fix: cache config enum

[kaeun97]

This will be followed by a sequence of work on improving the l1 cache/shared memory configuration implementation as stated here.

This PR fixes the following bug:

Traceback (most recent call last):
  File "/home/gmarkall/numbadev/issues/discourse-3080/repro.py", line 13, in <module>
    cufunc.cache_config(prefer_shared=True)
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/cudadrv/driver.py", line 2418, in cache_config
    flag = attr.CU_FUNC_CACHE_PREFER_SHARED
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: type object 'CUfunction_attribute' has no attribute 'CU_FUNC_CACHE_PREFER_SHARED'

with a test that would ensure that this won't regress.

[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.

[greptile-apps]

Greptile Overview

Greptile Summary

Fixed AttributeError in cache_config() by correcting the enum type from binding.CUfunction_attribute to binding.CUfunc_cache.

• Changed line 2412 in driver.py:2412 to use the correct CUfunc_cache enum that contains cache configuration constants (CU_FUNC_CACHE_PREFER_SHARED, CU_FUNC_CACHE_PREFER_L1, etc.)
• Added comprehensive test coverage with test_cuda_cache_config that validates all four cache configuration modes
• Test ensures the kernel executes correctly after setting cache configuration

Confidence Score: 5/5

• This PR is safe to merge with minimal risk
• Simple one-line fix correcting an enum type error with comprehensive test coverage added. The change aligns with the existing CtypesFunction class pattern and directly addresses the reported bug
• No files require special attention

Important Files Changed

File Analysis

Filename	Score	Overview
numba_cuda/numba/cuda/cudadrv/driver.py	5/5	Fixed enum type from CUfunction_attribute to CUfunc_cache for cache configuration flags
numba_cuda/numba/cuda/tests/cudadrv/test_cuda_driver.py	5/5	Added comprehensive test for all cache config modes (prefer_shared, prefer_cache, prefer_equal, default) with functional validation

Sequence Diagram

sequenceDiagram
    participant User
    participant Kernel as CUDA Kernel
    participant CudaPythonFunction
    participant Binding as cuda.bindings.driver
    participant Driver as CUDA Driver
    
    User->>Kernel: kernel.overloads[sig]
    Kernel->>CudaPythonFunction: get_cufunc()
    User->>CudaPythonFunction: cache_config(prefer_shared=True)
    CudaPythonFunction->>Binding: CUfunc_cache.CU_FUNC_CACHE_PREFER_SHARED
    Binding-->>CudaPythonFunction: flag value (0x01)
    CudaPythonFunction->>Driver: cuFuncSetCacheConfig(handle, flag)
    Driver-->>CudaPythonFunction: success
    CudaPythonFunction-->>User: configuration applied
    User->>Kernel: kernel[grid, block](args)
    Kernel->>Driver: launch with cache config
    Driver-->>User: execution complete

Loading

[greptile-apps]

_{2 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

_{2 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[gmarkall]

/ok to test a23a794

[gmarkall]

/ok to test

[copy-pr-bot]

/ok to test

@gmarkall, there was an error processing your request: E1

See the following link for more information: https://docs.gha-runners.nvidia.com/cpr/e/1/

[greptile-apps]

_{2 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[gmarkall]

/ok to test 3c14514

[gmarkall]

Many thanks! I've checked with NSight Compute that setting the attributes takes effect.