Interface["C Runtime and JIT Interface"]
    n_Continuous_Integration_and_Documentation["Continuous Integration and Documentation"]
    n_Build_and_Static_Analysis_Tooling["Build and Static Analysis Tooling"]
    n_CUDA_Standard_C_Library -- "references execution disposition types" --> n_Experimental_CUDA_C_Extensions
    n_CUDA_Standard_C_Library -- "invokes parallel algorithms" --> n_Parallel_Algorithm_Template_Library
    n_CUDA_Standard_C_Library -- "uses device runtime primitives" --> n_C_Runtime_and_JIT_Interface
    n_Experimental_CUDA_C_Extensions -- "uses synchronization barriers and views" --> n_CUDA_Standard_C_Library
    n_Experimental_CUDA_C_Extensions -- "dispatches parallel query operations" --> n_Parallel_Algorithm_Template_Library
    n_Experimental_CUDA_C_Extensions -- "invokes device intrinsics and state helpers" --> n_C_Runtime_and_JIT_Interface
    n_Parallel_Algorithm_Template_Library -- "delegates error termination and iterators" --> n_CUDA_Standard_C_Library
    n_Parallel_Algorithm_Template_Library -- "routes memory error handling" --> n_Experimental_CUDA_C_Extensions
    n_GPU_Collective_Compute_Primitives -- "queries device and execution context" --> n_Experimental_CUDA_C_Extensions
    n_GPU_Collective_Compute_Primitives -- "dispatches parallel kernel invocations" --> n_C_Runtime_and_JIT_Interface
    n_C_Runtime_and_JIT_Interface -- "uses standard memory management utilities" --> n_CUDA_Standard_C_Library
    n_C_Runtime_and_JIT_Interface -- "controls stream and task execution" --> n_Experimental_CUDA_C_Extensions
    n_C_Runtime_and_JIT_Interface -- "manages dynamic buffer storage" --> n_Parallel_Algorithm_Template_Library
    n_C_Runtime_and_JIT_Interface -- "applies type trait modernization rules" --> n_Build_and_Static_Analysis_Tooling
    classDef added fill:#1f883d,stroke:#0b5d23,color:#ffffff;
    classDef modified fill:#bf8700,stroke:#7d4e00,color:#ffffff;
    classDef deleted fill:#cf222e,stroke:#82071e,color:#ffffff,stroke-dasharray:5 3;
    class n_CUDA_Standard_C_Library,n_Parallel_Algorithm_Template_Library,n_C_Runtime_and_JIT_Interface modified;

Loading

_{download artifacts · run 42647988120}